Advances in Electronics and Electron Physics, Volume 45

ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS

VOLUME 45

CONTRIBUTORS TO THISVOLUME

A. A. Barybin H. E. Bergeson George L. Cassiday Herbert F. Matare G. V. Trunk

Advances in

Electronics and Electron Physics DlTED BY

L. MARTON Sniirhsoiiicta Instilritioii, Wushinyron. D.C Assiwtitr Ecliror

CLAIRE MARTON EDITORIAL BOARD

T. E. Allibone E. R. Piore H. B. G. Casimir M. Ponte W. G. Dow A. Rose A . 0. C. Nier L. P. Smith F. K . Willenbrock

VOLUME 45

1978

ACADEMIC PRESS

New York San Francisco London

A Subsidiary of Harcourt Brace Jovanovich, Publishers

COPYRIGHT @ 1978, BY ACADEMIC PRESS, INC. ALL RIGHTS RESERVED. NO PART OF THIS PUBLICATION MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER.

ACADEMIC PRESS,INC.

111 Fifth Avenue, N e w York, N e w York 10003

United Kingdom Edition published by ACADEMIC PRESS, INC. (LONDON) LTD. 24/28 Oval Road, London N W I 7DX

LIBRARY OF CONGRESS CATALOG CARD NUMBER:49-7504 ISBN 0-12-014645-2 PRINTED IN THE UNITED STATES OF AMERICA

CONTENTS CONTRIBUTORS TO VOLUME 45 . . . . . . . . . . . . . FOREWORD . . . . . . . . . . . . . . . . . . .

vii ix

Electrodynamic Concepts of Wave Interactions in Thin-Film Semiconductor Structures I1 A . A . BARYBIN

.

IV . Boundary Conditions on Carrier Stream Surfaces in Nondegenerate Semiconductor Plasmas . . . . . . . . . . . . . . . . V . Normal Modes in Thin Semiconductor Films without Magnetic Field . VI . Conclusion . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . .

1 19 36 37

Light-Emitting Devices. Part 11: Device Design and Applications HERBERT F . MATARE 1 . Device Design and Technologies . . . References for Section 1 . . . . . . 2 . Progress in Laser Technology . . . . References for Section 2 . . . . . . 3. Device Types and Technological Progress References for Section 3 . . . . . . 4 . Measurement Techniques . . . . . References for Section 4 . . . . . . 5 . Survey of Important Applications . . . References for Section 5 . . . . . .

. . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

40 69 70 101 102 137 138 158 158 200

Radar Signal Processing G . V . TRUNK I. I1 . III . IV . V.

Introduction . . . . Coherent Processing . . Noncoherent Detection . Tracking System . . . Summary . . . . . References . . . . .

. . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . . . . .

203 204 224 236 249 250

On the Teaching of Electronics to Scientists H . E . BERGFSONAND GEORGEL . CASSIDAY

I . Introduction . I1 . Organization . 111. First Quarter . IV . Second Quarter

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V

253 258 260 278

vi V . Third Quarter . V1 . Conclusions . Appendix . . References . .

CONTENTS

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

AUTHoRINDEX . . . . . . . . . . INDEX . . . . . . . . . . . . SUBJECT 1-45 CUMULATIVE AUTHORINDEX. VOLUMES CUMULATIVE SUBJECT INDEX. VOLUMES 1-45 .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . 331 . 337 . . 344 . . 373

301 322 322 329

CONTRIBUTORS TO VOLUME 45 Numbers in parentheses indicate the pages on which the authors’ contributions begin

A. A. BARYBIN,Department of Electron-Ion Processing of Solids, V. I . Ulyanov (Lenin) Electrical Engineering Institute, Leningrad, USSR ( 1 )

H. E. BERGESON, Department of Physics, The University of Utah, Salt Lake City, Utah (253) GEORGE L. CASSIDAY, Department of Physics, The University of Utah, Salt Lake City, Utah (253) HERBERTF. MATARE,ISSEC, International Solid State Electronics Consultants, Los Angeles, California (39)

G. V. TRUNK,Radar Analysis Staff, Radar Division, Naval Research Laboratory, Washington, D.C. (203)

vii

This Page Intentionally Left Blank

FOREWORD

Starting the present volume is the second part of A. A. Barybin’s review on “Electrodynamic Concepts of Wave Interactions in Thin-Film Semiconductor Structures” (the first part appeared in Volume 44). It deals first with the boundary conditions on the side surfaces of thin-film structures, as well as on the interfaces between layers of such structures. The knowledge of these boundary conditions allows a theoretical analysis of the wave processes in nondegenerate semiconductor plasmas. The last part of the review considers the behavior of normal modes in such thin films without the presence of a magnetic field. Two years ago, in our 42nd volume, we published the first part of a review, “ Light-Emitting Devices,” by Herbert F. Matare. Whereas the first part was devoted to “Methods,” the second part, presented herein, is entitled: Device Design and Applications.” A short listing of the chapter headings illustrates best the general structure of the review : Device Design and Technologies, Progress in Laser Technology, Device Types and Technological Progress, Measurement Techniques, and Survey of Important Applications. In the next contribution, G. V. Trunk reviews “Radar Signal Processing.” He considers the three broad areas of coherent processing (processing of amplitude and phase), noncoherent processing (processing of amplitude), and track-while-scan systems. Coherent processing seems to Trunk the most promising and he expects the construction and testing of such systems in the near future. Noncoherent systems have been built and tested and little more research is expected in that direction. Track-while-scan systems, while needing a solution of track initiation in a dense environment, appear to offer advantages. The review ends by calling attention to the problem of adaptively controlling the surveillance radar. The last contribution to the present volume is, in a way, an innovation. We normally limit these volumes to the publication of critical reviews in the fields of electronics and electron physics. It may, however, be useful to have a look, for a change, at how electronics is taught. The review of H. E. Bergeson and George L. Cassiday entitled “On the Teaching of Electronics to Scientists” is not a critical review. The authors found too little material on which to base a critical review and decided to limit their account to their personal experiences. Their philosophy may not meet with universal approval and some teachers may not agree with their statement : “Let us accept the existence of a device, forget about what makes it tick, “

ix

X

FOREWORD

and explore its possible consequences.” For this reason a short editorial footnote is added to the beginning of their paper. Nevertheless, it is felt that the review is most valuable in presenting one approach to the problem of teaching electronics, because it may stimulate attempts at further improvements on which we may base some time later a critical evaluation. As usual we list here the critical reviews planned for our forthcoming volumes : The Lifetimes of Metastable Negative Ions I/r Sitrr Electron Microscopy of Thin Films High Injection i n a Two-Dimensional Transistor Physics of Ion Beams from a Discharge Source Physics of Ion Source Discharges Tcrininology and Classification of Particle Beams The Gunn Hilson Erect A Kcview of Applications of Superconductivity Minicomputer Technology Digital Filters Physical Electronics and Modeling of MOS Devices Thin-Film Electronics Technology Characterization of MOSFETs Operating in Weak Inversion Electron Impact Processes Sonar Microchannel Electron Multipliers Electron Attachment and Detachment Noise in Solid State Devices Electron-Beam-Controlled Lasers Amorphous Semiconductors Electron Beams i n Microfabrication. I and II Photoaco tist ic Spectroscopy Design Automation of Digital Systems. I and II Magnetic Liquid Fluid Dynamics Fundamental Analysis of Electron-Atom Collision Processes Auger Spectroscopy Electronic Clocks and Watches Review 01‘ Hydromagnetic Shocks and Waves Beam Waveguides and Guided Propagation Recent Developments i n Electron Beam Deflection Systems Seeing with Sound Wire Antennas Electron Microdiffraction

L. G. Christophorou A. Barna, P. B. Barna, J . P. Pocza and I.Pozsgai W. L. Engl G .Gautherin and C. Lejeune G .Gautherin and C. Lejeune B. W. Schumacher M. P. Shaw W. B. Fowler C. W. Rose S. A. White J. N. Churchill, T. W. Collins, and F. E. Holmstrom T. P. Brody R. J. Van Overstraeten S. Chung F. N. Spiess R. F. Potter R. S. Berry E. R. Chenette and A. van der Ziel Charles Cason H . Scher and G. Pfister P. R. Thornton A. Rosencwaig W. G. Magnuson and Robert J. Smith R. E. Rosensweig H. Kleinpoppen N . J . Taylor A. GnBdinger A. Jaumotte and Hirsch L. Ronchi E. F. Ritz, Jr. A. F. Brown P. A. Ramsdale J . M. Cowley

FOREWORD

Ion Beam Technology Applied to Electron Microscopy Microprocessors in Physics Time-Resolved Laser Fluorescence Spectroscopy The Edelweiss System A Computational Critique of an Algorithm for the Enhancement o f Bright Field Electron Microscopy Large Molecules in Space Recent Advances and Basic Studies of Photoemitters Application of the Glauber and Eikonal Approximations to Atomic Collisions Josephson Effect Electronics Signal Processing with CCDs and SAWS

XI

J . Franks A. J . Daviea J . Delpech J . Arsac

T. A. Welton M . and G. Winnewisser H. Tiinan F. T. Chan. W. Williamson. and G . Foster M. Nisenofi’ W. W. Brodersen and R. M. White

s l i / ~ / ~ l ~Iti,.J. i l l ~vo i l IW1IC.Y

linage Transmission Systems Computer Techniques for Image Processing i n Electron Microscopy High-Voltage and High-Power Applications o f Thyristors

W. K. Prati

W. G . Saxton C . Karady

All the volumes of A h ~ ~ a c t .ins Elrctmiics crnd Eltwrori PIiysic.~reflect the friendly cooperation of many people. Our best thanks go to the authors. to our editorial advisors, and, last but not least, to Academic Press, for their assistance in both the editorial and production aspects of this and all the other volumes.

L. MARTON C. MARTON

This Page Intentionally Left Blank

Electrodynamic Concepts of Wave Interactions in Thin-Film Semiconductor Structures. 11* A. A. BARYBIN Department of Electron-Ion Processing of Solids V . I . Ulyanov (Lenin) Electricai Engineering Institute Leningrad, U S S R

IV. Boundary Conditions on Carrier Stream Surfaces in Nondegenerate Semiconductor Plasmas . . . . . . . . . . . . ................................................ A. General Discussion .......................................... .... B. Semiconductor Surface Models. Real and Equivalent Surface Sources . . . . . . . . . . . C. Boundary Conditions for Particular Models of Carrier-Stream Surface.. . . . . . . . . ... V. Normal Modes in Thin Semiconductor Films without Magnetic Field A. General Discussion .............................................. ... .................. B. Dispersion Equations for Quasistatic Waves . . . . . . . . . . C. Longitudinal Propagation of Waves in Anisotropic Fil Zero Diffusion . . . VI. Conclusion ............................................................... References ....................................... ............

1 1

4 9 19 19 20 29 36 31

IV. BOUNDARY CONDITIONS ON CARRIER STREAM SURFACES IN NONDEGENERATE SEMICONDUCTOR PLASMAS A . General Discussion

Theoretical analysis of wave processes in thin-film structures requires a knowledge of boundary conditions on their side surfaces and interfaces between layers of the structures. The question of a correct choice of boundary conditions on carrier stream surfaces in semiconductor layers proved to be rather nontrivial. Attention was first paid to this by Kino (1965, 1970) studying carrier wave propagation in semiconductor plasmas with no diffusion. With taking into consideration a thermal motion of carriers most authors use, as a boundary condition, the constraint of zero normal current component on lateral surfaces of a crystal (n * j, = 0). Physically, this means Part I of this article was published in Volume 44. Numbering of text sections and equations continues from Part I. 1

2

A. A. BARYBIN

that the carriers, moving under the influence of ac forces, feel a real crystal surface which inhibits their movement normal to the surface and thus eliminates a normal current component (in the absence of surface state effect). This condition follows, in natural way, from the usual phenomenological concept of a mutual compensation of conduction and diffusion currents on a crystal surface. However, in the zero diffusion limit this boundary condition turns into a zero normal electric-field condition, which gives rise to an identically zero solution for a semiconductor film without static magnetic field (Barybin, 1975a). An explanation of this result can be only an incorrectness of the boundary condition n * j, = 0 in the zero diffusion limit. Indeed, with 9 = 0 the equation of carrier motion [written in the form (31) for the collisiondominated situation in a semiconductor when W T < 1 and k l < 13 determines a local connection between a current and an electric field. This means that for solving any boundary-value problem, it is enough to apply the ordinary electrodynamic boundary conditions (24)-(27), and the necessity to have an additional constraint (in particular, on n j,) fails. In case of 9 = 0, the solution to the boundary-value problem provides n j, # 0 for a semiconductor surface. The normal current component n j, delivers a charge from the bulk onto the semiconductor surface where it concentrates in the form of an infinitely thin sheet of mobile carriers with 9 = 0. The thermal carrier motion gives rise, as was shown by Sumi (1967), to a diffusion spread of this charge in an undersurface layer of a finite thinness (comparable to the Debye length) and thus provides the equality n j, = 0 directly on the crystal boundary (neglecting a surface state effect). To take into account the surface mobile charges when solving the boundary-value problems in the zero diffusion limit Kino (1965) suggested to use the Hahn boundary conditions (Hahn, 1939). These conditions were previously introduced for a free surface of electron beams in vacuum (Bobroff et al., 1962; Chodorow and Susskind, 1964). Kino (1968,1970) and after him other authors (Hartnagel, 1969; Hofmann, 1969b, 1972; Gueret, 1970~;Masuda et al., 1970) have applied the Hahn boundary conditions to analyze wave propagation and the instability of thin semiconductor films without diffusion. Attempts to include diffusion in thin-film semiconductor problems were made by Heinle (1971), Metz and Gandhi (1974), and Freire and Marcante (1975). The analysis of electrodynamic boundary conditions for plasma media with drifting carriers was carried out by Mikhailovskii and Pashitzkii (1965), neglecting diffusion and collisions. A collision influence on boundary conditions of Mikhailovskii and Pashitzkii was phemenologically made by Khankina and Yakovenko (1967) [see also Kaner and Yakovenko (197511. This analysis is related to the conventional operation of Maxwell's equations

-

-

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

3

integrated over a volume enclosing a transitional layer on the plasma boundary, and it determines the discontinuities of the electromagnetic field components. For free surface of gas plasmas, the obtained discontinuities (Mikhailovskii and Pashitzkii, 1965) define the Hahn surface sources. An application of these boundary conditions far semiconductor surfaces, as was made by Khankina and Yakovenko (1967), requires a special consideration because conditions on a crystal boundary differ from the free-surface model. The discrepancy between physical models used by Hahn and Kin0 was remarked by the author (Barybin, 1975a). Kino (1965, 1970) emphasized that the Hahn boundary conditions were reasonable approximations for a semiconductor when diffusion and surface recombination can be neglected. He did not give a detailed discussion of a semiconductor-surface model. However, as can be seen from his considerations, the physical situation corresponds to that wherein the surface of the carrier stream coincides with the real surface of the semiconductor. In this case, there is an obvious disagreement with the model within which the Hahn boundary conditions are valid. As is known (Hahn, 1939; Bobroff et al., 1962; Chodorow and Susskind, 1964), the Hahn boundary conditions were introduced for a free surface of electron beams in a vacuum admitting a transverse displacement of electrons. The transverse displacement of a beam boundary under the influence of ac forces is equivalent to an introduction of surface charges and currents on an unperturbed stream boundary (the Hahn surface sources). The appearance of the equivalent surface sources is a natural consequence of the polarization description of the hydrodynamic model of charge-carrier streams (Bobroff et al., 1962). On the contrary, in the Kino model (when the boundaries of a crystal and a carrier stream coincide) the crystal surface hinders an electron from undergoing normal displacements because for an electron inside the conduction band, the crystal surface is perceived as a potential barrier with height equal to the work function. Therefore in this model there can be only real charges and there must not be the Hahn equivalent surface sources. As was remarked in the preceding discussion, the appearance of a surface charges in this model (which will be referred to as the Kino surjace sources) is due to another physical reason, namely, to a charge accumulation in the form of an infinitely thin sheet as a result of a charge flow from semiconductor bulk in the zero diffusion limit. The generalization of the Hahn equivalent sources to the quasi-free carrier stream surfaces in semiconductors including the space-charge layers (depletion, accumulation, and inversion) near the semiconductor surface was carried out by the author (Barybin, 1975a) on the basis of the polarization description of a nondegenerate solid-state plasmas. Attention was first given to the free-surface assumption in a semiconduc-

4

A. A. BARYBIN

tor model by Dean with collaborators. Disagreement between their experimental data (Dean et al., 1970, 1973; Dean and Matarese, 1972) and the theoretical results obtained on the basis of the stiff-surface model (Heinle, 1971)led Dean and Robinson (1974) to reject the rigid-boundary constraint in favor of the free-surface assumption corresponding to the practical situation of a thin epitaxial film with depletion layers on the side surfaces, wherein the carrier stream boundary recedes from the crystal boundary and can be displaced in transverse direction. Diffusion, being a nonlocal phenomena, requires an additional boundary condition (to the Hahn boundary ones). Such a condition for the stiff-surfacemodel is n j, = 0. For the free-surface model an analogous condition was unknown to Dean. So Dean and Robinson (1974) have got rid of the necessity of an additional constraint on the transverse motion of the quasi-free carrier stream boundary, by using the quasi-equilibrium assumption that the charge-carrier density is taken to be the Boltzmann distribution. However, in the case of an R F excitation in the carrier stream, this approximation is not always correct and the use of it requires specific substantiation in each particular case. Therefore, in the general case, it is necessary to have an additional boundary condition on a quasi-free surface of the carrier stream subjected to the diffusion effect. Such a condition was proved by the author (Barybin, 1975a),analyzing the applicability of boundary conditions to different physical situations and determining the features related to the existence of diffusion, surface states and space-charge layers, and transverse displacement of an effective carrierstream boundary in semiconductor thin films.

B. Semiconductor Su$ace Models. Real and Equivatent Surface Sources The problem of the electronic structure and the electrical properties of real semiconductor surfaces is extremely involved even under steady-state conditions, quite apart from further complications due to nonequilibrium conditions in the presence of an ac excitation. Thus a task is to choose a fairly simple and satisfactory surface model which would allow one to take into account the influence of charge in surface layers (not obeying the bulk model) on the small-signal behavior of the charge-carrier stream inside the semiconductor bulk. As far as is now known (Many et al., 1965), in the vicinity of the real crystal surface there are localized surface states arising, firstly, because of the existence of the surface as a two-dimensional defect even for the termination of an ideal crystal lattice (Tamm and Shockley states) and, secondly, from proper defects and impurities in or upon the real surface (or within an absorbed layer of foreign material on the semiconductor surface). These

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

5

surface states are associated with discrete (or continuously distributed) energy levels lying in the forbidden gap of the energy band diagram of the crystal. They may, in principle, act as traps, recombination centers, or both to yield a change of the mobile carrier concentration in the underlying semiconductor bulk region, as a result of their charge exchange with the allowed bands. The surface recombination is characterized by a phenomenological parameter s which is usually referred to as the surface recombination velocity and connects the recombination component of the normal charge flow (n jl)recwith the nonequilibrium (excess) volume density p1of mobile carriers at the surface by the relation (Many et al., 1965):

-

(n * jlLec= S P ~ .

(92) The charged surface states (traps), together with an electric field outside the semiconductor (established by an external source or by the proximity of another solid with a different work function), may produce space-charge layers near the semiconductor surface: a depletion layer (with the dearth of the majority carriers), an accumulation layer (with the excess of the majority carriers), or an inversion layer (with the excess of the minority carriers). In the case of considerably depleted layers the carriers recede from the crystal surface, so that the carrier stream boundary does not coincide with the real semiconductor boundary. In the case of accumulation (or inversion) layers where there are mobile charge carriers the situation is more complicated. The properties of the mobile charges differ from those of the bulk mobile charges because the distribution of carriers in the band may become degenerate, with considerable bending of the band edge within accumulation (or inversion) layers. Besides that, the crystal boundary affects predominantly the carriers moving close to the surface to give rise to additional scattering beyond the normal bulk scattering. Thus the carrier mobility in accumulation (or inversion) layers is generally less than that in the bulk (Many et al., 1965).

For these reasons we consider the carriers in accumulation (inversion) layers moving parallel to the crystal surface as a two-dimensional stream of charged particles with charge q, (the sign corresponds to that of the majority or minority carriers in layer), effective mass m, # m (as a result of a possible degeneracy), and mean free time t s# T (as a result of an additional surface scattering), so that the surface mobility p, = (qs/m,)t,= p. Moreover we assume, for generality, that the carriers taking part in the two-dimensional movement beneath the surface are subjected (in analogy with a bulk movement) to chaotic thermal motion characterized by the thermal velocity uTs (or surface diffusion constant 9, = u:, t,).The directed carrier motion inside a surface layer along the crystal boundary proceeds under the force of the tangential component of the electric field E, and of the normal component of

6

A. A. BARYBIN

the magnetic field B, (which are always continuous on any semiconductor surface). Then the Lorentz force F, = q,(E, + v, x B,) exerted on a carrier moving along the boundary with a velocity v, by electromagnetic fields, is parallel to the boundary surface and does not lead a carrier out of the surface layer. We neglect the transverse motion of carriers in accumulation (inversion) layers. This is fairly valid because of the existence of a strong normal dc electric field (arising as a result of band bending) which suppresses the normal mobility of carriers moving under the influence of ac forces. Consequently, rather than consider in detail the complex behavior of charges in surface states and layers, we can replace them by surface sources introduced on an effective boundary of the carrier stream which separates the semiconductor bulk from the surface layers (Fig. 1). The effective boundary

FIG 1. Schematic diagram of the surface model of a semiconductor with space-charge layers on the crystal surface, (a) without and (b) with an ac perturbation of the quasi-free emective boundary of carrier stream: (1) real crystal boundary with surface states; (2) surface space-charge layer; (3) unperturbed boundary of carrier stream; (4) perturbed boundary of carrier stream; ( 5 ) semiconductor bulk; (6) effective boundary of carrier stream with real surface sources ( p , , j,); and (7) effective boundary of carrier stream with real ( p , , L ) and equivalent (pzq, j:q) surface sources.

of the carrier stream can coincide with the real crystal boundary or can recede from it, being quasi-free in the case of space-charge layers on the real crystal surface. The first case corresponds to the stiff-surface model and the second case corresponds to the free-surface model. The choice of position of the quasi-free effective stream boundary is certain to be arbitrary to some extent. This arbitrariness, however, gives us the feasibility of a flexible

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

7

approach in each particular case. Introduced in this way the surface sources: the surface charge density p\ and the surface current density jsare characterized by such phenomenological parameters as qF,m,,t, (or p,), vTP (or 9,) which are analogous to the corresponding bulk parameters. These surface sources located on the effective boundary of the charge carrier stream are completely adequate to the real surface states and layers located outside this effective boundary [Fig. l(a)]. Thus they will subsequently be referred to as the real surface sources ( p s , js). Transverse ac perturbation of the effective boundary of a carrier stream as a result of ac excitation effect is taken into account in the surface model under consideration (Barybin, 1975a) with the help of the equivalent surface sources (pEq,j:q) which, like the real surface sources ( p s ,j,), are also located on the unperturbed effective boundary [Fig. l(b)]. In the general case, the real surface sources can contain both dc ( p s o ,jso) and ac or small-signal (pI1,j.l) components while the equivalent surface sources are small-signal ones according to their definition. The real and equivalent surface charges and currents, introduced in such a way, enter ordinary electrodynamic boundary conditions. These boundary conditions are derived from Maxwell's equations (1)-(4) by using the conventional procedure of their integration over a volume element V (which contains two media in contact with a portion AS of the boundary surface) and subsequent shrinking it to a point on the boundary surface (Fig. 2). The results of this procedure are (Barybin, 1975a)

where the superscripts '' " and '' - " mean the fields values taken at the points lying on different sides of the boundary surface which are marked by the inward normal unit vectors 'n (Fig. 2). The surface densities of charge pSland current jsl are defined by the following limiting equalities (Barybin, 1975a): +

1 .

pSl = lim v-0

I p1 d V

AS.V

I

=

1 . jsl = lim - j, d V = v-0

AS.v

lim ( p : An+

+p;

An-)

All+-+O Atn--O

lim (j: An' hf-0

+ j;

An-)

(97)

(981

8

A. A. BARYBIN n* ..

i

n ..

Insulator medium with no mobile charges

(p;=o,)I+=o,pl+=o)

volume V with

effective boundary

t n-

i

-n

semiconductor medlum with mobile charges ( p ; = p , .1 - 3 j,, P,- =I

FIG.2. Schematic illustration of the volume containing two media in contact, as used to derive the boundary conditions.

Application of Eqs. (97) and (98) to (43)and (44), giving the polarization representation of the Eulerian variables p1 and jl, reduces to the equivalent surface sources in the form (Barybin, 1975a), pEq = n p1 = n . porl j:q

= p:qvo = (n po r )vo 3

where n = - n- is the outward unit vector normal to the effective boundary of a carrier stream (Fig. 2) and rl is the small-signal displacement vector on the unperturbed-stream boundary. The expressions (99) and (100) exactly coincide with those first given by Hahn (1939) and treated for vacuum electron beams (Bobroff et al., 1962). These expressions are derived from the original formulas (43)and (44)which are valid in the general case of nondegenerate semiconductor plasmas (Barybin, 1970). Thus we can conclude that the equivalent surface charge density pEq= n p1 and the equivalent surface current density jzq= (n * pl)vo are always determined only by the transverse carrier displacement (n * r l ) on the quasi-free carrier-stream boundary, either in the absence or in the presence of the bulk processes of collisions and diffusion. In the zero diffusion (9+ 0) and zero real surface charge (psi = jsl = 0) limits equations (93k(96) with (99) and (100) are completely sufficient to solve any boundary-value problem for the free-surface model. Otherwise, it is necessary to have additional boundary conditions taking into account the influence of diffusion and real surface charges which are derived on the basis of the continuity equation (35) and the equation of motion (36) (Barybin, 1975a) and analyzed in Section IV,C.

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

9

C . Boundary Conditions for Particular Models of Carrier-Stream Surface

To derive additional boundary conditions, it is necessary to apply the preceding conventional procedure (the integration over a volume element V crossing an interface subsequently shrinking it to zero) to the continuity equation (35) (rewritten in the abbreviated form as C = 0) and the equation of motion (36) (rewritten in the abbreviated form M = 0, wherein all the terms are transferred in the left-hand side of the equation). The result of applying this mathematical operation to the continuity equation (35) assumes the form of a new boundary condition containing both the real and equivalent surface sources (Barybin, 1975a):

n jl

= C,

+ CEq + sp,

(101)

where, in analogy with the continuity equation C = 0, the notation

is introduced (V, is the two-dimensional vector differential operator). The boundary condition (101) expresses the law of charge conservation on the effective stream boundary. The charge delivered onto the effective boundary of the carrier stream out of the semiconductor bulk by the normal component of a bulk current (n * j,) is, firstly, transformed into a real immobile (in traps) or mobile (in accumulation or inversion layers) surface charge: (n jdreal= Cs (103) Secondly, it supplies the transverse displacement of the effective boundary of the stream, (n * j i )dispI = Cq (104) The third part of it may ultimately disappear owing to surface recombination [see Eq. (92)]:

(n * jlIrec= S P , (105) The case of an effective stream boundary with no surface charges (C, = 0) and without surface recombination (s = 0) corresponds to a practical situation with depletion layer on the crystal surface. In view of Eqs. (99), (loo), and (102), the boundary condition (101) assumes the form

10

A. A. BARYBIN

Using the expression of jl in the polarization variables (a), it is easy to see that the equality (106) represents an identity (Barybin, 1975a). Consequently, the boundary condition (101),obtained from the continuity equation, does not impose any additional constraint on the movement of the quasi-free stream surface on the boundary with the depletion layer (when Cs = s = 0). Such a constraint should be obtained from the equation of motion. When a carrier-stream boundary coincides with a real crystal surface without surface-charged layers the crystal surface interferes with a transverse displacement of carriers on the boundary so that n * r, = 0. This means the absence of the equivalent surface sources (pEq = n * p, = 0 and jzq = (n * pl)vo = 0). If on the crystal surface there are localized surface states in the form of traps and recombination centers carrying electric charges then the boundary condition (101) rewritten in the form

-

n j, = C ,

+ sp,

(107)

must be supplemented with equations of kinetics of filling and emptying the surface states (Many et al., 1965). If one attempts to express the left-hand side of (107) in terms of p1 with the help of (44),then one obtains a result disagreed with the condition n * p1 = 0 valid for the real crystal surface. Indeed, a term n j,, calculated with the help of (44),is determined by (106) that gives n * jl = 0 when n * p1 = 0. This contradicts to the boundary condition (107). The found contradiction allows one to conclude that the polarization representation of current in the form (44) is not able, in principle, to take into account the processes of charge disappearance on a surface related to the transition of charges into new states (at the surface traps or recombination centers). The Eulerian current representation (37) is correct in the general case, while the polarization current (44) is a component (104) of the total current which accurately reflects the transverse displacement of the effective stream boundary without consideration of the charge transition processes on a surface. It is completely evident that the one boundary condition (101) is not quite sufficient to separate the total normal current component n * jl into the three components (103), (104), and (105).For this purpose it is necessary to have extra equations, in particular, the boundary conditions derived from the equation of motion (36) in the Eulerian variables. The application of the preceding integrating and limiting operation to Eq. (36) gives a new boundary condition (Barybin, 1975a):

-

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

11

with the notation M, + (vw - VO)C, - SP, vo (10% Here pro and v,,, are dc (unperturbed) values of the charge density and velocity for real surface charges, respectively. These values, together with ac (small-signal) values psl and uql, determine the ac surface current density

s=

P\O

icl = ps0uS1+ pClv,.

(110)

The quantity M,, appearing in (lW), is a surface analog of the bulk equation of motion (36) written in the form M = 0. Analogously, the equality M, = 0 represents the surface equation of motion for real surface charges (in the form of an infinitely thin charged sheet) in the absence of their coupling to a semiconductor bulk. For the collision-dominated case of the surface-charged sheet (when oz, < 1 and kl, 4 1) the general expression of M, (see Barybin, 1975a) gets simplified so that the surface equation of motion M, = 0 gives an expression for u , ~in the form (Barybin, 1975a) us1 =

Ps(El., + Vso x B1.J + u,1 x PsBo,n - 9 s -

VSPSl

Pr,

(111)

analogous with a corresponding bulk expression. The boundary condition (108) connects the bulk charge density p 1 on the effective stream boundary with properties of real surface charges taken into account by the value S. In the presence of the real surface charges there is usually a bending of energy bands so that, generally speaking, on the effective stream boundary we have ap, / a n # 0. However, in most practical situations this surface can be assumed to be homogeneous, i.e., Vs(dpo/an)= 0. In this case, Eq. (108) assumes the simplified form

v: v , p ,

=

1 dP0 --s Po an

The boundary conditions (101) and (1 12), together with the electrodynamic conditions (93)-(96), are employed for different physical situations distinguishing by a degree of the influence of diffusion, surface states and charged layers, and transverse displacement of a carrier stream boundary. The boundary conditions for different particular cases are below analyzed. 1. Real Crystal Boundary with N o Surjace Sources and without Surjace Recombination This case corresponds to the physical situation in which the carrier stream boundary coincides with an ideal crystal surface with no traps and recombination centers (psi = 0, jpl= 0, s = 0). Because the crystal surface

12

A. A. BARYBIN

-

impedes a normal carrier motion the condition n rl = 0 takes place on the boundary. Then the electrodynamic boundary conditions (93)-(96) are n x E: = n x E; nx

H:

= n x H;

n

D:

=n

(113)

- D’;

n.B:=n.B;

(116) and (101)gives the boundary condition on a normal component of bulk current n.jl=O (117) which corresponds to a completely rejecting boundary. The additional condition obtained from the motion equation which connects a movement of the carrier stream boundary with real surface charges is not required in this case because both the former and the latter are absent. The boundary condition (117) “works” only with nonzero diffusion (9# 0) as an additional constraint arising as a consequence of a nonlocal effect of diffusion. In the zero diffusion limit (52 + 0) the necessity of the additional condition (1 17) vanishes. But the electrodynamic boundary conditions (113)-(116) turn out to be incorrect in this limit. This proceeds from that an mobile space charge, concentrated within the Debye length close to crystal surface under 9 # 0, is compressed into an infinitely thin sheet in the zero diffusion limit. This charged sheet plays the role of surface sources in corresponding boundary conditions. Consequently, the boundary conditions (1 13b(117) are valid only with 9 # 0. The zero diffusion limit can be realized in subsequent case.

2. Real Crystal Boundary with Surface Sources and Surface Recombination This case differs from the previous one only by the presence of charges on the semiconductor surface which, as before, impedes a transverse carrier displacement, i.e., n r l =.O and p S l # 0, jsl # 0, s # 0. Surface sources on a crystal boundary (neglecting an effect of depletion, accumulation, or inversion layers) can arise in the following physical situations:

-

a. localized surface states (traps and recombination centers) carrying electric charges which can be exchanged with the semiconductor bulk; b. infinitely-thin charged sheet of mobile carriers due to surface accumulation of charge in the limit of negligible diffusion; c. infinitely-thin charged layer of mobile carriers due to the Hall effect in the presence of a static magnetic field.

13

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

For all three situations the boundary conditions on components of electromagnetic field (93)-(96) are

n x E: = n x E; n x H:

=n x

(118)

+ jsl

H;

(119)

n-D:=n-D;+p,,

(120)

n.B:=n.B; (121) and the boundary condition (101) on a normal component of bulk current is written as

The system of boundary conditions (118)-( 122) is not complete because it contains two extra variables psl and jsl coupled by one equation (122). An additional couple must depend on the preceding physical situation. a. Localized charged-surfacestates. Usually, the conductivity along crystal surface due to electron transition between localized surface states is negligibly small so that jsl = 0. In order to express the dependence of charge at surface states psl on “external” excitation, it is necessary to use equations for the kinetics filling and emptying the surface states (Many et al., 1965). However, in practice the relaxation times of surface states are much more than a period of ac signal at the microwave-frequency range so that psl = 0. Finally, neglecting of surface recombination (s = 0) the boundary conditions (118)-(122) turn into (113)-(117), and the influence of localized surface charges can be neglected. b. Surface sheet of mobile charges in the zero diffusion limit with no magnetic field. In the zero diffusion limit, the additional boundary condition (112) reduces to the equality S = 0 which, in view of (102), (109), and (122) with s = 0, assumes the form psoMS+ (vw - vo)(n

i l )= 0

(123) The dc surface charge is assumed to be equal to zero in the absence of ac excitation, that is, p s o = 0. The ac surface charge psl appearing on the crystal surface under negligible diffusion ( 9= 0) can move along the surface with a velocity vso. According to (123) with p s o = 0, the value v, is equal to the bulk carrier drift velocity vo . Then the surface current density (110) is *

(124) Equations (122) and (124) give a connection between psl and n j, in the form (with s = 0): Js1

= Psl

VO

-

14

A. A. BARYBIN

While studying a wave process propagating along the direction of homogeneous carrier drift (vo = e,, uo = const)'with a wave factor exp(iwt - yz), the surface charge density psl is expressed in terms of n * j, by means of Eq. (125) as

Expression (126) was just used by Kin0 (1965, 1968,1970) in the boundary condition (120) which he called the Hahn boundary condition. As was remarked, the physical situation under consideration does not correspond to the Hahn model. The surface charge psl in the form (126) (subsequently, referred to as the Kino surface charge) arises really on the crystal surface only in the negligible diffusion case. The Hahn surface charge is equivalent one, as distinct from the Kino charge, which is artificially introduced on a quasi-free unperturbed surface of carrier stream as a result of its transverse ac displacement. As can be seen from comparison (126) and (139), both surface charges (Kino and Hahn) coincide with each other only in form having completely different physical meanings. c. Surface sheet ofmobile charges in the zero dgusion limit with magnetic Jield. As is well known, a static magnetic field Bo leads to the appearance of the dc Hall electric field Et = -vo x Bo . The Hall field induces on lateral sides of a semiconductor sample a dc surface charge with the density

Eg) = e(v0 x Bo) n (127) where E is the permittivity of semiconductor, n is the outward unit vector normal to a semiconductor side surface. As distinct from the previous case (where pso= 0), the dc surface charge psomoving along the surface with an ac velocity uslgives a contribution to the ac current (110). If the ac surface charge psl is assumed (as before) to move with the velocity vso = v o , then (123) with pso # 0 yields the equation of motion of surface charge M, = 0, which is necessary to find usl. The expression for uS1is given by (111) which can be transformed neglecting surface diffusion (gS = 0) to the form (Barybin, 1975a) pso= --E(n

us1

= Pr

*

(E1,Z + vo x

B1.J

(128)

where

is the surface mobility tensor in the presence of dc magnetic field; b, = pS(n Bo). The substitution (128) into (110) gives

-

isl = 4 * (El,T+ vo

x B1.J + pS1vo

(130)

15

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. 11

where

is the induced (Hall) surface-conductivitytensor which turns into the scalar ~ ps0p5on surfaces parallel to a dc magnetic field (where quantity C T = n * Bo = 0 and b, = 0), and which is equal to zero on surfaces perpendicular to a dc magnetic field (where n x Bo = 0 and ps, = 0). Equations (122) and (130) give a connection between pSl and n j, in the form (with s = 0):

-

-

ap,, V, P,~V,, V, [cry * ( E l . r+ vo x Bl,J] (132) at For a wave process propagating along the direction of homogeneous carrier drift (vo = eZuo = constant), Eq. (132) yields the expression

n j,

=

~

+

1

+ -

From the comparison of (126) and (133), it is seen that a dc magnetic field can give an additional contribution to the ac surface charge owing to an induced surface conductivity. The electrodynamic boundary conditions (1 18)-( 121) with surface sources j,l and psl written in the form of (124) and (126) (for the case in Section IV,C,2,b) or of (130) and (133) (for the case in Section IV,C,2,c), constitute the complete system of boundary conditions in the zero diffusion limit. In the case of the quasistatic approximation (V x El 'v- 0) this system is reduced to Eqs. ( 1 18) and (120).

3. Ejfective Boundary of' Carrier Streams with No Real Surface Sources and Without Surjace Recombination This case corresponds to the physical situation in which there is a depletion layer on a crystal surface so that the effective stream boundary is removed from the crystal boundary and can accomplish transverse oscillations. In this case n * rl # 0 and psl = 0, jsl = 0, s = 0, and the equivalent surface sources pzq and j:" are defined by Eqs. (99) and (100). Then the boundary conditions (93)-(96) are nxE:=nxE; n x H: = n x

-

H; + (n pl)vo

n.D:=n-D;

+n.p,

n.B:=n.B;

(134) (135) (136)

(137)

16

A. A. BARYBIN

-

In order to express the equivalent sources p:q and jEqin terms of n j, let us take advantage of the expression (44) for j, valid in all the points of carrier stream including its effective boundary either in the presence of or in the absence of diffusion. The final result is (Barybin, 1975a) n * j* l =-a(n ' p l ) at

+ (vo - V)(n

pl)

+ (V

*

-

vo)(n p l )

(138)

For a wave process propagating along the direction of homogeneous carrier drift (vo = e, uo = constant), Eq. (138) yields the expression,

Equation (139) defines the Hahn (equivalent) surface charge and coincides in form with the analogous expression (126) for the Kin0 (real) surface charge. As is now clear from the above, the Hahn and Kin0 surface charges, in spite of their similar forms, correspond to completely different physical situation. The Kino surface charge corresponds to the stiff-surface model of stream boundary and appears on a real crystal surface only with negligible diffusion. The Hahn surface charge corresponds to the free-surface model and arises on the unperturbed effective stream boundary (distant from crystal boundary), as a result of its ac transverse displacement, always independently of a degree of diffusion effect (i.e., either with 9 = 0 or with 9 # 0). For the solution of boundary-value problems on the basis of the freesurface model with 9 = 0 use is made of the boundary conditions (134)(137) with surface sources determined with the help of (139). Comparison of them with the analogous equations (118)-(121), (124) and (126) for the stiff-surface model with 9 = 0 shows their complete identity. This ascertains that the two models are indistinguishable in the zero diffusion limit. Nonzero diffusion eliminates the " degeneracy " of these models. Taking into account nonzero diffusion requires an additional boundary condition. As was previously ascertained, the boundary condition (101) derived from the continuity equation turns into an identity and does not impose any constraints on the movement of a quasi-free stream surface. The necessary boundary constraint is given by Eq. (112) obtained from the bulk equation of carrier motion. In the absence of real surface sources and surface recombination so that S = 0, Eq. (112) yields with uT # 0, VSPl

=0

(140)

-

For a boundary with one-dimensional variation of all the physical quanyp Then Eq. (140) tities in the wave propagation direction we have V, p

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

17

gives a necessary boundary condition on the quasi-free effective stream surface with 9# 0 in the form P1 = o (141) This boundary condition is opposite to the condition n - j, = 0 for the stiff-surface model, i.e., it corresponds to a completely absorbing boundary. The physical situation is such that all the ac charge p 1 which flows from the bulk onto the unperturbed effective stream boundary ultimately vanishes there owing to its conversion into the equivalent surface charge piq due to the transverse displacement of the quasi-free effective stream boundary. The boundary condition (141), in essence, is analogous to the condition on a surface of the ideal adsorbent for which the adsorption rate is much more than the rate of substance transport from an external medium: the adsorbent absorbs all substance transported to its surface independently of substance mass supporting thereby the zero mass density on its surface. The boundary condition (141) did not meet in electronics up to now and was first obtained by the author (Barybin, 1975a). It was applied to the analysis of normal mode problem in a semiconductor film without magnetic field in the quasistatic approximation (Barybin, 1975b). In this case (V x El z 0), it is sufficient to use the boundary conditions (134) and (136) with the equivalent surface charge (139) which, with 9 # 0, are supplemented by the constraint p1 = 0. As distinct from the stiff-surface model with the constraint n j, = 0, a solution for the zero diffusion case is obtained from the general solution for 9 # 0 as a result of limiting operation 9 + 0. 4. Effective Boundary of Carrier Streams with Real Surface Sources and Surface Recombination This case corresponds to the physical situations in a semiconductor crystal with accumulation (or inversion) layers which can exchange charges with localized surface states (traps and recombination centers) on the crystal boundary. The effective boundary of drifting carrier stream is removed from the real crystal surface because of the presence of a charged layer and can undergo, in general case, transverse oscillations. Then n rl # 0, psl # 0, jsl # 0, s # 0 and the equivalent surface sources pZq and jZq are defined by Eqs. (99) and (100). Then the electrodynamic boundary conditions (93)-(96) assume the most general form: n x E: = n x E; (142) n x H: = n x H; + jsl + (n pl)vo (143)

18

A. A. BARYBIN

In this case the additional boundary condition, obtained from the continuity equation, '' works " without any simplification in the general form (101), namely n * j, = C ,

+ CZq+ sp,

where C, and CZq are given by (102). The value CZq contains a transverse displacement of effective stream boundary n rl which gives in the presence of real surface charges, as was above remarked, only a partial contribution in the value n j,. So it is necessary to express n * r, in terms of Eulerian variables. The result of transition from n * rl to p , gives (Barybin, 1975a)

-

The second boundary condition (112) derived from the equation of motion is used for the replacement of (vo * S) in (146) by p , . Then the substitution of Eq. (146) thus transformed into (101) gives the boundary condition, expressing a connection between the real surface sources (psi, jsl) and the bulk values ( p , , jl), in the following form (Barybin, 1975a):

This boundary condition differs from the analogous one (122)for the real crystal boundary with surface sources and surface recombination only by the presence of the additional term in brackets on the left-hand side of (147), depending on p , . Consequently, this term takes into account the transverse movement of the quasi-free effective stream boundary. For a wave process propagating along the direction of homogeneous carrier drift (vo = ezvO= constant), Eq. (147) yields (with s = 0) PSI

=

n * jl

- vs * [4' (El,

+ _______ vso x B1, + (iw - ~ u o ) [ ~ o / ( d ~ o / J n ) l p 1 iw - yuso

(148) where for jsl use was made of (130) neglecting surface diffusion (9,= 0). The formula (148) is similar to (133) for the real crystal boundary in the zero diffusion limit but differs from (133) by the extra term in numerator contain1% P1.

The equivalent surface charge density pzq = n p, appearing in (143)and (144) is also expressed in terms of p1 with the help of the following relation (Barybin, 1975a):

WAVE INTERACTIONS IN THIN-FILM

SEMICONDUCTORS. I1

19

Thus Eqs. (142)-(145) with the real and equivalent surface charges in the form (148) and (149) form the complete system of boundary conditions. Equations (147)-(149) contain po and d p o / d n which are found from the static (dc) solution for a semiconductor surface with an accumulation (or inversion) layer. the simplest approximation po(x)= po exp( -x/lD), wher:cm% is the Debye length and the x-axis (ex = n) is directed outside the effective stream boundary, we have Po --

aPo/an

-

-ID

This relation can be substituted into (147)-(149) and, in particular, from Eq. (149) one can find that pEq = n p, = l D p , (150) In the zero diffusion limit ( l D - O ) , Eq. (150) yields pzq+O while Eqs. (147) and (148) for the model under consideration turn into the corresponding formulas (122) and (133) for the stiff-surface model of stream boundary coinciding with a real crystal boundary. The difference between the two models consists in that the dc surface charge psoarises in the given case as a result of a surface accumulation of majority (or minority) carriers at the expense of bending of energy bands while in the stiff-surfacemodel the charge psowas produced by the Hall effect. In the given case with vT -+ 0, the effective stream boundary approaches to the crystal boundary (ID -,0) and its transverse oscillations are suppressed -+ 0) so that the physical situation begins to correspond to that realized in the stiff-surface model with surface sources (see Section IV,C,2).

-

V. NORMAL MODESIN THINSEMICONDUCTOR FILMS WITHOUT MAGNETIC FIELD A . General Discussion

Experimental results on the suppression of Gunn oscillations in n-GaAs samples by reduction of their thickness (Kumabe, 1968; Schlachetzki and Mause, 1972) or by dielectric surface loading (Hofmann, 1969a; Kataoka et al., 1969; Kuru and Tajima, 1969; Hofmann and 'tLam, 1972; Anderson and Robinson, 1974) have showed a practical possibility to stabilize an electric field profile in n-GaAs thin films biased above the transferred-electron threshold (Dean, 1972a; Dean and Schwartz, 1972). Epitaxial layers of n-GaAs with stable field profile were used for design of thin-film semiconductor traveling-wave amplifier (Dean, 1969, 1972b; Dean et al., 1970, 1973; Dean and Matarese, 1972; Bianco et al., 1973; Kanbe et al., 1973; Dean and Robinson, 1974; Fleming, 1975).

20

A. A. BARYBIN

A theoretical analysis of processes in such semiconductor structures (Kino, 1965, 1968, 1970; Kino and Robson, 1968; Hartnagel, 1969; Hofmann, 1969b, 1972; Gueret, 1970b; Masuda et al., 1970) compels to reject the one-dimensional model (Gueret, 1970a; Engelmann, 197la) and to take into consideration the effects due to finite transverse dimensions of an active layer. Apparently, the first such work was that of Kino (1965). Later, Kino (1968, 1970) has given a detailed treatment of the propagation of space-charge waves (called by him as currier waves) in nondegenerate semiconductor plasmas including finite-thickness effects but neglecting the diffusion effect. Attempts to include diffusion in analysis were undertaken by Blotekjaer (1970),Heinle (1971),and Freire and Marcante (1975). Blotekjaer (1970) has pointed out that in the limit of an infinitely thick plate the results of Kin0 (1965, 1968, 1970) are in disagreement with those of the infinite semiconductor medium theory (Blotekjaer and Quate, 1964). The onedimensional analysis gives two plane waves (one of which, backward, vanishes in the zero diffusion limit) propagating along a carrier drift direction. However, none of waves obtained by two-dimensional analysis turned into these two plane waves. Blotekjaer explains this discrepancy, however, via consideration of the initial value problem rather than of the boundary value problem which was used by all the previous authors. Thus his results correspond to a transient behavior of an initial perturbation rather than to a steady wave propagation. Most authors used the stiff-surface model of carrier stream boundary with zero transverse current. The opposite free-surface model was first mentioned by Dean and Robinson (1974) to explain the disagreement between their experimental results (Dean et al., 1970; Dean, 1972b; Dean and Matarese, 1972) and the theoretical results of the stiff-surface model. Dean and Robinson have denied this model but they were not able to introduce rigorously a new model of quasi-free surface of a carrier stream for the case of nonzero diffusion. Instead of an additional boundary condition imposed by nonzero diffusion (now in the known form p 1 = 0), they have used the quasi-equilibrium assumption (Boltzmann distribution of mobile carriers), the applicability scope of which in the given case requires special treatment. The consistent and rigorous analysis of quasistatic wave propagation in thin semiconductor films was carried out by the author (Barybin, 1975b)for the two models of carrier stream, and there was ascertained an influence of diffusion and differential mobility anisotropy on a wave spectrum. In Section V,B, we shall briefly discuss the main results of this work. B. Dispersion Equations for Quasistatic Waves In thin semiconductor films, charge carriers can accomplish oscillations either in a film plane or normal to it. Because a dc electric field Eo producing.

21

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

a carrier drift is applied tangentially to a film plane and in the transverse direction a dc electric field is absent (when B, = 0), then the small-signal carrier mobilities along and transverse the drift direction are different. Electrical properties of an isotropic semiconductor medium are characterized by the velocity-field characteristic u(E) which determines the driftvelocity vector in the form Vdr = p(E)E, where p ( E ) = v(E)/E = (q/m)z(E).In linear approximation, representing T ( E ) as (39), it is easy to obtain the small-signal (ac) carrier velocity (Barybin, 1977a):

where p, = p ( E o ) = [u(Eo)]/Eois the static (dc) mobility at the bias field E o , p d = ( d ~ / d Eis) ~ the~ differential mobility at the same field value. Then for the arbitrary direction of electric field Eo the RF (small-signal) mobility tensor p, is equal to (Barybin, 1977a):

I

pc=pc

EOy Eox (x-l)-Eo Eo

EOy Eoy l+(x-1)-Eo Eo

( x - I)--EOy Eo, Eo Eo

(152)

where x = pd/p, is the anisotropy coefficient. In the particular case when Eo = e, E, , the tensor p, is essentially simplified: Pe=pe

[:," :I 0

1 0

(153)

Such a form was just used by some authors (Kino, 1965, 1968, 1970; Hofmann, 1969b, 1972; Gueret, 1970c; Masuda et al., 1970; Heinle, 1971; Dean and Robinson, 1974; Freire and Marcante, 1975; Barybin, 1975b) to analyze two-dimensional effects and longitudinal propagation (along a drift direction) of waves in thin semiconductor films. As is seen from (153), the transverse R F mobility p, is equal to pe and the longitudinal RF mobility is pz = xpe = p d . At low bias field corresponding to the ohmic part of u(E)curve there is x = 1 and p, = p r , i.e., all the semiconductors are isotropic. With increasing field we have 0 < x < 1 (that is, pt > pz > 0) and at high fields x = 0 (that is p, = 0 valid for semiconductors like Ge and Si with drift-velocity saturation) or x < 0 (that is p, < 0 valid for semiconductors like n-GaAs having NDM). The latter provides a convective instability of carrier waves giving an amplification effect of microwave signals.

22

A. A. BARYBIN

FIG.3. Schematic illustration of the semiconductorfilm of thickness 2a and permittivity E , , embedded in a dielectric medium of permittivity E ~ .

Now we discuss the results of our previous work (Barybin, 1975b)spread to the case of quasistatic wave propagation in the film plane at arbitrary angle 8 to a drift direction with the velocity vo = e,,uo = u,(e, sin 8 + e, cos 0) where the z-axis corresponds to the propagation direction (see Fig. 3). In this case the tensor pe given by (152) has the form

+ (x - 1) sin2 8

0

[ (X - 1) sin O 8 cos 8

0

1 Pe

=Pe

1

(x - 1) sin e cos e 0 1 + (X - 1) C O S ~8

]

(154)

The propagation of quasistatic waves for which El = - V q , , in the absence of magnetic field, is described by the equations: a. Inside the semiconductor film ( - a < y < a),

vzqy = - P1

(155)

El

V * j, = -imp,

(156)

-

(157)

j, = -be Vq\” - g e V p , + plvo b. Outside the semiconductor film ( 1 y 1 > a), v2qy = 0

(158)

In Eq. (157) obtained from (31) by linear approximation in view of (151) a field dependence of diffusion coefficient is omitted (that is, B(E)= ge= constant) and the tensor qe= pope takes into account the anisotropy of RF conductivity. The tensor form Qe is considered in our paper (Barybin, 1975b). Substitution of (154)into (155)-( 157) reduces to the system of the twocoupled equation :

v;cp‘ll + yzq\”

= -& 81

v;p,

+ k ; p , = (1 - X k l S r S d Y 2 cosz ed”

(159) (160)

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

where Vf is the transverse Laplacian, y = XI tion constant; k: = y z

$- y b d

23

+ i/3 is the 1ongitudina.lpropaga-

cos 0 - B r p d - i p e o d

(161)

With the help of introducing the new normal variables u1

= Cp:”

+zip,

u2

= q:”

+

z2p1

(162)

Eqs. (159) and (160) are transformed to the normal mode form (Barybin, 1975b): v:u,+C;u,=o vfu, + ( f U z= 0 (163) where the notations

r:, t = Y 2 -

P r P d w1.2 =

k:

+ p, b d w2, 1

(164)

were introduced. [In our previous paper (Barybin, 1975b) the quantity analogous to A was noted as 0.1 Equation (163) with Laplace’s equation (158)outside the film in the form

vfcpy’+ y2q‘:’

=0

(168) consists of the initial equations system, the solution of which is sought as a sum of symmetric and antisymmetric distributions (Barybin, 19758): U1,2(y)= A1.t cos

Cl,ZY

+4

2

sin C 1 . 2 Y

1169)

Here and in the following, the upper signs correspond to forward waves with /3 = Im y > 0 and the lower signs to backward waves with p = Im y < 0. Such a choice of signs in (170) provides for an exponential decrease of field as one recedes from the semiconductor surface. For the given value y as a function of w there are, in general case, two values of transverse wave number and defining a complicated distribu-

c2

24

A. A. BARYBIN

tion of q\') and p1 over film cross section in the form of the linear combinations of U , and U,:

A particular simplification when A = 1 and W2 = 0 is achieved, according to (167), in the following cases: a. isotropic semiconductor film with x = 1; b. zero diffusion limit with ge-+ 0 and Pd-, co; c. transverse wave propagation with cos 8 = 0 which will be discussed for two stream-surface models: the model of stiff (completely reflecting) surface with the boundary condition n * j, = 0 and the model of quasi-free (completely absorbing) surface with the boundary condition p , = 0. 1. Stiff-Surface Model With coinciding a stream boundary with a crystal boundary without surface charges the boundary conditions of quasi-static approximation (113), (115), and (117) are expressed in terms of U , and U 2 in the form (Barybin, 1975b): 2 2

u, - 2, u2 = - (2, - z2)qy

(172)

Substitution of (169) and (170) into (172)-(174) at y = ? a gives the dispersion equation of quasistatic waves. By virtue of symmetry of the structure under consideration (Fig. 3) the total dispersion equation separates into two independent equations for symmetric (s) and antisymmetric (a) modes (Barybin, 1975b): sin Cla . sin Cz a cos ( , a . cos T2a

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. !I

25

The particular cases of isotropic film ( x = 1)and transverse wave propagation (cos 8 = 0) simplify the dispersion equations (175) but d o not introduce specific changes into a character of wave process (in contrast to the free-surface model considered in Section V,B,2) (Barybin, 1975b). As was remarked in Section IV,C, the zero diffusion limit are not realized in the stiff-surface model without surface sources. Because in the zero diffusion limit the Kino surface sources (for the stiff-surface model) and the Hahn surface sources (for the free-surface model) coincide in form with each other, then a result of the limiting operation ge-+ 0, applied to a quasi-free surface problem solution, will hold for both models. In other words, the spectrum of normal modes of thin semiconductor film with ge= 0 does not depend on a choice of the model of carrier stream. 2. Free-Surface Model In the presence of a depletion layer on a crystal surface the boundary conditions of quasistatic approximation (134), (136), (139), and (141) are expressed in terms of U1 and U , in the form (Barybin, 1975b):

After the substitution of (169) and (170) into (176b(178) at y = +a, one obtains the dispersion equations for symmetric (s) and antisymmetric (a) modes (Barybin, 1975b): sin lla ' cos lza cos a . sin C2 a

cl

k T

c2 y

iyg

cos 8 -

flr

ise

cos

I

l1a . sin T2a

cos C1a . cos C2a sin a . sin C2 a

cl

The solutions of Eqs. (175) and (179), obtained with the upper signs, have a physical meaning only for /3 > 0 and with the lower signs only for fi < 0. The rest solutions must be omitted as not having physical meanings. With ge= 0 the film does not support the propagation of backward waves (because the only physical reason providing for a wave propagation against carrier drift is diffusion) and in Eq. (179) only the upper signs remain, while Eq. (175) has no physical meaning at all. Let us analyze particular cases for the free-surface model.

26

A. A. BARYBIN

a. Isotropic Jilm. The condition of isotropy x = 1 is realized in all the semiconductor materials at low bias field corresponding to the ohmic part of v ( E ) curve. In this case we have (Barybin, 1975b):

u1= P1

u2

= k,

=

cpp + zz p1

(180)

(181) The total dispersion equation (179) separates into two independent branches corresponding to the two uncoupled modes U , and U 2 . The Jirst branch of Eq. (179) answers to an existence only of the normal mode U1 = p , with Ti = k, while U2 = cp\') Z 2 p l = 0 (Barybin, 1975b). So the waves satisfying this branch are nonsolenoidal ( p , # 0 )ones in which cp\l) and p1 are strictly related to each other as (Barybin, 1975b) 51

52

=Y

+

The feature of these waves is the absence of electric field outside semiconductor film, that is, cp\') = 0. This is physically explained by that the electric field inside film is shorted on the equivalent surface charge pEq = po y , (due to transverse oscillations of stream boundary) and does not get out the film. Mathematically this is provided for the equalities (Barybin, 1975b), cos k,a = 0 sin k,a = 0 (183) for symmetric and antisymmetric waves, respectively. Equation (183) determine the values of transverse wave number = k,, namely nn = k, = (n = 1,2, ...) 2a Then the expression (161) giving a relation between y and k, allows one to obtain the dispersion equation of nonsolenoidal waves in the form

cl

In the limit of infinitely thick plate (a + co)the right-hand side of (185) is equal to zero and this equation with 8 = 0 turns into the dispersion equation for two plane waves obtained by Blotekjaer and Quate (1964) in onedimensional case. Consequently, the new found nonsolenoidal branch (185) describes just the modes which were unsuccessfully sought by Blotekjaer (1970) endeavoring to find out them among the solutions obtained by Kino (1968, 1970). Indeed, Kino's solution, as we shall see, correspond to the second (solenoidal) branch of the total dispersion equation (179). Kino (1970) mentioned the possibility of an existence of nonsolenoidal waves with cp'i2) = 0.

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

27

However, neither he nor other authors can obtain these waves because all they have used the stiff-surface model within which the nonsolenoidal branch does not exist. The obtained nonsolenoidal modes described by the dispersion equation (185) have a purely periodical transverse distributions in the form of trigonometric functions with = nnj2a. As there is no electric field outside the film these modes do not couple to external systems and can be excited only by means of an ac current flowing through the end boundaries of a crystal. The zero diffusion limit (ge-+ 0) does not change transverse distributions in nonsolenoidal modes but makes them “degenerate.” It means that with -+ 00, as is seen from (189, all the modes with different transverse distributions have the same propagation constant given by the relation

cl

sd

y cos 8 = fl,

+ ise

(186) which does not depend on the film thickness and coincides with a corresponding expression for the plane wave with ge= 0. The particular case of transverse wave propagation is obtained from (185) with cos 0 = 0. The solution of the obtained equation represents two waves (forward and backward) for each mode. These waves propagate due to diffusion to opposite directions with equal phase velocities and the same high attenuation. In the zero diffusion limit, as is seen from (186), these waves disappear because the physical mechanism of their propagation vanishes. The second branch of Eq. (179) answers to an existence only of the normal mode U 2= pi1)+ Z2p 1 with rz= y, while U,= p1 = 0 (Barybin, 1975b). So the waves satisfying this branch are solenoidal ( p l = 0). The field inside and outside the film is supported by the equivalent surface charge p:‘ = po y , . The solenoidal modes in thin semiconductor films are given rise to surface waves of an interface and do not have any analogs in the onedimensional approximation. The transverse distributions of field both outside or inside the film are described by the same constant equal to y, what should be expected from Laplace’s equation for solenoidal waves. The dispersion equation of the solenoidal modes have the form (Barybin, 1975b):

With 8 = 0 Eqs. (187) exactly coincide with the analogous equations obtained by Kino (1974). Because of solenoidalness of these waves = 0) diffusion does not affect them that is the dispersion equation (187) remains its form with ge--* 0 (however, the lower signs lose a physical meaning).

28

A. A. BARYBIN

b. Anisotropicjlrn with zero diJiision. The RF mobility anisotropy is achieved at high bias fields when y # 1. Without diffusion the anistotropic film supports only one normal mode U1= cp\') (in which Z2 = 0), while U , = cp\" + Z,p, = 0 (Barybin, 1975b).It follows that in the given mode U 2 the charge density p 1 is strictly connected with the potential cp\" by the relation (Barybin, 1975b):

Thus the anisotropic film supports only the propagation of nonsolenoidal modes ( p , # 0). The dispersion equations of these modes are (Barybin, 1975b): 1) cos

e - (pr + ipe)

y cos 8 - ipe

.

-rya--

e2 cot

T2a

El

120

E~

tan t2a

El

12a

+iya----

and the transverse wave constant c2 is connected with the propagation constant y by the relation (Barybin, 1975b):

When 8 = 0 Eqs. (189) and (190) coincide completely with analogous formulas obtained by Kino (1968, 1970).When x = 1 Eq. (190) gives c2 = y and then Eq. (189) turns into the dispersion equation (187) for the isotropic film. The particular case of transverse wave propagation is obtained from (189) and (190) with cos 8 = 0. As one should expect, in this case the influence of anisotropy disappears because from (190) one obtains c2 = y and Eq. (189) turns into (187) with cos B = 0 for isotropic case. Below we shall discuss the effect of the differential mobility anisotropy on a longitudinal wave propagation (8 = 0), following the main results obtained by Engelmann (1971b) and the author (Barybin, 1977b), in order to elucidate a mechanism of wave amplification in semiconductor films with NDM underlying an operation of thin-film semiconductor TWA (Dean et al., 1970,1973; Dean, 1972b; Dean and Matarese, 1972).The use of the zero diffusion limit makes an analysis valid, according to the above, for both the free-surface and stiff-surface models of carrier-stream boundary.

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

29

C . Longitudinal Propagation of’ Waves in Anisotropic Film with Zero Difusiort For lossy media the concept of wave holds only when a 5 P. So for subsequent analysis the simplification of Eqs. (189) and (190), related to the approximation about smallness of the attenuation (amplification) constant as compared with the phase constant, i.e., I a I < I P 1, is natural. In this case a phase velocity of waves does not differ almost from a carrier drift velocity, i.e., p IIbe(Kino, 1970; Engelmann, 1971b). In this approximation Eq. (190) assumes the following form:

It follows that depending on a relationship among a, PI, and xPr the values [ can be either purely real or purely imaginary. In the first case, according to (169), the transverse distributions cp\”(y) and p l ( y ) are described by trigonometric functions defining trigonometric modes. In the second case when [ = i I [ I the transverse distributions are described by hyperbolic functions cosh 1 ( 1 y and sinh 1 [ 1 y defining hyperbolic modes. The hyperbolic modes, according to (191), satisfy unequalities:

Then the dispersion equation (189) for the hyperbolic modes assumes the following form (Barybin, 1977b):

I

Because coth I [a 1 and tanh [a I are one-value functions there are only two hyperbolic modes: symmetric (HS mode) and antisymmetric (HA mode). The right-hand sides of (194) are always positive while the left-hand side can be negative under some conditions. It means physical irrealization of such conditions. Indeed, the wave attenuation state (u > 0) is possible only when (193) takes place because with (192) the left-hand side of (194) is negative. Analogously, the wave amplification state (a < 0) can be never realized. Hence, in the NDM region the hyperbolic modes are not supported by a semiconductor film and only the trigonometric modes can be amplified in this case.

30

A. A. BARYBIN

The trigonometric modes, according to (19l), satisfy the inequalities,

<

Because cot 1 [a I and tan 1 [a I are multivalue functions there is infinite number of trigonometric modes. When x > 0 all the trigonometric modes are attenuating ones with a > xp, while with x < 0 separate modes can either attenuate or grow (u < 0, I u 1 < 1 x 1 Br). Hence, the maximum value of the amplification constant of a growing wave is equal to Iu,,,I = lxl&. Because with x < 0 the sign of 0: can be changed (attenuation turns into amplification), let us find a threshold condition when u = 0. In this case, Eq. (191) gives 1 1 = and the dispersion equation (196) reduces to the desired condition:

r Bern

where even numbers of n correspond to the symmetric modes and odd numbers to the antisymmetric modes. Thus, with the fixed value x = - (1I < 0 an attenuation of the trigonometric n mode turns into an = (Bea)Ih uo /a amplification at the quite certain threshold frequency oIh given by (197). To ascertain the features of the hyperbolic and trigonometric modes one makes a graphic analysis of the dispersion relations (191), (194), and (196) which are convenient to be rewritten in the following forms (Barybin, 1977b): a. For the hyperbolic modes (with 0 5 x < 1)

where, according to (191) and (193), ( l a 1’ < X(Bea)’ < (Bea)’.

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

b. For the trigonometric modes (with

31

x 3 0),

where, according to (191) and (195), with x = - / x i < 0 one has lCal’ > Ix 1 (Be for attenuating modes and I l a 1’ < I x I (fiea)’ for growing modes. The graphic solution of (198) and (200) allows one to find out the values

1 l a I as a function of Be a, and then by using (199) and (201) to plot curves x > 0 and x < 0. Let us elucidate the features of mode propagation for the particular cases

a/& =f(B,a) for different modes with both

of

1. isotropic mobility (x = l), 2. positive anisotropy (x > 0), 3. negative anisotropy (x < 0). 1. Isotropic Mobility

As was pointed out in Section V,B,2, with x = 1, a semiconductor film supports the propagation of degenerate nonsolenoidal modes having the same propagation constant y = 8, + [see Eq. (186)]. The transverse distributions of field and charge in these modes are described by trigonometric functions with = nn/2a (n = 1,2, . . .) and have no electric field outside the film. Together with these degenerate trigonometric modes in the film there are also two hyperbolic modes: symmetric (HSmode) and antisymmetric (HA mode) with no ac charge ( p i 3 0). For these solenoidal modes the dispersion equation (187) admits the analytic solution (with approximation a 6 B ) (Barybin, 1977b):

ise

(202)

The curves u/Br =f(&a) for the trigonometric (nonsolenoidal) and hyperbolic (solenoidal)modes of the isotropic film are shown in Fig. 4. The HS mode and HA mode have various frequency dependence of the attenuation coefficient.

32

A. A. BARYBIN

4 FIG.4. Normalized attenuation constant ct//3, versus /3.a for dimerent modes of an isotropic film (x = 1): Tr modes, the trigonometric modes; HS mode, the hyperbolic symmetric mode; HA mode, the hyperbolic antisymmetric mode.

2. Positiue Anisotropy

This case corresponds to fields in n-GaAs not exceeding the threshold of intervalley transfer but giving rise to nonohmic behavior of the semiconductor (0 5 x < 1). As is clear from the above, in this case the film supports the propagation of decaying hyperbolic and trigonometric modes. The dispersion relations for the hyperbolic modes are obtained by solving Eqs. (198) and (199). The left-hand side of (198) Fl(l(ul) increases monotonously from the value Fl(0) = (cl / E ~ ) ( ~ / -x l)/Bea at l[al = 0 to infinity at Ira 1 = m B e a . As the value F,(O) decreases with increasing peathe crossing between the curve Fl( I [a 1 ) and the right-hand side of (198) in the form of tanh I[a 1 / I [a 1, values of which are always less than the unity, becomes only from the cutoff value when Fl(0) = 1, namely

Therefore, the HA mode has the low-frequency cutoff, i.e., it does not exist when Be a < (Bea)culand appears only with 8, a 2 (Bea)cuI(Fig. 5). At the same time, the HS mode, obtained from (198) by crossing of Fl( I[al) with the curve coth I[a I / I(a 1 exists with all the values Be a. As the graphic analysis of (198) shows, for both hyperbolic modes the magnitude of I [a I increases monotonical with increasing Be a from zero (when Be a = 0 for the HS mode and flea = (/3ea),uI for the HA mode) to infinity (Barybin, 1977b). Then (199) allows one to plot qualitatively the curves a/B, versus Bea for the hyperbolic modes shown in Fig. 5. For the H$ mode, a/B, increases mono-

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

(Pa0 4 U l

33

40

FIG. 5. Normalized attenuation constant alp, versus pea for different modes o f a semiconductor film with positive anisotropy (0 s x < 1): Tr modes, the trigonometric modes; HS mode, the hyperbolic symmetric mode; Ha Mode, the hyperbolic antisymmetric mode.

tonically from zero and for the HA mode, a//?,decreases from the magnitude a//?,= x at /Ica = (Bea)cu, to the high-frequency magnitude (a//?r)mtotal for both hyperbolic modes. It is easy to evaluate that this magnitude depends differently on x, namely (Barybin, 1977b):

The dispersion relations for the trigonometric modes are obtained by solving Eqs. (200) and (201) with x 2 0. The left-hand side of (200) F2(I ( a I ) decreases monotonically from the value F,(O) = l/x - 1)/pea at [ l a ]= 0 until zero at Ira] + 00. For antisymmetric modes the right-hand side of (200) in the form of tan [ [ a[ / [ [ a[ has the lowest branch n = 1 (where 0 < I l a 1 < 4 2 ) values of which exceed the unity always. So the crossing of this branch with the curve F2(Il a I ) takes place only when Bea (Beu)~,,, so that F 2 ( 0 ) > 1. The value (Bea)culis given by (203),as before for the hyperbolic modes. Thus the antisymmetric trigonometric branch n = 1 has the highfrequency cutoff: it exists when flea< (/?ea)c,,,and disappears with pea 2 (/?ca)cUl (Fig. 5). As at the cutoff frequency, we have Ira I = 0, then, according to (201), a//?,is equal to x. Hence, this trigonometric mode n = 1 turns into the HA mode at the cutoff frequency. Other trigonometric modes (n = 2, 3, . . .) exist in all frequency range and for them the value a//?, decreases with increasing bea as before for the lowest mode n = 1, from 1 to x (Fig. 5).

-=

34

A. A. BARYBIN

It follows from Fig. 5 that for all the trigonometric modes in the limit 1 one has a/& 1, that is there become their degeneracy. In accordance with (203) and (204), in this limit one obtains 0 and (a/fl,), -+ (1 + '. Therefore, in the limit x 1 the curves of Fig. 5 turn into the corresponding curves of Fig. 4 for the isotropic case. In other words, an introduction of mobility anisotropy eliminates the degeneracy of nonsolenoidal waves of the isotropic film and they turn into the nondegenerate trigonometric modes of the film with positive anisotropy. In the other limit case x - + O (corresponding to the threshold field in -+ 0. This means n-GaAs) Eqs. (203) and (204) give (8, a)cu, co and that the HA mode disappears at all while the HS mode exist in all frequency range and has no attenuation. For the trigonometric modes (n = 1,2,3, . . .) the value ./fir decreases from 1 till 0 with increasing Bea. The change of sign x must change a sign of a converting attenuation into amplification.

x

-+

-+

-+

-+

-+

3. Negative Anisotropy and AmpliJication Mechanism of Thin-Film Semiconductor Arnplijier

This case corresponds to fields in n-GaAs exceeding the threshold value, i.e., to the NDM range. Under these conditions the semiconductor film supports only the propagation of trigonometric modes among which there are both decaying and amplifying ones. These modes satisfy the dispersion relations (200) and (201) with x = - I x I < 0. The left-hand side of (200) represents the function F2(1[a1) have a discontinuity at [ [ alpole = a p ea : the branch I of F, corresponding to 0 II [ a 1 < I [ a I p ol e has negative values and by crossing with the right-hand side of (200) gives amplifying modes; the branch I1 of F , corresponding to I [ a I > 1 [ a Ipole gives decaying modes. The curves I [ a I versus Pe a for different trigonometric modes are qualitatively plotted in Fig. 6(a) by taking into x account that with increasing fie a the value I F2(0)1 = (1/ 1 x 1 + l)/fle a decreases and the pole point ((alpole = 8, a recedes from zero. Because the crossing of the branch I of the curve F,( 1 [a I ) with the zero branch (n = 0) of the curve -cot 1 l a I/ I [ a I takes place at all values Be a, then the fundamental symmetric mode ( n = 0) is amplified in all frequency ranges [Fig. 6(b)]. For other modes (n = 1,2,. . .) in the low-frequency range there is a crossing of the right-hand side of (200) only with the branch I1 of the curve F2( I [ a I ) that defines these modes as decaying. The attenuation of n mode is replaced by an amplification when the pole point 1 [ a I pole = m P e a of the function F, coincides with the n pole of the functions tan 1 [a 1 / 1 [a I and -cot 1 ( a 1 / 1 [ a I equal to n7t/2. This gives the threshold value (Bea),,, in the form (197) defining the conversion of attenuation of the n

Jd

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

35

(b)

FIG.6. Qualitative dependences of (a) 1 [a I and (b) a/Sr on Seafor dinerent trigonometric modes of a semiconductor film with negative anisotropy (2 < 0).

mode into an amplification. A qualitative forms of alp, as a function of Be a for different modes, plotted with the help of (201) and the curves of Fig. 6(a), are shown in Fig. 6(b). The analysis of the curves of Fig. 5 and 6 allows one to understand a wave amplification mechanism in thin films of n-GaAs. Under subscritical bias fields (when 0 < x I 1) all the modes (hyperbolic and trigonometric) decay in all frequency range (Fig. 5). As the value x approaches to zero the attenuation of all the modes in high-frequency range decreases. At the critical bias field (when x = 0) all the trigonometric modes remain decaying while the HA mode disappears at all and the HS mode becomes nondecaying in all frequency ranges. The latter just gives rise under supercritical bias fields (when x < 0) to the fundamental trigonometric mode ( n = 0) which is amplified at all frequencies beginning from the zero value of Pea [Fig. 6(b)]. For other modes (n = 1, 2, . . .), there are the threshold values (pea),),= n n / 2 m such that at /?,a 2 (&a),,,the given n mode becomes growing. When Ix is small the values (/.?,a)rh are high for all the modes so that on a fixed frequency the signal gain proceeds only at the expense of the fundamental symmetric mode (n = 0). Then all the higher-order modes (if they

I

36

A. A. BARYBIN

are excited in the film)decay and thus decrease a net gains and gives an additional contribution to the noise factor of amplifier. As 1 x I increases the successive inclusion of the higher order modes into the signal gain at the fixed frequency must be manifested. In this case the net gains must increase and the noise factor must decrease. Let us estimate the lowest value of fth (n = 1) assuming a = 1 m, vo = lo7 cm/sec, Ix I = 0.3 (Dean and Robinson, 1974):ft, = vo /4a Ix I = 45 GHz. Therefore for the band X we have f < J h . This means that the signal gain proceeds only from the fundamental modes (n = 0).An excitation of the higher-order modes by an input coupling element of amplifier is undesirable. First, this decreases the net gains of the amplifier and, second, enhances the noise factor because of a coupling between the fundamental mode and decaying higher-order modes. ignoring Let us estimate the maximum gain G,, = 8.681amuX)L(dB) the higher order modes effect. According to Fig. 6(b), we have

s"

For numerical calculation, we assume no = 1014 cm-3, p e = 5000 cm*/V sec, el = 12.5&,,uo = lo7 cm/sec, I x 1 = 0.3. Then for the active element length L = 50 pm we obtain G,, N 100 dB. Because the experimental data of net gain did not exceed 20-25 dB (Dean et al., 1970; Dean and Matarese, 1972), this confirms, apparently, an essential effect of the higher-order modes. This can explain also high value of the noise factor measured experimentally (Dean et a!., 1970; Dean and Matarese, 1972). VI. CONCLUSION The analysis of electrodynamic concepts of wave interactions showed that up to now the basic theoretical problems necessary for the study of wave processes in thin-film semiconductor structures (TFSS) are sufficiently developed :

1. The boundary conditions on charge carrier stream surfaces in nondegenerate semiconductor plasmas for different physical situations distinguishing by a degree of the influence of diffusion, surface states and charged layers, and transverse displacement of stream boundary; 2. The normal mode orthogonality for different subsystems (electromagnetic, ferromagnetic, acoustic ones and carrier streams) included in TFSS; 3. The excitation of normal mode by external sources and the mutual coupling between the subsystems of TFSS; 4. The power relations and the normalization of normal modes.

WAVE INTERACTIONS IN THIN-FILM SEMICONDUCTORS. I1

37

Dispersion properties of thin semiconductor films without magnetic field are studied and allow one to understand the mechanism of operation of thin-film semiconductor TWA. Effects of magnetic field on carrier-wave spectrum for different models of carrier-stream boundary can be investigated on the basis of the obtained theoretical relations. Among nonelucidated problems one should note a definition of external circuit current of thin-film semiconductor sample necessary to analyze an electric instability of TFSS. This question is very nontrivial. In the given case, as distinct from the one-dimensional model, the total current (calculated as a sum of conduction, diffusion, and displacement currents) does not remain constant along the sample length because of the presence of electric field outside the film. So an identification of the external circuit current with the total current of sample (as was made in the one-dimensional model), in the given case, is of no proof and, possibly, incorrect. In whole, the developed theoretical method of normal and coupled modes allows one to analyze wave interactions in different kinds of thin-film semiconductor structures.

REFERENCES Anderson, S. J., and Robinson, G. Y.(1974). IEEE Trans. Electron Devices 21, 377. Barybin, A. A. (1970). Radiotekh. Elektrotr. 15, 1556. Barybin, A. A. (1975a). J. Appl. Phys. 46, 1684. Barybin, A. A. (1975b). J. Appl. Phys. 46, 1697. Barybin, A. A. (1977a). Electron. Lett. 13, 243. Barybin, A. A. (1977b). Radiotekh. Elektron. 22, 1680. Bianco, B., Chiabrera, A., and Ridella, S. (1973). Alta Freq. 42, 181. Blotekjaer, K. (1970). IEEE Trans. Electron. Deuices 17, 30. Blotekjaer, K., and Quate, C. F. (1964). Proc. l E E E 52, 360. Bobroll. D. L., Haus, H. A., and Kluver, J. W. (1962). J. Appl. Phys. 33, 2932. Chodorow, M., and Susskind, C. (1964). “ Fundamentals of Microwave Electronics.” McGrawHill, New York. Dean, R. H. (1969). Proc. l E E E 5 7 . 1327. Dean, R. H. (1972a). IEEE Trans. Electron. Devices 19, 1144. Dean, R. H. (1972b). l E E E Trans. Electron Devices 19, 1148. Dean, R. H., and Matarese, R. J. (1972). Proc. IEEE 60,1486. Dean, R. H., and Robinson, B. B. (1974). IEEE Trans. Electron Devices 21, 61. Dean, R. H., and Schwartz, P. M . (1972). Solid State Electron. 15, 417. Dean, R. H., Dreeben, A. B., Kaminski, J. F., and Triano, A. (1970). Electron. Lett. 6, 775. Dean. R. H., Dreeben, A. B., Hughes, J. J., Matarese, R. J., and Napoli, L. S.(1973). IEEE Trans. Microwaoe Theory Tech. 21, 805. Engelmann, R. W. H. (1971a). IEEE Trans. Electron Devices 18, 587. Engelmann, R. W. H. (1971b). Arch. Elektr. Ubertragung. 25, 357. Fleming, P. L. (1975). Proc. IEEE 63, 1253. Freire, G. F., and Marcante, A. (1975). Int. J. Electron. 38, 443. Gueret, P.(1970a). Electron. Lett. 6, 197. Gueret, P. (1970b). Ekctron Lett. 6 213.

38

A. A. BARYBIN

Gueret, P. (1970~).Electron. Lett. 6, 637. Hahn, W. C. (1939). Gen. Electr. Rev. 42, 258. Hartnagel, H. L. (1969). Electron. Lett. 5, 303. Heinle, W. (1971). Electron. Lett. 7, 245. Hofmann, K. R. (1969a). Electron. Lett. 5, 227. Hofmann, K. R. (1969b). Electron. Lett. 5,469. Hofmann, K. R. (1972). Electron. Lett. 8, 124. Hofmann, K. R., and ’tLarn, H. (1972). Electron. Lett. 8, 122. Kanbe, H., Shimizu, N., and Kumabe, K. (1973). Electron. Lett. 9, 29. Kaner, E. A., and Yakovenko, V. M. (1975). Usp. Fiz. Nauk 115, 41. Kataoka, S., Tateno, H.. and Kawashima, M. (1969). Electron. Lett. 5, 48, 114. Khankina, S. I., and Yakovenko, V. M. (1967). Fiz. Tverd. Tela (Leningrad) 9, 578, 2943. Kino, G. S.(1965). “Carrier Waves in Semiconductors,” ML Rep. 1353. Stanford Univ., Stanford, California. Kino, G. S. (1968). Appl. Phys. Lett. 12, 312. Kino, G. S. (1970). IEEE Trans. Electron Devices 17, 178. Kino, G. S., and Robson, P. N. (1968). Proc. I E E E 56, 2056. Kumabe, K. (1968). Proc. IEEE 56, 2172. Kuru, I., and Tajima, Y. (1969). Proc. IEEE 57, 1215. Many, A., Goldstein, Y., and Grover, N. B. (1965). “Semiconductor Surfaces.” North-Holland Publ., Amsterdam. Masuda, M., Chang, N.S.,and Matsuo, Y. (1970). Electron. Lett. 6, 605. Metz, L. S.,and Gandhi, 0. P. (1974). IEEE Trans. Electron Devices 21, 118. Mikhailovskii, A. B., and Pashitzkii, E. A. (1965). Zh. Eksp. Teor. Fiz. 48, 1786. Schlachetzki, A., and Mause, K. (1972). Electron. Lett. 8, 640. Sumi, M. (1967). Jpn. J . Appl. Phys. 6, 688.

Light-Emitting Devices. Part 11: Device Design and Applications* HERBERT F. MATARE ISSEC. International Solid State Electronics Consultants Los A n g e l a . Califorttra

1. Device Design and Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Progress in Junction Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Progress in LED Technology . . . . . . . . . . . . . . ....................... References for Section 1 . . . . . . . . . . . . . . . . . . . . . . . ................................. 2. Progress in Laser Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Gain and Quantum Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Radiation Pattern ..................................................... C. Modulation Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Comparison of Laser and LED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Visible and Far-Infrared Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References for Section 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Device Types and Technological Progress . . . . . . . . . . . . . . . . . . . . . . . . A . Optical Problems. .............................................................. B . Adaptation to Integrated Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Thermal and Contact Problems., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Passivation and Degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References for Section 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 . Measurement Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . General Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Measurement of LED: Output Power and Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . C . Risetime Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Bandwidth Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E . Measuring the Thermal Impedance of LEDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F . LED Output Power as a Function of Wavelength ............................... G . Emitters and Current Sources., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................................................. References for Section 4 5 . Survey of Important Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A . Displays and Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B . General Applications of Infrared LEDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Photocouplers (Isolators) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Optical Radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E. Optical Communications and Integrated Optics ................................. References for Section 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . .............................

Part I of this article. on Methods. was published in Volume 42. 39

40 40 55

69 70 71 80 89 93 94 102 117 125 131 137 138 138 142 151 152 153 154 156 158

163 174 177 189 200

40

HERBERT F. MATARE

1. DEVICE DESIGNA N D TECHNOLOGIES A. Progress in Junction Performance

First consideration in the field of visible light-emitting devices (LEDs) is given to the human eye response with its particular maximum of sensitivity in the green spectral range and contrast enhancement due to the absorption of the adjacent spectral colors by the cover or background of the LED. As LEDs in many shades of colors can be made, the main problem is the external quantum efficiency which is a function of the internal quantum efficiency. Luminous eflcacy is a complex function of external quantum efficiency and in turn of the internal quantum efficiency and the eye response curve. Much work has been done to understand the effects of the isoelectronic traps like nitrogen in GaAs,-,P, or In,_.Ga,P, so important for the efficiency of the visible LEDs. We have stated in Part I of this article (Volume 42) that such a trap is formed by a replacement of the column V atom by nitrogen, an atom with a much greater electronegativity than, e.g., phosphorus (3 versus 2.1). Also, there is a hydrostatic deformation at the site of the nitrogen atom and a resulting high capture cross section for electrons. These trapped electrons can bind a hole by Coulomb attraction and form excitons. These excitons can decay via no-phonon radiative transitions because their electronic wave functions are diffused in k space. Therefore, such isoelectronic traps are plotted along a horizontal line in the energy diagram covering the range of direct to indirect transitions. Due to these trap levels, the range of efficient radiative recombination is extended into the indirect compositional range of the compound (InGaAs, -xP, for x > 0.45 at 300°K). The independence of the k vector for this trap leads to what is called the quasi-direct crystal. It is important to note that this band-structure enhancement of the radiative recombination is accentuated by orders of magnitude as the compositional range of the semiconductor is shifted toward the indirect side (see Part I, Figs. 4.14-4.17) or from GaAs GaAs InP

-

GaP AlAs GaP

InP AlP to name only a few. J. C. Campbell et al. ( 1 . 1 ) have calculated the band structure enhancement due to nitrogen using the Koster-Slater one-band one-site model. They found that the modulus of the wave function I$(k) Izof an electron bound to

LIGHT-EMITTING DEVICES. I1

41

I

FIG. 1 1. Modulus of the wave function I b(k)lzof an electron bound to an isoelectronic trap (N) in GaAs,_,P, along the A-symmetry line in the Brillouin zone r to X. Parameter is stoichiometric ratio x. For x near the direct-indirect overlap, Ir#~(k)l’ increases sharply at (k = 0) and exceeds the value at X. After Campbell P I al. ( 1 . 1 ) .

the isoelectronic trap in GaAs, -xP,:N increases considerably in the indirect compositional range (x > 0.45) and even exceeds the modules for x = 0.45 near the point X on the symmetry line of the Brillouin zone (indirect valley). Iqh(k) 1’ is a measure for the oscillator strength. On the other hand, the magnitude of (4(k)I2near T(k = 0) increases also sharply as the crystal composition x decreases and at x = 0.45 exceeds the value at X (Fig. 1.1). When plotted against crystal composition x, 14(k)12 shows that a remarkable gain is achieved when the E r - E x overlap point is approached (.Y = 0.45). (Fig. 1.2). For this composition, the magnitude of I4(k) 1’ at the r point ( k = 0) even exceeds the value at X.

-K

FIG. 1.2. l$(k)I2 for the trapped electron as a function of x at k = O for GaAs, _,P,:N. The strong increase in l#(k)l’ with decreasing x shows the strong influence of nitrogen for an increased recombination rate as Er approaches EN.

42

HERBERT F. MATARE

For the enhancement of the luminous efficiency of LEDs, the shift from the dark red to orange has become very decisive due to the higher eye sensitivity for the shorter wavelength. Therefore, it was an important step to enhance the overall efficiency by an increase in radiative recombination for x > 0.45.

FIG.1.3. The different recombination cases in GaAs, -,P,:N. (a) Radiative lifetime Tr, and nonradiative lifetime T,-" in direct processes between bands without involvement of trap levels. (b) Nonradiative indirect transition T~~ and electron transition from conduction band to trap level EN with electron capture time T ~The ~ filled . state N; can release an electron thermally (T,J or via nonradiative recombination ( T ~ J Hole capture into a level E , with a time constant T~~ is also shown. (c) Conversion of a bound exciton state by thermal emission of a hole ( T ~ " ) Or . NY becomes a neutral NP state by radiative ( T ~c )~ or , nonradiative ( T ~")~ bound , exciton transition. El, and ENare binding energies ofthe excitonic hole states and the N isoelectronic trap states, respectively.

As l4(k)l2 is a measure for the oscillator strength for radiative boundexciton recombination, its value is indicative for the observed radiative recombination rate. In order to understand the processes involved, we refer to Fig. 1.3, wherein the electron-hole recombinations in the different cases are shown. Campbell et al. (1.1) developed the recombination kinetics for these cases and derived the steady state solutions for these. From these expressions for the direct and trap-assisted quantum efficiencies, qr and qN can be derived.

LIGHT-EMITI'ING DEVICES. I1

43

FIG. 1.4. Energy band diagram of a two-valley semiconductor. Population transfer between the direct and indirect valleys is governed by the lifetimes T rx and f r r . Radiative transitions have lifetimes r R r and z R Z .Non radiative transitions are indicated as zNr and T ~ nF~and. n, are the carrier concentrations in the direct and indirect valley, respectively.

We will shortly describe this procedure, since we use the results on numerous occasions. The definition of q is Rate of photon generation - L -= Rate of carrier generation I

(1.1)

Referring to Fig. 1.4, where nr and n, are the direct and indirect valley carrier densities, we can write

and 1

1 (1.2)

As the balance for intervalley transitions has to be maintained, we also have

and

44

HERBERT F. MATARE

or (1.6) Since the transition time balances are

_1 --- 1 Tx

1

TNx

1

-

T r

TNr

+- 1

(1.7a)

+- 1

(1.7b)

?Rx

TNX

we have also

With

and

we get then with (1.8)

The usual expression for the efficiency (2.2). This expression can be varied in numerous ways by appropriately introducing the direct and indirect efficiencies of transition and the intervalley transfer time. Under conditions of thermal equilibrium and for Er < E x the carrier densities are given according to a Boltzmann distribution:

_ n~ -Txr = nr

(5) 312

e-AE/kT

(1.12)

Trx

From Eqs. (1.4) and (1.12), one obtains under simplifying conditions (1.2) (1.13)

LIGHT-EMITTING DEVICES. I1

45

Here N r / N , is the ratio of the state densities in the two valleys and AE = E x - E r , the energy separation between the two valleys. We see here that for T x r 2 T~ the electron concentration increases in the r valley beyond the value given by a Boltzmann distribution (see also Fig. 1.1). Calculations of the efficiency using this basic formalism [formula (1.1l)] have resulted in a good match with the measured data. The efficiency r] can be split into the efficiency of recombination in the r valley and the one in the x valley due to a trap level: r] = r]r

+r

(1.14)

] ~

q r , the direct contribution is shown in Fig. 1.5 as calculated (dotted line). The measured direct efficiency q,,, matches this curve at least in the important higher values. The indirect efficiency qN due to the N-isoelectronic trap is also shown (measured values). The calculated values can be fitted to follow this curve. We see that the indirect efficiency is high for the higher x values or, in other words, that the isoelectronic trap lifts the total efficiency (1.14) well beyond the direct-indirect overlap point. The importance of this

- a

FIG. 1.5. Quantum efficiency qr (calculated: dotted curve) and q m measured for GaAs, - .P, versus crystal composition x (300°K)resulting from electron-hole recombination external to the N isoelectronic trap. The calculated contribution to q from excitons bound to N traps, qN has been fitted to measured points q)y”’by varying the N concentration. After Campbell et al. (1.1).

HERBERT F. MATARE

46

lies in the fact of the increased overall efficiency considering the human eye response or the photopic response curve. In this connection, the results of efficiency calculations made by Archer are of interest as they have shown the importance of materials like Ga,In,-,P and AI,Inl-,P for visible light emitters. Archer (1.3) has compared the efficiencies of diffetent materials for LEDs. The materials of greatest interest and potential are Direct: GaAsP, AIGaAs, GaInP, AlInP Indirect: Gap, GaAsP, AlAsP, Alp, AlAs A consideration of infrared-pumped rare-earth phosphors is added later.

When calculating the external quantum efficiency,the following simplifying assumptions were made by Archer: (a) Forward current results only from recombination of electrons injected into a homogeneously doped p region. (b) r electrons recombine radiatively with a lifetime zR and nonradiatively with a lifetime 7 N . (c) X electrons recombine only nonradiatively with the same lifetime t N . With these assumptions, the efficiency formula as, e.g., derived by Campbell et al. (1.1) [similar to (Lll)] reduces to the equation used by Archer (1.3): (1.15) f is a proportionality constant between internal and external efficiency.The emission is assumed out of a planar p-n junction. For the electron distribution between the valleys, Eq. (1.13) is used with the assumption of 7 x r / f N = 0. Therefore, AE defines the basic form of this function. As E - E x increases, 1 decreases. The band-gap energies E r and Ex are given as functions of stoichiometry:

,-

E ( x ) = Eo

+

(El

- E ~ ) x+ 0.3(x2 - ~ ) / [ 0 . 5 ( E+, El)]"'

(1.16)

The indices 0 and 1 refer to the direct and indirect compound. This function represents a slightly bowed curve and can be used for most compounds. The result of these calculations is shown in Fig. 1.6. We have also plotted the photopic eye response curve (dotted) over the spectral range from 1.7 to 2.4 eV. It is obvious that Al,Ga, -,As is least and AI,In, -xPmost appropriate with respect to luminous efficacy. Also GaJn, -xP has a high efficiency within the range of the eye response curve. If one combines these two efficiencies to form the luminous efficacy in fL/Acm-', one obtains the forms of Fig. 1.7. Maximum efficacy is obtained for AlInP in the orange color

LIGHT-EMITTING DEVICES. 11

47

FIG. 1.6. Calculated quantum efficiencies of LEDs of indicated compounds versus photon energy, assuming equal radiative and nonradiative lifetimes [after Archer (1.3)Jplus photopic eye response curve (dotted line).

hu (eV)

FIG. 1.7. Calculated luminous efficacies of light-emitting diodes in the indicated alloys versus photon energy ( T = ~ T ~ ) .After Archer (1.3).

48

HERBERT F. MATARE 10'-

GaAs,-,P, : N

LUMENS 10"

-

3000K

DIODE OUTPUT

-

lo-'

~0.85 =0.65 10'C

-

I

10-1

100

101

10'

10'

I(mA)

-

FIG. 1.8. LED output (lumens) as a,function ofjunction current. Junctions are not encaplo-' cm2. The VPE curves with x = 1.0, 0.85, and 0.65 have been sulated with area of obtained on junctions prepared by Zn diffusion. The dashed lower curve for GaAs,_.P, was obtained from VPE-grown junctions (no nitrogen). The dot-dashed curve corresponds to an LPE-grown GaP:Zn,O LED. After Campbell et al. (1.1).

range. For these compounds, the overlap point Er = Ex is already beyond the area of maximum efficiency which is closer to the k = 0 point. If one compares the output characteristics of LEDs with and without isoelectronic trap (see Fig. 1.8), it is apparent that the efficiency is increased

1.8

1.9

2.0

2.1

-

2.2

2.3

2.4

(eV)

FIG. 1.9. Probability density 14(k)(* of the bound electron at k = 0 (r)in GaAs,-,P,:N and In, ..,Ga,P: N shown as a function of energy (ev). Eye response curve and color scale are added. The optimum brightness of In, _,Ga,P:N therefore occurs at higher energies, i.e., in the most sensitive portion of the photopic response as compared to GaAs, -,P,:N. After Campbell ef a/. ( J . 1 ) .

LIGHT-EMITTING DEVICES. I1

49

in a remarkable way even compared to GaP:ZnO grown by liquid phase epit axy. The same situation is visible when one plots the probability density I #(k) 1' of the bound electron at k = 0 (r)(see Fig. 1.9). From the foregoing considerations, one concludes that Ga,In -,P and AlJn, -,P are the logical choice for visible emitters with high luminous efficacy. At this point, a few metallurgical considerations have to be made. As Craford and Groves (1.4) have pointed out already, a good lattice match has to be accompanied by a small difference between the free energies of formation of the two compounds. If the growth temperatures for the compounds differ strongly, a growth temperature for the ternary compound cannot be established and the result is similar to one for very different lattice parameters. The two materials considered have the following parameters:

Compound

Lattice constant

InP GaP

5.869

AIP

(4

Expansion coefficient at 300°K (x "c-')

5.451 5.430

4.5 5.3 9

From the point of view of lattice mismatch, Ga,In, -,P and AlJn, -xPhave about equal chances to grow from a GaP seed crystal. But the large disparity of the free energies of formation in the case of Al,In,-,P: 650°C for InP 1250°C for AIP makes CVD a difficult process, as both compounds have to be formed at one particular substrate temperature. A discrepancy of 600°C makes deposition of the ternary compound difficult. Even when molecular beam deposition is carried out, monocrystalline growth requires substrate heating. Strong difference in free energies of formation will lead to disturbance of stoichiometry as one compound will disassociate or one element evaporate while the other compound grows normally. Another problem with the aluminum compounds stems from the strong chemical activity of reactive gases like aluminum chlorides. In this case, the quartz reaction system has to be used with protective liners. The performance of junctions in compound material depends on the usual parameters as known from elemental semiconductors. There is the condition of an abrupt p-n change within a zone of low dislocation density and the attendant condition of a sufficiently wide graded layer to decrease

50

HERBERT F. MATARE

FIG. 1.10. Profile of LED based on GaAs, -*P,deposited by CVD on either GaAs or GaP substrate. The top layer is grown by adding NH3 to the dopant gas stream. After Craford et al. (J.4).

the dislocation density in the upper epitaxial layer where the junction is positioned, Figure 1.10 shows the typical layer sequence: 0through The substrate @ is adapted to the ternary compound crystal 0 and @ by the graded section The change of x is plotted on the right. This section is the most important part of this structure (20- to 40-pm thick). The gas flow system in the CVD apparatus is set to gradually add more of the third constituant of the compound to the level desired. Layer @ is the result of a longer growth with constant x to improve on lattice perfection. As is well known (1.5) perfection is improved by glide or lateral dislocation movement during growth. The last portion of this layer is growth under an NH3 atmosphere to introduce the necessary density of nitrogen atoms as isoelectronic trap. Finally, a p-type diffusion step places a junction into the upper layer. The purity of the grown GaAs,_.P, layer is limited due to processinduced impurities. Above all, oxygen, SiO, and eventually water vapor are introduced when quartz tubes are used and heated in the process. Carbon is also introduced in the 1-ppm level due to the use of graphite boats and sliders. Therefore, a dopant level in the range of 1 ppm (or 5 X 10l6~ m - is~ ) required as a minimum to form junctions. In general, impurity levels in the iO”-~rn-~ range are targeted. As SiO and oxygen have to be limited in CVD open flow systems, so the native defects or stoichiometric defects have to be minimized. These defects

0.

0.

LIGHT-EMITTING DEVICES. I1

51

can be vacancies of one species which form complexes with impurities, giving rise to interband recombination levels. For example, a gallium vacancy, VGarcan combine with three Se or Si atoms or the complex with Six] arsenic vacancies may form (1.4-2.6). Such complexes act as nonradiative recombination centers or add to the radiative transitions in spectral areas outside of the desired optical band. As we discussed in Part I, nitrogen as an impurity forms isoelectronic centers which enhance the luminescent power efficiency in the indirect compositional range for GaAs, -xP. For x 1: 0.4,in the direct gap a second peak appears due to the NN-pair line (Part I). This peak is located at a point of longer optical wavelength (about 500 A higher). Beyond this point, however, the efficiency due to nitrogen traps is so much enhanced that the r transitions are overshadowed by the NN-pair transitions (Fig. 1.5). The measured values of external quantum efficiency versus alloy composition show clearly the enhancement due to the isoelectronic trap in the indirect range (Fig. 1.11). The brightness measured in footLamberts (fL)also shows the enhancement in the compositional range beyond x = 0.5 (Fig. 1.12). We have discussed earlier the desirability to work in the x = 0.6 to 0.7 range due to the emission wavelength change into a range of higher energy or closer to the maximum of the eye response curve (see Fig. 1.6). Important considerations are (a) the impurity profile and (b) the dopant concentration near the junction interface. In most cases, the p-type layer is external and the n-type layer is the epitaxial material. The diffusion of zinc should be carried out so as to yield a high carrier density. As electrons enter the junction from the n-type side under forward bias, they must find an equhl number of defect electrons for radiative recombination. In general, this is in the 10'8-cm-3 ~ , the range. The surface concentration of Zn can be as high as 10'' ~ m - but diffusion has to be controlled so that no excessive dislocation density results. As their number increases, zinc atoms move from interstitial to substitutional places and create strain and lattice deformation, the zinc atom being smaller than either the gallium or the arsenic atom. The effect of the zinc doping on the quantum efficiency is seen directly by considering the radiative lifetime t R . According to Eq. (1.15), the internal quantum efficiency in the direct range (GaAs, -xPxfor x < 0.45 and GaAs) is given by P A S

(1.15a) As T N is given by the epitaxial layer, one can only influence t R by the diffusion. t R decreases with increased hole or Zn density, increasing qin,.

52

HERBERT F. MATARE

In liquid phase epitaxy (LPE), the junction is mostly formed during the epitaxial growth process. Here the higher junction performance is well known and one can generally assume a factor of 10 in luminous power efficiency ( I A). This astonishing superiority of LPE as compared to VPE is due to the improved crystalline perfection in the junction plane. As the epitaxial layer grows out and improves in perfection, the impurities are distributed regularly across the lattice and no diffusion-generated dislocations form I

I

R

\

\ \ \

\

r-l--x I

X,

I I

I

0.4

0.5

0.6

I

0.7

1

0.8

I

0.9

FIG. 1.11. External quantum efficiency qe for GaAs,-,P, diodes versus composition (x). After Craford et al. (1.4).

53

LIGHT-EMITTING DEVICES. I1

f L(10A/cmz)

80(

60C

/

400

/

//

I I

200

074

0:s

0.6

0.7

0.8

FIG.1.12. Brightness versus alloy composition x for GaAs,-,P, nitrogen doping. After Craford et a/. (1.4).

0.9

with and without

nonradiative centers. However, due to the problems of monolithic device techniques, the diffusion-formed devices on the basis of GaAS0.6P0.4

GaAs0.3SP0.6S : N GaAso.,,Po.*s : N

(red) (red-orange) (yellow)

will remain preponderant for applications for a long time while the green devices based on GaP:N will more and more be produced by LPE. The red

54

HERBERT F. MATARE

devices of the type GaP:Zn-0 made by LPE have highest efficiencies and can be used at very low current densities. They seem to be more sensitive if used in pulsed operation, however. Typical luminous power efficiencies for VPE devices are in the 0.1 to 0.5 range while they range between 1 and 4 for the LPE-grown devices. A comparison of devices made by VPE (respectively, LPE) can be summarized as follows: (1) Epitaxial layer grown by VPE, junction diffused: q (la) Epitaxial layer grown by LPE, junction diffused: = 2 x q (2) Epitaxial layer grown by VPE and junction grown by VPE: 2 x q (2a) Epitaxial layer grown by LPE and junction grown by LPE: 10 x q

If, for example, green devices are produced by diffusion of Zn into n-type LPE-grown Gap, one combines the advantages of the high-perfection LPEepilayer growth with the disadvantage of defect enhancement of the diffused junction. One takes advantage of LPE in this case only by adding zinc in a two-step LPE process. On the other hand, there are limitations for the doping level which can be attained in LPE, e.g., in the case of nitrogen. Yellow emitters based on GaP:N with the concentration of N > 10'' cm-3 can best be produced by VPE (1.4). One can produce a device which changes color with current density by growing an n-type Gap-LPE layer doped with N and a p-type layer doped with Zn-0. This device produces both a red and green emission. At low current densities, the red emission dominates, as recombination of injected electrons takes place within the p-type zone which is Zn, 0 doped. At higher current density, the electron depletion drives holes from the p layer into then layer. Recombination here in the N-doped zone leads to green emission. Both GaAs,-,P,:N and GaP:N can be made to radiate in the yellow spectral range. In the first case, GaAso~,,Po,,, will yield good efficiency in the 5900 A range with moderate to high N doping. In the case of Gap, a very high nitrogen concentration is necessary ( > lo2' cm-'). In both cases, the NNdonor-acceptor-pair band is responsible for the emission and not the socalled A line which extends the emission to higher frequencies close to the Ex line (see Fig. 4.14 of Part I) (1.7). The highest quantum efficiencies are measured on devices made by LPE, but so that a first LPE layer grows out of a melt and the junction is formed during the same process by dopant overcompensationwithout prior exposure to a gas stream, vacuum, or, worst, to air. In this way, the junction is placed into a layer of high minority-carrier diffusion length. The difference in q due to this important parameter can be as high as 2 : 1 for LPE versus VPE (1.8).

LIGHT-EMITTING DEVICES. I1

55

As the ratio of the internal efficiencies is a quadratic function of the diffusion lengths, 2

(1.17)

the photoluminescence intensity is very dependent on this parameter. The foregoing considerations concerning the advantage of LPE are especially applicable in all cases where monolithic technologies are less important, as, for example, in the case of infrared emitters. As we have pointed out (Part I), there are cases of mixed doping and junction formation where the complex process of balanced gradient growth and impurity segregation can be controlled in liquid phase epitaxy.

B. Progress in LED Technology In a chronological review of the state of the art in 111-V compound semiconductor technology, one has to mention the first device realizations based on the original work by H. Welker and co-workers [see Part I (1.411. Electrical properties of these materials originally were investigated more in the direction of the usual applications as rectifiers,mixers (1.9), and eventually transistors [see von Munch, Part I (6.2)].The decisive properties for microwave generation and light emission were discovered later. An important beginning in this direction was made by Rupprecht et al. (1.10)who first made LPE-grown Gal -,Al,As junction devices radiating in the 1.8 eV red-visible range with external quantum efficiencies up to 3.3%. At this time, it was found that LPE improves the epitaxial layer perfection and that the dislocation density in the epitaxial layer is reduced by at least a factor of ten as compared to the substrate. The dislocation " outgrowth " has been explained by lateral, dislocation glide during growth of the epilayer (1.5, 1.6, 1.11). Also red-emitting GaP electroluminescent diodes with external quantum yield of 7% were fabricated by LPE as early as 1969 (1.12) and later such diodes were made with efficiencies as high as 15% (1.4). Similar values were produced with silicon-compensatedA1,Gal -,As LEDs radiating in the 8 6 0 ~ 8 0 0A range (1.3). Comparison with Zn-diffused LEDs made from the same material showed also here that the external quantum efficiencies are a factor of 10 higher for LPE, with gradual equality toward the high optical frequency range, where this material is operating in the indirect range. Kressel et al. (1.13) invoke free carrier absorption in the Zn-diffused layer as a partial reason for this finding. However, the action of silicon as a deep acceptor center and enhancement of optical efficiency due to the action of the silicon

56

HERBERT F. MATARE

traps on both the n-type and p-type sides is evident from the measured difference between the band-gap energy and the radiated energy:

AE = E, - hv, At 300"K, a metallurgical band gap of 1.49 eV corresponded to hv, = 1.37 eV, a difference AE = 0.12 eV. With increased E, (more aluminum), this differencebetween actual band gap and radiated energy increases to A E = 0.3 eV. This shows that some interband levels become more effective as E, increases. These levels are probably due to silicon as amphoteric dopant. In fact, silicon is known to introduce a level E, + 0.1 eV near the valance band due to [V,, . Six] complexes and there is a donor level with varying position as the A1 ratio increases. The high efficiency in the case of the LPE Si-doped junction is therefore in all probability due to the donor-to-acceptor recombination process in which these levels serve as charge collectors with appropriate relaxation times for efficient pumping action by the junction field. (Compare importance of these levels in the case of stimulated emission). The advantages of

I

I 0

0.2 DISTANCE FROM INTERFACE

FIG. 1.13. Schematic of semispherical epitaxial-diffusedLED with compositional profile of Ga, -,AI,As epitaxial layer. After Dierschke et al. (J.15).

LIGHT-EMITTING DEVICES. I1

57

the continued LPE process for the growth of efficient LEDs became evident in the course of further studies. Ladany (1.14) has shown that in LPE a sequence of differently doped flat layers can be grown and that external quantum efficiencies of flat radiators of 3 ,' are feasible. With dome-shaped structures in which the acceptance angle is wif ,ened considerably due to matching coefficients of refraction, external quantum efficiencies of 14% were achieved. These structures were made so that a junction was diffused into the last grown layer of Gal -,AI,As so that the wider gap layer with the higher A1 amount was external in the sphere (1.15) (Fig. 1.13). In a more elaborate construction, the plane of the junction should correspond to the plane in the Weierstrass sphere (see Section 3.) GaP as substrate for Gal -xAIx As layers was also used to grow relatively thin Ga, -,AI,As layers with junctions radiating out of the transparent GaP substrate (1.16). The geometry of the structure is shown in Fig. 1.14. METAL

r CoNTACFl

FIG. 1.14. Schematic of Gap-based Ga, _,AI,As LED. After Woodall er al. (1.16).

It is important to grow a sufficiently thick first Ga,-,AI,As layer to avoid P contamination by autodoping and meltback from the substrate. A graded growth of

Gal -.AI,Asl -,,P, with y decreasing with thickness takes care of the lattice matching in this case. There is no use here of the principle of heterojunction injection.

58

HERBERT F. MATARE

Considerable technical progress was made by the sliding crucible method with aliquot (limited) melt. Such techniques have been discussed in Part I (Fig. 6.17~).In such a case, the dopants can be changed easily by a graphite-slider movement, bringing the dopants in contact with the aliquot melt (1.17). The problems of surface morphology are predominant in LPE and much work was done to elaborate on the best conditions for flat layer growth. It turned out that above all the melt limitation mitigates the problems of layer roughness and hillock growth. We have discussed this general problem in Part I (Section 6,C). Here we want to add that practical solutions have been found to minimize surface roughness while growing layers of several micrometers in thickness up to 25 and more micrometers. In establishing the correct amount of initial supercooling of the solution A = (TI - T’)”C, (Ti = saturation temperature, T2 = crystallization temperature) between 1”

0

50

100

500 GROWTH TIME I (red

1000

I

5000

FIG. 1.15. Layer thickness d as a function of growth time t. Curves I and I1 are calculated thickness for A = 3°C and A = 5°C and D = 4 x cm2/sec; tan a = 3 lines fit measured points up to t = 100 sec [see Eq. (22)]. After Toyoda et al. (1.18).

LIGHT-EMITTING DEVICES. 11

59

and lWC, one is able to maintain a surface roughness well below 0.01 pm (2.18). Hsieh (2.19) derived the necessary formalism for this case. For growth from supercooled solutions of a semi-infinitethickness, the layer thickness is a function of

m = Slope of the liquidus curve dT,/dC,. C, = Concentration of solute in the solid (2.22 x lo’’ cm-3 for As in GaAs). D = Diffusion coefficient of the solute in solution. R = Cooling rate. It turns out that the thickness d of a layer grown by supercoolingis given by the sum of two terms: a step-cooling term,

d,, z At1/’

(1.18)

(t = time, A = temperature gradient across melt), and an equilibrium cooling term, d,, z Rt’’’ (1.19)

Thus 2 D d = -(-j1”(dS, C,m a

+

2

deq)

(1.20)

or (1.21)

For small time values t 4 (3/2) A/R, the first term of Eq. (1.21)is dominant and (1.21) can be approximated by (1.22)

This equation is the same as the one for the step-coolingprocess. Measurements of the layer thickness of the growth time t have confirmed this equation for small growth times t < l min (2.18). For larger values oft (23A/2R), the second term of Eq. (1.21) has to be considered also, but for a finite solution. In this case, the layer thickness is given by (1.23) qs and qc are correction factors to express the finite value of the thickness of the solution (Fig. 1.15).

HERBERT F. MATARE

0.2

-

0.1-

I \

\0-

0-

I 0

5

I

10

I5

A T ("C)

-

FIG. 1.16. Surface roughness of grown layers as a function of the amount of initial supercooling of the solution. Error bars reflect the range of the observed values. After Toyoda et a/. (1.18).

The measured results of surface roughness of epitaxial layers as a function of the amount of initial supercooling of the solution show that the critical range for A T is between 1" and 10°C (Fig. 1.16).This relatively wide range is striking and shows that a growth process initiated with controlled initial supercooling within these limits should yield good surfaces. This result is certainly aided by a vertical temperature gradient with the substrate at the lower temperature. In the search for more highly efficient lamps in the green-yellow and orange spectral regions, much progress was made. Especially, the application of VPE (vapor phase epitaxy) as a method for mass production has shown considerable progress ( I .20). While LPE (liquid phase epitaxy) continued to be applied to the infrared high performance emitters and Gap, the materials choice for green, yellow, and orange emitters resulted in a continued interest in the development of VPE. Vapor grown layers of GaP on GaP seed crystals can easily be doped with nitrogen and, depending on the concentration of N, a wide range of colors can be produced. The comparison between the different nitrogen doping ranges with respect to color emitted shows that green to yellow to orange

61

LIGHT-EMITTING DEVICES. I1

can be emitted when the nitrogen concentration varies from 1 x lo'* to 1 x 10'' ~ m - Figure ~ . 1.17 is a plot of relative intensity of some LEDs based on Gap. The production of the LPE sample is done according to the two-step LPE growth with limited melt and aliquot melt (change of dopant from sulfur to zinc under NH3 atmosphere) (1.21). The application of In, -.Ga,P for this frequency spectrum has recently seen great strides. Extensive work has been done by RCA in this field and

10-

9-

8-

-

I

1.9

2.0

2.1

I

2.2

2.3

214

215

-

(ev)

FIG. 1.17. Cathodoluminescence spectra at 300°K of nitrogen-doped gallium phosphide. (a) Spectrum of a nitrogen-doped LPE layer. The peak at 2.225 eV is the A line. The shoulder (2.18 eV) is probably the A - 0 phonon replica or a bound-to-free-donor transition. (b) through (d) Spectra from layers with increasing nitrogen density. Some side peaks are due to phonon replica. The strong peak in (c) is probably due to an exciton decay at nearestneighbor NN, pairs. After Hart (1.20).

62

HERBERT F. MATARE

Nuese has described this progress in an excellent report (I .22) on this work. Here, two main roads were pursued: (1) VPE of In,-,Ga,P on GaAs substrates. (2) VPE of In, -,Ga,P on GaP substrates.

We have discussed the advantage of this material for visible emitters and the expected high brightness level. Diffused p-n junctions prepared by the melt-grown process had shown a low-energy infrared peak and low efficiency. In contrast, the luminescence spectra from p-n junctions formed during vapor deposition showed the desired near-band-gap visible emission and high brightness. However, the measured values of around 300 &/A cm-’ were about an order of magnitude lower than the calculated maximum value (see Section 1,A, Fig. 1.7). This is supposedly a consequence of imperfect material due to lattice mismatch and disturbed stoichiometry. The alloy inhomogeneity is dependent on gas-density fluctuations in the VPE system. Nuese shows that the system Ino,5Gao.5P on GaAs lends itself to relatively perfect growth as the lattice mismatch is minimized here. The lattice constant of GaAs (5.653 A) is approached closely by I n O . 4 8 SG

0.5 15

(compare individual lattice constants: InP: 5.869 A

Gap: 5.451 A)

and yields a band-gap energy of 1.905 eV or emission of 6580 A (red). Due to this precise setting, a color change would imply a change in graded stokhiometry and apparently also a loss in efficiency. Much work was done to eliminate dopant-variations due to gas-flow inhomogeneities.It is most important that the metal chlorides (HC1:Ga and HC1:In) be mixed sufficiently before they reach the deposition zone. Figure 1.18 is a schematic for such elaborate vapor mixing method as applied to these ternary compounds. Electron mobilities are consistently higher for a gas flow of intensely mixed metal chlorides. During the work on epitaxial growth of In, -,Ga,P, a comparison of the results of deposition on the two substrate types showed that it is by far easier to deposit the lattice matching In0.48SGa0.51sP compound on GaAs than to deposit a graded layer of In, -,Ga,P with x varying from 1 to 0.515 on Gap. The slow growth rate of 8 to 15 pm/hr observed for the vapor deposition of these layers means that compositional grading must follow a relatively fast grading rate (2-3 mole % GaP/pm) to avoid excessively long growth periods, another reason for defects. Grading to higher GaP mole fractions starting from the Ino.sGao.5Pcompound on GaAs is also feasible and has

63

LIGHT-EMITTING DEVICES. II

Ga

I

I

I

1

I

T. MIXING ZONE

I I

'

I

I I

I T,

I

1,

I

SOURCES

I I I

FIG.1.18. Scheme of vapor-phase growth system with mixing zone for In, _,Ga,P. After Nuese (3.22).

been done with NH3 as a nitrogen carrier to enhance radiative transitions in the indirect range, as we discussed for the case of GaAs,-,P,. Again, the advantage ofnitrogen doping is felt mainly in the indirect range of the energystoichiometry relation. This means that the emitted light frequency increases and moves toward the center of the photopic response curve around the 60 mole X GAP point, i.e., at or above the Ino,,Gao,,P composition (see Part 1, Fig. 4.16). There is no reason why this stoichiometry cannot be produced gradually from Ino.sGao.5Pon GaAs. Nuese (1.22) specifically compares these cases (Table 1.1). TABLE 1.1

Substrate I 11

III

GaAs GaP GaAs

Alloy composition x at interface 0.515

1 .oo 0.515

Alloy composition at p-n junction

Amount of grading Ax

Electroluminescence emission

0.515 0.65 0.65

0 0.35 0.135

6580 6Ooo 6Ooo

4A)

Color

Red Y ellow-orange Y ellow-orange

It was found that a deposition with grading Ax = 0.35 on GaP substrates results in misfit dislocations which also propagate through the constant composition layer. Therefore, GaAs is preferred as a substrate. The reason for this difference lies in the necessary close control of stoichiometry during grading, which is difficult to achieve by controlling the gas

64

HERBERT F. MATARE

FIG. 1.19. Relative photoluminescenceintensity as a function of In, -,Ga,P alloy composition for Se-doped vapor-grown layers deposited on GaAs. Mean values. Theoretical curve fits measured data from cathodoluminescence and photoluminescence measured on melt-grown material. After Nuese (1.22).

flow of the metal chlorides. Small density variations will result in compositional variations and create crystalline defects. This becomes evident when we compare the photoluminescence intensity across the mole fraction for In, -.Ga,P on GaAs with the theoretical values and the corresponding values for melt-grown material. Figure 1.19 shows the theoretical curve for In, -,Ga,P derived according to the assumption of a dependence of the direct energy band gap on alloy composition [see Eq. (1.16a)l as E = 1.34 + 0.668~+ 0 . 7 5 8 ~ ~

(1.24)

While solution-grown crystals show photoluminescence intensity values following this line (dotted line) all the way, Inl -,Ga,P/GaAs CVD grown matches the curve only at the 50/50 compositional point, a clear indication that stoichiometric and other defects are causing nonradiative transitions for the other compositions. There is certainly a chance to improve on this situation as mere technical problems of gas-flow systems are involved. It is obvious that the lattice mismatch is responsible for most of the degradation. If the x value is chosen only 2 to 3 mole y i from the optimal value x = 0.515

LIGHT-EMITTING DEVICES. I1

65

the material is under stress and even microcracks appear ( 1 .22). It was also found that for a compositional range of x between 0.65 and 1.00 (indirect range), the addition of nitrogen (by NH, doping) will enhance the highenergy near-band-gap photoluminescence peak by a factor of 20, both for nand p-type samples. Not affected is the low energy peak at 8000 A. Recent progress in LPE processing of G a P has resulted in higher efficiency devices for the green range of operation. The complex growth and doping problems are solved either in a sliding boat arrangement [Part I, (6.53)],or the dipping process (1.23).Care was taken to form both the n-type and p-type layers under clean (oxygen free) conditions while the dopant (Te, N ) is homogenized in the melt at 900°C. The p-type layer (Zn) is formed by addition of zinc to the melt before further growth. The junction is formed at a lower temperature than the n-type layer (85O'C).This apparently keeps the zinc from evaporating out of the melt for the junction formation during the next cooling cycle from 850 to 700°C. Another addition of zinc during this cooling cycle assures sufficient surface conductance for contacting. By this process average efficiencies of uncontacted dies ranged between 0.06 to 0.14 at 7 A / m 2 current density. In all of these LPE process techniques, one outstanding problem is the cleanliness of the melt during the growth cycles. Complexes like Si-0 and VG,-O can act as undesirable recombination centers competing with the desired N centers. Also the Zn-0 complex (red) has to be eliminated for the green diodes. Therefore, all elements like 0 2 ,SiO, and H20have to be eliminated. Some researchers have intentionally doped the melt with oxygen-gathering elements like Al, Mg,and Si for the purpose of binding free oxygen during junction formation (1.24). The use of pyrolytic BN crucibles instead of quartz helped avoiding crucible attack by these dopants. External quantum efficiencies,however, did not exceed those produced in graphite crucibles under normal conditions of doping. We have to mention recent progress in the LED field for the spectral range of 1.1 pm. This is the area of minimum fiber absorption beyond the 9.9- to 1.0-um absorption peaks which can be slightly lower than the optimum value of 2 to 3 dB/km at the 0.85 pm range. The compounds applicable in this range are (see Part I, Fig. 4.21) In,Ga -,As In,AI, -,Sb InAs,P, --x In,Ga - ,Sb In,Al, -,As GaAs,Sb, -, (especially GaAS,,,,Sb,, and also quaternary compounds such as Ga,In,-.As,P1-, Double heterostructure devices have been made successfully (4% external quantum efficiency) (1.25).

66

HERBERT F. MATARE

In distinction to the problems encountered with AIxGa, -,As on GaAs due to strain by lattice mismatch, the extra degree of freedom in quaternary compounds allows an exact adjustment of the band gap to the spectral region of interest while keeping the lattice constant fixed. The exact lattice match can be achieved when InP substrates are used (lattice constant a. = 5.869 A). Using Vegard's law, the matching compound is GaxIn1- x A S 2 x P 1 - 2 x The quaternary layer which also fits the required emission range is GaO.l

71n0.83AS0.~4P0.66

LPE growth is carried out in a multiple bin graphite slider system. The resulting structure is shown in Fig. 1.20.

FIG. 1.20. Schematic of InP/Ga,In -xAsyPl-,/InP double heterostructure LED. After Pearsall et al. (1.25).

In such multiple bin sliding crucible equipment, stringent requirements with respect to melt composition have to be met. But in view of the wide wavelength range covered by this compound, extensive work along these lines is justified. The avoidance of aluminum as a component is also desirable (oxide formation). Moreover, In, -xGaxP,- y A ~ can y be grown latticematched on various substrates: GaAs, InP, GaAs, -xPx; In, -xGaxP or In,-,Ga,As; and InAs,-,P,, etc. So far only LPE has produced good results. This especially when care is taken so that each compositional layer is grown from the correct melt and that the remainder of the preceding melt is wiped off. Some authors (1.26) showed that this can be achieved in a cylindrical slider boat configuration where the substrate located at one end of the boat can be rotated from one melt to the next while a temperature gradient is

67

LIGHT-EMITTING DEVICES. I1

maintained across the melt in the direction of the substrate. The other quaternary compound for the 0.95 to 1.1 pm range is Al,Ga, -,Asl -,Sb, The development was based on GaAs, -,Sb, and has lead to external quantum efficiencies in the 2% range (1.27). Two structures are compared, the homojunction with the following layer sequence: GaAsO.

8 .ISb0. 13( P )

GaAs0.87Sb0.

13fn)

GaAs, -,Sb,(n) GaAs(n), substrate and the double heterojunction AIO.

17Ga0.83AS0.87Sb0.

AIO.O 3

GaO.97

13(P) )(.

8 7SbOI 1 3

AIO. 17Ga0. 83AS0.8,Sb0.

13 ( n )

GaAs, -,Sb,(n) GaAs(n), substrate The DH structure is superior with respect to efficiency, even considering the somewhat higher pulse current used by the authors (see the accompanying table).

Home-junction (a) (b)

1.038 1.056

0.063 0.060

0.18 0.13

0.12 0.14

0.05 1 0.048 0.043

0.20 0.45 0.17

2.1 1.7 0.64

DH structure (a)

(b) (c)

1.004 1.007 0.987

Finally, we should mention the status of technology in the field of short wavelength emitters. After the work on blue emitters based on GaN (see Part I, Section 4,A) metal-insulator-semiconductor structures using GaN and magnesium as a dopant have been made which emit violet light. These LEDs have the ,following structure: metal contact/insulating Mg-doped GaN/undoped n-type GaN Light emission in the forward bias mode peaks at 2.9 eV and is apparently released at the i-n junction with forward bias (at the m-i junction at reverse bias).

68

HERBERT F. MATARE

The I-Y characteristic of such a device is not normal and does not follow a fixed exponential I = Y" plot. n is 2 to 3 in the region of light emission and > 3 for lower voltages (no similarity to the p-n junction equation). Maruska and Stevenson (1.28) have discussed the possible conduction mechanisms in these devices: ionic, electron hopping, Schottky emission, Frenkel-Poole emission, space-charge-limited flow, and tunnel emission. All of these processes, with the exception of space-charge-limited flow, are discarded for the explanation of the light emission. (No voltage dependence for the thermal activation energy is measured.) The operation of the deviceexponential increase in luminous intensity with voltage applied-makes the process of impact ionization a probable explanation. To this end, high fields have to be established locally while external voltages range between 8 and 35 V. Since the light is generated in the forward mode (metal positive) and light originates at the lieu of subgrain structures, the authors (1.28) suggest that the local fields near lo5V/cm which appear at these subgrains are responsible for the radiative output. The structure of this LED is shown in Fig. 1.21. Metal

I GaN FIG.1.21. Gallium nitride grown on substrate (AI2O3)with n-type and i layer grown out in facets and light emission points. After Maruska and Stevenson (1.28).

The individual crystallites a, fl, y, etc., are separated by grain boundaries and under bias these are the areas ( 10-pm diam) where light originates. The authors correctly invoke similar processes in ZnS and Sic, where defects, especially grain boundaries, are also most active in this respect [see also (I .29)]. Another material of potential use in this spectral range is AIN (1.30). This material was grown by R.F. reactive sputtering on tungsten and s a p phire substrates at 1ooo"C. It was contacted by A1 and Nb dots. Figure 1.22 shows the geometry. An important feature of the deposition method described by Rutz (1.30) is that the active layer was improved by a pyrolytic vapor transport scheme. The sputtered layer of AIN served only to adapt the substrate surface to the properties of the deposited A1N during the subsequent vapor transport. In this second step, the substrate was placed (face

LIGHT-EMITTING DEVICES. I1

69

Al contact

Al N crystal

(substrate)

!pm sputtered AIN SAPPHIRE <substrate)

5,,m grown AIN layer

Al contacts (alloyed)

FIG. 1.22. Schematic of ultraviolet LED made from AIN (sputtered and pyrolytically deposited) on niobium and sapphire substrates, respectively. After Rutz (1.30).

down) on a polycrystalline sintered AlN source. At a temperature of 1850°C and in a forming gas atmosphere, the AlN layer builds up to a few pm thickness and presents sufficient grain size for light emission around 350 nm. Finally, we mention that techniques have been found to build multicolor LEDs with double junction structure. For example, in Gap, the excitation of the red Zn-0 complex or the NN-exciton line can respond to different current densities and thus change the color. A more sophisticated method is the combination of an infrared pumped phosphor with a red GaAsP device. This device can be formed by LPE on either side of a GaAs substrate. One side receives an amphoteric (Si) doped junction, the other side a layer of GaAsP (red) and a zinc diffused junction. As enough light from the amphoteric junction will pass the die toward the GaAsP side, one can excite a phosphor layer deposited upon the GaAsP top layer. Such a phosphor is, for example, the green-emitting NaYF,:Yb plus erbium prepared from YF,:Yb, Er, and Na,SiF6. As the GaAsP-diffused junction operates at lower current densities, the red color is emitted only after the current has reached a certain value (e.g., 10 mA) at which sufficient infrared light is injected into the phosphor, which in turns emits green light and thus changes the visible color (1.31). The efficiency of phosphor excitation suffers from the fact that optical upconversion (infrared to green) is lossy, and even with 6% infrared LED efficiency, the overall efficiencies are in the to lo-* range.

REFERENCES FOR SECTION 1 1.1. J. C. Campjell, N. Holonyak, Jr., M. G . Craford, and D. L. Keune, J . Appl. Phys. 45( lo), 4543455 1 (1974). 1.2. B. W. Hakki, J . Appl. Phys. 42(12), 4981-4995 (1971). 1.3. R. J. Archer, J . Electron. Mater. No. 1 (1972). 1.4. M. G. Craford and W. 0.Groves, Proc. IEEE 61(7), 862-880 (1973). 1.5. H. F. Matare, Crit. Rev. Solid State Sci. 5(4), 499-545 (1975).

70

HERBERT F. MATARE

1.6. H. F. Matare, Solid State Technol. Jan./Mar., Parts 1-111 (1976). 1.7. J. I. Coleman, N. Holonyak, Jr., M. J. Ludowise, P. D. Wright, W. 0. Groves, D. L. Keune, and M. G. Craford, J. Appl. Phys. 46(8), 3556-3561 (1975). 1.8. M.Ettenberg and C. J. Nuese, J. Appl. Phys. 46(8), 3500-3508 (1975). 1.9. G. Zielasek, Arch. Elektr. Liebertragung. 8, 529-533 (1954). 1.10. H. Rupprecht, J. M. Woodall, and G. D. Pettit, Appl. Phys. Lett. 11(3), 81-83 (1967). 1.11. H. J. Queisser, J. Cryst. Growth 17, 169 (1972). 1.12. R. H. Saul, J. Armstrong, and W. H. Hackett, Jr. Appl. Plzys. Lett. 15(7), 229 (1969). 1 . 1 3 . H.Kressel, F. 2. Hawrylo, and N. Almeleh, J . Appl. Phys. 40(5), 2248 (1969). 1.14. I. Ladany, J. Electrochem. SOC. 116(7), 993 (1969). 1.15. E. G. Dierschke, L. E. Stone, and R. W. Haisty, Appl. Phys. Lett. 19(4), 98-100 (1971). 1.16. J. M. Woodall, R. M. Potenski, and S. E. Blum, Appl. Phys. Lett. 20(10), 375-377 (1972). 1.17. K. K. Shih and J. M. Blum, J. Appl. Phys. 43(7), 3094-3097 (1972). 1.18. N. Toyoda, M. Mihara, and T. Hara, J . Appl. Phys. 47(2), 443-448 (1976). 1.19. J. J. Hsieh, J. Cryst. Growth 27,49-61 (1974). 1.20. P. 8. Hart, Proc. IEEE 61(7), 880-884 (1973). 1.21. R. H. Saul and D. D. Roccasecca, J. Electrochem. Soc. 120(8), 1128- 113 t (1973). 1.22. C. J. Nuese, “Yellow Luminescent Diode,” Final Rep. AD767338. Off. Nav. Res., Dep. Navy, Arlington, Virginia, 1973. 1.23. 0. G. Lorimor. P. D. Dapkus, and W. H. Hackett, Jr., J . Electrochem. Soc. 122(3), 407-412 (1975). 1.24. P. C. Miiran and R. N. Bhargava, J . Electrochem. SOC. 123(5), 728-733 (1976). 1.25. T. P. Pearsall, B. 1. Miller, R. J. Capik, and K. J. Bachmann, Appl. Pkys. Lett. 28(9), 499-501 (1976). 1.26. J. J. Coleman, N. Holonyak, and M. J. Ludowise, Appl. Phys. Lett. 28(7), 363-365 (1976). 1.27. R. E. Nahory, M. A. Pollack, E. D. Beebe, and J. C. Dewinter, Appl. Phys. Lett. 27(6), 356-357 (1975). 1.28. H. P. Maruska and D. A. Stevenson. Solid-State Electron. 17, 1171-1179 (1974). 1.29. H. F. Matare, ’‘ Defect Electronics in Semiconductors,” pp. 234 ff.. 363. Wiley (Interscience), New York, 1971. 1.30. R. F. Rutz, Appl. Phys. Lett. 28(7), 379-381 (1976). J.31. T. Saitoh and S. Minagawa, I E E E Trans. Electron Devices 22(2), 29-32 (1975).

2.

PROGRESS I N

LASERTECHNOLOGY

The creation of the semiconductor laser grew out of a considerable effort to apply the general laser principle to semiconductors. Resonance phenomena in semiconductors had occupied the time of many researchers in the 1960s. Cyclotron resonance had been developed as a means to measure effective masses in semiconductors and to elucidate their band structure. Kroemer described the effects of concave energy contours and proposed the use of negative mass carriers (2.1). It turned out that the effect is too small to be useful and that scattering of carriers out of the negative mass cone dominate at room temperature. Other effects considered during this time were plasma effects in solids, which had already been studied in connection with cyclotron

LIGHT-EMITTING DEVICES. I1

71

resonance (2.2). H. F. Matare described as early as 1957 the equivalent of the magnetron or the solid state oscillator between electric field plates in a magnetic field and coupled to a cavity (2.3). The idea of the use of the MASER principle in semiconductors was clearly described at that time (2.4.). The generation of frequencies beyond the then known MASER frequencies in the kMHz range is described as an application of the energy levels in crystals: E2 - El = hv; E2 > E l generating the emitted frequency v of the radiation when transitions between level E z and E , are induced. At this time also Heywang described a semiconductor junction device coupled to a circuit or cavity and subjected to a pump source using the down-conversion principle (2.5). The step to the semiconductor laser was prepared by the existence of the solid state laser-externally pumped within a Fabry-Perot cavity-and the existence of the light-emitting junction. This development has been described in detail in the review papers by Loebner (2.6)and Hall (2.7) and will not be repeated here. A . Gain and Quantum EfJiciency

Considerable progress in semiconductor laser technology has been made in recent years, particularly with regard to the threshold current. The original simple schemes: diffused p-n junctions in GaAs, chemical vapor deposited (CVD), p on n layers or n on p layers, liquid phase epitaxy (LPE), junction layers-required 10,OOO and more A / m 2 to enter into the superlinear portion of the emitted output power versus current characteristics, which excludes room-temperature operation. In Part I, Section 3, we derived the Bernard-Duraffourg condition for population inversion. Equation (3.4) of Part I states that the injection process requires that the external voltage V, or the difference Fc - F , of the respective quasi-Fermi levels exceed the gap energy EG = E, - E , or the diffusion vdtage V,: e K 2 eVD = E, - E , (2.1) In the usual case of degenerate doping, the Fermi level is higher than the conduction band in the n-type side and below the edge of the valence band in the p side. Thus e q > Ec - E, and the conditions for the external voltage in both the n and p layers can be written

72

HERBERT F. MATARE

The threshold condition for lasing at a given temperature is given by e x p ( h ) = R exp(gL)

(2.4)

+ (1/L) ln(l/R)

(2.5)

or g =a

Here g is the gain per unit length in the laser cavity, R is the reflectivity of the cavity ends (mostly 0.32 for uncoated GaAs), c1 is the loss per unit length in the cavity, and L is the cavity length. [For a review and literature summary, see (2.8, 2.9).] The relation (2.5) is experimentally well established and the gain expressed in terms of the threshold current Jthr is basically a linear function: dJthr)

= PJthr = a

+ (l/") In(llR)

(2.6)

Equation (2.6) is valid for all types of laser structures and the lowering of the actual value of Jthr is due to the remarkable improvement of g and lowering of a as a consequence of the development: homojunction laser -+ single heterojunction laser --* double heterojunction laser -+ large optical cavity (LOC) laser (Fig. 2.1)

The general relation between the gain coefficient and the current J is obviously 9z

(rti/d)J (2.7) as the gain is proportional to yli, the internal quantum efficiency, and to the current J and inversely proportional to the thickness d of the recombination zone. There are, however, temperature-dependent factors involved which change (2.7) into the form = B(T)[(rti/d)J -

J'(01

(2.8)

where P(T) is the temperature-dependent form factor and J'(7') is the temperature-dependent leakage current. If we want to express the threshold current f t h r as a function of gthr (optical gain per unit cavity length at threshold) and ti (average absorption coefficient in the cavity, including free-carrier absorption within the inverted population), we can write Eq. (2.8) in the form: (2.9) where d is the width of the inversion region, r is the fraction of stimulated emission within the inverted population region, and B(T) is the gain coefficient (increases with decreasing T and increasing carrier density or doping N).In highly doped laser junctions in the lOI9 range, b N 1. It Jthr

= (d/Vr

r)[fhhr/@(T)ll'b

LIGHT-EMITTING DEVICES. I1

73

can increase to N 3 for light doping (2.2). b can be called the carrier density weighting factorial or l/b the dopant exponent. qr is the usual radiative efficiency, which can be written (see Section 1) (2.10) where T~ is the radiative lifetime and T~~ is the nonradiative lifetime. Considering Eq. (2.7), we see that d has to be small and obviously r and q, large. But we see also that .Ilhris directly proportional to g t h , (for b 1). Using expression (2.5), for glhr

=

+ ( l/L) In( 'lR)

we have (2.11)

an expression which shows that the main factors to contend with are ti, the average cavity loss, and /?(T),the gain coefficient. Cavity length L and reflectivity are also factors to consider but they are of less influence. We see that the heterojunction laser especially in the DHL form with eventually six layers (see Part I, Fig. 5.4) satisfies conditions to minimize J [ h r : Electrons injected into the p region should recombine within 1 or 2 diffusion lengths. This is effectively the case where d is small and also the two potential barriers at the heterojunction interfaces prevent drift into the bulk and thus bulk recombination. Concerning minimization of losses, the DHL form is a remarkable improvement over the homojunction. The optical confinement works because of the different indices of refraction in the layers adjoining the inner active layer. As the coefficient of refraction is smaller in these layers (GaAIAs)than in the active layer (GaAs), waveguiding takes place. [Velocity of propagation-u, z c/n, where c = light velocity and n = coefficient of refraction; thus u,(GaAlAs) > u,(GaAs) or the wave propagating at the interface will have curved constant phase wavefronts. Thus guiding and focusing occur.] As a consequence,the threshold value of the current density for lasing has gone below the loo0 A/cm2 mark. This compares with a lower limit of 8000 A/cm2 for the single heterojunction (SH) laser (2.9). An important point to consider is the width d of the recombination zone. By adding a second heterojunction at the p-n interface (Fig. 2.ld), wave confinement is not bound by the thickness d of the active region, but by the adjoining layers of a different n value. Thus the inner layer of thickness d can be reduced by a factor of 10 to 0.2 pm, as it now is limited to the function of a recombination zone only.

74

HERBERT F. MATARE

HOMOJUNCTION

c) Ga,-, At, As (P')

-- -

G a A s m

d
,,

Ga,- I Al As (N) Ga As (N)

DHJ

(DOUBLE HETEROJUNCTION)

E,

n

Aly Ga,-, As(P) SCH (SYMMETR. SEPARATE CONFINEMENT)

FIG.1.1. (a) Homojunction Laser. Schematic of band gap variation and refractive index. (b) Same for SHJ (Single Heterojunction Laser). (c) Same for DHJ (Double Heterojunction Laser). (d) Same for SCH (Symmetrical, Separate Confinement double heterojunction laser.)

In early fundamental work, Pilkuhn and Rupprecht (2.10) had already noted that the threshold current follows a law of the type in Eq. (2.9). They used the simplified form, Jthr

= (l/W/P)[ln (1/R) +

4

(2.12)

[see Eq.(2.6)]and found that for a fixed cavity length Land reflectivity R the internal loss per unity length c1 accounts less for the difference between epitaxial and diffused junction lasers than the gain coefficient j3 (gain coefficient per unit length and current density).

LIGHT-EMITTING DEVICES. I1

75

Threshold lowering and room temperature operation being the immediate goal, much work was done to improve fl and decrease a. A first essential threshold current lowering to 2300 A/cmz was reported by Panish et al. (2.11) with their DH laser. Kressel et al. (2.12) showed at the same time that such improvement is due mainly to the reduction of a by the heterostructure. While early values reported (2.10)were in the range: a = 14 to 92 cm-

fl = 5.7

to 3.94 x lo-’ cm/A

x

Kressel et al. indicate values for a between 8 and 27 cm- and fl between 3.3 x and 5 x lo-’ cm/A with a corresponding increase in efficiency. While the recombination zone of width d was more and more subject to high optical power and could not be reduced below 0.2-0.5 pm, catastrophic failures occurred in the DH lasers when driven at higher currents.The LOC laser (see Fig. 2.ld) was therefore a decisive step forward (2.12). Figure 2.2 shows a comparison of the main structures (SH, single heterojunction, and DH, double heterojunction) with respect to threshold current as a function of the active region thickness d. The external or power efficiency of these devices can be very high, up to 20%. The differential quantum efficiency can be even higher. It is derived only for the superlinear branch of the output characteristic. One defines ql(internal efficiency) x Nph(number of photons produced in cavity per injected carrier), and VJexternal efficiency) = Nph(number of photons emitted from cavity per injected carrier). Thus qext

= r]i(Nbh/Nph)

(2.13)

For stimulated operation only, we write Artex, = qi(Nbh/Nph)

(2.14)

According to Maxwell’s equations for the Fabry-Perot cavity, the relation for Nph is (2.15) (2.8), where o is the frequency, k is the extinction coefficient, and c is the speed of light, while Nph, the number of photons produced, covers the absorption coefficient. Therefore, Nph = Gain = g or, with Eq. (2.5), Nph = a + (l/L)ln (1/R)

(2.16)

76

HERBERT F. MATARE

FIG.2.2. Comparison of threshold current density versus the width of the recombination region for single and double heterojunction lasers. The lower values were reported for LOC structures. After H. Kressel et al. (2.12).

Thus (2.17)

With the actual efficiency measured, one can calculate a and R . The external quantum efficiency is Power out

flex1

= Power in

-

P(Watts) iV(Watts)

P/hv

=--

Z/e

(2.18)

LIGHT-EMITTING DEVICES. I1

77

while

P - P o --AP = (I - f o ) V - A f V

(2.19)

Plotting the optical output power, measured by an integrating sphere, over the current in the range of stimulated emission allows us to find Aqex,according to Eq. (2.19). With Eq. (2.17), other interesting values like q i , R, or a can be found using, e.g., expression (2.12) for the threshold current, which is easy to measure. Figure 2.3 is a plot of the values for the differential efficiency of DH-LOC lasers with varied thickness d of the central recombination zone (2.12).

60-

A 4 -

m-

A A

/ A* 40-

I

\ A

'

\

\

\

30-

20-

FIG.2.3. Differential quantum efficiency as a function of thickness d of active layer (Fig. 2.ld) at 300°K. After Kressel et a/. (2.12).

The idea to separate carrier confinement and optical confinement by placing the recombination layer between two oppositely doped GaAlAs layers, bounded by two more GaAlAs layers of higher A1 mole fraction (2.13) has resulted in threshold values well below even the lo00 A/cm2 mark (Fig. 2.4). [Lowest value published by Panish et al. is 650 A/cm2 (2.13).] The influence of the layer thickness of the optical cavity in DH-structure lasers has been studied also. For a structure as shown in Fig. 2.lc with the GaAlAs layers adjacent to the active zone of composition A10.3Ga0.7As

HERBERT F. MATARE

78

t 77,;

p+- GaAs

P -AID,. Ga,,As \r

P

'.',

Ga As, , / /

P

- A1.n G

L A6

- Aloar Ga,,

n +- G e AS

t-- --

4

--

0.78p

--

--

0.45/l azzp o.=

n - Ahl,GaM, n

-

AS

L7p--

-

SUBSTRATE

i

c

n

FIG.2.4. Layer structure for low threshold separate confinement laser with representative variation in refractive index n. After Panish er a/. (2.13).

and for a d value of 0.12 pm, Casey and Panish (2.14) found that Jthrand 9 decrease drastically for a layer thickness below 0.8 pm. Presently, the lowest values of threshold current seem to lie in the vicinity of 500 A/cm2, where carefully chosen doping levels in the DH structure are maintained and active regions below 0.1 pm are maintained (2.15). With such structures, the beam width OL ,perpendicular to the junction plane, ranges between 20 and 20" for a heterojunction spacing (d) between 0.05 and 0.5 pm. In this case, germanium was used as a p-type dopant in one of the adjacent ternary layers. The role of silicon as a dopant in GaAs LEDs was mentioned (Part I, Section 1). It is well known that this impurity has special features in the GaAs lattice. It introduces traps with a large capture cross section and can act as a donor (replacing Ga atoms) or as an acceptor (replacing As atoms). In the first case, shallow donor states eV below the conduction band are introduced while two types of acceptors exist, one at 0.1 eV, the other at 0.03 eV from the valence band. The silicon atoms are more stable in donor states at higher temperature. Rapid cooling keeps these atoms in the gallium sites; but upon slow cooling, especially when crystallization occurs out of a gallium melt, as in LPE, the GaAs crystal becomes p-type. This effect is used to produce high-efficiency LEDs. GaAs is grown out of a melt while silicon moves into the arsenic sites. This is the ideal junction. Here no difference in lattice constant occurs, nor is there a difference in dopants needed to produce p-n junctions.

LIGHT-EMITTING DEVICES. I1

79

/

Ga

Ga

Ga

lattice mismatch

NO

No difference in dopant (or ionic radius)

I

\

I

FIG.2.5. pn junction in GaAs with silicon as amphoteric dopant (e.g..p layer recrystallized out of a gallium-rich melt.) Ideal junction without lattice mismatch or difference in ionic radius of dopant.

The advantages of silicon as a dopant for LEDs were also applied to lasers by using the amphoteric p-n structure (Fig. 2.5) in the active region of a D H laser (2.16). While threshold current density values descended below 2000 A/cmZ for 2 pm active zone thickness [according to the general relation between active zone thickness d and Jrhr(see Fig. 2.2), one could conceivably lower Jrhreven beyond the optimum values around 500 A/cm2], the problem here is the relatively low optical frequency (9000-9300 A) and the switching or modulation speed of these lasers. From such measurements on amphoteric LEDs [GaAs:Si(p)GaAs:Si(n)], we know that relaxation times are long, i.e., in the hundreds of nanoseconds. While stimulated emission allows us to decrease the delay time significantly, there is still no competition with a D H laser operating at the absorption minimum of the silicon oxide fiber (850 nm) and with response times in the nanosecond range. An important factor in the overall power efficiency of LEDs and lasers is the series resistance of the device. This resistance R is composed of the contact resistances and the spreading resistance within the device structure. In formula (2.19), AP is then the sum of the electrical energy applied, I V , and the loss energy I’R. This leads to the power efficiency:

(2.20) If I > I,, then

(2.21)

80

HERBERT F. MATARE

Obviously, the resistive dissipation l R / V is most important at higher temperatures and high currents. Maximum power handling, therefore, is conditional upon efficientcooling of the device.This leads to stringent requirements with respect to contact quality and provision for efficient heat conduction. Especially for room temperature operation, the technology to guarantee sufficient thermal contact and heat dissipation has led to numerous approaches which also involve the prevention of device degradation under heavy load conditions. This is essential above all where high power is required, as in devices for infrared communications over distances in air or space and long fibers (see Sections 3,C and 3,D, on contact technology and degradation). It is of minor importance for LEDs in the display field because these devices work at relatively low current densities. B. Radiation Pattern As is well known, the radiation patterns of lasers are more complex than those of LEDs. The onset of stimulated emission results in a multiplicity of peaks from which the laser line emanates above threshold. The Fabry-Perot cavity of miniature dimensions is a complex structure with many possible flaws and reflection properties. A typical far-field pattern shows that transverse and parallel modes increase with increasing width d of the active zone, as the number of modes increases also with increase in current density. The wavelength spacing between adjacent modes can be obtained in a simple fashion from the general relation for longitudinal modes:

n l = 2l.p

or

A12Lp = l / n

(2.22)

where n is an integer, L is the cavity length, and p is the refractive index. As the refractive index is a function of A (dispersion),the mode spacing AA/A can be expressed as

(2.23) A more detailed association of the fine structure of these modes with the mode numbers for transverse perpendicular, transverse parallel, and longitudinal modes has led to a relatively good fit of these with actual laser patterns [see, e.g., Panish and Hayashi (2.17) and literature given]. With DH lasers, a remarkable improvement with respect to mode control was achieved as compared to the erratic emission spectra of early structures, in which the spectral distribution lost all systematic relation to the cavity form and dimensions (2.18).

LIGHT-EMITTING DEVICES. I1

81

A further improvement in mode control was achieved by the large optical cavity DH laser or the “separate optical and carrier confinement heterostructure” (SCH) described by Casey et al. (2.29). The essentials of this structure were described in connection with Fig. 2.4. The relatively thick and symmetrical optical waveguides at both sides of the recombination zone establish a predominance of the fundamental transverse electrical (TE) mode. The spacing of the longitudinal modes in the spectrum indicates dominance of a single filament. Obviously, with increased current density and output power, the satellite modes increase in strength and a number of undefined spectral peaks appear. INTENSITY (relat )

5

4

3

1

8

FIG.2.6. Typical laser output spectrum with identification of modes (see text). GaAsstripe, homojunction laser, J,,, = 150 mA, J = 215 mA. T = 77°K. After Panish (2.17) and literature.

In Fig. 2.6, a typical case is shown for which the dominant modes can be identified according to the relations

(A%

= L , 8 ,y -

L a + 1.8. y

= J2/2WC

(2.23)

(An),

= l a , p, y -

La, /I+ 1. y

=L 2 / 2 m

(2.24)

(A47

= kl,8. y - A@,8. y +

= A2/2nLF

(2.25)

1

Here al, PI, and y are the mode numbers for transverse perpendicular, transverse parallel, and longitudinal excitation, respectively. L is the cavity

82

HERBERT r'.

MATARB

length; p is the equivalent refractive index which accounts for the dispersion; c and c' are constants dependent on the focusing of the radiation perpendicular and parallel to the junction plane. With increased current density, unidentified satellite modes increase in intensity and it is a function of the cavity perfection if one achieves dominance of the desired peaks in the spectrum. Irregularities, flaws in the p-n junction,dislocation clusters,etc., cause multiple reflections andfor scattering. For the infrared optical region, the size of such inhomogeneities is easily within the range of the emitted wavelengths, and therefore the possibility arises of Tyndall- and Raman-type scattering and polarization as in molecular optics. The polarizability a is a sum of both effects and can be represented for harmonic excitation vo as a = a'

+ 2aR cos(2nvo+ 6)

(2.26)

where aTand aR are generally the components of the polarizability tensor or deformation tensor. The subscripts T and R designate the Tyndall and Raman components, the latter with the known cosine factor (2.20). With a light injection, e.g., represented by the electrical field component: E = Eo cos 2nvt

(2.27)

the induced electrical moment is p = aE = {a'

+ 2aRcos(2nvot + 6)}Eocos 2nvt

(2.28)

Since 1 2 cos a cos /I = cos(a + p) + cos(a - /I)

d

this is equivalent to

1p

= aTEocos

2nvt + aREo{cos[2~(v + vo)t

+ 61 + cos[2n(v - vo)t - S]} (2.29)

We see that the excitation with the light frequency v results in the formation of combination frequencies v + vo and v - yo. As in the case of highfrequency modulation and side-band formation, one must expect numerous combination frequencies, depending on the scattering mechanism. Obviously much work has been devoted to an improvement of the purity and linewidth of the laser radiation. At this time, it seems that the DH structure, either with LOC or separate optical and carrier confinement (SCH), allows us to control the far-field pattern to some degree. A highly polarized beam and a remarkable improvement in angular divergence can be obtained by the distributed feedback lasers. In this case, the optically active Ga,Al, -,As layer has periodic corrugations with a grating distance or corrugation period of a multiple of the wavelength emitted: A = P(J0/4

(2.30)

LIGHT-EMI'ITING DEVICES. I1

I

GaAl AS n=3.2

I

1

83

I

I FIG.2.7. Schematic of corrugated laser structure interface. Rays scattered from the waveguide with phase difference b A sin 0. For A + b = p ( A o / n ) ( p = 0, 1, 2, . . .), coherent wave front will form.Exit angle @ into air limited by condition sin = n sin B = p ( L o / A ) - n. =I

where A is the grating period, do is the free-space wavelength, p = 1,2,3,.. . , and n is the refractive index. One has cophasal scattering of the forming beam under the condition that A

+ b = p(Ao/n)

(2.31)

b = A sin 8 (see Fig. 2.7) is the phase difference on the forming plane wave front due to the difference in refractive index of the two sides (GaAs and GaAlAs (2.21). From Eq. (2.31), we derive the condition,

P A0 sin 6 = -A n

-

1I 1

(2.32)

For second-order Bragg scattering, 1,/n = A and sin 8 - p -

1

(2.33)

with p = 0, 1, 2, .... The exit angle into air (see Fig. 2.7) is given by the Snellius relation : sin # = n sin 6 = PRO//\ - n (2.34) and for GaAs ( n = 3.6) only beams with 18 I 5 16" will exit, as usual. For light exit orthogonal to the junction, the optimum grating spacing is A = l(&/n)

where I = 1, 2, 3 . . .

(2.35)

84

HERBERT F. MATARE

Corresponding to the Bragg condition,

(I

1 results in radiation only normal to the plane of the grating). For technical reasons, A is chosen a multiple of A0/n, for example, 4693 A in the case of a GaAs/Gao,,Al0,,As structure. This corresponds to a fourth-order DFB grating (for A. = 8447 A; A = 4693 = 4A0/3.6) (2.21). The results of this technique are remarkable with respect to polarization and linewidth. The output beam is found to have an angular divergence of only 0.35" in the dimension determined by the grating and is virtually loo?,', polarized, with the electrical field vector parallel to the grooves of the corrugated feedback structure. Beam divergence in DH structures with small active layer thickness (0.18 pm) is generally much greater. Parallel to the junction plane, B , , = 9" was measured and perpendicular to the junction plane, e, = 50" (2.22). The latter value corresponds to the angle calculated for the TE mode emanating from a slab waveguide into air [see also (2.23)].Also with respect to linewidth, the distributed feedback laser (DFL) is superior to the DH and SH lasers. While individual modes in the DH structure can have a similarly small linewidth of a few Angstroms (see Fig. 2.6), it is the envelope of the main peaks which gives an actual linewidth of often 50 A or less (see also (2.17), Figs. 39,411. After the technique of the preparation of periodic corrugations was sufficiently developed, more work was done on the distributed feedback (DF) laser. The periodicity generated by an interference fringe pattern on photoresist and an ion milling or etching process is used to carve out the fine grooves. As a next step to the single heterojunction structure (Fig. 2.7), it appeared that the DH laser and the separate confinement heterostructure would allow a further decrease in threshold current or room-temperature operation while early structures had to operate at 77°K (2.24). Figure 2.8 gives a typical layer sequence for the grating-coupled (DF) DH laser. The layer sequence can be changed without much difference in performance. For example, the sequence (starting with the substrate) =

n -GaAs +

3Ga0. (2 pm) n-Alo. 2Ga0,88A~(0.36 pm) p-GaAs(O.13 pm) active zone P-A'O.,,G'O.~~AS(O.~~ pm) 1 (corrugated interface) p-A10.3Ga0.7As(2.0 pm) p+-GaAs(l.0 pm) contact n-AIO.

85

LIGHT-EMITTING DEVICES. I1

1

ACTIVE LAYER

/////////

n-type

GaAs

n-Ga,,,

Al,,, As

~

-155p.m

.-

n-Ga,,Al,,As

-

0.6rp.m

-

&GaAs

1.6p

0.23pm

‘CON TACT

FIG.2.8. Schematic of a separate Confinement heterostructure laser (SCH) with periodic corrugation in the optical cavity. After Casey er al. (2.25).

(2.25)leads to similar results. Here the corrugated structure is entirely within the GaAlAs layer. This is different from the original concept, where the corrugated layer is part of the active region. It became apparent that the filtering action of the distributed feedback is also operating as a structure within the light path or even as a reflecting layer. In this way, the crystal damage and resulting nonradiative recombination centers within the active layer due to the ion milling or etching of the corrugation can be avoided. This is important for low-threshold and roomtemperature operation. Thus the sequence n+-GaAs n-Gao,7Alo,3As p-GaAs P-Ga0.83A10,1 7As

pmGa0.93A10 .O p-Ga0.7A10.3As

(active zone)

I

1 (corrugated interface)

clearly separates active and corrugated areas (2.26).While these structures are still sequentially grown on the substrate such that the entire optical confinement layer is used for the DF action, Reinhart et al. (2.27) showed that the distributed feedback layer can be externally coupled to the lightemitting layer. In this case, it works as a Bragg reflector. Here the corrugated layer is taper-coupled to the optical cavity. A schematic is shown in Fig. 2.9.

86

HERBERT F. MATARE

FIG.2.9. Construction of the taper coupled laser (TCL) with corrugated emission region (Bragg reflector, c( = Bragg angle). Layer sequence, type, and thickness are indicated. After Reinhardt et d.(2.27).

The light from the active region is wave-guided into the corrugated layer and beamed out from there. Figure 2.10 shows the excellent monochromaticity or optical linewidth of the taper-coupled laser as compared to a normal DH laser. While the DH structure has a half-power spectral bandwidth of more than 300 A (2.17),the TCL laser has a half-power spectral bandwidth of 5 1 A. Also with respect to the angular distribution of the far-field pattern, the TCL laser represents considerable progress. Perpendicular to the junction plane, a half-power of I0.3"was achieved.

INTENSITY (arb units) 5--

4

--

3

--

2

--

I

--

DH

LASER

T=

~

O

1

8300

K

1

1

85M)

1

1

8700

1

.

8poO

1

,

9lW

-

A'

FIG.2.10. Comparison of spectral purity of DH laser and TCL laser.

87

LIGHT-EMITTING DEVICES. 11

6-

5-

4-

3-

2-

1-

t

- I0

1

-5

I

-1

l

l

0 I

I

5

I

c

10

DEOREES

0,

FIG.2.1 1. Comparison of far-field pattern for DH and TCL laser. DH laser: B l l = 9". SO"; TCL laser: OL = 0.3".

2

Figure 2.1 1 shows a comparison of the far-field pattern of a DH laser and the TCL laser described (2.27). For the DH laser, 8,, is 9", whereas eL= 500 (2.22). The method of using Bragg reflectors for mode suppression and to decrease spectral linewidth is being perfected. For example, a recent structure is a grating-coupled ring laser based on a single heterostructure laser. Here a layer sequence (n)GaAs

(P)GaAs (P)Ga0.6A10.4As

\ (corrugation) ( p ) GaAsI is used, but the grating and laser filament are not orthogonal or the grating period A is not an integral multiple of 4 2 (3, = guide wavelength). In this case, distributed feedback operation is avoided (no coupling of oppositely traveling waves) and the grating will only couple out power in collimated polarized beams (2.28).

88

HERBERT F. MATARE

A number of new forms and methods to use this technique appear while the questions of degradation and lifetime of the device loom in the background. Finally, we mention a development which is relatively new and has allowed us to lower the threshold current values once more. This is the method of the buried heterostructure. Here a mesa-etched laser is surrounded by higher bandgapiGaAlAs material which serves to confine both the injected carriers and light in two dimensions. No carriers are lost by diffusion parallel to the junction plane as in usual stripe geometry lasers and the transverse mode is also guided (2.29).The schematic of this structure is shown in Fig. 2.12. These structures show also a small temperature dependence of the threshold current as compared to DFB structures [and a small wavelength variation with temperature (0.72 °, as compared to 3.4 A/deg)] (2.29).A variant of this structure, but fabricated by a single LPE process, was also indicated by these authors (2.30). Here a groove is etched into the double heterostructure and later filled by LPE with an active double layer GaAlAslGaAs. The role of nitrogen as an active impurity in AI,Ga,-,As lasers has recently been confirmed even for the direct transition range. For x = 0.39, it was found that higher intensity is measured when nitrogen in a density of 1 x 10'' cm-3 is added (2.32). IP)

p-3Pm9

Ga, Al, ,As

\

I

I

I

I

I

I p

- Ga,,AI,,As

ACTIVE LAYER

--FIG. 2.12. Schematic cross section of buried heterostructure distributed-feedback diode laser. After Burnham et a/. (2.29).

LIGHT-EMITTING DEVICES. I1

89

C . Modulation Frequency As in the case of the Gal -.Al,As/GaAs LED, the search for an extension of the modulation frequency has led to numerous efforts to shorten the response time of these devices. In general, a doping type which enhances the output power increases the response time. Silicon as an impurity, as mentioned earlier, is known to enhance output power, but modulation capability is better with germanium in the DH or homojunction [modulation capability for Si Zn dopant is 50 MHz and for Ge dopant is up to 250 MHz for DH structures (2,3211. In the case of stimulated emission, the lifetime of electrons for carrier densities above the degeneration limit may already become smaller than 1 nsec. In the injection laser, the recombination time is inversely proportional to the energy stored in a natural oscillation of the resonator and accordingly decreases with increasing density of the injected carriers. Measurements of the time constant t of laser diodes confirm that z decreases with increasing current density in the device (2.33). For very high modulation or bit rates, the laser can therefore outperform the LED by a factor of 10 and more, but the price to be paid in device complexity and survival time has to be considered in all trade-in studies in actual optical communication systems. Recent investigations of the delay times in external cavity-controlled GaAs lasers have confirmed the fact already mentioned that the delay time is a function of the laser intensity. For coupling of the laser to an external optical cavity, fiber, or other optical equipment, one diode side, e.g., one side of the internal Fabry-Perot cavity, is anti-reflection-coated. The coating lowers the intensity of the electromagnetic wave in the diode cavity and thus increases the delay time between the current pulse and light pulse. In addition, there is a strong dependence of the laser emission on temperature. The band gap energy decreases with temperature such that the emitted photon energy declines as T is increased, e.g., from 1.45 eV to 1.35 eV or over 600 A for T from 77" to 300°K.However, the emission peak for increased drive current moves toward higher photon energy for T = constant. Since, in general, higher drive currents involve junction heating, these two effects-band gap decrease and higher energy pumping action involving intraband states-are in opposition. For a tripling of the diode current from 10 mA to 30 mA, the output center frequency changes; e.g., from 1.365eV to 1.435eV or from 9084 to 8461 A this is over 600 A [see also (2.11), Fig. 41~1.These two effects will partially cancel each other when the shift to lower frequency due to heating is of a similar strength as the shift to higher frequency by increased current. It was shown that the latter effect is found only in heavily doped and compensated material and only upwards

+

90

HERBERT F. MATARE

from a minimum current density. Casey and Bachrach (2.34) have shown that the peak shift follows an exponential law: J, = J, exp Shv,

(2.36)

where J, is the unchanged intensity, S is the logarithmic slope constant (170 f 10 eV-'), and hv, is the amount of shifted energy. For the DH lasers measured, the shift is found upwards from 10 mA diode current. 0.1-

FIG.2.13. Time resolved spectra for a DH laser. p-Type GaAs layer doped with Si. After Casey and Bachrach (2.34).

91

LIGHT-EMITTING DEVICES. I1

psec pulses were used above 10 mA to avoid sample heating. U p to I = 10 mA, the integrated intensity is proportional to 12, whereas it is proportional to I for higher current values. This is in agreement with the predominance of the space charge recombination current for small drive current and with the diffusion current or a switch from a dependence as exp(qV/kT) to exp(qV/2kT) for high drive current. The influence of the drive current density on the delay time q,can be shown directly by measuring the time decay from steady state at a particular emission energy or as time-resolved spectra. The p-n junction is pulsed to the desired current level and then left short-circuited with the pulse (turn-off time). Due to the confinement of the carriers within the active zone in the DH structure, carriers remain localized and cannot diffuse into the bulk during recombination. Therefore, the delay of the radiative decay after pulse decay indicates directly the response of the active region with respect to recombination time as a function of different injection (pulse) levels. Figure 2.13 shows a set of intensity (optical frequency) versus energy curves for increasing current. From 10 mA the shift J , = J , exp Shv, toward shorter wavelength is observed, and at the same time the delay measured after pulse interruption decreases. The time delay between the application of the current pulse and the onset of lasing is of equal importance (turn-on time). Figure 2.14 shows the current pulse and the delayed light pulse with 'sd (turn-on delay time) and

t

N

Current

I

I

time t

I

I

'

4-474

FIG.2.14. Definition of delay time

T,,

and decay time

T:

T: (turn-off delay time). The turn-on delay Td has been measured for diffused junction (DJ), single heterostructure (SH), double heterostructure (DH), and large optical cavity (LOC) lasers. It appears that Td is a function of temperature only for DJ and SH structures and that a critical temperature T, will increase Td by a factor of 10 or more. Below T,, the delay time Td can be represented by

(2.37)

92

HERBERT F. MATARE

where I is the current pulse height, l t h r is the threshold current, and T~ is the spontaneous lifetime of injected carriers. If I/Ithr‘v 1.02 as the smallest controllable value, then for a spontaneous lifetime of 1 to 2 nsec, ‘Cd should be the 4 to 8 nsec range. ‘Cd (see Fig. 2.14) has been measured as follows: T
Laser type DJ, SH DR, LOC

T~ 5 5d

T>T,

8 nsec

T~

5 15 nsec

-

50-100 nsec

~~

Rossi et al. (2.35) point out that formula (2.37) gives good results for T c T, only in the case of homojunctions, diffusedjunction, and single heterojunction lasers. It is valid at all temperatures for DH and LOC lasers. There are some basic differences between these laser types with regard to threshold current versus lasing energy, also called gain projle. Figure 2.15 is a schematic of the two different llhr(eV)curves for DJ or SH and DH or LOC lasers. These functional relations are measured with the laser mounted in an external cavity, i.e., between a mirror and a grating which can be rotated. A spectrometer takes power via a beam splitter. In this way, one can I

-.I

I

I

I

,

DJ rHJ

br 8

.~

I/

FIG.2.15. Gain profile for DJ and SH structure, and for DH and LOC structure with response (light pulses) to current pulses at different levels.

LIGHT-EMITTING DEVICES. I1

93

measure the threshold current for different grating positions or lasing frequencies. The functional form of the gain profile for the DH and LOC structures is basically different from the one for single homo- or heterojunction lasers. While the former are single valued with respect to l , h r , we see that the gain profile of single homo- or heterojunction lasers has double-valued lower frequency threshold currents. It is in this range also that abnormal delay effects occur. In Fig. 2.15 we have plotted the measured effects. For the DJ or SH lasers, two energy values, hv, and hv, ,are singled out, for which the current is changed from the lowest value I, to the highest, 14,On the low-frequency side, hv,, we notice a continuous decrease in t d as I increases, until for I4 one has exceeded the gain profile and no lasing occurs. On the higher optical frequencyside, h v 2 , lasing does not start at I1 but at I 2 and continues equally for all I values upwards. I, is the critical current value at hv, since lasing stops here with increasing current. The external cavity (EC) operation for DJ and SH devices is more complex than is the case for DH or LOC devices, where the basic device structure is already mode suppressing and where higher spectral purity is achieved. A variation in timing of the EC for single junction lasers probably results in mode distortion and delay effects which require a frequency change for continued laser operation (abnormal delay region!). An increase in device temperature, moreover, results in a decrease of the optical frequency, in accordance with the band gap change: -0.4 meV/"C. In DH and- LOC structures, the timing by EC variation is single-valued and allows continuous timing for 40 meV (300 A). It is interesting that temperature increase acts strongly on the gain profile in DJ and SH lasers. At a sufficiently high temperature [e.g., 64"C, see (2.35)], the gain profile of these devices resembles that of the DH or LOC lasers, i.e., it has no abnormal delay regions. By the use of external cavities, long delays can also be produced in DH and LOC devices. This is probably due to a reduction in electromagnetic wave intensity within the active region. Saturable absorption can produce Td values over 10 nsec in DH lasers with both sides cleaved (no EC). With EC, T~ ranges up to 30 nsec for increased ] / ( I Especially, the use of antireflection coating (AR) for efficient coupling to the external cavity will cause a further increase in T,,, up to 60 nsec (2.35). D. Comparison o j h e r and LED With these findings in mind, we briefly review the status of the infrared LED field as compared to the laser, in terms of operational gain or figure of merit : watts/second corresponding to the gain-bandwidth product. This is

94

HERBERT F. MATARE

an important figure for all applications in infrared communications. As threshold current values for lasers have decreased and room-temperature operation is common, the application of lasers to high data rate transmission can solve numerous problems in optical communications via air and fibers. One obvious advantage of the laser is the possible response speed in the nanosecond range and an optical energy in the watt range in pulsed operation. With differential quantum efficiencies between 40 and SO%, the laser has an obvious advantage also in power handling and optical purity. For drive currents in the 5 to 20 kA/cm2 range, output powers from 1 to 5 W are produced. This is possible with delay times in the nanosecond range, as we have seen. However, in external cavity couplers, this value can be considerably higher and delay times of 20 or more nanoseconds are usual. In this instance, the figure of merit would be in the 10' W/sec range. However, the pulse repetition rate is at best in the several kHz range to avoid laser heating and rapid degradation. The pulse duration or pulse width of a few nanoseconds has to be maintained by a trigger-pulsewidth of 300 or more nanoseconds. Relatively high-voltage power supplies are necessary to drive the device into the desired current mode. The consequence is that lasers are good transmitters for digital information of high data rate, but that the limitation of the pulse repetition rate excludes their efficient use in, e.g., TV transmission. For a transmission capability of, e.g., 5 MHz, a three times higher repetition frequency, i.e., 15 MHz is required. Here the LED has remarkable advantages, since C.W. operation at high power levels also approaching the watt level is feasible. Modem high efficiency GaAlAs/GaAs LEDs, sufficiently heat-sinked and correctly mounted, can have output power in the 50 to 100 mW range for current densities of a few 10oO A/cmz. In pulsed operation, higher duty cycles are possible. In addition, the problem of side modes and cavity resonances is not present and the spectral linewidth (25 nm or 250 A) is no impediment to the optics used in most communication sets. In the case of optical fiber communication, it is obviously a problem to efficiently couple into a small fiber. Here junction emission, stripe geometry, and LOC-type layered devices are used. Progress in this field is very decisive at this time and will be mentioned later (2.32). With external quantum efficiencies in the 10% range (2.36), infrared LEDs have become a major part of the transmitter for the growing field of optical communications and complement effectively the possibilities offered by the semiconductor laser. E. Visible and Far-Infrared Lasers

To complete the laser survey, we have to mention the intense work done in the field of GaAs,P, -,, In, -,Ga,P, Gap, In,Ga, -,As, GaAs, -,Sb, and

LIGHT-EMITTING DEVICES. I1

95

related quaternary compounds, as well as in the area of PbS,Se,-, and Pb, -,Sn,Te for the far infrared. The CVD or vapor phase deposition of GaAs,P,-, on GaAs was already sufficiently developed when laser structures were tried. One such device has the layer sequence (2.37): (n)GaAs:Te/(p)GaAs:Zn/(p)GaAso.gPo,:Zn active junction In this case, the use of the ternary compound GaAsP as top layer was motivated as an electronic and optical confinement layer. Craford et al. (2.37) state that improved laser action (as compared to a H.J. structure) should be possible for the case that the ternary layer provides confinement and that the minority carrier diffusion length is short compared to the width of the confinement region. But even with a graded GaAs-GaAsP interface, no improvement in laser efficiency was found. The authors wondered why the confinement works in GaAIAs-GaAs lasers and not in this case. One important aspect is, however, that in the GaAIAs-GaAs case the large-gape side is used for the injection. The consequence of this for the injection efficiency was pointed out earlier (Part I, Section 5). In addition, GaAlAs can be grown on GaAs due to the very small lattice mismatch (0.14% difference in lattice constant as compared to 3.6% for GaAs-Gap) without prohibitive dislocation density. Thus a structure like ,(Te) (P)GaAs0.9Po.1:Zn/(n)GaAs(Te)/(n)GaAso.~po, would be better, but could also not be a useful device as the active p-n junction would be loaded with nonradiative centers. A graded layer cannot help here since in this case the active junction would no longer be a good injector. Therefore, it is not astonishing that this type of device was not superior to a GaAs-GaAs homojunction laser.

GaP:N In indirect material, the isoelectronic trap introduces trapped electron wave functions with properties of the conduction band Bloch wave functions for the central region (r)of the Brillouin zone (see Part I). Stimulated emission is therefore in the realm of such “quasi-direct semiconductors and there have been cases of high gain value in such materials. Donors and acceptors behave differently and it seems that acceptors introduce less nonradiative centers than donors. Figure 2.16 sums up some results of photoluminescence decay, diffusion length (L), and cathodoluminescence efficiency (CL-q) measurements on N-doped Gap. One set of curves is for the donor-type (Te:N) material, the

”

96

HERBERT F. MATARE

- ---

ACCEPTOR: Zn,N DONOR: Te.N

lo1

10'

i

/*-

/ /

1oa

/-

/* 1Q

10

FIG.2.16. Photoluminescentdecay (PL) in nanoseconds. DiITusion length ( L )and relative cathodoluminescent efficiency (CL-q) in percent for GaP:N (-.-) with Zn doping, (---) with Te doping.

other for the acceptor-type (Zn :N). The cathodoluminescent efficiencies are close to values found for the photoluminescent efficiencies (air). It can be seen that the maximum CL efficiency is reached with Zn, N-doped material. At the peak (at about lo'* cm-3 doping) the diffusion length is still in the 3-pm range and the decay time is in the 30-nsec range (2.38).It is in this area that stimulated emission is most probable. Wolf et at. (2.39) have pointed out that measured gain values of up to lo4 cm- cannot be explained by the A-line activity (excitan bound to an isolated N trap). It is also pointed out that the enhancement of the ",-line emission, measured under higher

'

LIGHT-EMITTING DEVICES. I1

97

excitation, is indicative of a new recombination process, especially if combined with a reduction of the radiation lifetime 7,. In this context, it should be pointed out that nitrogen as an impurity in direct gap Al,Ga, -,As is effective in enhancing emission intensity (2.40) and leads to laser emission by optical pumping, similar to the case of GaAs,-,P, (2.41). This demonstrates that nitrogen adds a new recombination mechanism even if the Er transitions dominate. The measured lowering of the emission energy by 14.4 meV indicates that shallow energy levels are involved and that the role of N as an impurity is to enhance emission via phonon-assisted transitions otherwise inactive in Al,Ga, -,As (x = 0.39). In1 - .Gu,P

Laser operation in direct ternary compound material with a useful optical output at wavelengths shorter than 6400 A is of importance for many optics applications. In,-,Ga,P is direct over a large range of x values (0 I xI 0.74) and should operate from 900 nm to 580 nm (yellow range). Early, low-temperature lasers were of pure quality due to the lattice damage caused by the diffusion of a p-type layer from an In + 10% Zn source (2.42). Considerable progress was made by using the heterojunction In,Gal -,As/In,Ga, -,P. In this case, a wide optical frequency range (0.9 to 1.15 m) can be covered (see Fig. 4.20, Part I) and in keeping In,Gal -,P in the direct composition range (y 5 0.74), all components are direct semiconductors. A very close lattice match can be produced with these ternary compounds with the respective lattice constants: InAs : 6.0571 GaAs : 5.653 1

(6 2: 6.9%)

I ~ P : 5.869) GaP : 5.457

(6 2: 7.2%)

(see Table 4.Q Part I) because each ternary compound covers a wide range of the lattice constant of the other ternary compound. For ideal roomtemperature lattice matching, one should have, according to Vegard’s law: y = 0.483

+ 0.969~

(0 5 x I0.533; 0.483 I yI 1)

The actual structure, lasing at room temperature, had a layer sequence as follows (2.43): n - GaAs(substrate)/n - In,Ga, -,P(graded)/n - Ino,,6Gao.32P /In,. 16Ga0. ty ) / p - In0 .6eGa0. 32 p

/P - 1n0.16Ga0.84As(~p)

and was produced by CVD.

98

HERBERT F. MATARE

At an optical frequency of 1.025 pm, the linewidth was 10 A. Threshold currents are as low as 15 kA/cm2 in the best cases. In,Gal-x had been deposited earlier by CVD on GaAs substrates and lasing was achieved at 1.06 pm, but only at 77°K for Zn-diffused junctions (2.44). The frequency range is of interest in connection with the second minimum Si02 fibers. The increase in layer perfection and laser efficiency, possible in LPE process-controlled structures, can also be produced with In, -,Ga,P grown by LPE on GaAsl-,P, or GaAs. The major problem to solve here is the LPE deposition of the second layer when some lattice mismatch forms dislocations in the neighborhood of the active junction. It has been found that lattice matching between In,-,Ga,P on GaAsl-,P, substrates can be improved when the ternary compound is rendered a quaternary by the incorporation of a small amount of As in the LPE melt. The quaternary, In,-,Ga,P,-,As, (z 0.01) can be used as a wide gap emitter on GaAs,-,P,. It permits LPE growth on commercially available material (CVD, GaAsP) and has led to a significant improvement with respect to laser operation at low threshold values. The substrate layer mole-fraction relation for lattice match is given by

-

x = 0.52

+ 0.48~

(z being very small, it is not considered here). Lasers made by LPE on VPE-substrate surfaces have operated at wavelength I < 6300 A, with threshold currents J t h r 2: 6.2 x lo4 A/cm2 at 77°K (2.45). Further progress was reported by LPE growth of double heterojunctions of Inl -%Ga,P, -rAs, on GaAs, - ,P,. Layers of varying As mole fraction and therefore varying energy gap can be grown sequentially while maintaining the lattice match. This is apparent from the fact that the lines of constant lattice parameter in the x-y-z scaling between the compounds intersect with the lines of constant energy gap (Fig. 2.17). The energy gap data for the binary corners are known and the energy gap curves have been drawn as smooth translations of the ternary boundaries while the lines of constant lattice parameter are obtained by applying Vegard’s law across the alloy system: a, 2: 5.869 - 0 . 4 2 ~+ 0.189~+ 0 . 0 1 5 ~ ~

Lasers made according to this scheme using double heterojunctions showed threshold values of Jthr N 3.6 x lo3 A/cm2 at a wavelength of I Y 5920 A (yellow) at 77°K. Operation in the orange-yellow region of the spectrum was extended to 200°K. Room-temperature lasing was achieved in pulsed operation with In, -,Ga,P1 -,As, DH structures for x N 0.77 and z = 0.21 at threshold values Jthr2: 2 x lo4 A/cm2

LIGHT-EMITTING DEVICES. I1

99

FIG.2.17. Nomogram for lattice constant and energy gap as a function of composition of the quaternary alloy system In, -xGa,P, -,As, and the ternary system GaAs, - ,P, (77°K). After Coleman et a/. (2.46).

but, for a somewhat longer wavelength (2.47)

Operation to longer wavelengths, in particular beyond the red toward the infrared range, is easier as one can finally apply binary compounds instead of ternary and quaternary alloys for the wider gap confinement layers (InP). For a double heterostructure GaO,

1

*In,, 8.SAS0.23

77

/InP

emission at 1.1 pm with Jthr'v 2.8 kA/cm2 was measured (2.48). For this frequency range, the stoichiometry is simplified as the lattice constant of the quaternary easily matches the one of I n P over the range from 0.92 to 1.7 pm. (See Fig. 2.17 at the corner for small x and small z, the 1.2 eV line cuts the InP lattice constant line while y values are high or near the GaP comer). and the related heterojunction cases GaAs,Sb, -, Gal -.In,As/Ga, -,In,P and GaAs, -,Sb,/Ga, -,,AI,As, -,Sb, have a150 been used in this frequency range around 1 pm.

100

HERBERT F. MATARE

Again, the growth of heterojunctions is improved in perfection when the active layer (GaAs, - ,Sb,) is sandwiched between quaternary compound layers, allowing to adapt the lattice constant while increasing the band gap without complex grading procedures (2.49). Lowest threshold values were 2100 A/cm2 and differential quantum efficiencies reached 9%. In a subsequent publication (2.50), these authors reported C.W.room temperature operation of these lasers and thus added one more laser structure of importance to the GaAs/GaAlAs DH laser used between 800 and 950 nm. As the second and lower optical fiber (SiOz) minimum is near 1.1 pm, such a device fills a definite need. Layer sequence chosen was as follows: p-AI,Ga, -,As, -,Sb,(Ge)/p-GaAs, -ZSb,/n-A1,Ga, -,As, -,Sb,(Te) Typically, y = 0.4; x = z x 0.12; active layer thickness = 0.45 pm. The layers are grown on a GaAs substrate overgrown with three layers of GaAs, -,Sb,(Te) step-graded with z = 0.025,0.058,and 0.093 for stress relief. Jthrwas down to 2.1 kA/cm2. Differential quantum yield qD 2: 12% and C.W. - q = 11% are excellent values. PbSn T e ; PbSSe Finally, we direct our attention to the important work which was carried out with narrow gap semiconductor compounds to achieve laser action in the far-infrared optical range. Lasers made from PbSl -,Sex were operated C.W. as early as 1971 at liquid helium temperature (x = 0.18, and also, 0.39).Magnetic field tuning over a range from 2108 cm-' to 2120 crn- was achieved for a 4.7 pm PbSo,82So,lslaser. Wavelength measurements showed lasing capability at 10°K between 3.33 and 5.4 pm with the related compounds Pb,_,Ge,Te and Pbl-.Ge,S (2.51). The application of the LPE process and the growth of heterojunctions in this material have greatly enhanced the power output, originally in the microwatt range. Present output is a factor of 10 to 100 higher. Threshold reduction was achieved by the growth of heterojunctions : n-Pbo.87Sno.,,Te/p-PbTel p-Pbo. Sn o . ,Te/n-PbTe

,

(substrate)

and double heterojunctions:

,

n-PbTe/n-Pbo,,,Sn0. ,Te/p-PbTe p-PbTe/p-Pbo.,,Sn0, ,Te/iz-PbTe \

(substrate)

Doping ranges are between 10' and 6 x 10' cm-,. To avoid the diffusion of Pb into the substrate when undoped or lightly doped PbTe is used, one can start with a thallium-doped substrate.

LIGHT-EMITTING DEVICES. I1

101

Threshold values as low as Jlhr2 150 A/cm2 at 8°K and JthrY 1.5 kA/cm2 for 77°K have been achived (2.52). Continuous wave operation was recently extended to 114°K (2.53)with devices made by molecular beam epitaxy (MBE). Even distributed feedback lasers have been fabricated from this material with response in the 750-cm-’ range. The DF laser has a much smaller frequency variation with heat-sink temperature change than the FP laser (usual Fabry-Perot type) (2.54). Cadmium-diffused PbSnTe lasers have also successfully operated in the 940-cm- range (2.55). REFERENCES FOR SECTION 2 2.1. H. Kroemer, frog. Semicond. No. 4. 1-34 (1960). 2.2. See, e.g., J. Bok, ed., Plasma Egects in Solids; Inr. Con5 Phys. Semicond., 7th, No. 1 (1964-1965). 2.3. H. F. Matare. U.S. Patent 2,944,167 (filed October 21, 1957) (patented July 5, 1960). for Sylvania Electric Products, Inc. 2.4. H. F. Matare, ’‘ Fortschritte der Hochfrequenztechnik” (M. Strutt, F. Vilbig, and F. Riihmann, eds.) pp. 347412. Akad. Verlagsges. Frankfurt a. Main, 1960. 2.5. W. Heywang, Ger. Patent No. 1,248,826 (filed April 30,1958) (patented March 14, 1968). 2.6. E. E. Loebner, l E E E Trans. Electron Devices 23(7), 675-699 (1976). 2.7. R. N. Hall, l E E E Trans. Electron Deuicrs 23(7), 700-704 (1976). 2.8. M. J. Adams and P. T. Landsberg, in “Gallium Arsenide Lasers” (C. H. Gooch, ed.), pp. 5-79. Wiley (Interscience), New York. 1969. 2.9. H. Kressel and H. Nelson, Phys. Thin Films 7, 115-256 (1973). 2.10. M. H. Pilkuhn and H. Rupprecht, J. Appl. Phys. 38(1), 5-10 (1967). 2.11. M. B. Panish, I. Hayashi, and S. Sumski. Appl. Phys. Lett. 16(8), 326-327 (1970). 2.12. H. Kressel, H. F. Lockwood, and F. Z. Hawrylo, J . Appl. Phys. 43(2), 561-567 (1972). 2.13. M. B. Panish, H. C. Casey, Jr., S. Sumski, and P. W. Foy, Appl. Phys. Lett 22(11), 590-591 (1973). 2.14. H. C. Casey, Jr. and M. B. Panish, J . Appl. Phys. 46(3), 1393-1395 (1975). 2.15. H. Kressel and M. Ettenberg, J . Appl. Phys. 47(8), 3533-3537 (1976). 2.16. F. H. Doerbeck, D. M. Blackerall, and R. C. Carroll, J. Appl. Phys. 44(1), 529-531 (1973). 2.1 7 . M. B. Panish and 1. Hayashi, Appl. Solid State Sci. 4, 306 (1974). 2.18. H. J. Henkel, E. Klein, and H. Kuckuck, Solid-state Electron. 8, 475-478 (1965). 2.19. H. C. Casey, Jr., M. B. Panish, W. 0. Schlosser, and T. L. Paoli, J. Appl. Phys. 45(1), 322-333 (1974). 2.20. See, e.g.. M. Born, “Optic,” p- 392 IT. Springer-Verlag, Berlin, 1933. 2.21. D. R. Scifres, R. D. Burnham, and W. Streifer, Appl. Phys. Lett. 26(2), 48-50 (1975). 2.22. H. C. Casey, Jr.. M. B. Panish, and J. L. Merz, J . Appl. Phys. 44(12). 5470-5475 (1973). 2.23. D. Marcuse, “Light Transmission Optics,” Bell Lab. Ser. p. 305. Van NostrandReinhold, New York, 1972. 2.24. R. D. Burnham. D. R. Scifres,-and W. Streifer, Appl. Phys. Lett. 26(11), 644-647 (1975). 2.25. H. C. Casey, Jr., S. Somekh, and M. Llegems, Appl. Phys. Lett. 27(3), 142-144 (1975). 2.26. K . Aiki, M. Nakamura, J. Umeda, A. Yariv, A. Katzir, and H. W. Yen, Appl. Phys. Lett. 27(3), 145-146 (1975). 2.27. F. K. Reinhart, R. A. Logan, and C. V. Shank, Appl. Phys. Lett. 27(1), 45-48 (1975). 2.28. D. R. Scifres, R. D. Burnham, and W. Streifer. A p p l . Phys. Lett. 28(11), 681-683 (1976). 2.29. R. D. Burnham, D. R. Scifres, and W. Streifer, Appl. Phys. Lett. 29(5), 287-289 (1976).

102

HERBERT F. MATARE

2.30. R. D. Burnham and D. R. Scifres, Appl. Phys. Lett. 27(9), 510-511 (1975). 2.31. Y. Makita, S. Gouda, and H. Jjuin, Appl. Phys. Lett. 29(5), 309-311 (1976). 2.32. M. Ettenberg, J. P. Wittke, and H. Kressel, High Speed Light Emitting Diodes,” Final Rep. 4-1-73-5-30-75, Off. Nav. Res., Contract No. N00014-73-C-0335 (AD-A018 757-NTIS). Dep. Navy, Arlington, Virginia, 1975. 2.33. G. Winstel and K. Mettler, Int. Congr. Phys. Semicond., 7th Sess. 4. pp. 183-193 (1964). 2.34. H. C. Casey, Jr. and R. Z . Bachrach, J. Appl. Phys. 44(6), 2795-2804 (1973). 2.35. J. A. Rossi, J. J. Hsieh, and H. Heckscher, I E E E J. Quantum Electron. 11(7), 538-545 (1975). 2.36. E. A. Ulmer, Jr., Solid-State Electron. 14, 1265-1273 (1971). 2.37. M. G. Craford, W. 0.Groves, and M. J. Fox, J. Electrochem. Soc. 118(2), 355-358 (1971). 2.38. R. N. Bhargava, IEEE Trans. Electron Devices 22(9), 691-701 (1975). 2.39. H. D. Wolf, K. Richter, and C. Weyrich, Solid State Commun. 15, 725-727 (1974). 2.40. Y. Makita, S. Gouda, and H. Jjuin, Appl. Phys. Lett. 29(5), 309-311 (1976). 2.41. D. J. Wolford, 8. G. Streetman, R. J. Nelson, and N. Holonyak, Jr., A p p l . Phys. Lett. 28(12), 711-713 (1976). 2.42. H. M. Macksey, N. Holonyak, Jr., D. R. Scifres, R. D. Dupuis, and G . W. Zack, Appl. Phyz. Lett. 19(8), 271-273 (1971). 2.43. C. J. Nuese and G. H. Olsen, Appl. Phys. Lett. 26(9), 528-531 (1975). 2.44. C. J. Nuese. M. Ettenburg, R. E. Enstrom, and H. Kressel, Appl. Phys. Lett. 24(5), 224-227 (1974). 2.45. J. J. Coleman, W. R. Hitchens, N. Holonyak, M. J. Ludowise, W. 0. Groves, and D. L. Keune, Appl. Phys. Lett. 25(12), 725-727 (1974). 2.46. J. J. Coleman, N. Holonyak, Jr., M. J. Ludowise, and P. D. Wright, J. Appl. Phys. 47(5), 2015-2019 (1976). 2.47. J. J. Coleman, N. Holonyak, Jr., M. J. Ludowise, P. D. Wright, R. Chin, W. 0. Groves, and D. L. Keune, Appl. Phys. Lett. 29(3), 167-169 (1976). 2.48. J. J. Hsieh, Appl. Phys. Lett. 28(5), 283-285 (1976). 2.49. R. E. Nahory and M. A. Pollack, Appl. Phys. Lett. 27(10), 562-564 (1975). 2.50. R. F. Nahory, M. A. Pollack, E. D. Bube, J. C. Dewinter, and R. W. Dixon, Appl. Phys. Letk 28(1), 19-21 (1976). 2.5 I . I. Melngailis and T. C. Harman, “Solid State Research,” No. 1, pp. 5-8. Lincoln Lab. (MIT), Lexington, Massachusetts, 1972. 2.52. L. R. Tomasetta and C. G. Fonstadt, Appl. Phys. Lett, 25(8), 440-442 (1974). 2.53. J. N. Walpole, A. R. Calawa, T. C. Harman, and S. H. Groves, “Solid State Research.” No. 1, pp. 1-3. Lincoln Lab. (MIT), Lexington, Massachusetts, 1976. 2.54. J. N. Walpole, A. R. Calawa, S. R. Chinn, S. H. Groves, and T. C. Harman, in “Solid State Research,” No. 2, pp. 1-4, Lincoln Lab. (MIT), Lexington, Massachusetts, 1976. 2.55. W. LO, Appl. Phys. Lett. 28(3), 154-156 (1976). ‘I

3. DEVICETYPES A N D TECHNOLOGICAL PROGRESS A. Optical Problems The great difference in refractive index between air and the compound semiconductor is at the root of the complex technologies to increase the light output of light emitting devices. According to the laws of optics, each point at the junction plane emits light which meets the semiconductor surface at

LIGHT-EMITTING DEVICES. 11

103

an angle 0 and leaves the semiconductor at the angle of refraction 4. Their relation is given by Snellius' law:

sin 0/sin 4 = l/n

(34 where n is the refractive index with respect to air. Thus, by differentiation, n cos 0 d0 = cos 4 d$

(3.2) The elementary cone of radiation d0 d0 emanating from each junction point will emerge as a cone of solid angle: d4 d4

The total flux within each cone being constant, we have

I(0)/n is the intensity of radiation normal to the device surface: lo. For large values of n, cos 0 is close to one and we obtain from (3.4): I ( + ) d 4 = l o d ( 0 ) cos 4

(3.5)

Since the total flux within the cone is constant,

I(4)= l o cos 4

(34

as is typical for a Lambertian source. The total emitted light power L is found by integration of (3.6), first for an element dw of the angular aperture of the unity cone (see Fig. 3.1).

1

L(w)= I, cos 4 dw

(3.7)

.u

= 2111,

1

'0

cos 4 sin

= X I , sinz u

4 d4

(3.8) (3.9)

For the plane radiation (u = W"), L = nl,

(3.10)

104

HERBERT F. MATARE

uJunction Plane

FIG.3.1. Snell's law of refraction, high index of refraction n, to low index (air, n = 1). Cone of exit with opening angle w. Angle of incidence is 0; angle of exit is 4.

For the refractive index of 3.6, the critical angle 0, for which there is total reflection (into the semiconductor) or sin b, = 1, is, from (3.1), sin 8, = 113.6 N 0.278

(GaAs)

or

8,- 16" For other semiconductors of the 111-IV type, the situation is similar. Compound

Refractive Index

GaAs GaP InP InAs InSb GaSb

3.6 3.37 3.4 3.4 3.75 3.9

For the 11-VI compounds of the type ZnS, ZnTe or CdS, CdSe, n ranges between 2.3 and 2.7. If one considers the total power at the semiconductor surface S and introduces the power P emitted from the unit junction area, then the power

105

LIGHT-EMITTING DEVICES. I1

which meets the surface at an angle less than 0, is

P, =

P SBO, sin 0 d0 2 J;” sin 8 d0

(3.11)

(3.12) The transmission coefficient T is 1 - R , where R is the coefficient of reflection. R = - (n - I)’ (n Thus

+

(3.13) The value of the transmission coefficient changes little with the angle of incidence up to 0,. The power emitted from the dielectric-air interface at the top of the device is, therefore, with (3.12),

L = P,T=

P(l - cos 0,) 4n 2 (n 1)’

(3.14)

+

From these relations we easily find, e.g., that in the case of GaAs (n = 3.6) only 2 % of the emitted power P at the junction is contained within the subcritical zone and, according to (3.13), only 707” of this amount is emitted into air. We see that one important aspect of LED (and laser) production is the increase in light output by technical steps which can either increase 0, or increase T and decrease absorption. The loss of light from the radiation in other than the desired direction is also an aspect which has prompted the construction of reflecting surfaces or mirrors within the device. The total emitted light power can in fact be improved by a factor 2n2 (= 26 for GaAs) when a spherical surface is chosen. From (3.14) and (3.10), we can express the intensity normal to the surface:

I0 -- -L. = P(l - cos dc)4n H

For small 8,, sin 0,

=

I/n can be equated with the angle and as

6: 1 c o s e c %1 - 1 -2 2n3 Thus : zo

[see Gooch (3.1)].

(3.15)

2n(n +

it follows that

= P/nn(n + 1)2

1 1 - cos ec 2 -

2n2

(3.16)

HERBERT F. MATARE

106

The intensity in the spherical case is I0

= 42n

(3.17)

[instead of (3.15)], and L is now equal to: 2nP/(n

+ 1)’

(3.18)

which is an increase by a factor of 2n2 while the intensity is now, with (3.17),

I,

= Pn/n(n

+

(3.19)

(see Fig. 3.2).

nSEMICOND. OR PLASTIC,

(nd)

FIG. 3.2. Semispherical lens of high index of refraction (GaAs), d =junction plane;

D = lens diameter.

The critical angle of incidence 8, is given by the condition tan 8, = d/D

(3.20)

The wider the diameter of the junction plane within the sphere, the more 8, increases. There is another geometry which allows an even higher degree of beam utilization, namely, the “ Weierstrass sphere.” This is originally a construction to find the refracted beam when light falls onto a sphere of higher index of refraction. It is based on the fact that two concentric spheres have to be drawn around the refracting sphere of radius r with the respective radii, rI = (n’/n)r

r2 = (n/n’)r

n’ > n

(3.21)

(see Fig. 3.3). According to Weierstrass (3.2), one has to prolong the incoming ray A P to the outer circle at Q and connect Q with the center of the spheres. The point R , where this connection cuts the inner sphere, is the locus through which the refracted ray from the point of incidence (P) has to pass. Points Q and R are the aplanatic points. As the triangles OPR and

LIGHT-EMITTING DEVICES. I1

107

FIG.3.3. Construction according to Weierstrass to find the refracted beam for a light beam AP incident on a sphere with radius r . Concentric circles are r , = (n'/n)r

and r2 = (n/n')r

or r , / r 2 = (din)'

and r , r z = r z

Refracted beam P R is found at intersection of r l with QO. Triangles PQO and PRO are similar due to this construction; sin 4: sin 0 = n': n (see text).

OQP are similar, the angle of refraction 8 is also found at Q and the angle of incidence is also OPQ. Thus, sin 4:sin 8 = 0 Q : O P = r J r

= n'/n

which means that Snells law is satisfied. The radius product and quotient are, according to Eq. (3.21), r 1 r 2 = r2 (3.22)

r , / r 2 = n2

(3.23)

The construction of the Weierstrass lens has to satisfy these conditions. Starting with the refractive index n and the desired size of the lens, r is given and the circles r/n and m can be drawn (see Fig. 3.4; in this example n = 2). The outer circle defines the aplanatic points F and P . The critical angle is 8 = sin-' ( l / n ) and the junction plane is assumed to cover a part of the lens-flat which is tangential to the inner sphere at P'. Ae we carry out the construction of Fig. 3.3 for the rays emanating from points along PP', we see that all rays drawn from F through these points are parallel to the refracted beam as constructed according to Fig. 3.3. Therefore, all rays emanating from points along the junction A B will exist within a

HERBERT F. MATARE

108

PARALLEL

\

T r.n

1

sine = I /n

r,- 2xr

FIG.3.4. Weierstrass-sphere construction for junction plane A-E: Circle r limits spherical lens with refractive index n. Small circle with radius r/n, large circle with radius rn give Q and R as aplanatic points (see Fig. 3.3) and F as origin for all rays emanating from A B as virtual center (conjugate points).

cone with the solid angle 2 4 1 - cos(l/n)] ‘u .in2. The intensity of radiation is here proportional to n3/n(n 1)2 instead of n/n(n 1)* in the spherical case [see (3.19)]. The enhancement of the output by this much has prompted the development of commercial devices based on the Weierstrass sphere, and excellent efficiencies have been achieved (Texas Instruments). However, there are some problems inherent in this structure. Aside from a demanding lensforming and polishing procedure, it was found that such a spherical surface

+

+

LIGHT-EMITTING

DEVICES. I1

109

will show sensitivity to thermal shock and that microcracks will form in the sphere which propagate at high pulse power. In addition, the larger body of GaAs produces scattering losses aside from absorption, as the light passes through a thicker layer of the original GaAs wafer with swirls and striations. To decrease the absorption losses within the Weierstrass sphere, one proposal is to replace the dome by GaAsP which has a smaller absorption or higher transparency. LEDs produced with GaAsP epitaxial layers, lapped and polished into spheres (radius 1 mm) (3.3), showed increased light output power compared to those with GaAs only. A factor of 2 to 3 was gained (see Fig. 3.5). (mW)

I

I

LIGHT POWER OUTPUT

0.1

0.1

0.3

0.4

05

-1

DIODE

CURRENT ( A )

FIG. 3.5. Measured (integrated) light power as a function of diode current for epitaxial (GaAsP) and (GaAs) dome; 7' = 293°K.

110

HERBERT F. MATARE

Other possibilities to enhance the light output are: collection of the light emitted in other than the top-layer direction by reflective surfaces, contrast enhancement and in the case of devices for optical fiber transmission, devices with implanted optical fibers or epoxy-encapsulated spherical GaAs lenses and, in the case of lasers, stripe geometry. The first case has been treated mainly in connection with GaP devices. The light emanating from the LED junction plane downward into the die can be reflected from the bottom of the die when a method is applied to combine the ohmic contact with a reflective or partially reflective surface. To enhance the effect, it is useful to place the junction into epitaxial material whose forbidden band is smaller than the one of the substrate. For example, GaAs, -xPx(0.3 I xI 0.5) can

FIG.3.6. LED construction with reflecting base layer: (a) mesa structure and (b) planar structure. M = metal electrodes; D = dielectric (transparent)layer.

LIGHT-EMITTING DEVICES. I1

111

be the material where the junction is positioned and the substrate could be GaP which is basically transparent to the light frequency emitted by the junction (650 to 700 nm). The reflective properties of the base layer can be produced by coating the semiconductor with a dielectric (transparent) layer onto which a reflective metal-layer is deposited. To produce a base contact, the dielectric layer is interrupted regularly to allow direct contact between the base metal and the die. Such structures have been made (3.4) and have shown enhanced light output (see Fig. 3.6). The top junction has the form of a mesa structure or can be of the planar type to reduce the top-layer absorption for the light reflected from the base layer. Difficulties with these structures stem from the fact that the multipie base contact is not ideal for a low spreading resistance of the device, the larger part of the base area being needed for reflection. In the case of low-current visible LEDs for displays, the method can have some application. But in cases where appreciable power is necessary, as in infrared LEDs for communications, the spreading resistance has to be minimized and one cannot afford to loose any part of the base layer to achieve good ohmic contact. Another method to enhance light output is the reduction of the individual junction size and radiation from a multitude of junctions appropriately spaced below the eye discrimination (1.8 mils). As the decisive factor for the light output is the current density, the actual current range can be reduced. Some proposals have been made (3.5)to place a multitude of small junctions on a semiconductor and to place a contact layer on the diffused area with openings for the light emitting junction areas (Fig. 3.7a). This method distinguishes itself from the usual multiple contact methods, as shown in Fig. 3.7,b and c. However, the branching in parallel of a multiplicity of junctions will lead to a similar light output versus current. The difference is that in case c each junction area has its separate spreading resistance, whereas in b the junction base currents overlap. The ideal method to reduce electrical loss while improving the voltage handling (higher voltages) is the branching in series of equal p-n layers, as can be achieved with lasers (Fig. 3.7d). In this case, the current through all junctions is the same and if the individual devices are all of the same internal resistance and power characteristic, the group will evenly radiate and form a relatively narrow beam. Contrast enhancement is essential for visible LEDs because these devices are supposed to be well recognizable (or readable in the case of digits) under external light level variations from zero to over lo5 lux. As the LED is a narrow band emitter, the use of filters is an efficient method to absorb incident light adjacent to the main emission peak and eliminate the competing reflected frequencies. T h e contrast for the unfiltered LED is defined by c1 =

(Ll

+ LZ)/LZ

(3.24)

112

HERBERT F. MATARE

where L , is the luminous flux of the activated LED, and L2 is the luminous flux of the unactivated LED due to reflected ambient illumination. If one defines the spectral power distribution curve of the LED as I(1) and the photopic eye response as L(1) and considers the reflectance R(1) of the LED display surface and K(A), the spectral power distribution of the

FIG.3.7. (a)MultijunctionLED with common contacts over part of the junction areas; (b) and (c) multicontact scheme on one and on severaljunctions; and (d) stacked n-p laser diodes for equal current flow.

LIGHT-EMITTING DEVICES. 11

113

ambient light, one can define

L,

=

1 I(1)L(1)d l

(3.25)

L2 = (. K(1)R(A)L(I)d1

(3.26)

L3 = (. K(I)R’(A)L(I)dA

(3.27)

and

with R’(1) the reflectance of the area surrounding the LED. L3 expresses the luminous flux due to the reflectance of the surrounding area. The contrast between LED and its surrounding area is

cz = (Ll + Lz - L3)/L3

(3.28)

In actual measurements of C1 or C2,R(1) does not need to be measured explicitly, but can be considered simply by measuring the luminescence contrast of the unfiltered display at the ambient light level and for the activated LED (3.6). For a neutral density filter with a flat transmission TN throughout the visible spectrum, C can be calculated from

c = (TNL, + T;L,)/T;L,

(3.29)

( T i because the ambient radiation passes through the filter twice). For a bandpass filter, one has C = (BL,

+ SLz)/SLz

(3.30)

with

B=

s W ) T ( W ( 4dJ Z(A)L(I)d 1

(B = fractional luminance of the filtered source), S=

s K(I)TZ(I)L(J)dJ s K(W-44d l

(3.31)

(S = fractional luminance of the ambient radiation returned to the observer). A neutral density filter with TN= 0.2 improves the contrast C already on the average by a factor of 4.Bandpass filters and circular polarizers with bandpass characteristic will improve C to a high degree. This is evident from a partial list of contrasts with different filters and under increased ambient illumination (3.6). Table 3.1 shows a remarkable contrast enhancement for filters of higher quality:

114

HERBERT F. MATARE

TABLE 3.1 Red or green LED

Ambient light: 300 lux 500 lux (normal room) lo00 lux Sunlight: 5000 lux 1oooOlux 5oooO lux

LED Luminance out:

Green light

Red light

No filter

Filter 2

Filter 5

Filter 6

2.57 1.94

8.69 5.61

12.5 7.9

23.8 13.7

559 336

1109 666

1.47

3.31

4.45

7.9

168

333

1.63 1.31 1.06

4.08 2.54 1.31

6.15 3.58 1.52

25.7 13.4 3.47

100 50.7 10.9

483 242 49.2

1/4

* 113

-

-

* 113

1

113

Filter 8

113

Filter 9

Values for contrast C , using different filters: Filter 2: Neutral density filter, TN= 2. Filter 5: Bandpass filter 40% peak transmission for 570 nm. Filter 6: Circular polarizer with bandpass characteristic of filter 5, fraction of light specularly reflected from display, F = 0.8. Transmittance T(1)of the filter to ambient light, specularly reflected from the display surface after passing through the filter: T ( 1 )= 3 x lo-’. Filter 8: Bandpass filter, 400,; peak transmission at 660 nm. Filter 9 : Circular polarizer with bandpass characteristic of filter 8: F = 0.8, ~ ( 1=)3 x 10-3.

The LED luminance is of course also a function of the filter type and decreases generally with increasing filter complexity. In Table 3.1, the LED output goes from 100 (no filter) down to about 30 with filter 9. Worst case contrast calculations for LEDs at typical driving currents have shown that bandpass filters are of marginal benefit to enhance contrast for green LEDs in distinction to red LEDs. For high specular reflection of the LED surface, a circular polarizer in combination with a neutral density bandpass filter and antireflective coating provides adequate contrast enhancement for ambient light levels up to 100,OOO lux for both green and red displays. In the infrared spectral range, the optical problem of efficient fiber couplers is a basic one. LEDs and lasers have to be coupled to optical fibers so that a minimum of losses occur at the injection point. This is not a trivial problem as the lasers and LEDs have a relatively wide angle of exit and the fibers have a highly reflecting (polished) plane of acceptance.

LIGHT-EMITTING DEVICES. I1

115

LIGHT

FIG.3.8. Cross section of LED prepared for optimized fiber coupling. GaAs-CaAlAs mesa structure with etched well and fiber attached with epoxy.

Some constructions have been indicated for efficient LED fiber couplers. One such coupler incorporates the fiber end into the LED (3.7) (Fig. 3.8). This method has its practical difficulties as the device plus fiber has to be exchangeable and therefore another fiber-fiber coupling point will be necessary. More refined is the method to use a spherical lens system for the focusing into the fiber. But if multimode fibers of small diameter are to be used, this construction will also have to be made separable for device exchange. Abram et al. (3.8) have calculated the percentage of coupled power versus source radius for a spherical lens system. They found that this system can be at least two orders of magnitude more efficient than a close coupling system for fibers with a numerical aperture of 0.14. Numerous other systems for fiber optics applications have been developed where mainly simple mechanical constructions are sought to incorporate LED chips and detector chips into couplers such that fiber ends can be equipped with standardized metallic endpieces or jacks to facilitate completion of fiber optic communication lines (Fig. 3.9) (3.9). Finally, the famous stripe geometry, mainly used for laser structures, has to be mentioned. In distinction to the broad area laser, the stripe contact has the advantage to limit the active region to the central and most perfect portion of the active layer and to eliminate complex mode pattern. In addition, the current flow is homogenized within the active region due to a low spreading resistance. The point here is to confine the junction to the stripe area and not to select a stripe contact on a larger junction plane. In the latter

116

HERBERT F. MATARE

I

r ENCPrPSUL

I

S\ource,(LED)

I FIG.3.9. LED with Weierstrass sphere embedded into encapsulation (epoxy) carrying fiber.

case, a greatly increased threshold current results as current spreading across the diffused upper layer occurs (3.10). Figure 3.10 shows the general case of a complete surface contact as it is prevalent in DH and LOC structures (Fig. 3.10a) and the case of a stripe contact covering a portion of the top layer (Fig. 3.10b). This case is of course much less efficient than the isolated stripe geometry, inasmuch as the parasitic current flow along the diffused layer or upper junction layer (resistances R2and R , ) will detract from the stripe flow and increase the threshold value. This is avoided for a stripe geometry in planar form or in mesa-type structures (Fig. 3.10~and d). Lasers with stripe geometry show uniform near-field patterns and single axial mode operations in relatively wide ranges of current levels (3.11). In this context it is important to apply good insulator separation layers on top of the active layer. Here Si02 layers are generally used, produced by thermal decomposition of tetraethylorthosilicate at 650°C or by silane decomposition at 325°C. Also Si3N, layers are applied when diffusion masking is required. In this case, NH, and SiH4 as reactive gases are used. There is another method which works at room temperature and allows deposition of a good oxide layer. It is based on the use of a pH-adjusted hydrogen peroxide solution. Ammonium hydroxide is used as a pH adjusting agent (pH of the solution between 5 and 10) (3.12).

117

LIGHT-EMITTING DEVICES. I1

(b) JUNCTION

-

/STRIPE

\ R,

CURRENT FLOW LINES

I METAL

(c)

CONTACT

INSULATOR

DIFFUSED JUNCTIO METAL CONTACT

,METAL

FIG. 3.10. Laser geometries: (a) full area contact; (b) stripe contact on top junction with R , normal spreading resistance, R , resistance of diffusion layer, R , parasitic spreading resistance, and R , R , (see Ref. [3.10]);(c) stripe contact over masked-off diffused junction area; (d) mesa structure with stripe geometry.

+

B. Adaptation to Integrated Optics We restrict this text to the active device itself and have to refer the reader interested in “integrated optics” to an increasing amount of special literature [see, e.g. (3.1311. However, the adaptation and constructional detail of active devices (lasers and LEDs) with respect to their incorporation into integrated optics will be touched upon here. In the quest for a monoblock buildup of the active and passive devices needed for numerous electrooptical tasks (e.g.. heterodyning,multiplexing, up and down conversion, pulse formation, detection and amplification), one desires laser and LED structures

118

HERBERT F. MATARE

which are easily incorporated into integrated optics monoblocks or which are suited for fiber optics applications. We have discussed briefly some methods for fiber-optic transitions. Here the stripe geometry has some advantages. The losses of coupling light from a device into low-loss optical glass fibers are mostly due to the smallness of the acceptance angle of the fiber. Here the stripe geometry allows us to adapt the angle of exit of the device and the angle of acceptance of the fiber. For this case, it is useful to apply fibers with spherical-ended silica-core glass, coupled to a mesa-stripe geometry device (e.g., a double heterostructure laser) as the angle of exit for the stripe structure can be adapted to the fiber acceptance angle (3.14). The half acceptance angle 6 of a sphericalended fiber is (3.32) In the case of r = co,formula (3.32) degrades to the form for the acceptance angle of the flat-ended fiber. d is the core diameter, r is the radius of curvature of the fiber end, n, and n2 are refractive indices of the core and cladding, respectively. 6 increases first slowly with d/2r, i.e., from 0 to 0.6. From here on (6 = 40")there is a steep rise of 6. One is thus interested to keep d/2r between 0.3 and 0.6 (6 = 10" to 30") or r 2 d (see Fig. 3.11).

--ID+

I I ADJUSTABLE DISTAW.

FIG. 3.1 1. Stripe laser adjusted to spherical ended optical fiber: r = fiber-end radius, d = core diameter, 0 = half acceptance angle, D = distance (adjustable).

A step further in integration is the laser structure coupled to a GaAs waveguide in monoblock form. To separate the laser electrically from the waveguide, one deposits Si02 photolithographically on the edges of the junctions before CVD of GaAs on the substrate. The waveguide is a n- on n+ (substrate) structure which accepts the laser emission and guides it along

LIGHT-EMITTING DEVICES. 11

119

--hv4 n--

guide

n+ GaAs SUBSTRATE

1 Au-sn

Fic. 3.12. Schematic of cross section of GaAs-GaAlAs-DH laser integrated with high resistivity GaAs waveguide. n--guide structure is grown onto laser structure by CVD. Electrical separation by SiO, layer.

the substrate surface (3.15). In this way, one can generate a monoblock laser waveguide and then combine the waveguide with other desired circuit structures (Fig. 3.12). There are other means for the preparation of light waveguides which are preferable from a technical point of view than the slow and difficult way via CVD. For example, the growth of junctions and heterojunctions by CVD and LPE forms guides, also thin films, deposited by vacuum evaporation and stripes of different materials, e.g., GaAlAs embedded in GaAs (:Cr) can very well serve as waveguides. The latter materials combination, i.e., DH structures from GaAs and AI,Ga, -,As can also serve as light modulators (3.16).The coupling of active devices to guides either by prisms or gratings is combined with more losses than a direct junction coupling. Coupling losses between LED or laser and the different types of light guides have been the subject of research, as has the use of guides for modulation purposes (3.1 7). There are numerous beginnings to develop complete optical microcircuitry to serve in optical communications as microcircuits serve in normal electronics today. The use of thin films has been a great help from the technological point of view. A wide range of values of the refractive index can be covered and materials can be adapted to the needs of the circuit designer. In addition to the variations in refractive index within semiconductor waveguides, where ternary and quaternary compounds can be embedded into binary compounds like GaAs:(Cr), a number of materials that can be evaporated or electron-beam deposited are useful in this context (see accompanying table) :

120

HERBERT F. MATARE

Material

n

SiO MgF, CaF, Li F BaF, La,F, SO,

1.770 1.406 1.388 1.398 1.480 1.573 1.464

Material MgO A1203

,

CeO La203 DBF glass Vicor glass

n 1.708 1.603 2.211 1.80 1.555

1.464

The construction of optical waveguides in GaAs is also possible by external means without metallurgical processes like epitaxy : Ion and proton implantation has been used successfully for this task (3.18).The development of truly integrated optics is via the opto-hybrid circuit technology because the monoblock is not always an easy proposition. We now briefly describe some important developments in this field. Appendix

Interesting possibilities arise from the fact that GaAs is also the basic material for detectors, UHF (Gunn) oscillators, and amplifiers. We will mention a few device combinations which are being studied specifically with respect to integration in monoblocks. An optical detector with sufficient high frequency response is the bulk photoconductor based on the resistance

HIGH RESISTIVITY MONOCRYSTAL

7=?

CONTAC'I

FIG.3.13. Photoconductor as high frequency modulated light detector.

change due to carrier density change (see Fig. 3.13). While sensitive operation of such a device is quite possible, especially when light mixing is desired, the transit time limitation makes this device as difficult to use as the P I N diode. The equivalent resistance of a P I N diode is

(3.33) (where R , is the series resistance and C is the diode capacitance) and decreases sharply beyond 10 GHz, resulting in a difficulty to adapt to the

LIGHT-EMITTING DEVICES. II

12 1

microwave circuit impedance. The transit time reduction factor for a photoconductor is defined by the quantity:

where to is the transit time and t, is the recombination time. For the usual case t o b t,, this expression reduces to

(3.35) where f , is the transit time cutoff frequency ( f , = l/nto, which is twice the P I N diode value). For a reasonable diode size, e.g., 1-mm thick, the transit time cutoff frequency will be about 30 MHz, whereas we need responsivity to 100 GHz. Recent advances in Schottky-barrier diodes for UHF make this technology most competitive with another type of structure: the GaAs traveling wave amplifier [see Dean and Matarese, Prac. IEEE 60(12), 1486-1502 (1972)l. In this device, one can use the excellent high-frequency properties of the Gunn oscillator to build a wide (low-capacity) region for signal amplification (traveling wave amplification.) As the GaAs used for Gunn devices is a high-resistivity, pure and perfect crystal slab, made by epitaxy, we can use the same structure as a traveling wave phototransistor. In this operation, the Gunn device is similar to the subcritically operated Gunn oscillator as photodetector [see P. Guetin, J . Appl. Phys. 40(lo), 41 144122 (1969)l; however, in this case we do not limit our operation to a threshold device function, but combine a continuous high sensitive signal detection with a very high frequency amplifier capability. The basic scheme of this structure is shown in Fig. 3.14. As there is gain (at least voltage gain) at the input region (2), the traveling wave region (3), and at (4), the output region of the traveling wave channel, the overall voltage gain is basically the product of these gain factors. Experimentally, a power gain of 10 dB at 12 GHz has been measured and a peak stable net gain of 28 dB at 9.7 GHz. Our proposal to use the traveling wave Gunn device as an optical detector has two reasons: (1) The input Schottky barrier represents an ideal high frequency detector with minimal transit time and recombination time losses. (2) The well-known Gunn plasma is a medium for amplification which is rather independent of contact capacity and relaxation phenomena, as it is based on a fieldgenerated electron plasma wave. In looking for a fast but sensitive photodetector with microwave subcarrier amplification capability, one is led to this device in preference to the subcritically biased Gunn oscillator. The

122

HERBERT F. MATARE

FIG.3.14. Integration of LED (hv), Schottky barrier and subcritically operated Gunn device as photodetector-amplifier:(a) device form; (b) schematic of operation. (1) Input signal excites Schottky barrier, (2) electrode voltage induces charge on depletion capacitance, (3) charge drifts toward anode as exponentially growing traveling space-charge wave, (4) traveling wave current drives load, (5) reflection from anode drives load, and (6) output signal feeds back to input.

input power gain of the GaAs traveling wave amplifier is A = (Yo/Ys)G2

(3.36)

where yo is the characteristic admittance of the traveling space-charge waves, y, is the source admittance, and G designates the input voltage gain,

Here g, = Cb u/x2 = intrinsic input transconductance per unit width; lo7 cm/sec; x2 is the extension of the input electrode across the channel; all y’s are per unit width; y o is the characteristic admittance of the traveling space-charge waves; y, is the source admittance; and yc is the contact admittance between cathode and input electrode. The width x2 of the input electrode should be small (Fig. 3.15), as one can see. In the case of optical signal injection, this is desirable also from the point of view of optical efficiency, as the total channel width b2 is 50 to 100 pm. x2 is generally between 4 and 8 pm long. The metal layer can be made extremely thin, to the point where optical transparency would result (Fig. 3.15). Eventually, the input electrode can incorporate the Schottky-barrier light detector and the device geometry will look like that shown in Fig. 3.15. Figure 3.15 shows also the representative scheme of the arrangement, as discussed previously. uN

123

LIGHT-EMITTING DEVICES. I1

RF-out

Au THIN FILM SCHOTTKY CONTACT

nu

FOR LIGHT INJECTION

FIG.3.15. (a) Traveling wave semiconductor amplifier (epitaxial GaAs): x 2 = electrode width; b, = channel width: b, = outer electrode spacing; L = inner electrode spacing (see text). (b) As in (a) but with larger metal layer as Schottky barrier for light injection. (c) Circuit diagram for integrated light detector traveling-wave amplifier.

As this device structure looks very promising, we have still to consider that it is not yet available commercially and may take some time to develop. Therefore, other detector types are considered for the system outlay. The next best and promising device for high-sensitivity detection is the Schottkybarrier diode as a light-pumped converter. Here we have a system which allows very high frequency response and can in addition be pumped with band gap radiation to yield signal power amplification. This is due to the fact that injected hole-electron pairs are separated at the metal-semiconductor barrier, the holes being trapped near the surface of the semiconductor at the metal interface, the electrons being carried in the opposite direction and eliminated at the base. Thus, under efficient illumination, injected carriers maintain a barrier height increase and barrier narrowing against continuous leakage and relaxation.

124

HERBERT F. MATARE

The narrowing of the depletion region permits additional electrons to be tunneled into the photoconductor, thus giving rise to a gain factor for signal injection [see also R. R. Metha and B. S. Sharma, J . Appl. Phys. 44(1), 325-328 (1973)l. From the foregoing, the following scheme for signal amplification results: The optical signal is directed toward a surface barrier (Schottky barrier) diode, which simultaneously is subjected to a band gap pumping radiation (see Fig. 3.16).

-THIN

-OHMIC

FILM [AU)

CONTACT

FIG. 3.16. Injection of modulated light beam: h(w, + 0,)[wL= light frequency (carrier), frequency] onto LED pumped semiconductor Schottky barrier.

w, = signal

As the silicon forbidden band is close to the optical frequency of the light emitting devices of high quantum efficiency, it is easy to combine the Schottky detector with a LED as a pump source.In general,the 0.9 pm emitter matches well the silicon in its main optical gap. The LED may radiate either into the top Schottky barrier or sidewise into the bulk silicon. There exists another very efficient way to combine the high optical efficiency of ternary compounds and the efficient light generation in LEDs. This scheme is shown in Fig. 3.17. Here the LED and the photodetector are part of one and the same crystal. Both are part of the surface heterojunction made by diffusion, gaseous epitaxy, or, better, by liquid epitaxy. The most efficient ternary compound used (GaxAl,-xAs) can yield both an extraordinarily sensitive detector and a high-efficiency light emitter. As both junctions are identical in stoichiometry and are only separated by an etch groove, optimized pumping action can be obtained. Another interesting device for optical detection is the avalanche photodiode. In this case, we can use the internal current gain for amplification. Several combinations are possible: 1. Photodetector + broad-band amplifier --t mixer + IF amplifier, etc. 2. Photodetector + tunable mixer + fixed IF amplifier. 3. Photodetector (as mixer) -t IF amplifier.

LIGHT-EMITTING DEVICES. I1

125

FIG.3.17. Monoblock LED photodetector modulator.

The third method has been described in detail by Kulcyk and Davis [IEEE Trans. Electron Deoices 19(11), 1181-1190 (1972)l. In spite of the fact that measurements on such devices resulted in satisfactory signal/noise ratios (to 15) for reasonable local oscillator voltages, there is the undeniable fact that avalanche devices have a limited life and are prone to degradation. Finally, we have to mention that optical parametric devices allow us to detect and amplify as well as transpose in the frequency domain. We will return to these questions in the section on Applications (Section 5 ) . C . Thermal and Contact Problems To complete the analysis of the device technology, we have to describe briefly the intricate problems arising from the need for device efficiency and stability. In a general sense, the contacts are supposed to be ohmic in nature, to have high mechanical stability and good heat conduction and to minimize thermally produced strain in the crystal. The foremost difficulty in working with semiconductors like GaAs is the fact that their thermal conductivity K strongly decreases with increasing temperature in the range of interest. In Fig. 3.18, we have drawn a curve for the thermal conductivity versus temperature for a typical 111-V compound, GaAs. The different branches of the curve have been derived theoretically on account of different scattering processes. We have marked some of the major processes on the curve. The most important aspect of the thermal conductivity is that its maximum lies at about 10°K.From there on, K decreasesrapidly. This decrease and negative temperature coefficient causes heat accumulation when such material is under thermal stress. At about 10"K, GaAs is a good heat conductor, whereas K at

126

HERBERT F. MATARE

K

(W/cm degree)

FIG.3.18. Thermal conductivity versus temperature for a typical Ill-V compound: (a) corresponds to GaAs cm-'); (b) corresponds to GaAs (10"' cm-').

300°K has fallen to 1/100 of the value at 10°K. There are changes due to doping which in general decrease the thermal conductivity (3.19). While pure GaAs (5 x 10'' cm-3) has a value of about K = 50 W/cm "K at the 10°K point, this value drops to the 0.3 range for a dopant density of 10" cmT3. For such dopant density, the maximum of K shifts toward higher temperature (100"K), but this maximum is still a factor of 10 below the one for pure GaAs (Fig. 3.18). A metal like copper (see curve Cu) has, e.g., a much smaller change in thermal conductivity across the temperature range considered and does not go into a negative range at higher temperatures (the negative temperature coefficient is typical for semiconductors). It is essential therefore to reduce the die thickness in devices which are used at higher load and which develop heat (power LEDs for communications). As far as the lattice thermal conductivity is concerned, the elemental semiconductors (Si, Ge) range higher than GaAs. If we plot the main semiconductors according to the values of their Debye temperature (0 = k o / k )

127

LIGHT-EMITTING DEVICES. II

TABLE 3.11

Semiconductor

Si AIP AlAs GaP InP Ge AlSb GaAs GaSb InAs InSb

Debye temperature 0

(OK)

Lattice thermal conductivity K (3WK)(W/cm OK)

647.8 588.0 4 17.0 435.0 321.5 374.0 292.0 344.2 265.5 249.0 202.5

1.412 0.9 0.8 0.77 0.68 0.606 0.57 0.455 0.390 0.273 0.166

and tc(W/cm OK),we have the sequence shown in Table 3.11. We see, e.g., that the ternary compound GaAlAs has an improved thermal conductivity as compared to GaAs. The usual methods to form degenerate contact areas by an alloy-diffusion step after metallization reduce the width of the semiconductor, becausea highly doped layer with metal inclusions will improve the thermal conductance. Another consideration that enters into the prescriptions for contact formation is the thermal expansion. Strong differences in thermal expansion between contact layers, dies, and stems can induce strain and damage to the crystal, with complete destruction over longer time periods. The linear expansion coefficients of 111-V compounds are well known, and when plotted versus temperature, show saturation of values at higher temperature (Fig. 3.19) (3.20).Germanium has been added here to show the close fit to GaAs and because it is used extensively for ohmic contact formation on GaAs. The linear expansion coefficients of most metals used for device contacting are a factor of 3 to 4 higher, especially at temperatures above 300°K. The problems of the contact barrier between metals and semiconductors have been dealt with in detail [see, e.g. (3.21)]. Here we are mainly interested in the contact technology as applied to LEDs and lasers. The usual method applied to n-type GaAs and C a P is the goldgermanium alloy formation. Other group IV impurities, such as Sn and Pb, are also applied together with silver as contact metal. For p-type GaAs and GaAlAs, gold, aluminum, indium, and beryllium are used. There is a consensus that the formation of good contact bonds is a many-faceted problem and depends to a high degree on the individual process steps applied. This explains why contact formation is still a topic for

A 6

a x lo6 deg -1

- I

5 4

-

3 m

2 -

P

1 cl

s

0 -1

-2 -3

k

I

I

20

40

I

60

I

I

80

100

1

I

I

1

140

180

2 20

I

I

260

I

I

1

I

300

340

T

(OK)

FIG.3.19. Temperature dependence of the linear expansion coefficient for GaAs, Ge, GaSb, InSb, and AISb.

*

129

LIGHT-EMITTING DEVICES. I1

symposia of wide technological interest. We confine our considerations to the elucidation of the main problem areas. Especially germanium as alloy material has found wide applications and is known to form N + regions on n-type GaAs. Since Ge is amphoteric in GaAs (like Si), it is conjectured that the deposition on lapped surfaces (defects)and under some strain due to the metal bond (Au) distorts the shape of the valance band sufficiently so that the acceptor level disappears (3.22). However, one may also argue that the alloy formation will preferentially absorb Ga atoms (Au-Cia) and thus free Ga sites to be taken up by free Ge atoms. TABLE 3.111"

Alloy composition

Crystal orientation

Epitaxial layer resistivity cm) 0.25 0.60 0.11 0.15 0.60 1.2 2.6 0.14 0.35 0.50 0.11

After Mehal

et al.

Specific contact resistance (Q m') 1 1 <1 1 1

x 10-~

x ~ o - ~ x 10-~

x 10-4 ~10-3 x 10-~ x 10-3 x 10-4 x 10-~

9 1 3 4 6 x 1 x 10-3

(3.23).

Many other contact combinations have been used and their applications and success are dependent on other process steps and device features. Some of the combinations used for GaAs epitaxial layers and their specific contact resistance are shown in Table 3.111. Particular contact combinations using tunsten + zinc (3.24) and tantalum + palladium (3.25) were patented because of the good match of the coefficient of expansion of this material to TABLE 3.IV Metal combination

Alloy temperature

Au-Zn (PtYpe) Au-Ge (P-tYpe) Au-Sn (n-type) Au-Si (n-type)

700°C 360°C 700°C 600°C

130

HERBERT F. MATARE

GaAs. A rather complete survey of the 111-V contact technology was made by Rideout (3.26). In the case of Gap, the metal combinations used are mostly as shown in Table 3.IV. Beryllium was also recently used in connection with gold (3.27). Appendix

Calculation of luminous flux as a function of surface coverage by contact area on a flat die: (Fig. 3.20).

‘I I

I

FIG. 3.20. Simple contact geometry. A : top area (radiating), C: contact pad (opaque).

The luminous output is a function: I P =q-(A C

-

C)

+ I2R

(3.38)

where q is the external quantum efficiency, I is the current, C is the contact area, A is the top surface area, and R is the spreading resistance. We neglect the loss component Z2R and consider the following function: P=qI(g

- 1)

(3.39)

Equation (3.39) is valid under the conditions that: (1) only radiation from the top surface is considered; (2) C 6 A and q independent of I; (3) the current density is governed by the pad contact area C ; and (4) C is opaque to emitted light. Since Eq. (3.39) is of the hyperbolic form: P = const

i: ) -

-

1

131

LIGHT-EMITTING DEVICES. I1

9 -

A

a 7 -

6 5 4 -

3 -

2 1 -

0.1

0.2

0.3 0.4

0.5 0.6

0.7 0.0

X

Luminous output (arbitrary units) versus x (= c / A = relative coverage).

P(C) decreases rapidly for increasing C . P more than doubles for C varying from 0.5 to 0.3 ( A = 1). P plotted versus I with C as parameter shows that for small drive current requirements, C should be small. In cases of power LEDS, this means that the surface conductance has to be high, since only limited bond areas should lead to full surface illumination. It also means that the p-n junction has to be flat and free of faults (see Figs. 3.21 and 3.22, where the output power for an IR power LED is plotted for different CIA values). D . Passiuation and Degradation

1. Passiuation In LEDs and lasers, the problem of passivation is solved to some extent simultaneously with the optical encapsulation. As we discussed, the need to adapt the high coefficient of refraction of the 111-V compound to air requires a casting or lens-making process during which in most cases the die is also covered with a thin layer of silicone to avoid heat-generated strain between die and lens material (epoxy). These silicones exist in different chemical forms with different degrees of polymerization and are excellent insulators. Numerous experiments have

132

HERBERT F. MATARE

X -0.05

0.98

OUTPUT (mW)

80

7[

6C

5c

4c

3(

21

II

I

2

3

4

5

10

20

3 0 4 0

FIG.3.22. Output (mW)of an infrared power device (IAV) versus pulse current with x as parameter. x = 0.1 corresponds to actual IAV unit. IAV = International Audio-visual, Inc., Van Nuys, California.

been done to cover GaAs and GaP with a protective oxygen layer. The usual oxidation at higher temperature in oxygen is not applicable to GaAs because of a loss of arsenic and subsequent weight loss due to gallium oxidation. This method works, however, with Gap. Here a weight increase is observed when GaP is heated above 1O0O"Cin oxygen (3.28). The temperature has to remain below 11 10°C.Between 11 lo" and 1150"C,an energetic exothermic reaction takes place, leading to a formation of the cristobalite form of GaP04. A minor compound formed is /?-Ga203. Passivation of GaAs is generally possible by CVD or sputtering of Si02 or Si3N4.The latter compound can be deposited at relatively low temperatures from NH3 and SiH4 (silane), as is known from silicon technology. Other starting materials for Si02 deposition are organic compounds such as Si(C2HS0)4,which decomposes over GaAs at 700°C in the presence of a carrier gas (92% N2 8 % H2). Other passivation methods are based on A1203, deposited by anodization in an aqueous solution of tartaric or citric acid with glycol and a

+

133

LIGHT-EMITTING DEVICES. I1

subsequent annealing at 250°C. Also, slow evaporation of A1 through an 0,-atmosphere (5 x torr) will form A1,03 on GaAs. Interface state density was measured to be 10" states/cm2 (eV)- (3.29). Protective oxide layers can delay or avoid degradation of LEDs and lasers and are extensively applied in microwave applications. Anodic oxidation is also studied. The need for a low temperature operation suggested electrolytic methods (H20, electrolytes).Improved methods are based on a mixture of carboxylic acid plus polyhydric alcohol. Tartaric and citric acid work as well when mixed either with ethylene and propylene glycol or polyhydric alcohol. A 3x-57: aqueous solution of the acid mixed with the alcohol is found to work well. Maximum thickness is 7000 A. Control of the oxide thickness is possible by the interference colors, as shown in Table 3.V. The color of the anodic native oxide on GaAs is observed perpendicularly under fluorescent light.

Thickness 500 800 loo0 1300 1600

(A)

Color Brown Dark violet Royal blue Light blue Light green

Thickness 1700 1800 2200 2400 2700

(A)

Color

Yellow Gold Red purple Dark blue Green

Other methods of oxidation are based on evaporationAeposition, e.g., e-gun evaporation of A1203.

2. Degradation When one speaks of device degradation, one covers a great number of mechanisms which contribute to the effect of a loss in performance. In the case of light emitters, it is above all the loss in external quantum efficiency which devalues the devices. The reasons for device degradation and failure can be the following: (1) Contact degradation ( 2 ) Junction degradation (3) Bulk material degradation (4) Thermal effects resulting in strain (1) Contact degradation. The top and base contacts are involved. Here we have to consider the methods of metallization and whether sufficiently deep degenerate regions have been formed under the contact pads.

134

HERBERT F. MATARE

Wire bonds can degrade easily when aluminum wire is used in ultrasound bonding equipment and especially in contact with gold. Therefore, some modern power LEDs are made by an alloy-bonding process without gold or by the known gold-ball bond technique when gold pads are foreseen. The base contact has to be heat-treated sufficiently to form deeper alloy regions. Such forming in a slightly reducing atmosphere cannot be too prolonged, due to contact metal absorption (a few minutes at the eutectic temperature.) (2) Junction degradation. All defects, also those resulting from dicing (laser or diamond-scribing, breaking), will have a tendency to grow into the die center and form the known dark-line defects which favor nonradiative recombination. Defects present within the junction area will grow under strain and heat generated by the current flow. (3) Bulk material degradation. In higher power operation, the electrical field generated and the current through the device can move impurities like zinc. This will result in junction degradation, soft breakdown, and finally loss of injection efficiency. Other possible effects, especially those under (4), are increased strain and, in the case of amphoteric impurities such as silicon, a migration from A to B places and change in type. In Part I (see Volume 42) we described some of the effects of c1 and p dislocations on light output. Here we will confine our comments to the inherent bulk material changes. A rise of the nonradiative current component suggests tunneling and introduction of deep levels in the space charge region and, thus, also a change in junction profile. In the forward bias direction, the junction field is substantially reduced and ionized impurities can move across the barrier. If the flux of ion migration at zero bias is F(,,, the flux at a forward bias voltage V, is F(YF)= F ( 0 ) exp(4l/,/kT)

(3.41)

Other mechanisms are based on defect formation due to atomic rearrangement, vacancy migration, formation of interstitials (Frenkel and also Schottky defects). There is consensus, however, that other degradation mechanisms are possible, espe'cially those based on metallic impurities such as copper (3.30). Intentional copper contamination certainly has produced degradation of LEDs. In the case of the GaP : Zn-0 red emitter, the light emission is based on the recombination of excitons on nearest-neighbor Zn-0 centers in the p-type semiconductor side. Copper will allow the charge carriers to bypass this mechanism and thus increase the nonradiative current component. But even without copper and careful gettering (Ga),degradation is still a fact. Arrhenius plots (temperature versus time to reach 50% output) show that copper-free processing can extend the lifetime of devices from lo3 to lo5 h (3.30).

LIGHT-EMITTING DEVICES. I1

135

In the case of Si in GaAs, there is the danger of the place change Sic, -,SiAs.We have mentioned the fact that silicon changes place (transition from n type to p type) when slow cooling (annealing) is effected and especially in the presence of gallium surplus. This effect can be prevalent in GaAlAs/GaAs structures where the GaAs side has to maintain good n-type conductivity. Very fast cooling after epitaxy, up to ZO"C/min, has resulted in improved lifetime of SH and D H diodes for communication LEDs (Ge doping, equal in this respect to Si doping) (3.31). While cooling rates between 0.2"C/min and 8"C/min resulted in a 50% degradation after 200 hr and 500 hr, respectively, the 20"C/min fast cooling resulted in no degradation at all during lo00 hr measuring time (Fig. 3.23).

FIG.3.23. (a) Single heterostructure laser, and (b) degradation curves for different cooling cycles. After Ettenberg and Kressel (3.31).

The explanation of the influence of the cooling rate on degradation can be found via the formation of nonradiative centers. Amphoteric impurities especially can move easily, and a preset condition by annealing will enhance further place changes due to direct current flow and current heating. It is also assumed (3.31) that vacancies initially present due to the meltback become mobile during diode operation. Apparently, a fast cooling

136

HERBERT F. MATARE

cycle during LPE decreases the vacancy concentration and the formation of interstitials of the dopant with attendant transition from A to B places during operation of the device. Devices with Ge doping like these have great importance in modern optical communications due to their high frequency response (up to 200 MHz). In this case, high pulse power is applied and especially in free-air infrared communication sets, the LED is subjected to high pulse power in the several watts range. Here one has to consider that the die will heat up and that contact layers of insufficient stability will detach or bum out. A good base contact has to fulfill the following conditions: (1) Decrease the bulk layer width of the die. (2) Increase the heat conduction. (3) Form a graded alloy-contact zone adjacent to a degenerately doped zone.

In general, the operation of LEDs and lasers, with their strong forward bias condition, results in the formation of defects and especially the growth of defects already present. This has been shown to be the case in both GaP LEDs (red or green) and GaAIAs-GaAs (DH) lasers. I t was shown that: (i) The forward biasing of the device is a prerequisite to the occurrence of degradation. (ii) Mechanical stresses have a strong influence on the rate of degradation. (iii) The degradation process is thermally activated with an effective activation energy 0.6 < E, < 0.8 eV (3.32). It is therefore important to minimize the defect density in the original substrate and epitaxial layer. . It was found that dislocations in both the substrate and epilayer mostly have a Burgers vector b = *a( 110) and are parallel to the (21 1) or (1 10) directions. They mostly occur in pairs (3.32). These structures grow out into large-area defect complexes during degradation. It must be assumed, however, that new dislocations form during operation, especially when mechanical strain occurs (importance of strain-free, soft-solder, die bonding). In fact, fundamental changes in the level scheme also occur and recently it was shown by “deep-level-capacitance spectroscopy” that adecrease in Zn-0 concentration in GaP is due to the appearance of deep levels at a concentration of 1015cm-3 (3.33).

LIGHT-EMITTING DEVICES. 11

137

REFERENCES FOR SECTION 3 3.1. C. H. Gooch, Injection Electroluminescent Devices,” p. 63 8.Wiley (Interscience), New York, 1973. 3.2. See, e.g., M. Born, “Optic,” p. 58 ff. Springer-Verlag, Berlin, 1933. 3.3. H. J. Henkel and G. Ziegler. Solid-State Elecrron lo(?), 158-160 (1967). 3.4. R. W. Soshea. U S . Patent 3.790.868 (February 5. 1974). for Hewlett-Packard, Palo Alto, California. 3.5. H. D. Edmonds, US. Patent 3.g06.777 (April 23. 1974). for IBM. 3.6. J. Pucilowski. R. Schumann, and J. Velasquez, “Contrast Enhancement of Light Emitting Diode Displays,” Tech. Rep. ECOM-43 18 (NTIS-ADA010761). Electron. Technol. Device Lab., US. Army Electron. Command, Fort Monmouth, New Jersey, 1975. 3.7. C. A. Burrus and R. W. Dawson, Appl. Phys. Lett. 17(3), 97-99 (1970). 3.8. R. A. Abram. R. W. Allen, and R.C. Goodfellow, J. Appf. Phys. 46(8). 3468-3474 (1975). 3.9. R. S. Speer (Spectronics, Inc.), “Optoelectronic Devices Packaged for Fiber Optics Applications,” Vol. 11, Final Tech. Rep. Nav. Avion. Facil., (Contract No. N00163-74-C-0308), Washington D.C. 1975. 3.10. W. P. Dumke, Solid-State Electron. 16, 1279-1281 (1973). 3.11. S. Iida, K. Takata, and Y. Unno, IEEE J. Quantum Electron. 9(2), 361-366 (1973). 3.12. R. E. Albans and J. C. Dyment, U S . Patent 3,791,862 (February 12, 1974), for Bell Laboratories. 3.13. D. Marcuse, ed., “Integrated Optics,” Selected reprints. IEEE Press, New York, 1973. 3.14. D. Kato, J . Appl. Phys. 44(6). 2756-2758 (1973). 3.15. C . E. Hurwitz. J. A. Rossi, J. 1. Hsieh, and C. M. Wolfe, Appl. Phys. Lrtt. 27(4), 241-243 (1975). 3.16. F. K. Reinhart and 9. 1. Miller, Appl. Phys. Lett. 20, 36-38, (1972). 3.1 7. H. G. Unger, “Optische Nachrichtentechnik.” Elitera Verlag, Berlin, 1976. 3.18. E. Garmire, H. Stoll, A. Yariv, and R. G. Hunsperger, Appl. Phys. Lett. 21(3), 87-88 (1972). 3.19. M. G. Holland, in Semiconductors and Semimetals” (R. K. Willardson and A. C. Beer, eds.), Physics of 111-V Compounds, Vol. 2, pp. 3-31. Academic Press, New York, 1966. 3.20. S. 1. Novikova, in ”Semiconductors and Semimetals” (R. K. Willardson and A. C. Beer, eds.), Physics of 111-V Compounds, Vol. 2. pp. 33-48. Academic Press, New York. 1966. 3.21. H. F. Matare, Crit. Rev. Solid State Sci. 5(4), 499-545 (1975). 3.22. M. Jaros and H. L. Hartnagal, Solid-state Electron. 18, 1029-1030 (1975). 3.23. E. W. Mehal et a / . , “Gallium Arsenide Functional Electronic Blocks,” Tech. Rep. AFAL-TR-6C316. Texas Instruments, Dallas, Texas, 1966. 3.24. J. C. Marinace, U.S. Patent 3,768,151 (October 30, 1973), for IBM. 3.25. S. Schwartzman, U S . Patent 3,686,539 (August 22, 1972), for RCA. 3.26. V. L. Rideout. Solid-State Electron. 18, 541-550 (1975). 3.27. W. A. Brantley, V. G. Keramidas, B. Schwartz, M. H. Read, and P. M. Petroff, J . Electrochem. SOC. 123(10), 1582-1584 (1976). 3.28. M. Rubinstein, J . Electrochem. SOC.113(6), 540-542 (1966). 3.29. H. L. Hartnagal, R. Singh, H. Hasegawa, and K. E. Forward, “New Passivation Methods for GaAs,” Newcastle-Upon-Tyne Univ., Rep. AD-A009780 for Army Res. Dev. Group Europe (NTISAd-AW9780). January, 1975. 3.30. A. A. Bergh, f E E E Trans. Electron Devices 18(3), 166-170 (1971). 3.31. M. Ettenberg and H. Kressel, Appl. Phys. Lett. 26(8), 478-480 (1975). 3.32. P. M. Petroff, 0. G . Lorimor, and J. M. Ralson,J. Appl. Phys. 47(4), 1583-1588 (1976). 3.33. C. H. Henry and P. D. Dapkus, J . Appl. Phys. 47(9), 4067-4072 (1976). “

I’

138

HERBERT F. MATARE

4. MEASUREMENT TECHNIQUES A. General Relations

A light-emitting crystal surface can be regarded as a diffuse radiator. If the radiating surface is da (see Fig. 4.1), the radiant intensity 1 in space about da depends upon the angle 8, measured from the surface normal:

i(e) = ~ ( ocos ) e

(44 where Z(0)is the radiant intensity in the direction of the normal (Lambert’s law). To find the radiance in the direction of 8, consider an annulus on the surface of a sphere and concentric with the normal to the surface. The surface da of the annulus is da = (2x1 sin e)r d8 (44 The solid angle dw subtended by this area is (4.31

d o = da/r2

sin 0 de The radiant flux (power) passing out through this solid angle is = 2n

d 4 = r(e)dw = i(0)cos 8 27t

(4.4) (4.5)

sin 8 d8

(44 As a Lambertian source generates a radiant flux proportional to its radiance N(O), we can write d 4 = N ( 8 ) aa cos 8 or

i.e., the radiance is the same in all directions (independent of 0 ) about a diffuse radiator. The radiant emitted exitance is formed by integrating Eq. (4.6) from 0 to n/2:

.n/2

4 = 2ni(0)‘10 cos e sin e de = ~ I ( o ) Thus the radiant emitted exitance is M = 4/aa = d ( O ) / d a

(4.9)

(4.10)

Most radiative sources approach in their characteristics those of a blackbody radiator for which the spectral radiant emitted exitance is well known as Planck’s radiation law:

m(T)dv = (2nv2//cZ)[exp(hv/kT)- 13-’ dv

(4.11)

LIGHT-EMITTING DEVICES

139

X

2

FIG.4.1. Diffuse radiator and geometry.

M , ( T ) defines the number of photons passing through an area element within the blackbody at temperature T and within a frequency range v and v + dv. The energy flow is obtained by multiplying (4.11) with the photon energy hv : M , ( T ) dv = (2nhv3/c2)[exp(hv/kT)- 13-' dv

(4.12)

since v = L-'c

C

ldvl = 1 ldll l

(4.13)

(4.12) can be written M , ( T ) d l = (2nhc2/15)[exp(hc/klT) - 11-l

(4.14)

Sometimes M,,(T)or M,(T) are written in abbreviated form by contracting the nonvariables into first and second radiation constants; e.g., M , ( T ) = c,v3[exp(c,/AT) - I]-'

(4.15)

140

HERBERT F. MATARE

As is well known, the energy flow according to (4.11) through (4.15) IS a maximum at a particular frequency for each temperature.The wavelength of an LED is associated with a specific temperature of the blackbody radiator. All maxima lie on the line given by A,,,T = 2898 pm O K , (see Fig. 4.2). A,

(M,,,

ImT- IIV$K)

. 10s lo' 10'

Id

. lo' ,

Id

.Id

. Id 10

I0

,

Id

,

Id

'Id 0.02

001

01

02

05

I

2

5

10

20

50

lo2

Here Planck's radiation law ( M A )is plotted over the wavelength scale. Wien's displacement law is visible by the fact that the spectral radiance peaks at higher frequencies for higher temperatures. These points are connected by the line for I at M,,, or the condition: 5,T = 2898°K pm; which follows from a differentiation of Planck's law (4.1). From Eq. (4.14)one derives easily the number of photons emitted per second, or the radiance: R = MA/hv

(v = radiation frequency)

R = (2nc/A4)[exp(hc/IkT) which has a similar trend (Fig. 4.2).

11-l

(4.16)

LIGHT-EMITTING

DEVICES.

I1

141

As one sees, all luminous sources can be associated with a blackbody radiator at a particular temperature. In the spectral range of interest here (0.5 to 1 pm), the 1, line covers a temperature range from 3000" to 6000°K. The luminance of a blackbody at a temperature T is the integral of the luminance values L,. for each wavelength. Luminous efficiency is defined as the effectiveness of a given light source in producing lumens compared to the effectiveness of a monochromatic source of the same power (watts) which radiates at the wavelength of the photopic eye response peak (see Part I). For a definition of the many photometric units, we have to refer the reader to the literature (4.1). What is important to point out is the difference in measurement between infrared radiators and visible LEDs. In the first case, an absolute measurement of the total optical power emitted (integration) is used for a qualification, whereas in the second case the photopic eye response curve with its particular maximum at 555 nm and a bandwidth of about 80 nm (from 520 to 600 nm) must be considered. Here the candela is used as a unit of luminous intensity, i.e., one lumen per steradian. If the photopic function is V(1), one calculates the luminous energy Q,. as (4.17)

where K, is a constant determining the size of the Q units; U1 is the distribution function of the energy over the various wavelengths, or .m

Q =K,

1

V(A)UAd 1

(4.18)

'0

With (4.18), the value of the constant K, can be established. With the luminous exitance or the luminous flux 4, leaving an infinitesimal element of area divided by that area (Im cm-'),

M , = a6,ias

(4.19)

Mu,. = Km V ( 1 )d6,Ids

(4.20)

(4.17) becomes

or .m

M , = I<,

I

V ( l ) M Ad1

(4.21)

'0

Calibration of visible light sources thus has to consider not only the emitted energy U , but the photopic weighting function. In measuring equipment for red, yellow, or green emitting LEDs, this factor is calibrated in. In all cases, it is, however, possible to make a comparison of emitted power purely on the basis of an integration of the exitance, which also

142

HERBERT F. MATARE

TABLE 4.1 Photometric unit

Symbol

Luminous energy

Q

Q=,(4dt

lumen x hour (Im x t )

joules

Luminous flux

4

4 = dQ/dt

lumen (Im)

W (at 1 = 555 nm IW = 694 lumen)

1 lux = 1 Im/mz 1m/cm2 = lo4 lux

Wim'

lumen/steradian = 1 candela

W sr-'

Definition

International unit

(t = time)

Luminous flux density

F

Luminous intensity

I

Luminous flux density per unit solid angle

F = a+/as (as = surface element) I = a4jaw w = unit solid angle

L = a24ias.a cos e

L (I/cm')

M.K. S. units

Luminance

(Lambertian source)

TABLE 4.11

CONVERSION TABLE ~

~

Luminous flux density or luminance 1 Lux 1 Phot 1 Foot candle

1 Im/m'

1 Im/cm2 1 Im/ft2

9.3 x lo-' ft candles 9.3 x 10' ft candles 10.76 x phot

phot lo" lux 10.76 lux

Luminance or luminous intensity/area 1 Lambert 1 Foot lambert 1 Stilb 1 Nit

l/n cd/cm' l/n cd/ftz I cd/cm2 1 d/m'

9.29 x 10' ft lambert 1.1 x lambert 2.919 x lo3 lambert 0.2919 ft lambert

0.318 stilb 3.43 x stilb n lamberts n x lo-" lamberts

allows us to calculate the external quantum efficiency, and to compare visible and infrared devices. Table 4.1 gives the symbols for the main photometric units. Table 4.11 is a conversion table for the usual units. B. Measurements of L E D : Output Power and Eficiency One defines the output power in general terms as (4.22)

143

LIGHT-EMITTING DEVICES. I1

where 4 is the luminous flux, A is the area of a sphere of integration of ,Iz is the wavelength interval in which the radiation luminous flux, 1, 5 1 I is emitted, and d$/dll is the distribution of luminous flux over the wavelength scale. In the frequency domain, P can also be defined as (4.23) where .f defines the interval where the main radiation occurs. C = d#/df represents the average value of flux within the band between fi and f 2 . This value can be set constant in most cases, leading to (4.24) ‘I1

‘It

and f 2 +fl -

P =h7 c(.fz -

(4.25)

4

The first term

is the average energy of all the photons emitted and C Af is the number of photons emitted within the band Afper second or 4. Thus one may write p = i?fM $

(4.26)

The driving force for 4 is the electron injection current I L through the L E D and the conversion to photons takes place via the external quantum efficiency: Total number of externally emitted photons per second qexl = Total number of electrons flowing through the device per second (4.27) In photometric equipment, the LED photon flux is sensed by a detector exposed to the total flux within an integrating sphere (see Fig. 4.3). The detector works also with an efficiency VD =

‘

(4.28)

in transforming the photon back into the sense current I D .Therefore, VLED

= (l/VD)(lD/IL)

(4.29)

144

HERBERT F. MATARE

n

FIG.4.3. Integrating sphere (IS) with LED: D = detector, C = LED characteristic, and

S = screen.

Now with (4.27) and (4.28), one has (4.30) (4.31 ) Since fM

= c/AM

P = (We)(1/b)(lD/tl~) V A sec’ (Planck’s constant) with h = 6.624 x c = 3.108m sec-’ (speed of light) A (electron charge) e = 1.602 x and A M = 900 nm,as typical value, and a detector efficiency qD = 0.8, one gets P = 1.72ID(watts)

(4.32)

(4.33)

in A). This result can also be found by the simple argument that the output power P of the light emitter is proportional to the detector current divided by the detector efficiency times a conversion factor E(eV)/photons (ID

P=

ID E(eV) qD(electrons/photons)photons

(4.34)

and since E(eV) = (1.24/A)(pm),one has P=

ID (mA) q,(electrons/photons)

(4.35)

LIGHT-EMITTING DEVICES. I1

145

or

p=-- I D(mA) 1.24 ( m w ) 0.8

0.9

P (mW) = 1.72 x I, (mA)

(4.36)

In the usual readout calibration, a factor 1 / 0 3 is used for the detector. This would lead to a value,

1

1.8 x ID

(4.37)

For precise calibration, the detector within the integrating sphere has to be replaced by a thermopile and a monochromatic (laser) injector from a source close to the frequency to be used should be the activator. As integrating spheres have low transfer ratios and thermopiles are not very sensitive, lock-in amplifiers should be used for amplification. When the output power is known, one replaces the thermopile by the detector (mostly a silicon PIN detector) and measures 1., q , is, of course, a function of the wavelength and we have to consider the frequency band under consideration. If infrared LEDs and lasers close to the maximum sensitivity of silicon detectors are used (see Fig. 4.19, Part I), i.e., in the range of 0.9 pm, one may disregard this factor. In the visible spectral range, the detector slope is such that from 900 nm to smaller wavelengths a 12% loss in efficiency is measured for 0.1 pm change: dqD/dh = 12%/0.1 pm

(from 0.9 pm down)

In the range above 0.9 pm, the slope is even higher. For radiators with different radial distribution, the integrating sphere may give different values if the reflective properties and geometry are not optimized. In Fig. 4.3, the usual construction of the integrating sphere is shown. Independently of the individual radiation characteristic (C) the light output is measured as a result of total reflections at the inner surface of the sphere* and not due to direct irradiation on the detector. In many cases, a screen s is used to shield off direct incidence. For an efficient use of the LED, the relationship between output power and forward current across the junction is important. In Fig. 4.4, the 4(I) characteristic (average value of 10 devices with a chip size of 0.02 in. x 0.02 in.) is plotted. Three basic ranges are apparent. Painted with MgO coating or Eastman K u d a k BaSO,

+ binder + solvent.

146

HERBERT F. MATARE

rnlpwl IY) A

100

-

10

-

FIG.4.4. Output (light-flux)characteristicof a Ga,,,A1,,,As/GaAs ( p on n ) heterojunction LED with the different exponential ranges and the linear portion L of the 4(I)characteristic.

(1) The parabolic increase at low current according to a law y = x y

(Y = 2).

(2) A linear portion: y = lox. (3) A saturation range where: y = x y with y = 3.

In applications where the solid state emitter is modulated, preference is given to the linear range. If the LED is to be used to transmit analog signals over optoelectronic channels, IF = 150 mA would be the best value to the LEDs in Fig. 4.4, to obtain transmission fidelity, i.e., low distortion and phase variation. The output power measurements and efficiency measurements are often carried out with a relatively simple setup consisting of a large-area solar cell facing a conic mirror with the LED mounted within the cone. Figure 4.5 gives a schematic of the setup. For the electrical-opticalmeasurements, the

LIGHT-EMITTING DEVICES. I1

147

SI Photodlode

I

FIG.4.5. Mirror reflector with solar cell for LED output and efficiency measurements:

PL = 1.72 x I D(mA) (for a photodiode efficiency of 0.8).

q,. = 1, (lllA)/qDIL(mA) 1 1.25(ID/lL)X 100%

LED (or laser) is operated while current and voltage are read. The detector (or solar cell) is shunted by a 1042 resistor and the voltage across the latter is used for reference. The reading in mV or pV divided by 10 gives the diode current (mA or PA). The external efficiency of the LED is VL

= ID/VD

'

I L

(4.38)

qD is mostly of the order of 80%. The power efficiency is:

D=

Total optical power output Total electrical power input (4.39)

The latter value is less favorable, as P , includes the I Z R losses within the LED. For measurements at very different wavelengths, the integrating sphere or ellipsoid or cone has to be calibrated, using known frequency sources and a thermopile. In general, suppliers of such equipment give the calibration curve, i.e., sensitivity in pA/mW versus wavelength (see Fig. 4.6). The curve follows the silicon-photodetector curve with a somewhat higher slope toward the higher frequency end of the scale due to higher reflection losses. One is obviously dependent on the standards used, and over longer periods of measurement those standards have to be recalibrated. It has therefore been proposed to measure LED output or power spectra by a direct comparison with the power spectrum of a standard lamp on a pointto-point basis. The advantage is that the measurement is independent of a calibrated detector and independent of reflectance of the sphere and the response of the equipment (filters). In this case, one can use an integrating

148 FA -

mW

HERBERT F. MATARE of

Sensitivity

inregrating sphere

3

2

f

4

5

6

7

8

9

10 x 10'

x(nrn)

FIG.4.6. Example of a sensitivity-vs-wavelength curve of an integrating sphere with control points at different frequencies. Curve B is a manufacturer's calibration of an integrating sphere; curve A shows measured points (4.4).

sphere with three ports, one for the LED, one for the detector, and a third for the standard lamp (4.2). The standard lamp proposed is a loo0 W (D x W) tungsten-halogen lamp and is a standard for spectral irradiance (Eppley Laboratories). It is supplied with a table of power density (per spectral interval) at a 40 cm distance from the lamp. As the area of the entrance port of the integrating sphere is known, one has the spectral distribution of the total optical power entering the sphere. For recalibration of the tungsten lamp, one can direct the radiation from the exit port of the sphere on a monochromator with photomultiplier tube attached. In this way, a point-to-point comparison can be carried out between standard lamp and LED. The power spectrum of the diode is calculated from the point-to-point spectral data:

(4.40)

LIGHT-EMITTING DEVICES. I1

149

where P D is the diode spectral power density, P , is the tabulated lamp spectral power density, N o and N L are total measured signal counts (e.g., RCA 8852 photomultiplier plus photon counting instrumentation). All quantities are defined at energy E = hv. The efficiency per unit energy or the “differential quantum efficiency ” is

(4.41) with dq(E)/dE in eV-’ and I d (diode current in amperes). Over most of the spectral range in question, the standard lamp output is rather flat. The total quantum efficiency is obtained by numerically integrating the differential efficiency between chosen energy limits. In the case of LEDs with a more complex output spectrum (e.g., GaP LEDs), one has to choose the integration limits according to the desired output color. Thus, for a green GaP LED, one defines

(4.42) where El = 2.00 eV and E , = 2.34 eV. The lower limit eliminates the undesired Zn-0 bound exciton luminescence. For Zn-0-doped red-emitting LEDs, the limits in (4.42) are

El

=

1.37 eV

E Z = 2.10 eV

eliminating the residual green emission and the oxygen-related free-tobound recombination. To measure the actual illumination, one has to form the product with the photopic function to get “luminous current efficiency” or A in lumens/ampere.

(.

A=

E2

Pd(E)V(E)d E

(4.43)

Id ’ E l

where V ( E ) is the CIE (Comite International d’Eclairage) relative photopic luminous efficiency function (lumens/radiated watt, Fig. 4.7). The spectral intervals for the mostly used LEDs are as in the accompanying table:

Gap, green Gap, red GaAsP, red

2.00 1.37 < 1.70

2.34 2.10 2.00

In the visible range, the photometric unit used for LEDs is luminous intensity, measured in millicandelas or microcandelas. As shown in Table 4.1,

150

HERBERT F. MATARE RELATIVE EFFICIENCY

10

.-

lo

--

--

CIE Relate Luminosity Relate Response of Diode and Filter

10 *-

10

600

1

620

,

1

1

640

1

660

1

1

1

680

1

1

700

1

720

1

740

__t

X (nm) FIG.4.7. Commission International d’Eclairage (CIE) photopic curve and approximated diode plus filter curve.

luminous intensity is defined as the flux passing through a given angle in the limit where the angle approaches zero:

I = lim (4/0)

(candelas)

0-0

In actual measurement equipment (e.g., the Tektronix digital photometer Model J16), one actually averages over a finite angle, e.g., 0.01 steradians. The instrument measures the illumination, which is the flux 4, incident on the detector surface (foot candles or lux). The collection geometry of the setup has to be such that a known solid angle is defined by the detector at the source (LED far enough from the detector to appear as a point source). If the detector measures only the direct light (baffled for reflected light),

LIGHT-EMITTING DEVICES. I1

151

the illumination measured in foot candles or lux can be converted to an average luminous intensity of the source in the direction of the detector. To convert from illumination (foot candles) to luminous intensity (candelas, Cd, or lumens per steradian), one has to d o the following: (1) Multiply the reading in foot candles by the detector area in square feet. Result: luminous power in lumens (since 1 ft-cd = 1 lumen/ft2). (2) When luminous power (lumens) is divided by the solid angle (in steradians) subtended by the souce at the detector, the result is luminous intensity (lumens/steradian) in candelas averaged over the solid angle given (e.g., 0.01 steradian). The luminous intensity can also be derived from the quantum efficiency, and vice versa, if both the spatial and spectral light distribution of the LED are known. This is done in the following steps: (a) Measure luminous intensity at a given diode current. (b) Measure and integrate spatial distribution of the optical flux and obtain luminous power output in lumens. (c) Measure the spectral distribution and use the CIE curve (standardized photopic eye response curve, see Fig. 4.7) to convert luminous power (lumens) into optical power (watts). (d) Divide optical power by the weighted average photon energy to obtain the number of photons per second. (e) Divide by the current input (electrons per second) to obtain the quantum efficiency qext(4.4).

C . Risetime Measurements As we mentioned in Section 2 (lasers), risetime, delay time, storage time, and decay time are important quantities in LEDs (and also detectors), since they determine the amount of information which can be sent over an optical transmission or display system. All these quantities can be measured in a system which is adapted to a fast oscilloscope with trigger circuit (see Fig. 4.8). The dual trace allows us to compare input pulse shape and output pulse shape. In general, the risetime of the line components, LED and detector, are much higher than the risetime of oscilloscope and pulse generator (7, and T,,) which are generally neglected. With rL = risetime of LED and tD = risetime of detector, the total risetime of the system is T, = ( T i

If tp and

T,

+ 7;)1’2

(4.44)

are not negligible, the risetime of the system is 7,= (7;

+ T:, + T,’ + T

y

(4.45)

152

HERBERT F. MATARE

Tektronix 4 7 1

rlW v

FIG.4.8. Test circuit to measure risetimes of LED5 and detectors.

If these are all of the order of 100 nsec, 7 $ = 20 nsec. Since the risetime of PIN photodiodes is as low as 1 nsec and since T, and tp are of the same value, the LED is generally the limiting device and determines ts (4.2). D. Bandwidth Measurements

Figure 4.9 shows a system by which amplitude and phase of an optical signal are measured. As the bandwidth is defined as the frequency range where the amplitude of a sine wave decays for 3 dB from its maximum value, the indication of the phasemeter at the 3 dB point will be a 45" shift between input and output signal. In order to read the amplitude fast and accurately, rf voltmeters should be used. The oscilloscope is used to monitor the wave shape and check the fidelity of the transmission system: LED to detector. If the risetime of the optoelectronic system is 7, ,one should measure the 3 dB limitation frequency:

(4.46) If the system is used for digital data transmission, pulse repetition frequencies higher than fL are possible.

153

LIGHT-EMITTING DEVICES. I1

SINE-WAVE

("1)

cw

-

CH?

-

osc

L

PIW OElECTOfl

-

PHASEMETER

,

.t

FIG.4.9. Test circuit diagram for measuring bandwidth of LEDs and detectors.

E. Measuring the Thermal Impedance of LEDs The thermal impedance is defined as the ratio of temperature difference between junction and case T i - T, and the dissipated power Pdissof the LED : m

e,,

=

Ij

- 1, m

~

(4.47)

Pdiss

In Fig. 4.10, the peak emission wavelength of a GaAIAslGaAs heterojunction device is plotted versus case temperature. Parameter is the forward current; line A for 10 mA, line B for 100 mA (20 x 20 mil die). For a case temperature of 20"C, the emitted peak wavelength shifts by about 15 nm when the forward current of the LED is increased from 10 to 100 mA. This corresponds to a temperature increase of 50°C for the same LED current of 10 mA. Therefore, the conclusion can be drawn that a shift in forward current from 10 to 100 mA entails a temperature difference between case and junction of about 50°C. As the curves approach each other

154

HERBERT F. MATARE

'c

I

x 10

FIG.4.10. Peak wavelength of a GaAlAs/GaAs homojunction LED as a function of case temperature at two values of forward current.

with increasing temperature, the frequency change decreases at higher temperature and thus the relative temperature increases. With 1.6 V forward voltage, the thermal impedance is about 30O0C/W.The thermal impedance is a measure for the die thickness, the quality of the contacts and the heat sinking. In the case of lasers, the effects of doping and band filling through injection (Burstein shift) and thermal cavity-length changes also have to be considered (4.5). F . LED Output Power as a Function of Wavelength The optical bandwidth of light emitters is very much dependent on the type of emitter (homojunction, single heterojunction, DH laser, etc.), as we have described. In the case of infrared heterojunction LEDs, e.g., linewidths of 20 to 60 nm are normal. Figure 4.1 l a shows the measurement set for the spectral distribution of the light output. Light emitted from the LED enters the slit of the monochromator acting as a selectivefilter. Radiation passing through the exit slit is detected by a PIN (or other) silicon detector. The response of the detector is

LIGHT-EMITTING DEVICES. I1

155

flattened by a special filter (see Fig. 4.7). The adjustment of the monochromator should be checked with the aid of a sodium lamp. The sodium spectrum has two relatively strong lines very close to 5890 and 5896 A. Figure 4.11b shows a typical spectral distribution of radiation from a GaAlAs/GaAs heterojunction LED (test conditions are forward current I F = 100 mA, case temperature T, = 25°C). Here the optical bandwidth, measured between half-power points, is about 40 nm.

-.f'

qyklpdout

d!" Wave length Dial

d,;

Filter/ Detecror

0.5 m2

d = 1 mm2

It=100 ma

T,-

25'C

out

t

FIG.4.11. (a) Setup for measurement of LED output power versus wavelength. (b)Output power of single heterojunction LED (20 x 20 mil die) as a function of detector 1 x 1 mm surface).

156

HERBERT F. MATARE

G . Emitters and Current Sources

Since the impedance of light emitting diodes changes as a function of forward voltage and junction temperature, one operates LEDs from constant current sources. In Fig. 4.12a-f, some circuit schemes are drawn where LEDs are connected to current sources using operational transistors and amplifiers. In the circuit of Fig. 4.12a, the forward current equals V / R and is supplied by the regulated voltage supply. In this scheme only small drive currents can be used. In Fig. 4.12b, the reference voltage source works with practically no load, and high forward currents can be drawn. However, the LED cannot be

tv

FIG. 4.12. (a-f) Different circuit schemes for current drivers in the application of LEDs with operational amplifiers and transistors (see text).

LIGHT-EMITTING DEVICES. I1

( I

tv,-v,,

157

I

T

FIG.4.13. Substitution network for the circuit of Fig. 4 . 1 2 ~based on an infinite input impedance of the I.C. (operational amplifier).

connected to ground. One can solve this problem as shown in Fig. 4 . 1 2 ~Here . it is assumed that the input impedance of the operational amplifier is infinite. Thus I l = I, = 0 (see the scheme of Fig. 4.13) and the load resistor R L of the LED is driven by the output power of the operational amplifier: 1

I F = a(Vl - V2)-

R

+

(4.48) RL

where a is the open loop gain, or V, = Vl - (l/a)(R

+ RL)IF

(4.49)

Since the two voltage loops (Fig. 4.13) give the equation V, = a(Vl - V,) - V

(4.50)

V1 = R L I F

(4.51)

one gets, with (4.49), 1 Vl - (R a

01

+ R,)IF = l + a ~

Vl - J-V

l+cC

(4.52)

Inserting (4.47) into (4.48) leads to (4.53) If a B 1, (1

+ a)(l/a) will be equal to one and it follows that I F ‘Y

V/R

158

HERBERT

F. MATARE

In other words, the driving current of the LED is again determined by the voltage source and the resistor R . The voltage source does not have to supply current to the circuit, but cannot be grounded. If the LED and the regulation voltage supply both should be connected to ground potential, one can use circuit d (Fig. 4.12). In case the resistor R4 is selected according to the following equation :

R4 = R2R3IRI the LED current will be only a function of R , , R 2 , and R 3 , and the voltage V : (4.55) Figure 4.12e shows the case where a transistor is used as current source. In Fig. 4.12f, the reference signal V is obtained by the transistor power supply via a voltage divider.

REFERENCES FOR SECTION 4 4.1. C . S. Williams and 0. A. Becklund, “Optics,” A Short Course for Engineers and Scientists, p. 53 ff. Wiley (Interscience), New York. 1972. 4.2. I. M. Ralston and R. Z. Bachrach, IEEE Trans. Electron Deoices 20(1 I), 1 114-1 116 (1973). 4.3. Sections (c) through (9) have been adapted from work by H. Heintzeler, personal communication. 4.4. S. R . Korn. “How to Evaluate Light Emitters,” G.E. Appl. Note 200.59.General Electric Semiconductor Products Dep., Syracuse, New York. 4.5. H. Meixner and R. Unger, Siemens Forsch. Entwicklungsber. 3(3), 190-194 (1974).

5. SURVEYOF IMPORTANT APPLICATIONS A. Displays and Indicators

LEDs as indicator lamps and in digital readout applications are so numerous and of such general use in measurement instruments that only some basic developments need to be pointed out here. The combination of the necessary microcircuitry to transform the binary-coded information into appropriate drive pulses to the lamps is also state of the art. All functions in the sequence: Binary coded information + decoder + character selection +encoder +

lamp selection --t display

are conveniently handled by AND, NAND, OR, and NOR gates, commercially available. These systems have been described in detail (5.1) are are today off-the-shelf items (see, e.g. (5.2)).

LIGHT-EMITTING DEVICES. I1

159

Progress has been made in particular in the field of LED efficiency, integration, and large digital displays. In competition with numerous other display devices, the LED has found particular application where solid state compatibility is required and where the number of data points is limited (as compared to the C.R.T. display with a 10’-lo6 information point capability). Gas discharge tubes as displays are today largely replaced by LEDs and the competition with liquid crystal displays (LCDs)also seems to tend to an equilibrium, with LCDs used only in specific cases where stringent power limitations play a role (some digital watches). The low power consumption (m-watts), low voltage, ruggedness, brightness, and small size all are factors in favor of LEDs. But one of the most important features giving weight to this technology is the device life in combination with its low price. Modern mass production methods have decreased production costs to a level which cannot be reached by other techniques. Automated die sorting, bonding, and lead frame processing in connection with pressure molding of lenses have made this production so efficient that the recently achieved improvements in life have given the LED as a display device a decisive lead. Figure 5.1 shows typical lead frame sets of LEDs and a large alphanumeric display device.

FIG.5.1. Lead-frame mounted LEDs before separation, individual LEDs, and an aiphanumeric display device. (Photo, courtesy of AEG-Telefunken, Heilbronn, Germany.)

The techniques used for the different applications vary; for watch displays, monolithic numeric packages with individually separate (diffused) junction bars are sufficient. For larger size displays, a series of diffused regions is used, to define each segment. In the case of even larger numeric display devices, individual LED chips of higher efficiency are used (Gap instead of GaAsP) (5.3).In this case the transparency of GaP is an added advantage as the light emission will be effectively collected by the mirror mounting (Fig. 5.2). Also, the higher efficiency of GaP devices compared to GaAsP is important for these displays. In general, LPE LEDs will be used in

HERBERT F. MATARE

160

EFFECTIVE BAR HEIGHT

4

I DIFFUSER PLATE’ a FILTER

BASE LED’-DIE

MI&

( METALIZED)

FIG.5.2. LED alphanumeric package with individual disks arranged within reflector mount.

this case. As we have seen, the efficiency of LPE-grown devices is by a factor of 10 or more superior to those made by chemical vapor deposition (CVD). The step to monolithic device techniques, however, is facilitated by CVD and diffusionmethods. For example, matrix addressable arrays of LEDs can be made in monolithic form by the p-n junction electrical isolation, known from silicon-IC technology. For example, an n-type GaP substrate can be used and a pepiln-epilayer sequence is deposited by CVD. The Zn isolation diffusion can be applied to the n-epi layer to separate areas of discrete light output (5.4). The layer structure is shown in Fig. 5.3. Ohmic contacts are formed on the n-type cathode (B) and p-type anode (A), using Au-Ge and Au-Zn alloys, respectively. Instead of one n-type contact (B) per row, distributed n-type contacts can be foreseen. All column and row contacts can be connected to a sputtered Mo/Au/Mo metallization deposited on an insulating layer (e.g., sputtered glass). Measured luminous efficiencies were in the range of 20 fL/A/cm2 or 2.2 x lo-’ L/A/cm’ (see Table 4.11) or 342 pd/A/cm’. Interconnection metallurgy for LED arrays made from single chips can be made in hybrid form and will have some importance for large scale displays. For small character size the monolithic approach mentioned has the advantage of compactness and batch processing. Displays have been

LIGHT-EMITTING DEVICES. I1

161

G

Ga P (nJ Substrate

FIG.5.3. Schematic of GaP epitaxial monolithic matrix-addressable array of LEDs. L, = VPE Gap: Zn (p-type) layer; L, = VPE Gap: S' (n-type); L, = VPE Gap: S, N (n-type); A, A', A" = planar LED anodes (p-type); B, B , B" = evaporated ohmic contacts on n-type layer; G = zinc isolation diffusion stripes or laser-scribed grooves. Column (A) and row (B) connections can be evaporated metal stripes on deposited glass, SiO,, Al,O,, or other insulator.

fabricated from GaAsP wafers, using masking and localized diffusion. Addressing can be achieved either in an X-Y-multiplex mode or by a common cathode or anode configuration with a separate lead associated with each diode. SiOz and A1203 layers are then used for the aluminum interconnect overlay (5.5). Such alphanumeric displays are commercially available and are mostly combined with the appropriate logic circuitry (5.2). There is a general trend to replace all scales and indicators with electromagnetic drive mechanisms by LED indicators. Level indicators are replaced by LED lines which are illuminated to the level to be indicated. Such rugged and very clearly visible gauges are important for airplanes and cars. In the not too distant future, LEDs will indicate speed, oil pressure, battery charge, mileage, etc., continuously day and night, as their lifespan is infinite compared to the life of a car. The measurement instruments in industrial use are now largely working with digital readout and this trend will continue as more analog instruments are replaced by digital ones. There is, however, an area of instrumentation were analog readout is highly desired due to the fact that an immediate estimate of a trend, a remaining period, etc., is desirable. Especially in instrumentation used for functional plots of characteristics with possible negative portions of slope, such analog readout is essential. This can be realized with LEDs also when a segment of LEDs lights up to the level to be indicated (see e.g., Fig. 5.4).

162

HERBERT F. MATARE

LED LIT UP

To HERE

t

a

LI 17

-?

FIG.5.4. LED level indicator (analog readout).

Similarly in instrument readings a sequence of LED point sources can move with the signal from the logic circuitry, indicating the precise positioning of the signal to be measured. There are also tendencies to adapt the modem electronic watch to this condition, since one of the competitive features of mechanical watches is the analog readout, giving the viewer an immediate estimate, in geometrical terms, of a time past or an interval of time yet to cover. An electronic watch (Fig. 5.5) with a circle of green LEDs, spaced on a minute scale and red LEDs spaced on an hourly scale, can fulfill the

'bd LEDshifting each hour

FIG.5.5. LED watch with analog readout due to shifting green LED points indicating minutes, and shifting yellow LEDs indicating 12-min intervals,and red LEDs displaying hours.

LIGHT-EMITTING DEVICES. 11

163

conditions: The logic is very similar to the one used for digits as the clock circuit shifts the signal along the row input of the green LEDs and after 12 steps the yellow (12 min) and red hour LEDs are activated. For each yellow (or red) and green LED lighted, the time can be read in analog fashion.

B. General Applications of Infrared LEDs The high efficiency infrared light emitter is ideally suited to be used in all pulse and data transmission circuits where isolation and remote control are desired. Mainly due to the extraordinary ease of direct modulation up to very high frequencies with attendant advantage in sensitivity and low noise reception, these LEDs and matching sensors (PIN detectors made of silicon) find abundant use in the following systems: Data transmission Position indicators Intrusion alarms Systems isolators Communication links Safety devices Logic circuits Information displays Level indicators Programming control Relay operation Optical transmission Test comparator Fire detection Choppers

Comparison tircuit ry Gas-flow indicators Card/tape readers Smoke detectors Counters Sorters Optical radar Gas analyzers Door controls Size monitors Remote telemetry Electronic circuits Edge tracking IR illumination

LED use will also be enhanced in special equipment like the following: Telephones (illuminated dial indicators and numbers) Computers (display, couplers) Anemometers Appliances Annunciators Contactless potentiometers (photopotentiometers) Toll booths IR amplifiers SCR light triggers Production machinery Auto dashboards Vending machines Game machines

Medical instruments Auto ignition Remote control toys Digital voltmeters Measuring instruments Airplane panels Foul line indicators Space vehicles Level control equipment Numerical control equipment Go-no-go gauges Cameras, projectors High-voltage transmission lines TV sets

HERBERT F. MATARE

164

In general, the use of LEDs is enhanced by a pulsed operation, due to the fact that ac operation increases power output (less heating and higher efficiency) and facilitates signal amplification. A normal transistor oscillator circuit as shown in Fig. 5.6 can effect a very efficient relay over wide distances when the amplifier is sensitive enough. A power stage can be used at point A, for example.

FIG.5.6. Schematic of an ac coupler or burglar alarm system. The transistor circuit modulates the LED beam. The readout circuit activates a relay in case of beam (modulation) interruption.

'8"

.pv I

5PId

I5k

+

FIG.5.7. Multivibrator modulator for LED. Readout amplifier with power stage for relays. -6VO

1

IN

hu -W

,Signal OUT

Signal IN

FIG.5.8. (a) Simplified silicon readout amplifier. (b) LED with signal amplifier (appre priate for hybrid circuitry).

LIGHT-EMITTING DEVICES. I1

165

For a more efficient modulation of the LED, a multivibrator can be used and the amplifier can be improved as shown in Fig. 5.7. A very compact package can be made in hybrid form. Such a system is shown in Fig. 5.8, where (a) is the silicon readout sensor circuit and (b) is the GaAs (LED) emitter circuit. The signal may be derived from a modulator (microphone, telephone line, multivibrator, transistor oscillator, etc.). The functions carried through with such pairs are of great variety and have been described (5.6). Examples for inspection systems are shown in Fig. 5.9: (a) Control of parts on flow bands. The optical monitoring equipment is housed in one box with the LED power supply and the readout device. (b) Counting parts (e.g., screws). (c) Underfilled clear containers. (d) Counting of ejected parts. (e) Cutting rods to length. (f) Scanning of materials for imperfections. Here varied LED frequencies can be used depending on materials and type of imperfections. The basic principle is the change in absorption due to imperfections. As one of the LED sensor lines registers a change in output of given magnitude, a discarder system is activated. (g) Activation of cutting tool according to register marks optically sensed and imparted to steering gear. Further specific use of LED readout or LED sensor pairs are concerned with film reading or function generation. A few examples are given in Fig. 5.10: (a) Sound head in film projector. (b) Card or film or tape readers. (c) Shaft encoders. (d) Timing impulses. In the case of conveying systems, a number of new approaches are possible when efficient infrared LEDs are used in ac mode in conjunction with sensing silicon devices. The precise beam system and independence from stray light allows to add such light control systems to any machinery used in full daylight. A few examples are given in Fig. 5.11: (a) Sorting of pieces according to height. The chutes are activated by the corresponding light pairs. (b) Flow control of material. Here the continuous flow over crossing flow bands is maintained by avoidance of clogging. (c) Retroflective tape marks on packages deliver code to transport gear. In Fig. 5.12 a number of counting systems are described in which the interruption by the light beam or the change in output causes a counting

166

HERBERT F. MATARE READ OUT

FIG.5.9. (a) Control of parts on flow band (the reflex or missing reflex from a piece will release the control relay or counter), (b) screw counting system, (c) control of degree of filling of transparent containers, (d) ejected parts falling through transparent tube are counted, (e) acti-

LIGHT-EMITTING DEVICES. I1

167

vation of a cutting tool at the precise length of a work piece, (f) scanning of IR-transparent material for imperfections, and (g) activation of cutting tool as register marks pass by the optical beam.

168

HERBERT F. MATARE

ND

LEDs

U

Z

! PUNCHED TAPE

FIG.5.10. (a) LED and light sensor in film projector sound head, (b) punched tape (card) system with LED sources and readout, (c) shaft encoder with LED row and sensor row, and (d) forming of timing impulses on function generator.

FIG. 5.11. (a) LED readout pairs of decreasing height above the flow band sort pieces according to height and activate (via amplifiers) different chutes, (b) continued flow at flow-band intersection by light activated brake, and (c) special reflecting markers on packages sort out different pieces.

170

HERBERT F. MATARE CONVEYOR

01

CHUTE

EA

FIG.5.12. (a) LED counter for chute, (b) counting by changes of reflection, (c) counting part with brake for filling, and (d) conveyor counting.

system to be activated: (a) A counter for a chute. (b) A counting of paper cups. (c) A parts count in packaging operations. (d) A counting system for a conveyor. In Fig. 5.13, we see level systems as: (a) A flow control for boards having been subjected to paint and being readied for the oven. (b) An optical level meter. (c) A level meter working according to opaqueness of the liquid. (d) A level meter indicating filling heights (within bins). Figure 5.14 shows the following control systems: (a) An indicator for over- or under-sized wires. (The light level change is recorded.) (b) Elimination of doubly stacked parts. (c) Counting of small parts in a chute. (d) control of a liquid by refraction of the light beam. Further uses of LEDs are anticipated in computers. Coupling between subsystems is a specific function to be fulfilled ideally by LEDs. Due to their high efficiency, low voltage and current requirement, long life, and compatibility with solid-state circuitry, they are immediately usable as light connectors, especially when the light beam is pulsed or modulated. In this case, the light beam is especially suited for information

1LI

172

HERBERT F. MATARE

storage and transmission in data processing. The ac modulated light beam transmits signals between subsystemsin a low noise mode because of the low current and the elimination of a common ground. Special Circuits with LEDs

As the efficiency of LEDs decreases with increasing temperature or heat dissipation at dc operation, one is led to use these devices in pulsed operation. As discussed earlier, the operation by a multivibrator (Fig. 5.7) will considerably enhance the power output. A very efficient operation is by a relaxation oscillator, as shown in Fig. 5.15. Here, a four-layer diode D, is fired when the voltage at the capacitor C1 exceeds the breakdown level of the diode D,, passing a high current pulse through the laser or LED. The diode Dz protects the laser or LED from a reverse current.

FIG. 5.15. Relaxation oscillator as power supply for LED. Diode D, is protection against surge current in reverse. D, is a four-layer diode.

A very important use of the LEDs is made for the purpose of voltage isolation. High-voltage thyristors or SCRs can be operated by injected light from a GaAs light emitter. A new use of LEDs is important in the circuitry of oscilloscopes (5.7). Here LEDs are used in storage target tubes to allow a high potential of the storage target with respect to ground such as the overall accelerating potential of the CRT is, e.g., 18 kV.This is done by elevating to k 15 kV the circuits supplying the target waveforms and then optically coupling them to the ground potential control circuit (see Fig. 5.16). There are numerous other applications of LED and readout pairs, e.g., as null indicators. Here the LEDs are used, e.g., in the feedback loop of an operational amplifier. When both LEDs are dark, the input voltage is null (see Fig. 5.17). (A, is a differential amplifier which drives differential amplifier A2.)

173

LIGHT-EMITTING DEVICES. I1

-

Common

cornnwn

elevated to + I 5 kV

ground-

+5v

store

, store

erase generator

::Enhence +5v

P

1

i

FIG.5.16. Target-control block diagram for optical links used in store mode, store enhance, and erase functions.

Sensitivity Switch

I"'

8 U J f $

__

1 I

E,>OV

[

E,
I

, E,aOV I

(zero trim)

, I

LED 2 Visual Indication

FIG.5.17. Null indicator with LEDs. See accompanying table for visual indicator and correlation to level of input voltage.

174

HERBERT F. MATARE

A growing application of LEDs will be found in the areas of optical switching. Digital switching in telephone networks and in data transmission is bound to make great strides. With the enormous increase in LED efficiency, the time is near where one cannot afford metal contactors in all network and relay applications. PCM switching systems are already in use and one system already handles 240 digitized voice channels. Here the LED is the desired device.

C . Photocouplers (Isolators) One of the commercially developed applications of infrared LEDs is the isolator or the photocoupled pair of an infrared LED, emitting in the 900-pm range and a silicon diode, or a silicon transistor, or a siliconcontrolled rectifier (SCR). Couplers using light injected from neon lamps or tungsten lamps and light detectors have been used in the past to connect circuits which work at different potential or to activate a network at high voltage with respect to the control network. Frequency matching being perfect for GaAs LEDs and silicon diodes, the technique of photocouplers developed rapidly after the first infrared LEDs were available. Isolation voltages can be chosen according to the need given by the application. One has to balance transfer ratio and isolation in each case. If low voltage separation is sufficient (a few loo0 V), one can use a simple isolator sheet separator. That is, the LED and the detector are separated by, e.g., a glass slide or a plastic layer, facing each other. If higher separation voltages are required, one has to put a wider distance between LED and detector or use a fiber as coupling medium. In each case there is a balance to strike between isolation voltage and current transfer ratio. The latter can be defined as

Y = Ioutflin and decreases with separation of the devices. It is relatively small (61)for LED diode couplers and can be 1 for LED transistor couplers because of the built-in amplification. Figure 5.18 shows the different couplers with their main features. The case of Fig. 5.18a is the preferred mode for high speed response couplers. The photodiode can be a fast PIN device. Frequency response is therefore limited only by the LED. If a high-speed GaAlAs/ GaAs SH or DH structure is used, a 100 MHz response for data transfer can be achieved. In this case a high frequency amplifier is attached to the diode output. This microcircuit amplifier can also be part of the package. In some cases such couplers are working with fast red (GaAsP) LEDs and the amplifier is monolithically integrated with the detector.

=-

175

LIGHT-EMITTING DEVICES. I1

(output)

+V

aIin

high FIG.5.18. (a) LED Photodiode coupler with output characteristic lo-* > 7 > low frequency frequency response (to 100 MHz), (b) LED transistor coupler, 10- > y > response (to 100 kHz), (c) LED photo-Darlington coupler, 10 > y > lo-': low-frequency response (< 100 kHz).

'

In the case of Fig. 5.18b, the usual LED phototransistor coupler is shown. The speed is limited usually to 100 KHz. This is the most used package for simple isolation tasks. The use of the photo-Darlington (Fig. 5.18~)improves greatly the current transfer ratio of the package but cuts down further on speed. The output or load resistance RL influences the limit frequency response (increase over a factor of 10 when RL changes from 10 kQ to 100 Q). Couplers are produced in ever more sophisticated forms for varied transfer ratios, isolation voltages, and speeds. In some cases, a protective diode is aded to the LED input circuit because the LED can easily be damaged by a negative surge voltage. Couplers of the forms shown in Fig. 5.18b or c are used to activate relay circuits and controlled rectifiers (SCR).

176

HERBERT F. MATARE LOAD

FIG. 5.19. (a) LED Phototransistor coupler with separated power supply, driving load circuit SCR. (b) LED SCR coupler deriving the energy to trigger the load circuit SCR from its anode power supply. (c) LED SCR coupler with separate power supply.

The direct-coupled SCR, however, presents great difficulties due to the fact that the function of a light-sensitive gate region is contradictory to other desirable SCR properties as low dV/dt and high dl/dt. A photosensitive SCR obviously triggers at low gate currents and there are difficulties to avoid switching during undesired line-current surges. Therefore, it is customary to drive the main SCR with a phototransistor or other photo SCR. Figure 5.19 shows three cases: (a) Coupler with phototransistor and separated power supply connected to SCR gate. (b) LED-SCR coupler deriving the energy to trigger the load SCR from the latter’s anode supply. (c) LED-SCR coupler with separate power supply.

177

LIGHT-EMITTING DEVICES. I1

There are numerous ways to solve circuit problems with more sophisticated combinations of couplers with SCR load circuits, especially in ac applications (5.8). In some cases the opto-coupler is combined in one package with a complete microcircuit amplifier. Here and in the case of a Darlington amplifier, current transfer ratios can be much in excess of 100%. TABLE 5.1 CURRENT TRANSFER RATIO

PROPAGATION DEJAV

7? a'

PHOTO DIODE

3-fI

0.1- 0.2

%

10-20 n sac

5 MHz

< I nA

PWTO TRANIISTOR

2-125

3-9

%

< lO/,lseC

100-600KHz

<5nA

PHOTO DARLINQTON

100- 600

500 %

%

-

50 I50p.962

l/,l

88C

I-IOKHz

2OOKHz

.IpA

lOOnA

Table 5-1 gives a few typical values for the important coupler cases. Opto-couplers have become a workhorse for industrial equipment because they are solid state compatible, rugged and of low power consumption and because they can be adapted to all possible cases of isolation needs and transfer requirements (5.9).

D. Optical Radar The availability of high-intensity IR emitters and high-sensitivity IR detectors was the necessary condition for increased activity in the field of the optical radar and range finding systems. The earlier availability of the ruby laser, the C 0 2 laser and other high-power laser sources had prompted a development of the radar field into the optical spectrum. Laser guidance and ranging are techniques especially well developed for military purposes. The refinement of the LED and IR semiconductor laser has added new capabilities in this field.

178

HERBERT F. MATARE

The simplicity, low power handling, and small size of the semiconductor IR sources have produced new technical possibilities. Since beam powers in pulsed operation of more than 10 W are now achieved and of 100 W for stacked semiconductor lasers, a radar system is feasible. Especially since the sensitivity of silicon detectors is in the lo-’’ W range, there is a wide dynamic margin between transmitter and detector even when strong absorption losses and scattering losses prevail. As we have seen (Part I, Fig. 1.2), water vapor absorption (percent per beam path in precipitable water vapor) is low in the region between 0.8 and 0.9 pm and in some regions further and in the far infrared. In fog and dense rain or snow, water vapor absorption is not the main reason for losses, but scattering by suspended droplets and resonance absorption are. However, with the developed high-efficiency LEDs and lasers and newly developed III-V active IR cathodes for vidicons, as well as with a more developed IR receiver technology,the dynamic power range can be extended into the range of 1013to 1014 (10’W transmitter power, lo-’’ W detector mixer sensitivity), allowing for a wide margin of losses. Consideration of the optical radar shows that its advantages with respect to conventional centimeter-wave radar are (1) Higher resolution. (2) Small beam width (directivity). (3) Small equipment size and power consumption. The disadvantages are (1) Absorption and scattering losses in air. (2) Pulse modulation frequency limitation (especially with LEDs). (3) Resolution limitation due to detector bandwidth limitation. Consider the transmitted power W emitted into a beam of solid angle ( R = distance from radiator to target), or

nT.The target area AT therefore sees a power W if AT > nTR’

WAT&RZ if AT
wz AD

The product of (5.1) and (5.2)is the detector power:

LIGHT-EMITTING DEVICES. I1

179

In addition, the attenuation in the atmosphere is generally of the order of exp( - 2aR), where u is the absorption coefficient. For good conditions, a 40-km visibility can be assumed and u = 0.05 km-’ (5.f). Restriction of the detector field of view to eliminate background radiation and filtering is a limited means for correction. A restriction in the detector bandwidth would reduce ranging accuracy Ax, as the pulse length 7 must be of the order of Ax/c (c = 3 x 1 O * O cm/sec) or the bandwidth Af 2 c/Ax. For example a resolution of 10 cm requires a pulse length of 1 nsec or a detector bandwidth of 1 GHz. There are problems at the IR transmitter side to achieve this resolution. Pulse length in the 100 nsec range, feasible with fast LEDs, will result in resolutions of the order of 10 m or a bandwidth of 30 MHz. Laser transmitters will achieve higher pulse frequencies and smaller optical beam spread. The choice of the IR emitter depends on the range R to be covered, the target size, the allowable equipment, and size of the power supply. For smaller distances of the order of a few kilometers, LED transmitters are applicable and can result in relatively compact (portable) radar equipment. On the side of the detector there is a wide margin of technological development open for exploitation. It is based on the use of down conversion and heterodyne detection and leads to a more sophisticated integrated optics, compatible with small equipment size and small power consumption. Infrared parametric amplification is usually based upon the mixing of a signal in a nonlinear crystal with an intense monochromatic source, called the pump, which produces sum and difference frequencies. The sum frequency combination is used for up/down conversion and harmonic generation. The difference frequency combination is used for amplification oscillation and rectification. When sources are used where stimulated emission is dependent on particular natural resonant frequencies within the atomic structure (as in gas lasers), optical parametric interactions add a welcome freedom of selection of specific frequencies and bandwidths, since their optical interactions depend upon a reactive nonlinear phenomenon in the transparent band between atomic resonances. GaAs waveguides are used extensively as a nonlinear crystal medium for parametric conversion since they can be shaped in geometry and dimensions comparable to the operating wavelengths in the infrared range and because GaAs is transparent in the infrared and combines a lack of birefringence with a high index of refraction. The technique of parametric amplification and conversion is, however, based on a high pump power. In the case of the pulse-signal amplification in IR ranging and general radar equipment, the advantages of the frequency mixing or heterodyning can be realized without high pump power if the methods used in microwave mixer technology can be adapted.

180

HERBERT F. MATARE

We now detail such methods because they are important for future IR ranging experiments. The first step, before full integration can be accomplished, is always the hybrid approach or the combination of optical components with microwave technology in a way accessible t o present technologies. An important aspect of receiver technology is the sensitivity extension by superheterodyne detection. The direct detection of demodulation of a high frequency signal at a nonlinear device and amplification of the modulation frequency have been well analyzed (5.10). If the nonlinear characteristic of the detector is developed into a Taylor series: J ( t ) = J o + ( $ ) o A u + g (1 m a25 )oAu2 1 ia33\

the current can be expressed as J = Jo

+ S AU + -21T!

1 AU2 +- W A U 3 + . . * 3!

(5.5)

AU = signal voltage amplitude

Zr = U , + U , cos or = U , + ~ ( +1m cos p t ) cos or = modulated

carrier amplitude (o= carrier frequency)

U o = dc voltage

U , = ac amplitude of voltage m = degree of modulation p = modulation frequency

Expressing the rectification current by integration over a carrier frequency amplitude :

one obtains

J ( t )= (m/2)TU2 (5.7) If the detector (R,) is coupled to an amplifier with the input resistance R,

LIGHT-EMITTING DEVICES. I1

18 1

and equivalent noise resistance R e , the signal-to-noise ratio can be shown to be (5.10)

F* = equivalent noise factor

u = R/R, =

(

x = RJRD =

equivalent resistance external (input) resistance external (input) resistance detector resistance

Equation (5.8) can be reduced in most practical cases to

where q/Af = ratio desired per Hertz bandwidth. For q = 1 and mA/V

S

N

SIT

2:

1/20

Af = lo6 HZ

(m = 1)

one obtains

($)

= 3.75 x lo4

[kT,]

(5.10)

opt

The necessary signal strength is calculated to be between N = lo-' and W . sec). In the case of the superheterodyne W (kTo = 4 x detector receiver the sensitivity or signal-to-noise ratio can be improved considerably. Details of the reasoning can be found in Matare (5.10).We can only point out here that the oscillator frequency o,will produce the intermediate frequency 01,

= w,

- 0,

(5.11)

where o, is the signal frequency, which is in a lower range and easier to amplify. In addition, the ac exploration of the characteristic with a local oscillator amplitude greater than the signal amplitude allows us to adapt the

HERBERT F. MATARE

182

detector input impedance and to optimize the sensitivity on account of the reaction factor:

(5.12) flop1

= b(sc/s,)2

(See Ref. 5.10)

S, = conversion slope

S , = rectification slope x = RJR, y increases with x -,00 and 8 = arc cos(U&,)

decreasing.The current flux angle 8 cannot be made to decrease to 8 = 0 because of the reverse characteristic of the detector and its noise equivalent power; but there is a possibility to increase the mixer sensitivity to values between N , , ~= ,

10-12-10-13

w

for a bandwidth of several MHz [for details, see Matare (5.fO)I. The oscillator frequency wo can also be an optical signal produced by an LED and the intermediate frequency wIF= uL-t w i can be produced as the sum or difference frequency of two optical frequencies. However, if a modulation frequency w, (signal) is contained in an optical carrier of frequency w L, one can form a new signal frequency 0:as the sum or frequency difference between wL and uo,the local light oscillator: O L

f w, = w:

(5.13)

To generate 100 MHz or 10' sec- difference frequency means that wL- u, has to be in the range (aL'v l O I 4 range) to form an intermediate frequency wIF= w, k 0:in the 10' Hz range. However, there is always the possibility to form harmonics and subharmonics of the oscillator or pump frequency. Parametrically related frequencies generated are wp - w , = w j > w p

w p - w1 = 0

2

> w1 < w p

Between these and harmonics 2wp, 30, and subharmonics hp, &of, etc., new combination frequencies with w, can also generate an intermediate frequency. In the case of optical carriers with microwave subcarriers we have several options to separate the subcarriers or signal frequencies from the optical carrier. The general method is the incorporation of Gunn oscillators as active devices into a hybrid stripline circuit in such a fashion that the push-

LIGHT-EMITTING DEVICES. I1

183

pull input of the microwave oscillator works on the optical detector, which has one side grounded. In this way the mixer is balanced and works with suppressed oscillator noise in the intermediate frequency channel (5.10, 5.1 1). As is known in modern microwave integrated circuit technology (5.12), such circuits can be built in very compact form on metallized ceramic layers. Modern solid state microwave devices are compatible with this technology (5.12). A few examples of hybrid microwave optical circuits are briefly discussed because they convey insight into a growing field which will probably prevail for a long time before monolithic integrated optics have developed.

\METALLIZATION

FIG.5.20. Schematic of an opto-hybrid microwave circuit with Gunn oscillators: B = base capacitor, R = RF blocking impedance, C = Gunn oscillator, D = diode detector (semiconductor), (oL= light frequency, w s = signal frequency, and IF = intermediate frequency.

In Fig. 5.20 the scheme of such a circuit is shown. The Gunn oscillators (G) drive the oscillator circuit, which couples to the detector (D), subjected to the incoming light wave and microwave or high frequency carrier. The demodulation and mixing with the local oscillator frequency takes place within the nonlinear detector. Since the microwave subcarrier may have several modulation frequencies or bands, separation into I.F. channels can

184

HERBERT F. MATARE

be effected by filters or selective local oscillator injections from tuned Gunn oscillators at frequencies w h , a&,etc., to form subchannels Awl; Am2, etc.: A u I = Wsignal & wb

AuZ = w,ignal& W& Contacts B serve to bias the detector with a high impedance to the microwave signal w, . The I.F. side is grounded through I.F. output. A capacitive coupler C connects the stripline with the detector D to two microstrips tuned to the operating frequency w,. The Gunn oscillators are biased by contacts B (through R).

M I

I

I

I I

I

I

I

LEDnl

LA \ \ \ \ \ \ \ \ \ \ \ \ \ \

\ \

\\\\\\\\\\A

\/

\METAL

FIG.5.21. Schematic of an opto-hybrid microwave circuit with photoconductor (PC) as light mixer and LED as local oscillator: T = transistor, w L = light frequency, w s = signal frequency (microwave carrier), 0: = combination frequency to form I.F.,and Q, = local oscillator light frequency.

In another case of reduction to practice of the scheme proposed, the local oscillator is a light injector LED (see Fig. 5.21). In this case, application of the light mixing properties of a photoconductor is made and combination frequencies wiignal= wL f woScresult. Here optical gain can be achieved, restricted only by the gain bandwidth product limitation. The generated I.F. . as usual to a microstrip transistor signal of frequency w ~ .is~ processed

185

LIGHT-EMITTING DEVICES. I1

amplifier. The difference frequency between the optical frequency wLof the incoming light and the frequency of the local LED, w , , can be located into the microwave region, or parametric relations can be used. The nonlinear detector or photoconductor also forms an intermediate frequency between this difference frequency, 0:= w L k a,,and the microwave subcarrier 0,: w1.F. =

k

This frequency is amplified by the transistor amplifier T. Again, B designates all base capacitor layers and R represents the connections of high impedance for microwaves (wJor mi). In this scheme, two basic advantages of optical mixing are put to work: (a) Conversion gain, which is characteristic for optical mixing. (b) The fact that the optical difference frequency, which is located in the microwave region, is converted in the same nonlinear mixer into the I.F. frequency, amplified by the transistor T. Because the sensitivity to the incoming low level light signal has to be optimized, a few new approaches are proposed. In Fig. 5.22, we consider the use of a subcritically operated Gunn oscillator G as a sensitive and fast

0

-F

1.E >>

FIG.5.22. Schematic of opto-hybrid microcircuit wRh subcritically operated Gunn diode (G) as mixer amplifier and LED as local optical oscillator (me): T = transistor amplifier, oL= light frequency, and o,= signal (microwave frequency).

186

HERBERT F. MATARE

optical detector. A local LED modulates the Gunn wave at the same time as the light signal is focused on G. The optical combination frequencies 0:= 0 L

f w,

in the microwave region will be superimposed on the signal w, and mixed, to form intermediate frequency

The advantage of using a Gunn oscillator in a subcritical mode is that triggering into oscillation by light injected carriers results in signal enhancement (5.13). The process is efficient for frequency conversion when the triggering local LED is so adjusted that the Gunn diode is sensitized for the incoming signal us. The latter appears as a modulation of the combination frequencies oLi-0,. A disadvantage is that higher noise levels are expected due to the forming electron plasma as the subcritically biased Gunn diode is subjected to a high electric field. Therefore two new schemes making use of devices, recently described and measured, are given.

FIG.5.23. Traveling Gunn amplifier coupled to light signal and working through a filter F on transistor TR1. Here the demodulated signal carrier is mixed with a frequency w, generated by transistor TR2. C are capacitive decouplers and contact areas.

LIGHT-EMITTING DEVICES. I1

187

One of these is the traveling-wave Gunn ampliJer, which works in a similar fashion to the device mentioned before, but uses a long path in a lateral Gunn diode for signal enhancement. Figure 5.23 describes this setup in a semiperspective view. The optical signal: h(wL+ w,) is received and demodulated at the traveling Gunn device. From here, it is processed through a microstrip with filter action to a transistor TR 1, where it is mixed with a local oscillator frequency from TR 2. For very high microwave frequencies w, the transistor TR 1 is replaced by a Schottky diode mixer stage (see Fig. 5.24 for the equivalent scheme).

or

i

FIG.5.24. Circuit diagram for schemes of Figs. 5.23 and 5.25.

An aiternate opto-hybrid microwave circuit using a self-pumped optical detector is shown in Fig. 5.25. Here we take advantage of the fact that injection of bandgap radiation into a detector sets up an enhancement field in the junction and through pumping action allows signal amplification. This arrangement, in combination with a Schottky barrier diode mixer instead of TR 1 and a local Gunn oscillator at the place of TR 2, is very favorable. The last-mentioned combination of devices should result in efficient opto-hybrid microwave receivers, especially if one can combine the function of a Schottky barrier diode mixer with that of the GaAs heterojunction detector (Fig. 5.26) in a balanced mixer mode.

FIG. 5.25. Opto-hybrid microwave circuit using a self-pumped optical detector (labeling as in 5.23).

iAlA8

FIG. 5.26. Balanced mixer using self-pumped optical detector and Gunn oscillators (pushpull) with transistor amplifier. Local oscillator can also be one Gunn diode: I.F. = intermediate frequency, Coll = collectors, B = base, and C = capacitor.

LIGHT-EMITTING DEVICES. I1

189

E. Optical Communications and Integrated Optics There is no doubt that a substantial amount of all communications will in the future be based on optical channels. This will have a great impact on telecommunications as we have known it. Telegraph poles with hundreds of wires drawn across the landscape will disappear, and instead a small fiber or a fiber bundle within a protective skin a few feet underground will connect the cities and towns. Since even now with infrared LEDs the frequency response has been extended into and beyond the 100 MHz range, there is a basic capability here for 10,OOO telephone conversions per fiber or 10 color TV channels. Multiplexing systems may not be overextended when several fibers are used per cable. The literature on optical fibers is so voluminous that we cannot attempt here to cover it (see Part I). It should be stated, however, that with the advent of a fiber loss below 2 dB/km (for monomode fibers), the technical situation now calls for applications in various fields. There are two basic areas of development:

(1) Short range multipurpose transmission systems. (2) High data rate long range systems. In the first case, the distance covered will typically range between 30 m and 0.5 km, with eight or more input and output terminals and a 10 to 100 Mbits/sec information capacity. Such systems will consist of a multimode fiber optics bundle with LEDs of the fast type (to 100 MHz) and silicon PIN diodes as detectors. Where group connections between several buildings, hotels, industrial complexes, and government agencies are required, such compact systems will be the future solution. These systems can also cover connections where underground or overground fibers are not practical. In these cases a direct optical transmission via air is feasible to complete a complex system of telephone, picturephone, or TV intercommunications networks. High data rate long range systems can compete with long distance telephone systems, since repeater stations are needed at longer distances only than is actually the case in present long distance telephone systems. The necessary multiplexing technique or separation into channels for the different carrier frequencies is state-of-the-art. With the advent of integrated optics, this technology will offer compact and interchangeable units. Many problems were foreseen for the case of fiber optic interconnections, as it is a lossy process to couple into and out of optical fibers. It appears, however, that the application of the technique of beamsplitting can be minimized since one may combine a repeater station with each coupling point, where the coupler can be combined with a multiple LED or laser output of the amplified signal.

190

HERBERT F. MATARE

For short-haul systems of the first kind all components are available (5.14). The superiority of the GaAlAs-GaAs heterojunction LED for high speed optical transmission was demonstrated here as compared to the known GaAs-GaAs homojunction LED. Much work has been done to develop high speed LEDs applying this material combination. A rugged and relatively simple construction based on a process of meltback and regrowth has led to an LED with high power capability and speed (10 MHz) that is used in TV transmission systems via air and fiber (5.15). Other types with multiple layer structures are developed: Single heterojunction (SH) device:

Al,Ga, -,As A1,Ga1 _,As Al,Ga, -,As

(p-type) (p-type) (n-type)

and : double heterojunction (DH) device: Al,Ga, -,As Al,Ga, -,As Al,Ga, -,As

(p-type) (p- or n-type) (n-type)

or large optical cavity (LOC) device:

AI,Ga, -,As Al,Ga, -,As Al,Ga, _,As Al,Ga, -,As

(p-type) (p-type) (n-type) (n-type)

or four-layer heterojunction

Al,Ga, Al,Ga, Al,Ga, Al,Ga, Al,Ga,

(p-type) (p-type) (p- or n-type) (n-type) (n-type)

_,As -,As -,As -,As -,As

Measurements show that the SH and DH structures as LEDs are capable of satisfyingmost of the needs of high data rate transmission, that modulation capability can be extended into the 0.1 GHz range (5.16), and that there is a potential of higher frequency modulation. The compositional range (A1 concentration) and resulting stoichiometry as well as the dopant type (Si and Ge) are very critical for the performance of these devices, and it has been shown, e.g., that for a single heterostructure the A1 composition or the value of x in Gal -,Al,As should not exceed 0.2 for high frequency response (5.17). Further improvements with respect to frequency response and compatibility with fiber optics will certainly lead to devices with a 0.5 GHz response and high quantum efficiency and lifetime (5.18). By confining the active junction area by proton bombardment (p-type conversion), modulation cutoff frequencies of 170 MHz and a radiance of 11 W/sr cm2 (100 mA) were achieved, whereas DH structures with Ge doping already achieved 250 MHz (5.16).

LIGHT-EMITTING DEVICES. I1

191

Laser technology, on the other hand, has improved device lifetime into a range close to that for LEDs. With rise- and falltimes in the nanosecond range, there is here a GHz modulation capability and high optical beam power plus high spectral purity (optical losses); the only restriction is the duty cycle in pulsed operation. CW operation as a mode of reliable laser application is still a problem (5.19). LEDs are also specifically developed for fiber optic communications. Flat geometry for coupling to fiber ends and bundles is combined with a flask and housing to permit easy connection (5.20).Multimode fiber bundles have in many instances a 45 mil diameter. In these cases the Lambertian source of a 40 to 45 mil2 die is a useful signal generator (5.15). One of the first areas of application of fiber optic transmission has been the military aircraft (F-14, F-15, and B-1). A data bus, i.e., a single transmission Line carrying many different multiplexed signals servicing a number of spatially distributed terminals, is less expensive to install and maintain than conventional cable connections. Its lighter weight, smaller size, and reliability are attractive features. It also can easily be modified and expanded and is less vulnerable to damage than systems based on point-to-point links. Usual wire or cable connections are susceptible to reflections, ringing, cross talk, ground-level voltage shifts, fire damage, and have bandwidth limitation. The technology of electrical transmission lines reached maturity decades ago-the coaxial cable of today is not much different from that of 1950-and there is little change or major improvement in this area; while fiber optic transmission promises great improvements. The bandwidth of multimode fiber optics lines is, however, orders of magnitude higher than that of a coaxial cable of comparable size and weight. Ground shifts cannot occur because the terminals are electrically isolated from each other. Also, fibers will operate at temperatures considerably above the 300°C rating of high temperature electrical cables and are less subject to radiation damage. Under this aspect data bus lines have been developed for varied distribution schemes. It is important to balance the distribution scheme (the various pick-off fractions) and the overall optical losses (5.21).Calculations of different distribution schemes show that there are both.advantages and disadvantages for each scheme and that each case has to be carefully planned and evaluated. For instance, the tree or star distribution scheme (Fig. 5.27) with equal derivative points requires a relatively high overall power level, whereas the parallel distribution (Fig. 5.28) can operate at lower power. In the first case, one has, assuming equal length of fiber between the derivative data points, (5.13)

192

HERBERT F. MATARE

A -

8Ps/a3b3c2

t

4P,/a2b2c2

1

FIG.5.27. Tree or star configuration of data bus with power ratio at each derivative point:

P,= total power of emitter to cover three data extraction lines, and P, = output power of the last distribution point, a,b = interface losses at each data point, c = loss within a segment of length I = Ljn, and n = number of derivative points between emitter E and each data point.

n = number ofderivative points between emitter E and a data point P, = total power P, = power at the output of the last distribution point a,b = losses at interface of each data point c = losses in a segment between emitter E and one data point of length L/n (Lis the fiber length) For the scheme in Fig. 5.27 the emitter has to generate the power: 8Ps P, = a3b3c3

(5.14)

In the case of parallel feeding (Fig. 5.28) of data extraction points, one has (5.22):

+ ...

ps +-abc

(5.15)

Further development in optical communications is much dependent on progress in integrated optics. We have discussed the use of optical compon-

LIGHT-EMITTING DEVICES. I1

193

FIG. 5.28. Schematic of a parallel distribution system': P , = total emitter power, P , = output power at the distribution points, n = number of derivative points between emitter E and each data point, a,b = loss at interface of each data point, and c = loss within a segment of length I = 4 n .

ents such as LEDs and photodetectors in connection with microstrips and hybrid microwave circuits. In these cases, the optical signal was confined to the optical component, where it was transformed immediately into a useful microwave or I.F. signal. In the future, however, the use of optical circuits will enable the designer to process optical signals the way microwave signals are processed and to design circuits which give access to optical lines and compounds in the way this is the case at present with striplines. What are the advantages of optical circuits? To answer this question we only need to recall the reasons for the creation of distributed impedance microcircuits in the form of striplines. This step from lumped microwave devices to distributed parameter devices allowed optimization of impedance matching, low-loss coupling between individual devices, circuit functioning of distributed parameter waveguides, and prefabrication of entire circuits in batch processes. In the case of optical circuits, the reasons are similar. In building optical waveguides, the local excitation can be channeled to a given point in the optical circuit and transferred into a different type of mode, polarized or split into several beams to effect optical circuit functions. Most of the experience on optical waveguides has been gained on thin film material, especially GaAs and GaAlAs on GaAs substrates. A difference in dielectric constants between substrate and top layer is sufficient to confine optical excitations as long as the condition

, is the is fulfilled, where AE is the change in dielectric constant (= E~ - E ~ ) Lo free space wavelength, and t is the thickness of top layer or waveguide. Such a dielectric discontinuity may be produced in several ways. For example, if a surface layer of GaAs is epitaxially deposited (as monocrystal),

194

HERBERT F. MATARE

but with a higher carrier concentration, the difference AN introduces a dielectric difference E2-&,=-

ANe’ m*w2

where m* is the carrier effective mass, w is the radian optical frequency, and e is the carrier charge. The epitaxial layer may be doped with the same dopant and may be less or more doped than the substrate. Proton bombardment (or ion implantation) can be used to convert the top layer of a heavily doped GaAs crystal into a semi-insulatinglayer. In this case, however, sufficient annealing is necessary to decrease the number of defects introduced by energetic particles (3 x 10’ eV, for example). Losses decrease strongly for higher annealing temperatures. The annealing temperatures have to be in excess of 500°C. But as long as one depends on carrier density differences for guiding of optical waves, optical losses due to energy absorbing carrier density differences are high. Therefore, the best methods are those by which the difference in E is brought about by compositional differences. Here, the compound Ga,AI, -,As on GaAs is most interesting. First of all, it has been studied metallurgically, since it is the basis for the most efficient LED and laser structures. The difference in lattice constants is very small, GaAs (5.646 A) and AlAs (5.639 A), and can be decreased further by the ternary layer Ga,AI, -,As on GaAs with, e.g., x = 0.7. Furthermore, the ion radii of Al and Ga are equal (1.26 A) and the number of interfacial dislocations can be kept low (5.23). When two layers of GaAlAs are used, the difference between Gal-,Al,As and Ga,-,AI,As is sufficient (y > x, for example) to induce index discontinuity

Close confinement structures like those used in lasers can also be used to guide optical waves within the enclosed GaAs layer. An important feature of GaAs optical waveguides is the fact that an electric field will cause phase and/or amplitude modulation. This is due to the large electrooptic coefficient of GaAs and (GaA1)As. A thin layer of metal forming a Schottky barrier on the top layer can be used. Back-biasing the barrier results in a sufficient field to induce switching (Fig. 5.29a). Much higher modulation efficiency can be induced by the backbiasing of sufficiently different N - P structures (Fig. 5.29b). A wide range of additional technical features can be exploited as, e.g., coupling of optical energy. Two waveguides in a parallel position can be

195

LIGHT-EMITTING DEVICES. I1

Ga,-,AI,As(n) Ga,-,AI,As

,

t

z,

f

(n)E,

-

t

Ga,AII-, A s h ) Ga,Al,-,As(pl

,---7

FIG.5.29. Field-induced variation of intensity profile in optical waveguide formed by ternary compound junction: (a) different dielectric constants, and (b) n-p junction.

coupled by proximity to allow a spillover of energy from one guide to the other. Energy fed into one waveguide spreads as it propagates into other waveguides. A band-filter effect can originate between two waveguides with perfect wave synchronism, as the energy will oscillate between two modes at a spatial frequency” which depends on the coupling strength. Since this coupling can be controlled by an electric field, more complex functions such as switching, multiplexing, coupling variations, etc., can be achieved. Energy can be coupled into such waveguides from the outside by directing the light beam toward a surface grating structure with, e g , A/;! distance between the lines. Such gratings are produced by normal photoresist pattern deposition. Ion milling through photoresist masks has been applied, and, also, laser standing wave illumination has been used to induce such structures. Other materials have been used to produce waveguides, e.g., LiNbOJ and LiTaO, . In this case, one cannot grow different material combinations easily. Differences in index of refraction were generated by careful out-d flusion (5.24). “

196

HERBERT F. MATARE

Diffusion has also been used to produce waveguides in 11-VI compounds. In these, the natural transparency of the material within the interesting 1R spectrum helps to produce low loss guides. However, the emplacement of optical waveguides by diffusion in a masking operation has limitations (5.25). Diffraction limits the width x thickness/length ratio of bulk waveguides as modulators. As the modulation power P is proportional to Wtll

and wtll is Limited to A/4n (n is the coefficient of refraction), we see that there is a limit to decreasing P. In thin films, however, this power can be decreased by two to three orders of magnitude (e.g., t = 1 pm). In the case of directional couplers, lateral guide precision is of great importance. Here methods of electron beam masking and photoresist mask preparation using standing wave illumination by lasers can yield the necessary resolution (5.26). Optical waveguides have been made by ion implantation into quartz (5.27). But, as mentioned before, losses are generally high, due to defect generation in this case. The work on proton implantation is interesting insofar as a high resistivity layer of small thickness is created at the 1 pm for each 100 keV). Carefully directed beaming and evensurface tually masking can lead to a good line precision in this case (better than in diffusion); however, there is still the need for annealing over 500°C (5.28). More recently, channel waveguides have been induced in glass by laser heating (5.29). In developed optical circuitry, coupling and function injection into optical waveguides is an important process. Structures such as those shown in Fig. 5.30,or multiple structures, are developed in many techniques. Thin films have been used for wave propagation in the 10.6 pm region, applying AgBr. These films were epitaxially deposited on NaCl substrates (5.30). Another method used is migration of ions in glass plates. In this case, the migration can be induced electrically along a wire electrode, and the resultant material is rather perfect, leading to small optical losses (about 0.1 dB/cm or 10 dB/m) (5.31). To complete this survey, we add that light guiding structures have also been made in photoresist films (5.32) and in single-crystal garnet films (5.33). To assess the trend which emerges from these efforts, we first have to assess the situation concerning the loss in optical guides. For long distances, the only light guide available at this time is the glass waveguide, single or multimode. Progress in producing losses down to a few dB/km has been made and will start a new technology of optical data transmissions. Values of 4 dB/km seem to be realistic and will soon allow wide applica(h-

197

LIGHT-EMITTING DEVICES. I1

\

\

FIG.5.30. Coupled optical waveguides and energy distribution (schematic).

tion in interurban and even long distance optical communications. At 0.9 pm, the ultimate low loss seems to lie at 2-3 dB/km (5.34). Other authors working on bulk fused silica have even estimated a lower limit (5.35). Next Generation Optoelectronics

As losses are important for all optical guides, they are less important for short-length optical circuits. Here circuit technology (line precision, etc.) is more important than the loss figure. A damping of 10 dB/cm is good enough in many cases of circuit functions, as the lengths in those circuits are in the range of centimeters and below. We have to mention, however, that low-loss optical wave propagation and strong electrooptic coefficients for optical circuit functions are the most desired combination. At the time of this writing, such a combination is feasible with the ternary compound (A1Ga)As and, in particular, with the double heterostructure Ga,AI1 -,As-Ga,Al, -,AS, with y > x , for example. As we have mentioned, these compounds are very compatible with GaAs as substrate material and are easily grown by liquid epitaxy, which is the most efficient way to generate high-perfection junctions for high-efficiency light emitters and waveguides.

198

HERBERT F. MATARE

While future trends are certainly in the direction of further improvements of ion-implanted and proton-implanted structures, the next generation of electrooptic circuits for communications is of the “optohybrid” kind. In these circuits, the optically active material is in all probability GaAs and its ternary compounds, mainly Ga,AI, -.As. As we have seen before, such material combinations allow the generation of IR light emission, modulation, detection, and polarization in one circuit (5.36).

I

----

GUIDE

3

2 1

0

Ga,,Alo,,As

(3xlO%m~’)

P

Gao,,,AI,,o,As(~3x~O‘~)

N

Ga,, Al,,As(3~10’~)

N

G a As (N) ( 1 0 ’ 8 C d 1

}

0.4

-~p

4

-bias I

I I I

Structures like those shown schematically in Fig. 5.31 have been used to guide optical waves within a center layer of higher optical dielectric constants. Depending on the mode, TM or TE excited, the transmission varies with the bias and is also a function of the optical frequency (between 0.90 and 0.92 pm). The dependence on the applied bias (between 0 and - 16 V, for example) is very strong (transmission changes for two orders of magnitude) and promises excellent intensity modulation and also photocurrent variation (transmission lowered or absorption higher for strong reverse bias of the junction). One can now foresee optical circuits produced from such epitaxial material which incorporate in one block of 111-V material the following functions (5.37):

(1) Optical signal detection (2) Local light emission (local oscillator) (3) Light coupling (4) Polarization (5) Intensity modulation (6) Microwave subcarrier isolation (7) Coupling to a microstrip circuit

LIGHT-EMITTING DEVICES. I1

199

As future communications systems will be based on progress achieved in

optical waveguides and couplers, present-day systems are mainly a combination of microstriplines and microwave processing, with optical components included in the stripline buildup. As far as optical communications in the sense of interurban connections is concerned, the new fast and highly efficient LEDs are the best generator with easy modulability. Duty cycle can be much higher than in lasers and no cooling is desired. Transition to microwaves is not necessary for TV transmission and telephone distribution lines. In coupling to an LED via the necessary transistor amplifiers, we face the problem of maintaining the desired bandwidth in a frequency range sufficiently high to handle gigabits/second. It is an interesting fact that modem (fast) LEDs generally represent a device more adaptable to high-frequency operation than would correspond to the circuit performance on account of its impedance range. A device with a dc capacitance of about 100 pF (corresponding to a usual die size and junction bias near zero) should represent an impedance: 1 1 z=--=-wed

530I(m) 1 C(pF) d

where d = A f / f may be assumed to be 0.1 for an assumed frequency of f = 100 MHz. I is 3.0 m. Therefore,

z-

1600

a reasonable circuit value at the resulting bandwidth of 10 MHz. Approaching 1 GHz as basic frequency, the impedance value would drop to 16 0, but allow for a bandwidth Af= 100 MHz. But, as is well known, the strong forward bias of these devices changes the impedance from negative to positive (5.38).This change to inductive behavior accounts for a partial compensation of the capacitive loading. Another factor is that trap filling at higher current levels improves the device response speed such that rise- and falltimes for devices with a dc capacitance of several hundred picofarads is still in the nanosecond range. With these data, it is clear that systems design will heavily converge toward the use of LEDs when monochromaticity is not a factor per se. Details of the state of integrated optics are best taken from the recent summary by Marcuse (5.37). We conclude with the statement that optical communications are here now, as air-to-air systems are already marketed (TV transmission, data transmission), and fiber-optic systems are installed on ships and airplanes. The enormous flexibility of a system based on generators and sensors which allow either fiber coupling or free space connection only with a

200

HERBERT F. MATARE

change in optics is evident. Such line connections can cross rivers by air and are easily laid underground when necessary. It is foreseen that interurban telecommunications will one day be independent of all high frequency based systems and do away with antennas, telephone wiring, and even concentric cables. In fact, each house can be attached to a system which will allow us to channel all information (TV, radio, telephone, picture phone) to each point and allow two-way communications with the same systems components which promise to be so low in price that such a huge enterprise will be feasible at reasonable cost.

REFERENCES FOR SECTION 5 5.1. C. H. Gooch, ‘‘ Injection Electroluminescent Devices,” pp. 147- 169. Wiley (Interscience),

New York, 1973. 5.2. See, e.g., “The Optoelectronics Data Book for Design Engineers.” pp. 263-310. Texas

5.3. 5.4. 5.5. 5.6. 5.7. 5.8.

5.9. 5.10.

5.11. 5.12. 5.13.

5.14. 5.15. 5.16.

5.17. 5.18.

Instruments, Dallas, Texas; Hewlett-Packard, “Solid State Display and Optoelectronics Designer’s Catalog ”; Fairchild, “Optoelectronics Handbook ”; Telefunken, ** Semiconductor Catalog”; Siemens, “Semiconductor Catalog”; and many others. M. G. Craford and W. 0. Groves, Proc. I E E E 61(7), 862-880 (1973). D. L. Keune, M. G . Craford, W. 0. Groves, and A. D. Johnson, IEEE Trans. Electron Devices 20(11), 1074-1077 (1973). H. E. Edmonds and W. E. Mutter, IEEE Trans. Electron Deuices2O( 1 l), 1068-1073 (1973). H. F. Matare, Int. Elektron. Rundsch. 26(H-8), 177-184 (1972). K . L. Kounerth and B. R. Shah, IEEE Spectrum 7(Sept.), 37-46 (1970). See, e.g., S. R. Korn, R. E. Locher, and W. H. Sahm, 111,‘’ Photon Couplers,”G. E. Appl. Note 200.62.General Electric Semiconductor Products Dep., Syracuse, New York; E. K. Howell, “The Light Activated SCR,” G.E. Appl. Note 200.3411/74. General Electric Semiconductor Products Dep., Syracuse, New York. J. Seaman, Electron. Prod. 19(4), 27-32 (1976). H. F. Matare, “Receiver Problems in the UHF-Region,” p. 201 IT.Oldenbourg Verlag, Munich, 1951). (In Ger.) H. C. Torrey and C. A. Whitmer, “Crystal Rectifiers.” McGraw-Hill, New York, 1948. See, e.g., M. Caulton, In “Topics in Solid State and Quantum Electronics” (W. D. Hershberger, ed.), pp. 42e468. Wiiey, New York, 1970. See, e.g., R. F. Adams and H. J. Schulte, Appl. Pkys. Lett. IS(8), 265-267 (1969). P. Guetin, J . Appl. Phys. 40(10), 4114-4122 (1969). H. F. Matare, D. B. Medved, and G. L. Sandberg, Proc. Eur. Electro-Opt. Markets Technol. Con$, lst, Geneua (1972); see also Opt. Laser Technol. 4(5), 201-256 (1972). IAV/Standard, F.I.R.E. Division of International Audio Visual, Inc. 15818 Arminta St., Van Nuys, California 91406. M. E. Ettenberg, J. P. Wittke, and H. Kressel, “High Speed Light Emitting Diodes,’’ RCA Lab. Rep., Final Rep., Vol. 1, AD-A018757. Off. Nav. Res. (Rep. period, 1 Apr. 1973-30 May 1975), Contract No. N00014-73-C-0335.Dep. Navy, Arlington, Virginia, 1975. H. Namizaki, M. Nagano, and S. Nakahara, IEEE Trans. Electron Devices 21(11), 688-691 (1974). J. Heinen, H. Westermeier, W. Harth, and K. H. Zschauer, IEEE Trans. Electron Devices, 23( lo), 11861 187 (1976).

LIGHT-EMITTING DEVICES. I1

20 1

5.19. 1. Ladany, J. P. Wittke. and H. Kressel, Injection Lasers for High Data Rate Optical Communications,” RCA Lab. Rep., Vol. 2, ADIA-018358. Off. Nav. Res. (Rep. period, 1 Jan.-3 Sept. 1975). Dep. Navy, Arlington, Virginia, 1975. 5.20. R. S. Speer, “Fiber Optic LED,” Spectronics, Nav. Electron. Lab. Cent. Final Rep., AD-A010356. Offic. Nav. Res., Contract No. 00123-74C-2024. Dep. Navy, Arlington, Virginia, 1975. 5.21. H. F. Taylor er a/.,“Fiber Optics Data Bus System: Current State-of-the-Art in Suitability of Fiber Optics for Multiterminal Data Communications,” Nav. Electron. Lab. Cent. Rep. NTlS AD/A-002222. Off. Nav. Res., Dep. Navy, Arlington, Virginia, 1974. 5.22. G. De Corlieu and T. A. Hawkes, Rev. T e c h . Thornson-CSF 6(4), 1205-1224 (1974). 5.23. H.F. Matare, Solid State Techno/. 15, 41-45 (1972). 5.24. I. P. Kaminow and J. R. Caruthers, A p p l . Phys. Lett. 22(7), 326-328 (1973). 5.25. W. E. Martin and D. B. Hail, Appl. Phys. Len. 21(7), 325-327 (1972). 5.26. S. Somekh, E. Garmire, A. Yariv, H. L. Garvin. and R. H. Hunsperger, Appl. Phys. Lett. 22(2), 46-47 (1973). 5.27. T. David, Y. Wei, W. Lee, and L. R. Bloom, Appl. Phys. Lett. 22(1), 5-7 (1973). 5.28. E. Carmine. H. Stoll, A. Yariv, and R. G . Hunsperger, Appl. Phys. Lett. 21(3), 87-88. (1972). 5.29. D. Chen, B. Koepke. J. D. Zook, and E. G . Bernal, Appl. Phys. Lett. 29(10), 657-659 (1976). 5.30. J. H. McFee, J. C. McGee, T. Y. Chang, and V. T. Nguyen, Appl. Phys. Lett. 21(11), 534-536 (1972). 5.31. T. lzawa and H. Nakagome, Appi. Phys. Letr. 21(12), 584-586 (1972). 5.32. H. W. Weber. R. Ulrich, E. A. Chandross, and W. J. Tomlinson, A p p l . Phys. Lett. 20(3), 143-145 (1972). 5.33. P. K. Tien, R. J. Martin, S. L. Blank, S. H. Wemple, and L. J. Varnerin, Appl. Phys. Lett. 21(5), 207-209 (1972). 5.34. D. B. Keck, R. D. Maurer, and P. C. Schultz, Appl. Phys. Lett. 22(7), 307-309 (1973). 5.35. T. C. Rich and D. A. Pinnow, Appl. Phys. Lett. 20(7), 264-266 (1972). 5.36. F. K. Reinhart, Appl. Phys. Lett. 22(8), 372-374 (1973). 5.37. See, e.g., D. Marcuse, ed., “Integrated Optics.” IEEE Press, New York, 1973. 5.38. See, e.g.. H. F. Matare, Receiver Problems in the Ultrahigh Frequency Range.” Oldenburg, Munich, 1951. (Detector Impedance Measurements, Figures 28 and 29.) ”

‘I

This Page Intentionally Left Blank

Radar Signal Processing G.V. TRUNK Radar Analysis Stafi Radar Division, Naval Research Loboratory Washington, D.C.

........... ..................... 203 11. Coherent Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 A. Sidelobe Cancelers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 B. Adaptive Arrays and Radars ..................................... 210 ...................... 219 C. Moving Target Indicator ...................... 222 D. Doppler Processing . . . . . E. Noncoherent MTI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 111. Noncoherent Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Classical Theory ................................. 225 B. Integrators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. False Alarms ... ..................... 230 D. Sequential Detectors ....................................... IV. Tracking System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 A. System Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 B. Tracking Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Maneuver-Following Logic . . ......................................... 240 D. Track Initiation ................................................. 242 E. Correlation Logic.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 F. Radar Integration . . . . . . . . . . . . . . . . . . . ................................ 248 V. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

I. INTRODUCTION During the last decade considerable progress has been made in radar signal processing. This progress is directly traceable to the lowered cost and increased speed of digital hardware and computers and to more sophisticated techniques in the areas of adaptive processing and tracking systems. In this survey of radar signal processing, the author has chosen to neglect the subject of waveform design and include the subject of the track-whilescan systems. This decision has been made because wave-form design has received considerable attention, with the books of Rihaczek (1969) and Cook and Bernfeld (1967) covering the subject in detail. On the other hand, while track-while-scan systems properly fall under the heading of radar data processing, it does not make sense to have an automatic detection system 203

204

G. V. TRUNK

unless it is accompanied by a tracking system. Therefore, since tracking is a necessary part of the entire system, it was included in the survey. Thus, this survey of radar signal processing will consider the three broad areas of coherent processing (i.e., processing of amplitude and phase), noncoherent processing (i.e., processing of amplitude), and track-while-scan systems. The subjects will be discussed in the same order as the radar signal passes through the radar system. Specifically, in the area of coherent processing the subjects of sidelobe cancelers, adaptive antennas, and moving target indicator (MTI) will be covered. In the area of noncoherent detection, methods of obtaining a constant false-alarm rate (CFAR) using either adaptive thresholding or nonparametric detectors will be emphasized. The section on the tracking system will cover the tracking filter, correlation logic, track initiation, maneuver following logic and a basic overview of an entire tracking system. 11. COHERENT PROCESSING

In the coherent processing, the area of adaptive processing will receive considerable attention. There are two approaches to adaptive processing: the method of maximum signal-to-noise ratio ( S I N ) due to Howells (1976) and Applebaum (1976) and the least mean-square (LMS) method due to Widrow and Hoff (1960). The two methods, while appearing quite different, yield almost equivalent results. For completeness, the LMS method will be used to discuss sidelobe cancelers and the method of maximum SIN will be used to discuss adaptive arrays and radars. For adaptive radars, special consideration will be given to the problem of convergent rate. Finally, MTIs will be discussed and the Moving Target Detector (MTD) system will be used as an example of doppler processing. A. Sidelobe Cancelers

The basic idea of a sidelobe canceler (a device that attempts to eliminate interference entering through the antenna side-lobes) is shown in Fig. 1. The signal S of interest enters through the main lobe of the antenna and the jamming (interfering signal), which is much stronger than the signal of interest, enters tnrough the sidelobe of the main antenna. The auxilary antenna is an omni and it will be assumed that the signal entering the omni is much smaller than the jamming J , and can be neglected since the signal and jamming now have the same antenna gain [the treatment of the signal in the auxilary channel can be found in Widrow et al. (197511. The adaptive filter produces an output Y which is as close as possible to the input jamming J . The filter output is then subtracted from the main input producing an

205

RADAR SIGNAL PROCESSING MAIN SIGNAL SWRCE

JAMMING SOURCE

-

i

OUTPUT

S + J

2

FILTER Y OUTPUT

4

/

AUXILIARY ANTENNA

ERROR E

FIG. 1. Adaptive noise canceling concept.

output Z = S + J - Y . Obviously, if the filter output is an exact replica of J , the output is the desired signal S. The filter is controlled by adjusting its parameters to minimize the output power. To show that this minimization will force Y to be a replica of J , a development in Widrow et al. (1975) is repeated. First, assume S , J, and J , are zero-mean random variables, S is uncorrelated with J and J , , and J , (and hence Y ) is correalated with J . The expected output power is

E{Z2} = E(SZ} + E((J - Y ) 2 )+ 2E{S(J

-

Y ) }= E{S2} + E { ( J - Y)2} (1)

Now, adjusting the filter to minimize E ( Z 2 } is equivalent to minimizing E { ( J - Y)’) since Y is uncorrelated with S, i.e., Y is the best least squares estimate of the jamming J. Furthermore, since Z - S = J - Y , minimizing E{(J - Y)’) causes 2 to be the best least-squares estimate of the signal S. The adaptive filter for obtaining a least-squares estimate of a desired signal S can be described by a weighting vector W, where WT = (W,, W2, . . . , Wn) and T denotes the transpose, operating on the input J , = X, XT = (xl, x 2 , ..., xn). Thus, the filter output is

Y = xTw (2) and the error, defined as the difference between the input signal and the filter output, is c=S

+ J - XTW

(3) The least-mean-square (LMS)adaptive filter adjusts the weighting vector W to minimize the mean-square error. The squared error is E2

= (S

+ J)2 - 2(S + J)XTW + WTXXTW

(4)

206

G . V. TRUNK

Taking the expected value of Eq. (4)and letting the vector P be the cross correlation between J and X, P = EIJX}, and the matrix K be the covariance matrix of X, K = E{XXT), one obtains

E{E’} = E{S2} + E { J 2 } - 2PTW + WTKW

(5)

To find the minimum of ( 5 ) with respect to W, the gradient V of ( 5 ) is set to zero, yielding the optimal weight vector W = K-’P (6) The LMS adaptive algorithm is an iterative method of finding an approximate solution to Eq. (6). The algorithm has the advantage of not requiring an explicit measurement of the correlation function or inversion of the covariance matrix. Specifically, the LMS algorithm uses the method of steepest descent to solve Eq. (6), i.e., the next weight vector Wj:, is equal to the old weight vector plus a step in the direction of the negative gradient, Wj+’ = wj - p vj The gradient of the squared error on the jth iteration is

Vj = V E=~V(S + J

- X;Wj)* = - 2cjxj

(7) (8)

Thus, the next weight is given recursively by

wj+1 = wj + 2pEjXj

(9) and is known as the Widrow-Hoff LMS algorithm. The parameter p is a factor which controls the rate of convergence and the stability of the method. It has been shown that (Riegler and Compton, 1973; Widrow et al., 1967)

INDIVIDUAL LEARNING CURVE

ENSEMBLE AVERAGE OF 48 LEARNING CURVES

NUMBER OF ITERATIONS

FIG.2. Typical learning curves for the LMS algorithm. (From Widrow et tesy of the Institute of Electrical and Electronics Engineers.)

a!.. 1975, cour-

RADAR SIGNAL PROCESSING

207

Eq. (9) converges to the optimal solution as long as p is between zero and the reciprocal of the largest eigenvalue of the covariance matrix K.Shown in Fig. 2 is a typical learning curve and an average of 48 learning curves for the LMS algorithm. The average reveals the basic exponential nature of the learning curve. Obviously, for the radar case Xjrepresents the sample from the j t h range cell and consequently the number of iterations corresponds to the number of range cells. In principal, if the situation shown in Fig. 1 is correct (no uncorrelated noise in each channel and no signal in the auxiliary), the jamming can be completely canceled. However, if the situation is as shown in Fig. 3, total

FIG.3. Adaptive noise canceler with correlated and uncorrelated noises in the main and auxiliary antennas.

cancellation cannot be accomplished. Specifically, the performance of the canceler can be described by the ratio R of the SIN at the output to the SIN at the primary input (main antenna). Widrow et al. (1975) have shown that this ratio R for steady state (Le., after convergence) can be expressed as

where A ( z ) and B(z ) are noise-to-noise ratios

So, S , , and S, being the power density spectra of the noises m o , m,, and n respectively, and H ( z ) being the channel transfer function for the correlated noise (jamming). It is obvious from Eq. (10) that the cancellation is limited

208

G. V. TRUNK

by the uncorrelated noise components in the primary and reference channels. When the jamming is much stronger than the uncorrelated noise components, A ( z ) and B ( z ) are small and RZ

1

A(2)

+

B(2)

giving a large improvement in the output signal-to-jamming ratio. However, it should be noted that the improvement indicated by expression (13) is rarely achieved in practice. Factors limiting performance include (1) the finite length of time for the adaptive process, (2) the presence of signal components in the auxiliary channel, (3) multipath problems, and (4) misadjustment caused by gradient estimation noise in the adaptive process (Widrow et al., 1967). Furthermore, in theory N omni antennas (and associated cancellation loops) are needed to cancel N jammers. However, because of multipath, the energy from a single jammer can enter the antenna from several directions and for all practical purposes appears as several jammers. Therefore, in practice one requires several times as many cancellation loops as jammers. Recently, Kretschmer and Lewis (1976) have developed an improved algorithm for simulation of the Applebaum-Howells adaptive loop and for use in adaptive processing. The LMS algorithm already discussed is given by

This is commonly used to simulate and analyze the Applebaum-Howells adaptive loop in the form Wj+ 1 = kWj + G(1 - k)EjXT

(141 where k = 1 - l/r, r is the filter smoothing constant, and G is the gain term. Thus, in both algorithms the next weight is derived in terms of the present error and sample. Kretschmer and Lewis point out that for fast loops, Wj+ as given by Eqs. (9) and (14) is not the proper weight. Rather, for better cancellation and more realistic canceler loop simulation, Wj+ should be calculated from

wj, 1 = wj + 2pcj+ Ixj.,1

(15) In effect, by using the sample Xj to calculate the weight Wj+ 1, a phase shift is introduced which can result in loop instability. Kretschmer and Lewis have shown (for the Applebaum-Howells application) that the stability condition of the LMS algorithm is JG(1- k)IXj12 - kl c 1

and that their improved algorithm is unconditionally stable.

(16)

RADAR SIGNAL PROCESSING

209

Comparison of the LMS algorithm with the improved algorithm was made using computer simulations. Correlated Gaussian noise (mean zero, variance equal to two) was used as an input to the main and auxiliary channels of the sidelobe canceler. At the 250th-range cell a constant signal at a SIN = - 20 dB is introduced. The signal residue for both algorithms with canceler parameters of k = 1 - 27c (0.000124) and G = 100 is shown in Figs. 4 and 5. While the LMS algorithm had unstable performance, the improved algorithm had completely stable performance. It should, also be noted that for slow loops, there will be ringing in the LMS algorithm which will result in degraded cancellation performance. In a previous paper, Kretschmer (1975) investigated cascading sidelobe canceler stages as a method of obtaining improved cancellation ratios and transient responses by achieving a higher effective loop gain with low actual loop gains which are required for stable operation. In lieu of their later work, the improved algorithm provides another way of obtaining high loop gains. During 1976, Lewis and Kretschmer have been working on a open loopdigital implementation of a sidelobe canceler. The sidelobe canceler removes the jamming signal after it has entered the main antenna. Adaptive arrays, which require individual receiving elements, attempt to prevent jamming from entering the antenna receive pattern by placing a receiving antenna null in the direction of the jammer. Before

FIG.4. Adaptive canccler response of the LMS algorithm.

G . V. TRUNK

2 10

B

P

a+,

o

I

100

1

200 300 SAMPLE NUMEEA

Lioo

500

FIG. 5 . Adaptive canceler response of the Kretschmer-Lewis algorithm.

commencing with a discussion of adaptive arrays and radars, it should be pointed out that the September, 1976 issue of the IEEE Transactions on Antennas and Propagation is a special issue on adaptive arrays and contains many interesting articles.

B. Adaptive Arrays and Radars Qualitatively, an adaptive array is one where the received signal is the weighted sum of the signal at the individual receiving elements with the weights being a function of the received signal. While the theory of adaptive arrays was first discussed by Applebaum (1964), and while Widrow et al. (1975) have made major contributions to the theory, a latter development of Brennan and Reed (1973) will be followed. Their approach is similar to Applebaum’s in that they maximize the SIN which they show is equivalent to maximizing the probability of detection when the noise is Gaussian distributed. Let the radar be composed of N receiving elements and let the last M time samples from each element be processed. Thus, there are a total of n = N M space-time samples. Define S to be a complex (amplitude and phase) n vector which contains the desired signal components and X be a complex n vector containing the noise samples. The radar return Z is given by

z=s+x

(17)

RADAR SIGNAL PROCESSING

211

To detect the signal S, the radar output is passed through a linear filter described by a weighting vector W. Thus, the output of the detector (i.e., the filter) is

Y = W'Z

(18) It can be shown (Brennan and Reed, 1973) that the SIN at the output of the filter is

where the asterisk indicates the complex conjugate and K is the noise covariance matrix; K = E { X * X T } , X having zero mean. Consequently, what is required is the value of W that maximizes Eq. (19). Using the Schwarz inequality, it can be shown that the maximum value of Eq. (19) is S'K- 'S* and that this value is obtained when

W = a'K- I s *

(20)

where a' is an arbitrary, nonzero complex number. This criterion has been known for some time (Van Trees, 1965); however, it is rarely used since K is not known a priori; and if K is estimated, it has been extremely difficult to invert K in real time. What makes the Brennan and Reed (1973) approach different from other adaptive array processing is not the ability to place spatial nulls in the direction of jammers but rather the temporal processing that is equivalent to a motion-compensated MTI. The compensated MTI behavior is obtained by selecting the proper steering signal S. The selection of the steering signal S will be illustrated for the case of an airborne coherent pdsed radar. Assume that the return is range gated, there are N , range cells, and the return from the jth cell is

Z(j) = + S(j) (21) The return signal from the rth receiving element and mth time sample can be written as &(m) = b,eim; ( r = 1, ..., N ) (22) where y is the Doppler phase shift ( y = -4nVT/I), V is the radial velocity of the target, T is the time between transmitted pulses, and I is the radar wavelength. The quantity b, is

+ id)

( r = 1, ..., N)

(231 where A, is the signal amplitude at the rth element, d is a constant phase factor, and 4, is the relative phase between the target and that rth element. b, = A, exp(i4,

212

G . V. TRUNK

For a linear array with element spacing d, the phase angles 4, for a signal arriving at an angle $ with respect to the array normal are

4, = (2nrd/2) sin II,

(241 Thus, the expected signal for a linear array can be obtained by substituting Eqs. (23) and (24) into (22). Both clutter and the target will have returns of the form of (22). Since the velocity of the target is unknown (and consequently the radial velocity V ) , it is impossible to specify S for the optimal weighting given by (20). However, since (22) is computable for ground clutter as a function of the radar-clutter cell geometry, one selects a “steering” signal S which is orthogonal to the ground clutter vector S’. Thus, the purpose of S is to reject the clutter, not to detect the target. This is about as much of an optimal detector as one can obtain, since it can be shown that no uniform most powerful test exists when the target velocity is unknown (Van Trees, 1961). As an example, let M = 2 and assume one wants to detect a target located in a direction normal to the direction of the platform velocity (i.e., radar is sidelooking). Then, S,(m) = A, exp(ib) and for uniform amplitude taper ( A , = 1, r = 1, . . . , N ) , the clutter signal is s T = eib[l ... 1 1 ... 13 (251 (I=

1, ..., N )

The appropriate steering signal S which is orthogonal to S‘, STS* = 0, is S T = [I ... 1 - 1 ... - 11 (26) which corresponds to a target at one-half the blind speed of the radar,

(271 Thus, if (26) is used in (20), the detector is optimized for canceling mainbeam clutter. We now consider how (20) can be implemented adaptively. Brennan and Reed use the method of steepest ascent to maximize the I/ = +(A/2T)

SIN

F 4

1 wTs12 WT*KW

The recursive algorithm for steepest ascent is

W(j + 1) = W(j) + 3 A j ) VF[W)I

(29)

where VF[W(j)] is the “complex ” gradient of F evaluated at W(j), which has been shown to be equal to

VF = 2(

wTs ) IS* - (CT:i’)KW]

WT*KW

R A D A R SIGNAL PROCESSING

213

If K is assumed known and p ( j ) is chosen to be a constant, one can apply known theorems (Powell, 1970) to show W(j) approaches a critical point as a limit. Thus, if W(0) is sufficiently close to the optimal value, W(j) approaches a'K- 'S* in the limit. The trouble with using (30) in (29) is that VF is a nonlinear function of W(j), which in some adaptive systems can cause computational difficulties. Thus, the algorithm was linearized by noting lim j-m

WTS - 1 * =a WT*KW - 2

Thus, if p(j) is equal to a constant p, Eq. (29) reduces to

+

+

W ( j 1 ) = W(j) pa[S* - a*K(j)W(j)] (32) where K(j) is a statistical estimate of the unknown covariance matrix K. The best (maximum likelihood) estimate of K is

W) = z*(j)zT(j)

(33) Brennan and Reed (1973) then showed that (32) converged. Specifically,the expected value of (32) converges to a'K- 'S*,where K = E{K(j)} for all j, if Z(j) are independent and 0 < p < 2a"/max l i ,where li(i = 1,. . . ,n) are the eigenvalues of K. The block diagram of the adaptive radar is shown in Fig. 6 and the implementation of an adaptive loop is shown in Fig. 7. The steady-state antenna pattern can be calculated from (20) and the SIN improvement can

FIG.6. Schematic of adaptive AMTl radar using control loops. (From Brennan er a/., 1976, courtesy of the Institute of Electrical and Electronics Engineers.)

2 14

G.

V. TRUNK

FIG.7. Implementation of adaptive loop. (From Brennan et al., 1976, courtesy of the Institute of Electrical and Electronics Engineers.)

be found from STK-'S*. However, in many radar environments the clutter has a temporal and spatial variation and consequently the rate of convergence is important. To study this phenomenon, computer simulations were used. The basic parameters for a 10-element adaptive array using only one time sample (i.e., N = 10 and M = 1) are given in Table I. In the first simulation, 30 discrete clutter points were uniformly distributed in the two intervals [ 17", W] and [ - 17", - 90'1 and the radar was looking normal to aircraft velocity vector. The simulation results are summarized in Fig. 8, where the base of the plot is 45 dB below the peak gain. The back antenna TABLE I PARAMETERS ASSUMEDI N SIMULATION OF ADAPTIVERECEIVING ARRAY Ten-element linear array Element patterns isotropic over -n/2 5 0 I4 2 Half-wave spaced elements Uniformly illuminated transmit array Thirty scatterers in sidelobe region, equally spaced in angle No interference for -0, < 0 < 0, Each receiving element weight-controlled adaptively Simulate 1600 independent sets of input signals (range resolution cells) No receiver noise

RADAR SIGNAL PROCESSING

NORNRLLZCO GIIlN Of W T R R RRRRY 10 ELM. 39 SCRIT. 6 : looO.0 S

= .OOl

215

TRI -1WmOO

FIG.8. Projectograph plot of adaptive array gain: improvement in signal-to-sidelobe clutter ratio is 27.3 dB for steady state and 25.7 dB after 1600 iterations. (FromBrennan and Reed, 1973, courtesy of the Institute of Electrical and Electronics Engineers.)

pattern is the initial receiving pattern, the middle eight patterns are from range cells 200 to 1600 in 200 range-cell intervals, and the last pattern is the steady state pattern. Since there are 30 interference sources and only 10 elements, it is impossible to put a null at each interference angle. Rather, the adaptive array follows two strategies: ( 1 ) it widens the main beam and, consequently, lowers the general sidelobe level, and (2) it places receiver nulls at transmitter maximums and vice versa. After 1600 interations, all but 1.6 dB (27.3 - 25.7) of the maximum signal-to-clutter improvement has been obtained. In the second simulation, the 30 clutter points were placed in the interval [15", 45"]. The simulation results are summarized in Fig. 9. While the sidelobes are reduced in the proper angular interval, after 1600 iterations only 24.7 dB of the possible 44.1-dB improvement in the signal-to-clutter ratio has been obtained. Brennan and Reed (1973) have shown that the time behavior of the weights is a sum of exponentials of the form N

Wi =

C C,exp[I= 1

(GI, + l)t/r]

(34)

2 16

G . V. TRUNK

FIG.9. Plot of adaptive array gain for nonuniform clutter distribution: improvement in signal-to-sidelobe clutter ratio is 44.1 dB for steady state and 24.7 dB after 1600 iterations. (From Brennan and Reed, 1973, courtesy of the Institute of Electrical and Electronics Engineers.)

where 7 is the time constant of the low-pass filter. Thus, the rate of convergence is controlled by the smallest eigenvalue of K; and specifically, the effective time constant is T/(GA,,,~,,+ 1).This suggests that rapid convergence can be obtained by selecting G to be large and/or 7 to be small. However, this is not a useful solution to the convergence problem since Brennan et al. (1971) have shown that the total output noise power in the adaptive array is

where W is the average weight vector in the absence of loop noise (departure from steady state). The quantity WTKW is the noise power when W = K-'S*. Consequently, the output power has been increased by the factor G 1 , / 2 ~due to loop noise. Thus, when K contains both small and large eigenvalues, it is impossible to select a G and 7 which yield both rapid convergence and low loop noise. To avoid the convergence problem, Reed et al. (1974) have suggested a direct computation of the weights.

RADAR SIGNAL PROCESSING

217

The maximum likelihood estimate of K,assuming the noise is gaussian distributed. is I L K = - Z*(j)Z'(j) Lj,1

c

Since Z * ( j ) Z T ( j )is an n x n matrix of rank 1, L must be 2 n in order for the inverse to exist. Then the filter has the form \jir = K- IS* (37) The output SIN for (37) normalized by the maximum SIN, STK-'S*,which corresponds to (20), is

The expected value of (38) is

+

E{p(K)} = (L 2 - n ) / ( L+ 1 ) (39) Thus, the average loss can be kept below 3 dB (E{p(K)} 2 1/2) by letting L 2 2n. However, while the adaptive loops of Fig. 6 require n complex multiplications, the sample matrix inverse method requires approximately n3 complex multiplications. To reduce the complexity of the method, one can update the covariance matrix using

Kj = (1

- a)Kj-

+ aZ*(j)Z'(j)

(40) where a is the weight applied to the current sample. Then, the inverse of Kj, given Kj- 1, is (Shapard et al., 1971)

This method of updating the inverse requires approximately 2n2 complex multiplications. It should be noted that the average computation time for updating the weights W depends on how frequently they must be updated. For example, depending on the radar environment, updating the weights every PRF using (36) may be quite adequate; and consequently, the computation time may be less than that of the adaptive loops. Brennan et al. (1976) compared the convergent rates of the three methods using a computer simulation illustrating airborne, MTI performance. The results of the simulation are shown in Fig. 10. In both forwardlooking and side-looking instances, the two methods of calculating 8-' provide an excellent convergence rate. While Fig. 10 indicates an MTI gain of + 100 dB, in practice the MTI gain would be limited to a lower figure by internal clutter motion.

1

1wo

(b)

FIG.10. (a) Adaptive performance versus number ofsamples: 8 elements, 2 pulses, element spacing = 0.5, interpulse motion = 0.2, scan angle = O W , steady state gain = 108.1 dB. (b) Adaptive performance versus number of samples: 8 elements, 2 pulses, element spacing = 0.5, interpulse motion = 0.2, scan angle = 90.0",steady state gain = 125.5 dB. (From Brennan et a/., 1976, courtesy of the Institute of Electrical and Electronics Engineers.)

RADAR SIGNAL PROCESSING

2 19

Most work on adaptive arrays and radars has been limited to theoretical studies. However, there has been some experimental work at Ohio State University (Compton, 1976), the Naval Research Laboratory (Gabriel, 1974), and the Wide Aperture HF Radio Research Facility operated by Stanford Research Institute (Griffiths, 1976; Washburn and Sweeney, 1976). C . Moving Target Indicators

Moving target indicators (MTIs) were first investigated in the 1940s and detailed discussions of them can be found in Skolnik (1962,1970) and Nathanson (1969). The coherent MTI, the most common MTI, utilizes a internal coherent reference source to distinguish a moving target from fixed clutter returns. The MTI signal is obtained by coherently subtracting the returned voltages from successive transmitted pulses, i.e.,

ZXj) = Zi(j) - Zi- l(j) (42) where Zi(j) is the ith returned pulse in the jth range cell. Larger clutter attenuations can be obtained by using multiple pulses. The frequency (Doppler) response of the MTI is that of a band-pass filter. The most serious problems associated with MTI are limiting and blind speeds. The first of these can be covered very simply. In the classic paper of Ward and Shrader (1968), it was shown that MTI improvement could be degraded by 20 dB in a three-pulse canceler by limiting the clutter return. Their work showed that the degradation was fundamental to limiting, and consequently a large dynamic range is required to avoid limiting. The major problem with MTI is that blind speeds corresponding to Doppler frequencies higher than Nyquist rate occur at

v,=

11 2T

--

( I = 1, 2, 3, ...).

(43)

Thus, for an L-band (1.3 GHz) radar with a PRF of 300 pps, the blind speeds occur at multiples of approximately 70 knots. Taking into account the width of the clutter notch (rejection region of canceler), many air targets would not be detected. There are several solutions to the problem of blind speeds in MTIs. Among these are variable PRF, staggered-PRF MTI, and dual-frequency MTI. The simplest solution is to use a variable PRF system. For instance, if an interpulse period of T is used, a blind speed of V, is obtained. Then, by changing the interpulse period by a small fraction r, the blind speed changes by the same fraction r; and the smallest common blind speed is VB/(l - r). Thus, if an L-band radar has two PRFs, 300 pps and 270 pps, the blind. speed of the radar system is approximately 700 knots. There are two disad-

220

G . V. TRUNK

vantages of such a system: (a) second-time-around clutter (clutter beyond the unambiguous range, caused by ducting at sea, or high-altitude longrange clutter, such as mountains or chaff) passes through the MTI, and (b) the constant PRF for a two- or three-pulse burst makes the system more vulnerable to jamming. The simple solution to (a), using an extra filler pulse (i.e., transmitting three pulses but only using the last pulse out of a two-pulse MTI), makes situation (b) worse.

e0

ei

t

FIG.11. A staggered-PRF MTI filter. (From Hsiao and Kretschmer, 1973, courtesy of the Institution of Electronic and Radio Engineers.)

An elegant solution to the blind-speed problem is the staggered-PRF MTI. The basic MTI configuration is shown in Fig. 11. The interpulse durations zi are constrained by the relation

(44) where F B is the first blind Doppler frequency and Ii are integers for all i. Capon (1964) showed that the optimal weights {ai}for minimizing the output clutter residue while retaining some fraction of the average gain of the filter (this constraint avoids the trivial solution ai = 0 for all i) are the components of the eigenvector associated with the smallest eigenvalue of the clutter covariance matrix. This procedure ignores what happens in the filter passband. Hsiao and Kretschmer (1973) developed a procedure for setting the interpulse periods to minimize the RMS passband ripple while maintaining minimum clutter residue. A typical response is shown in Fig. 12. The basic trouble with this system is that second-time-around clutter will not be canceled. A third solution to the blind-speed problem is the dual-frequency MTI first discussed by Kroszczynski (1967, 1970) and later by Hsiao (1975). The system works by transmitting two frequencies whose ratio r is slightly less than 1, filtering out the sum signal and retaining the difference signal. The F B t i = Zi

22 1

RADAR SIGNAL PROCESSING

- 70

0

1

2

3 Y 5 6 7 DOPPLER FREQUENCY/PRF

8

9

10

FIG. 12. Seven-pulse staggered MTI filter frequency response curve. (From Hsiao and Kretschmer, 1973, courtesy of the Institution of Electronic and Radio Engineers.)

system performance is basically that of a low-frequency radar and hence the blind-speed problem is reduced. The detrimental factor is that the clutter improvement factor is reduced by several decibels. A typical filter response for a dual-frequency MTI is shown in Fig. 13. While the passband response is quite variable, no attempt has been made to reduce the variation by changing r. Hsiao indicates that the staggered-PRF MTI is preferable to the dual-frequency MTI. However, this author believes that the dualfrequency MTI should not be discarded that readily. Of course, an alternative solution, and possibly a better one, is to operate individual MTIs at the two frequencies.

m W LT I1I

-40

W

+ 1-50

-

LL

- 60

FIG.13. Hsiao, 1975,

0.89. (From

222

G. V. TRUNK

D. Doppler Processing

A MTI canceler provides near optimal target detection in clutter, but provides little or no improvement against receiver noise. McAulay (1972) formulated the problem as a classical detection problem and showed that the optimal detector could be structured (approximately) as an MTI canceler followed by a narrow-band Doppler filter bank. This structure has the practical advantage of greatly reducing the dynamic range required at the input of the filter bank. In this configuration, the MTI canceler provides improvement against clutter and the Doppler filter bank provides improvement against noise. The moving target detector (MTD), developed by Lincoln Laboratory for the FAA (O'Donnell et al., 1974; Muehe et al., 1974), uses this type of processing. During 1976 the MTD was tested with a modified FPS-18 radar at the FAA facility in Atlantic City, N. J. The modified FPS-18 radar is an S-band radar instrumented to 48 nmi. The range cell is approximately 1/16 nmi, beamwidth is 1.5", the scan rate is 15 rpm, and 20 pulses are returned as the radar sweeps past the target. A block diagram of the MTD signal processor is shown in Fig. 14. An azimuth cell is defined as a half beamwidth (3/4")and contains 10 pulses referred to as a coherent processing interval (CPI). In a CPI, the 10 pulses are passed through a three-pulse MTI canceler and the eight output pulses (two pulses are needed to load the MTI) serve as an input to an eight-point FFT, the points being weighted in order to provide low-filter sidelobes. The

10 BITS 2.6 MHr RATE

-7 1

MEMORY

36 BITS

VELOCITY FILTER

MAGNITUDE

CLUTTER RECURSIVE

FIG. 14. Block diagram of the MTD signal processor.

WEATHER LEVEL MEASUREMENT

THRESHOLDING

RADAR SIGNAL PROCESSING

223

radar PRF is changed from lo00 to 1150 pps on alternate CPIs to avoid the blind-speed problem. The 2.9 x lo6 range-azimuth-Doppler cells (760 x 480 x 8) are individually thresholded. In this process, a clutter map is generated by weighting the radar return in the zero Doppler filter over the last eight scans (32 sec) using a digital filter. Thus, tangential targets having zero Doppler can be detected if the target level exceeds the clutter map level by a specified constant. That is, tangential targets can be detected in spotty ground clutter by using the principal of inter-clutter visibility (Barton and Shrader, 1969). The thresholds for filters No. 2 through 6 are set using a mean-level threshold. Specifically, the threshold for a given filter number is based* on the average return in the filter number from the range cells f 1/2 mile (eight cells) on either side of the test cell. Since there exists clutter spilling over into filters (No. 1 and No. 7), two thresholds are generated for these filters. One threshold is based on the map, a second threshold is based on the mean level over a range interval, and the higher of the two thresholds is used. The MTD represents a great improvement in signal processing for FAA air surveillance radars. A good design of matching processor to radar has been accomplished and component technology has made the processing practical to implement. Presently a second generation MTD is being designed. This MTD uses no MTI but rather each filter is optimized to obtain the maximum signal-to-clutter-plus-noise ratio for an assumed clutter spectrum.

E. Noncoherent M T l Noncoherent MTIs are described by Skolnik (1962, 1970). They differ from coherent MTI by not using an internal coherent reference source but rather mixing the received signal with itself. Thus, when both clutter and a target are present, the beat between them yields a return at the target Doppler. On the other hand, when only a target is present, the signal return is at zero Doppler and cannot be detected. Consequently, for noncoherent MTI to be useful, gating circuitry is required for passing the noncoherent MTI output when clutter is present and passing the regular video when clutter is not present. Generally, fringe areas cause major problems for the gating circuitry, making performance unacceptable. A different kind of noncoherent MTI has been made possible by highpower microwave sources (Granatstein et al., 1975). Lewis and Cantrell (1975) propose transmitting a short pulse and subtracting successive noncoherent pulses. This is similar to an area MTI discussed by Skolnik (1962), Details about various thresholding techniques can be found in Section I11

224

G.

V. TRUNK

except that the short pulse enables the subtraction to be made on a pulse-topulse rather than a scan-to-scan basis. Thus, with a 1-nsec pulse and a PRF of 200 pps, all moving targets above 60 knots can be detected, i.e., there are no blind speeds. 111. NONCOHERENT DETECTION

The earliest noncoherent signal processing was performed by radar operators using visual inputs from PPIs and A scopes. While operators can perform this detection task very accurately, operators are easily saturated and become quickly fatigued. To remedy this situation and to provide quick reaction times, automatic detection and tracking (ADT) systems have become quite popular during the 1970s. The statistical framework necessary for the development of ADT was introduced to the radar community in the 1940s by Marcum (1960) and later Swerling (1960) extended the work to fluctuating targets. They investigated many of the statistical problems associated with the noncoherent detection of targets in Rayleigh' noise. Their most important result was the generation of curves of probability of detection (P,) versus signal-to-noise ratio (SIN) for a detector which sums N enveloped detected samples (either linear or square law) assuming equal signal amplitudes. However, in a search radar, as the beam sweeps over the target, the returned signal amplitude is modulated by the antenna pattern. Many authors investigated various detectors (weightings), comparing detection performance and angular estimation results to the optimal values. The detectors investigated included the moving window, feedback integrator, two-pole filter, binary integrator, and batch processor. The original work on these detectors was based on the assumption that the environment was known and homogeneous so that fixed thresholds could be used. However, a realistic environment, containing land, sea, and rain for example, will cause an exorbitant number of false alarms for a fixed threshold system. Two approaches, adaptive thresholding and nonparametric detectors, have been used to solve the false alarm problem. Both solutions are based on the assumption that homogeneity exists in a small region about the range cell that is being tested. The adaptive thresholding method assumes that the noise density is known except for a few unknown parameters. The surrounding reference cells are then used to estimate the unknown parameters and a threshold based on the estimated density is obtained. Nonparametric detectors obtain a constant false alarm rate (CFAR) by ranking the test sample with the reference cells. Under the hypothesis that all the samples (test and reference) are independent samples from an unknown density function, the test sample has a uniform density function, and, consequently, a threshold yielding CFAR can be set.

RADAR SIGNAL PROCESSING

225

A. Classical Theory

The radar detection problem is a binary hypothesis testing problem, i.e., H o : No target present. H , : Target present. While many criteria can be used to solve this problem, the most appropriate for radar is the Neyman and Pearson (1928) criterion. This criterion maximizes PD for a given probability of false alarm ( P J by comparing the likelihood ratio (L) to an appropriate threshold T. A target is declared present if

where p ( x , , . . . , x, 1 H,) and p ( x , , . . . , x, I H , ) are the joint densities of the n samples under the conditions of target presence and target absence, respectively. For a linear envelope detector and white Gaussian noise, the samples have a Rayleigh density under H , , a Ricean density under H,, and the likelihood detector reduces to

where lo is the Bessel function of zero order. For equal-amplitude (A, = A ) small signal pulses (A, 4 a), the detector reduces to the square-law detector i x 2I -> i=l

T

(47)

This detector and the linear detector were first studied by Marcum (1960) and were studied in succeeding years by numerous people. The most important facts concerning these detectors are 1. The detection performances of the linear and square-law detectors are similar and are very close to the performance of the optimal detector (Marcum, 1960). 2. Since the signal return of a scanning radar is modulated by the antenna pattern, only 0.84 of the pulses between the 1/2-powerpoints should be integrated and the antenna beam shape factor (ABSF) is 1.6 dB (Blake, 1953). The ABSF is the number by which the mid-beam SIN must be reduced so that the detection curves generated for equal signal amplitudes can be used for the scanning radar. 3. The collapsing loss for the linear integrator can be much greater than

G. V. TRUNK

226

the loss for a square-law integrator (Trunk, 1972). The collapsing loss is the additional signal required to maintain the same PDand Pfawhen unwanted noise samples along with the desired signal-plus-noise samples are integrated. Besides detecting targets, most signal processors are required to make angular estimates of the azimuth position of the target. Swerling (1956) calculated the standard deviation of the optimal estimate by using the Cramer-Rao lower bound. The results are shown in Fig. 15, where a normalized standard deviation is plotted against the SIN per pulse. This result holds for a moderate or large number of pulses' integrated and the optimal estimate involves finding the location where the correlation of the returned signal and the derivative of the antenna pattern is zero. While this estimate is rarely implemented, its performance is approached by simple estimates, e.g., the maximum value and threshold-crossing procedures, as can be seen in Fig. 15. 10 0

70 5 0

3 0 20

kJP

10

07 0 5

0 3

-- -

MAXIMUM OF 2-POLE FILTER

THRESHOLD CROSSING OF 9.POLE FILTER

02

01

0

2

4 6 S/N (dB)

B

10

FIG. 15. Comparison of angular estimates with the Cramer-Rao lower bound. n is the standard deviation of the estimation error and N is the number of pulses with the 3-dB beamwidth, which is 28.

RADAR SIGNAL PROCESSING

227

B. Integrators

Almost all signal processors use linear rather than square-law detectors, since a linear detector is easily built by using a matched filter and a halfwave rectifier followed by a low-pass filter. However, many different integrators are used to accumulate the linear envelope-detected pulses. A few of the

----?-?INTEGRATORS

WINDOW MOVING

...

FEEDBACK INTEGRATOR

BINARY INTEGRATOR

T,

FIG. 16. Block diagram of the common integrators.

most common integrators are shown in Fig. 16. Some advantages and disadvantages of these integrators are as follows (Palmer and Cooper, 1964; Dillard, 1967; Cantrell and Trunk, 1973):

1. Moving Window The moving window performs a running sum of N pulses; as the latest pulse is added to the sum, the pulse N PRFs in the past is subtracted from the sum. The detection performance of this detector is only 0.5 dB worse

228

G . V. TRUNK

than the optimal detector which weights the returned signal by the fourth power of the voltage antenna pattern. The angular estimate is obtained by either taking the maximum value of the running sum or taking the midpoint between the first and last crossing of the detection threshold. Both methods have a bias of N / 2 pulses which is easily corrected. The standard deviation of the estimation error of both estimators is about 20% higher than the Cramer-Rao lower bound. The major disadvantage of this detector is that the last N pulses for each range cell must be saved. For radars with large beamwidths and thus many pulses, the moving window yields large hardware requirements. However, with the lower cost and size of memory, this disadvantage is rapidly disappearing. 2. Feedback Integrator

The amount of storage required can be reduced significantly by using a feedback integrator which requires the storage of only one number. While the feedback integrator applies an exponential weighting into the past, its detection performance is only 1 dB less than the optimal integrator. Unfortunately, difficulties are encountered when using the feedback integrator to estimate the azimuth position. While the threshold crossing procedure yields estimates only 20% greater than the lower bound, the bias is a function of the SIN and must be estimated. On the other hand, the maximum value, while it has a constant bias, has estimates which are 100% greater than the lower bound. It is this author’s opinion that this detector has limited utility. 3. Two-Pole Filter The two-pole filter requires the storage of an intermediate calculation in addition to the integrated output. However, with this rather simple device, a weighting pattern similar to the antenna pattern can be obtained, and consequently, good perfmnance would be expected. The detection performance is within 0.15 dB of the optimal detector, and its angular estimates are about 20% greater than the Cramer-Rao lower bound. If the desired number of pulses integrated is changed (because of change in rotation of the radar or use of another radar), it is only necessary to change the feedback values K , and K 2 .Their optimal values are set by

K,

=2

exp[ - {ad 5 / ( 1 -
K z = eXp[ - 2<0d 5 / ( 1 - 52)1’2]

(481 (49)

where 5 = 0.63, N o d t = 2.2, and N is the number of pulses between the 3-dB points of the antenna.

229

RADAR SIGNAL PROCESSING

4. Binary Integrator

The binary integrator is exactly equivalent to the dual threshold, M out of N, and rank detectors. The input samples are quantized to a zero or one, depending on whether or not they are greater than a threshold TI.The last N zeros and ones are summed and compared to a second (detection) threshold = M. The detection performance of this detector is 2 dB less than the moving window because of the hard limiting of the data, and the angular estimation error is 25% greater than the Cramer-Rao lower bound. There are several reasons why this detector is used: a. The binary integrator is easily implemented. b. The binary integrator ignores interference spikes which cause trouble with integrators that directly use signal amplitude. c. This detector works extremely well when the noise has a non-Rayleigh density (Schleher, 1975; Trunk, 1976).

6°'90W

DETECTOR

t-

10

12

14

16

18

20

22

24

SIN (dB)

FIG.17. Comparison of various detectors in log-normal interference.

A comparison of the binary integrator (3 out of 3), the median detector (2 out of 3), and the mean detector (moving window) in log-normal interference is shown in Fig. 17. It is obvious that the optimal binary integrator is much better than straightforward integration. The optimal values for the second threshold were found by Schwartz (1956) for Rayleigh interference and by Schleher (1975) for log-normal interference.

230

G.

V. TRUNK

5. Batch Processor

The batch processor is used when there is a large number of pulses in the 3-dB beamwidth. If K . N pulses are in the 3-dB beamwidth, K pulses are summed and either zero or one is declared, depending on whether or not the sum is greater than a threshold TI. The last N zeros and ones are summed and compared to a second threshold M. The batch processor has all the advantages of the binary integrator plus these additional advantages: a. Batch processor requires less storage than the binary integrator. b. Detection performance is better than the binary integrator (i.e., less than 2 dB from moving window). c. Angular estimate is more accurate than one obtained by binary integrator. The batch processor has been implemented by the Applied Physics Laboratory (1975) of Johns Hopkins University with great success. To obtain a more accurate azimuth estimate, they use

6 = C A i O i / CA i

(50)

where Ai are the amplitudes of the sums greater than TI and Oi are the corresponding antenna azimuth angles. When there are many pulses on target (N > 20), this detector is generally favored by this author. C . False Alarms

If fixed thresholds are used with the previously discussed integrators, the detectors will saturate the tracking computer associated with the system and disrupt the system. Three important facts should be remembered: a. It makes little sense to have an automatic detection system without an associated tracking system. b. The sensitivity of the detector should be as high as possible without saturating the tracking computer. c. False alarms and false targets are not a problem if they are removed by the tracking computer. It should be noted that tracking (scan-to-scan processing) is the only way to remove stationary point clutter or target MTI residues. Of course, one can reduce the number of false alarms with a fixed threshold system by setting a very high threshold. Unfortunately, this would reduce sensitivity in regions of low noise (clutter) return. What is required is a detector which will detect a target when it has a higher return than its immediate background. Two such types of detectors are adaptive thresholding and nonparametric detectors. Both of these detectors assume that the

23 1

RADAR SIGNAL PROCESSING

samples in the range cells surrounding the test cell (called reference or neighboring cells) are independent and identically distributed; and furthermore, it is usually assumed that the time samples are independent. Both detectors test whether the test cell has a return sufficiently larger than the reference cells. A survey of CFAR procedures can be found in Hansen (1973). 1. Adaptive Thresholding The basic assumption of the adaptive thresholding technique is that the noise density is known except for a few unknown parameters. The surrounding reference cells are used to estimate the unknown parameters and a threshold based on the estimated density is then obtained. The simplest adaptive detector is the cell-averaging CFAR investigated by Finn and Johnson (1968). If the noise has a Rayleigh density, only the parameter o needs to be estimated: the mean of a Rayleigh being o m and the variance being 0 2 ( 2 - 71/2). Thus, by estimating the mean, one obtains an estimate 6 which can be used to set a threshold T to yield the desired P , , . However, since T is set by an estimate 6, it must be slightly larger than the threshold one would use if G were known a priori. The raised threshold causes a loss in target sensitivity and is referred to as a CFAR loss. This loss has been calculated by Mitchell and Walker (1971) and some results are summarized in Table 11. As can be seen, for a small number of reference cells, the loss is large because of the poor estimate of 0. TABLE I1 “CFAR

LOSS” IN

Number of pulses integrated 1 3 10 30 100

dB

FOR

P,, =

AND

f,

= 0.9

Number of reference cells

1

2

3

5

10

a3

-

7.8 3.3 2.0 1.4

15.3 5.1 2.2 1.4 1.0

7.7 3.1 1.3 1.0 0.6

3.5 1.4 0.7 0.5 0.3

0 0 0 0 0

~

6.3 3.6 2.4

~

This thresholding technique is more effective in maintaining CFAR when it is applied to the binary integrator or batch processor as shown in Fig. 18. This is because when the number of pulses integrated by the binary integrator is moderate, the P,, on a single pulse is rather large, e.g., P,, = 0.1 for a single pulse yields P,, = lov5for a 7 out of 10 integrator. Thus, since most

232

G . V. TRUNK

-

INTEGRATE PULSES, THEN DUMP

K

FIG. 18. Cell-averaging CFAR implemented with the batch processor.

non-Rayleigh densities are Rayleigh-like to the 10th percentile, this type of processor will maintain a low PI, in most non-Rayleigh environments. This demonstrates a general rule: to maintain a low Pr, in various environments, adaptive thresholding should be placed in front of the integrator. For any noise distribution, CFAR can be maintained by counting the number of ones out of the comparator per scan and using this number to control K,i.e., if the number is too large, K is increased. Front-end thresholding, which maintains amplitude information by dividing the average reference value into the test cell, was investigated by Hansen and Ward (1972) and is shown in Fig. 19. This type of processing is especially effective when there is strong interference which is variable on a pulse-to-pulse basis. I

-

SNARE-LAW DETECTOR

qTt FIG. 19. Block diagram of front-end cell-averaging CFAR receiver.

RADAR SIGNAL PROCESSING

233

When the noise has a non-Rayleigh density, such as the chi-square density or log-normal density, it is necessary to estimate two parameters and the adaptive detector is more complicated. If several pulses are integrated with any of the amplitude integrators, the integrated output will be approximately Gaussian distributed. Then, the two parameters which must be estimated are the mean and variance. These estimates are given by

and 1

X=--cX,

N l

where the summation is over the N range cells surrounding the test cell. When successive pulses in the same range cell are correlated (as with returns from rain or sea clutter) many false alarms will occur if only the mean value (52) is estimated. A threshold of the form

T=X+KC? (53 1 will provide a low P, for the amplitude integrators, moving window, feedback integrator, and two-pole filter. There is nothing that can be done to the binary integrator to yield a low P, in correlated noise, and thus, it should not be used in this situation. On the other hand, if the correlation time is less than a batching interval, the batch processor will yield a low Prawithout modifications.

2. Nonparametric Detectors The most common way nonparametric detectors obtain CFAR is by ranking the test sample with the reference cells. Under the hypothesis that all the samples are independent samples from an unknown density function, the test sample has a uniform density function. For instance, referring to the rank detector in Fig. 20, the test cell is compared to 15 of its neighbors. Since in the set of 16 samples, the test sample has equal probability of being the smallest sample (rank equal zero or, equivalently, any other rank), the probability that the test sample takes on values 0, 1, ..., 15 is 1/16. A simple rank detector (Hansen and Olsen, 1971) can be constructed by comparing the rank (i.e., number of reference cells that test cell exceeds) to a threshold K ; and the output is a one if the rank is larger, a zero otherwise. The zeros and ones are summed in a moving window. This detector incurs a CFAR loss of about 2 dB, and is extremely effective if the time samples are independent. It should be noted that only certain values of Pfacan be ob-

234

G. V. TRUNK I

RANK M T E c r n u

TWO-POLE INTEGRATOR

FIG.20. Block diagram or a modified generalixd sign test processor. (From Trunk er ctl.. 1974. courtesy of the Institute or Electrical and Electronics Engineers.)

tained. Thus, if the number of pulses integrated is small, low values of Pfd cannot be obtained. If the time samples are dependent, the rank detector will not yield CFAR. A modified rank detector, called the modified generalized sign test (MGST) (Trunk et al., 1974), is an attempt to maintain a low P,, and is shown in Fig. 20. This detector can be divided into three parts: a ranker, an integrator (in this case, a two-pole filter),and a thresholding device. A target is declared when the integrated output exceeds two thresholds. The first threshold is fixed (equals p TI /K from Fig. 20) and yields CFAR when the reference cells are independent and identically distributed. The second threshold is adaptive and maintains a low PI, when the reference samples are correlated. The device uses the mean deviate estimate, where extraneous targets in the reference cells have been excluded from the estimate by use of a preliminary threshold T2 to estimate the standard deviation of the correlated samples. The rank and MGST detectors are basically two-sample detectors. They decide a target is present if the ranks of the test cell are significantly greater than the ranks of the reference cells. Target suppression occurs at all interfaces (e.g., land, sea), where the homogeneity assumption is violated. However, there are some tests [Hansen (1970) investigated the Spearman Rho and Kendall Tau tests] which only depend on the test cell. These tests

+

RADAR SIGNAL PROCESSING

235

work on the fact that as the antenna beam sweeps over a point target, the signal return increases and then decreases. Thus, for the test cell, the ranks should follow a pattern first increasing and then decreasing. While these detectors do not require reference cells and hence have the useful property of not requiring homogeneity, these detectors are not generally used because of the large CFAR loss taken for moderate sample sizes. For instance, for N = 16 the loss is 10 dB, and for N = 32 the loss is 6 dB. The paper by Hansen (1970) is worth noting because it introduced the concept of importance sampling for calculation of false-alarm thresholds. The fundamental principle of the importance sampling technique is to modify the probabilities that govern the outcome of the basic experiment of the simulation in such a way that the event of interest (i.e., the false alarm) occurs more frequently. This distortion is then compensated for by weighting each event by the ratio of the probability that this specific event would have occurred if the true probabilities had been used in the simulation to the probability that this event would occur with the distorted probabilities. Consequently, by proper choice of the distorted probabilities, the number of repetitions can be greatly reduced. Further details on importance sampling can be found in Trunk et al. (1974), Hansen (1974), and Hillier and Lieberman (1967). In summary, when only a small number of pulses is available (less than 8), amplitude information must be used, and this author favors the moving window integrator. When a moderate number (between 8 and 20) is available, a rank detector should be used if samples are independent; and a two-pole filter with thresholding of the form T = X + K6 should be used if the samples are dependent. If a large number of pulses (greater than 20) is available, the batch processor or MGST processor should be used. It should be noted that the above rules should only serve as a general guideline. It is highly recommended that a sample of the radar environment be collected and analyzed, and that various detectors be simulated on a computer and tested against recorded data.

D. Sequential Detectors

Sequential detectors, which can be used with phase array radars, are based on the idea that in many cases, depending on the returned samples, a decision can be made on a few samples. The sequential likelihood ratio test (SLRT) works as follows: Given independent samples xl,...,x, ,calculate the likelihood ratio

236

G. V. TRUNK

If L,,, 2 A , accept H, (target present); if L,,, I B, accept H o (no target present); and if B < L,,, < A , take another sample. The SLRT has the useful properties that the thresholds are set by the simple formulas A = P D / P f ,and B = (1 - PD)/(l - Pf,), and that for all tests with a given PD and Pfa,the SLRT requires the smallest average sample size. Further details about the SLRT can be found in Lindgren (1962). An early application of sequential detection to radar was discussed in a paper by Marcus and Swerling (1962). Unfortunately, in radar the application of SLRT is not straightforward since one is required to make a decision in every range cell before the test can be ended and the agile beam moved. The modified problem considered was:

H,: Noise present in all range cells. H,: Exactly one signal present in the ith range cell (i unknown). They performed some numerical calculations and came to the following conclusions: (1)The greatest savings in average sample size comes when there is no signal present (i.e.,H o true). (2) In comparison with the fixed sample size test, SLRT provides a greater savings when the number of range cells is small and when SIN is small. (3) It is not necessary to truncate the test.

IV. TRACKING SYSTEM In this section, track-while-scan systems (i.e., tracking systems for surveillance radars whose nominal scan time is from 4 to 12 sec) are considered. If the probability of detection (PD)per scan is high, if accurate measurements are made, if the target density is low, and if there are few false detections (a detection is a crossing of a threshold; no judgement is made on whether or not it belongs to a valid target), the design of the correlation logic and tracking filter is straightforward. However, in a realistic radar environment, these assumptions are never valid, and the design problem is quite complicated. White and Silberman (1975) list many problems encountered in actual situations. Among these problems are target fades (due to multipath, clutter masking, interference, blind speeds, and atmospheric conditions), false alarms (due to noise, clutter, interference, and jamming), and poor radar parameter estimates [due to noise, unstabilized radar platforms, unresolved targets, target splits (ie., two detections for a single target), multipath, and propagation effects]. A general outline of a track-while-scan system is considered first. Then, detailed discussions of the tracking filter, maneuver-following logic, track initiation, and correlation logic are presented. Finally, methods of integrating data from several radars are discussed.

RADAR SIGNAL PROCESSING

237

A. System Outlirie

Almost all track-while-scan systems operate on a sector basis. A typical series of operations is shown in Fig. 21. For instance if the radar has reported all the detections in sector No. 11 and is now in sector No. 12, the tracking program would start by correlating (defined as the act of trying to associate) the clutter points (i.e., stationary tracks) in sector No. 10 with detections in sectors Nos. 9, 10, and 11. Those detections that are associated with clutter points are deleted (are not used for further correlations) from the detection file and are used to update the clutter points. Updating clutter points usually implies replacing the old point by the associated detection.

::9

INII1AL12LTIO"

,<"IIT,"L

l"U"9

I ,#MI T I 1 C K I

1 0 CLUTTEl Fm",S

'?

FIG.

"ma" F.XJ1Inu

21. Various operations of a track-while-scansystem performed on a sector basis.

Next, firm tracks in sector No. 8 are correlated with detections in sectors Nos. 7, 8, and 9. By this time all clutter points have been removed from sectors No. 9 and below. Those detections which are associated with firm tracks are deleted from the detection file and are used to update the appropriate track. The filter for performing this updating is given in the next section. Usually, some provision is made for giving preference to firm tracks (with respect to tentative tracks) in the correlation process. In Fig. 21, by performing the correlation process two sectors behind firm track correlations, it is impossible for tentative tracks to steal detections belonging to firm tracks. In other tracking systems, the correlation for firm and tentative tracks is performed in the same sector; however, the generalized distance D between tracks and detections is incremented by AD if the track is tentative. Finally, detections which are not associated with either clutter points or tracks are used for initiation purposes. The most common initiation procedure is to initiate a tentative track; later the track is dropped or else made a firm track or clutter point. Cantrell et al. (1975) suggest that both a clutter point and tentative track be established. If the detection came from a stationary target, the clutter point will be updated and the tentative track will eventually be dropped. On the other hand, if the detection came from a moving target, the tentative track will be made firm and the clutter point will be dropped. The latter method requires less computer computation time when most of the detections are clutter residues. The correlation procedure is made in a sector framework to avoid the

G . V. TRUNK

238

necessity of correlating all tracks with all detections. The procedure can be implemented very easily by defining two computer arrays: a sector file and a track file. The sector file (for sector I) contains the first track number in sector I and the track file for track J contains the next track number in the same sector as track J or a zero indicating that the track is the last track in the sector. For further information about the details of a tracking system, see Cantrell et al. (1974), Wilson and Cantrell(1976), and Trunk et al. (1977).

B. Tracking Filters Before proceeding further, a discussion of the coordinate system in which the tracking will be performed will be given. The quantities measured by the radar are spherical in nature: range, azimuth, elevation, and, possibly, range rate. Thus, it may seem natural to perform tracking in spherical coordinates. However, this causes difficulties since motion of constant-velocity targets (i.e., straight lines) will cause acceleration terms in all coordinates. A simple solution to this problem is to track in a Cartesian coordinate system. While it may appear that the appropriate transformations (x = R cos 0, cos 8,, etc., where R is the range, 8, is the elevation angle, and Oa is the azimuth angle) will destroy the accurate range track, Cantrell (1973) has shown that the inherent accuracy is maintained. Quigley and Holmes (1975) note jhat maneuvering targets cause a large range error but a rather insignificant azimuth error and thus suggest using a target-oriented Cartesian coordinate system. Specifically, the x axis is taken along the azimuth direction of the target and the y axis is in the cross-range direction. Sklansky (1957) performed one of the first analyses of the tracking filter for a track-while-scan system. He considered the a-fl filter described by

+ 4 x m ( k ) - x#)l - 1) + B[xm(k) - x,(k)l/T

xdk) = Xp(k) 4 ( k )= x,(k

+ 1) = x,(k) + u,(k)T

(55)

(56)

(57) where x,(k) is the smoothed position, u,(k) is the smoothed velocity, xp(k)is the predicted position, x,(k) is the measured position, T is the scanning period (i.e., time between detections), and ct and fl are the system gains. The optimal filter for performing the tracking when the equation of motion is known is the Kalman filter, first discussed by Kalman (1960) and later by Kalman and Bucy (1961). The Kalman filter is a recursive filter which minimizes the least-square error. The state equation in x-y coordinates (see Castella and Dunnebacke, 1974) for a constant-velocity target is x(t

+ 1) = @ ( t ) x ( t )+ r(t)A(t)

(58)

RADAR SIGNAL PROCESSING

239

where

X(t) being the state vector at time t , consisting of position and velocity components x ( t ) , i ( t ) , y(t), and j ( t ) ;t 1 being the next observation time; T being the time between observations; and a,(t) and ay(t)being random accelerations with covariance matrix Q(t). The observation equation is

+

Y(t) = M(t)X(t)

+ v(t)

(59)

where

Y(t) being the measurement at time t consisting of positions x,(t) and ym(t), and V(t) being a zero mean noise whose covariance matrix is R(t). The problem is solved recursively by first assuming the problem is solved at time t - 1. Specifically, it is assumed that the best estimate X ( t - 1 I t - 1 ) at time t - 1, and its error covariance matrix P(t - 1 1 t - 1) are known. [In X(r Is), the caret signifies an estimate and X ( t I s ) signifies that X(t) is being estimated with observations up to Y(s).] The six steps involved in the recursive algorithm are: Step 1 . Calculate the one-step prediction, X ( t I t - 1 ) = O(t - l ) X ( t -

1 It - 1 )

(60)

Step 2. Calculate the covariance matrix for the one-step prediction,

P ( t l t - 1) = q t - i)P(t - 1 I t - ipr(t- 1 )

+ r(t- i)Q(t - i)rr(t- 1) (61 )

Step 3. Calculate the predicted observation, Y(tIt-

l)=M(t)X(t(t- 1)

(62)

240

G. V. TRUNK

Step 4. Calculate the filter gain,

1

+

1

A(t) = P(t t - l)MT(t)[M(t)P(t t - l)MT(t) R(t)]-

(63)

Step 5. Calculate the new smoothed estimate,

I

1

X(t t ) = X ( t t - 1)

+ A(t)[Y(t) - %‘(tI t - l)]

(64)

Step 6. Calculate the new covariance matrix, P(t It) = [I - A(t)M(t)]P(t I t - 1)

(65) To summarize, starting with an estimate X ( t - 1 I t - 1) and its covariance matrix P(t - 1 1 t - l), after receiving a new observation Y(t) and calculating the six quantities in the recursive algorithm, a new estimate X ( t 1 t) and its covariance matrix P(t I t) are obtained. It is fairly simple to show that for a zero random acceleration, Q ( t ) = 0, and a constant measurement covariance matrix, R(t) = R, the a-fl filter can be made equivalent to the Kalman filter by setting a = 2(2k - l)/k(k

+ 1)

(66)

and

p = 6/k(k + 1 )

(67)

on the kth scan (Quigley, 1973). Thus, as time passes, a and fl approach zero, applying heavy smoothing to the new samples. While this method is optimal for straight-fine tracks, modifications must be made to enable the filter to follow target maneuvers. C . Maneuver-Following Logic

Benedict and Bordner (1962) noted that in track-while-scan systems there is a conflicting requirement between good noise smoothing (implying small a and fl) and good maneuver-following capability (implying large a and p). While some compromise is always required, the smoothing equations should be constructed to give the “best” compromise for a desired noise reduction. Specifically, since the variance reduction ratio K , defined as the steady-state variance in the filter position output divided by the variance in the measured position, equals K=

2a2 + p(2 - 3a) a(4 - p - 2a)

the (a, /?) pair should be chosen to satisfy (68) and maximize the maneuverfollowing capability. Benedict and Bordner (1962) defined a measure of transient-following capability and showed that a and fi should be related by

B = a2/(2- a)

(69)

24 1

RADAR SIGNAL PROCESSING

Alternatively, an (a, 8) pair satisfying (69) can be chosen so that the tracking filter will follow a specified g turn. Cantrell (1972) developed a method of determining the probability that the target detection will fall within a correlation region centered at the predicted target position when the target is performing a specified g turn. Then, the (a, 8) pair yielding the smallest correlation region should be used. The trouble with the above method is that to follow high g turns, the noise performance is rather poor. To rectify this situation, a turn detector employing the two correlation regions shown in Fig. 22 is used. If the detection is in the nonmaneuvering correlation region, the filter operates as usual, MANEUVERING GATE

PREDICTED

NONMANEUVERING GATE

POSITION

FIG.22. Maneuver and nonmaneuver gates centered at the target’s predicted position

a and /3 being reduced according to (66) and (67). Usually it is worthwhile to bound a and /3 from zero by assuming a random acceleration [ Q ( t ) # 01 corresponding to approximately a 1-8 maneuver. When the target falls outside the inner gate but within the maneuver gate, a maneuver is declared and the filter bandwidth is increased (i.e., a and /3 are increased); Quiqley and Holmes (1975) increase the bandwidth by lowering the value of k in Eqs. (66) and (67). To avoid the problem of the target fading and a false alarm appearing in the large maneuver gate, the track should be bifurcated when a maneuver is declared. That is, two tracks are generated: the old track with no detection and a new “maneuvering” track with the new detection and increased bandwidth. The next detection is used to resolve the ambiguity and remove one of the tracks. Cantrell et al. (1975) suggested that the a-#l filter [described by Eqs. (55), (56), and (57)] be made adaptive by adjusting a and fl by a = 1 - e-2500T (70) f l = 1 + e - 2 t m T - & - t o o T cos w 0 T/(I - 52)1’2 (71)

+ (1 - e-ObT)E(k)E(k- 1) p , ( k ) = e-“bTp,(k - 1) + (1 - e-”bT)E(k)E(k)

pl(k) = e-OnTpl(k - 1)

wo = 0.5

I Pl(k)/P2(k) I

(72) (73) (74)

242

G.

V. TRUNK

where 5 is the damping coefficient (nominally0.7), T is the time since the last update, o,and obare weighting constants, and ~ ( kis)the error between the measured and predicted positions on the kth update. The basic principle of the filter is that p,(k) is an estimate of the covariance of successive errors and pz(k) is an estimate of the error variance. When the target trajectory is a straight line, p,(k) approaches zero since the expected value of E(k) is zero. Thus wo approaches zero and the filter performs heavy smoothing. When the target turns, p,(k) grows, since the error E(k) will have a bias (either positive or negative). Thus oo grows and the filter can follow the target maneuver. Another solution to the target-maneuvering problem is due to Singer (1970). He suggested using the Kalman filter with a realistic targetmaneuvering model. He assumed that the target was moving at a constant velocity but was being perturbed by a random acceleration. Since the target acceleration is correlated in time (i.e., if target is accelerating at time t, it is likely to be accelerating at time t + At), it was assumed that the covariance of the correlation was r(T) =

E{a(t)a(t+ 5 ) ) = o$e-+I

(75)

where a(t) is the target acceleration at time t, o i is the variance of the target acceleration, and a is the reciprocal of the maneuver time constant. The density function for target acceleration consists of delta functions at k A,,, each with probability P,,,, a delta function at zero with probability P o , and a uniform density between -A,,, and A,,,. For this density

G = ( A iI ax/3)(1+

4Pmax - P o )

(76) For this target motion, Singer then calculated the state transition matrix @ ( t ) and the covariance matrix Q(t), thereby specifying the Kalman filter solution. He generated curves which give the steady-state performance of the filter for any data rate, single-look measurement accuracies, encounter geometry, and class of maneuvering targets. D. Track Initiation Detections that do not correlate with clutter points or update tracks are used to initiate new tracks. If the detection does not contain Doppler information, the new detection is used as the predicted position and a large correlation region must be used. Since there is a large probability of false alarms being in the large correlation region, tracks should not be declared firm until a third detection (falling within a smaller correlation region) is obtained. The usual initiation criteria are three out of four and three out of five. The possible exceptions are when Doppler information is available

RADAR SIGNAL PROCESSING

243

(then a small correlation region can be used immediately) or for pop-up (close) targets in a military situation. Quigley and Holmes (1975) suggest using a sequential hypothesis testing scheme for initiating tracks. When a correlation is made on the ith scan, Aiis added to the likelihood; and when a correlation opportunity is missed, Aiis subtracted from the likelihood. The increment Ai is set by the state of the tracking system, being a function of the closeness of the association, the number of false alarms, the a priori probability of targets, and the probability of detection. While this method will inhibit false tracks in dense detection environments, it will not necessarily establish the correct tracks. The proper solution will probably be a method of generating trial tracks using detections from the last several scans and then eliminating the false tracks in an easily implementable manner. To initiate tracks with detections from unsynchronized radars, Trunk et al. (1977) suggest that two times, TD and Tk,be used. If time TD goes by between track updates, the tentative track should be dropped; and if the track is updated after a time TFafter initialization, the tentative track should be made firm. Thus, setting TF > TD insures that three detections are required for making a track firm. Firm tracks that are not updated in 30 or 40 sec are usually dropped.

E. Correlation Logic

In this section several procedures are given for associating detections with tracks. Of special interest are the cases where there are conflicting situations: Multiple tracks competing for a single detection or multiple detections lying within a track’s correlation gate (or region). First of all, to limit the number of detections that can update a track, correlation gates are used. A detection can never update a track unless it lies within the correlation gate which is centered at the track’s predicted position. The correlation gate should be defined in r-8 coordinates, regardless of what coordinate system is being used for tracking. Furthermore, the gate size should be a function of the measurement accuracy R ( t ) and prediction error P(t I t - 1) so that the probability of the correct detection lying with the gate is high (at least 0.99). In some tracking systems (Cook, 1974), the correlation gate is fed back to the automatic detector, and the detection threshold is lowered in the gate to increase the PD. When there are several detections within the correlation region, the usual and simplest solution is to associate the closest detection with the track. Specifically, the measure of closeness is the statistical distance

Dz= [(rp

-

rm)Z/a12]+ [ ( e p - 8m)’/d]

(77)

244

G.

V. TRUNK

where ( r , , 0,) is the predicted position, ( r , , 6,) is the measured position, 0; is the variance of (rp - rm), and 08’ is the variance of (0, - Om). Since the prediction variance is proportional to the measurement variance, O; and 0; are sometimes replaced by the measurement variances. It should be ernphasized that statistical distance, rather than Euclidean distance, must be used because the range accuracy is usually much better than the azimuth accuracy.

X PREDICTED POSITION

x

2 e7

09

x

3

Problems associated with multiple detections and tracks are illustrated in Fig. 23: two detections are within gate No. 1, three detections are within gate No. 2, and one detection within gate No. 3. Table I11 is generated listing all detections within the tracking gate, and the detections are entered in order of their statistical distance from the track. The closest detection is tentatively associated with each track, and then the tentative associations are examined to remove detections which are used more than once. Detection No. 8, which is associated with tracks Nos. 1 and 2, is paired with the closest track (in this case, No. 1) and then all other tracks are reexamined, eliminating all associations with detection No. 8. Thus, detection No. 7 is tentatively associated with track No. 2; a conflict is noted but is resolved by pairing detection No. 7 with track No. 2. When other associations with detection No. 7 are eliminated, track No. 3 has no associations with it and consequently will not be updated on this scan. Thus, track No. 1 is updated TABLE 111 ASSOCIATION TABLE

Closest association

Second association

Track number

Detection

D2

Detection

D2

1 2

8 8

7

4.2 5.4

3

1

1.2 3.1 6.3

1

-

~

Third association Detection

D2

-

9

7.2

-

-

RADAR SIGNAL PROCESSING

245

by detection No. 8, track No. 2 is updated by detection No. 7, and track No. 3 is not updated. An alternative strategy is always to pair a detection with a track if there is only one correlation with a track. As before, ambiguities are removed by using the smallest statistical distance. Thus, in the example, track No. 3 is updated by detection No. 7, track No. 1 is updated by detection No. 8, and track No. 2 is updated by detection No. 9. Singer and Sea (1973)were among the first to recognize and characterize the interaction between the correlation and track update functions. Specifically, three distinct situations can occur: (1) the track is not updated, (2) the track is updated with the correct return, and (3) the track is updated with an incorrect return. They generalized the tracking filter’s error covariance equations to account for the a priori probability of incorrect returns being correlated with the track. This permits the analytical evaluation of tracking accuracy in a multitarget environment which produces false correlations. Furthermore, using the generalized tracking error covariance equation, the filter gain matrix was optimized, yielding a new minimum error tracking filter for multitarget environments. Also, a suboptimal fixedmemory version of this filter was generated to reduce computation and memory requirements. A later paper by Singer et al. (1974) uses a posteriori correlation statistics based on all reports in the vicinity of the track. Again the mathematical structure is similar to the Kalman filter: state equation is (58), observation and corresponding covariance equation is (59), one-step prediction is (a), matrix is (61). The estimation error is denoted by a(tI t‘) = X ( t ) - X(t I t’) and has mean and covariance matrices denoted by b(t I t’) and P(t I t’). The correlation gate size and shape is based on the Mahalanobis distance (Singer and Kanyuck, 1971) and it is assumed n, sensor reports fall within the gate on scan k . Included in the number n, are extraneous reports whose number obeys a Poisson distribution and whose positions are uniformly distributed within the gate. The smooth estimate is given by X(t

1 t ) = X(t It - 1 ) + A(t)

(78)

where A([) is chosen to minimize the noncentral second moment of the filter estimation error. The problem is solved by using track histories. A track history a at scan k is defined by selecting, for each scan i I k , a sensor report Yj)(i) where 0 I ji I n,, ji= 0 corresponding to the hypothesis that none of the reports belong to the track. The number of such track histories is

n(1 + k

L(k) =

i= 1

ni)

(79)

246

G. V. TRUNK

Associated with each history a is the probability p,(t) that the history a is the correct one given observations through time t (scan k). The terms b,,(t I t - 1) and P& 1 t - 1) are the bias and covariance of the estimation error X ( t c - l), given observations through time t - 1 and given that track history a' at time t - 1 is the (only) correct one. Recursive equations are obtained for p, , b,, and P, and then it is shown that the optimal correction vector is given by

I

This solution not only minimizes the mean-square error but also is an unbiased estimate. The trouble with the optimal a posteriori filter is that it requires a growing memory. Thus, several suboptimal filters were suggested. The first suboptimal filter only considers the last N scans; track histories which are identical for the last N scans are merged. The second suboptimal filter only considers the L nearest neighbors in the correlation gate; essentially the gate

t

FILTERS USING POSTERIORI STATISTICS

FIG.24. Experimental and calculated variances in filtered position errors for optimal and suboptimal a posteriori and a priori tracking filters. (From Singer et al., 1974, courtesy of the lnstitute of Electrical and Electronics Engineers.)

247

RADAR SIGNAL PROCESSING

LEGEND

0.7

0.5

-

1

i1

FILTER U S E 0

FILTERSUSING A POSTERlORl STATISTICS

OPTIMAL 2 SCAN MEMORY 1 SCAN MEMORY 0 SCAN MEMORY

FILTERS USING A PRIOR1 STATISTICS

1

f

EXPERIMENTAL VALUES

: II A

1

OPTlMAL A N D 2 S C A N M E M O R Y I SCAN M E M O R Y o SCAN MEMORY KALMAN

/I

FIG.25. Experimental probability of divergence or lost track for optimal and suboptimal a posteriori and a priori tracking filters. (From Singer et a/., 1974, courtesy of the Institute of Electrical and Electronics Engineers.)

size is changed to limit number of reports to L The last method uses both techniques, considers only the last N scans, and restricts the number of reports on any scan to L Simulations were run to compare the optimal and suboptimal a posteriorifilters, the optimal and suboptimal a priori filters, and the Kalman filter. Some of their results are summarized in Figs. 24 and 25. In Fig. 24 the filter variance normalized by the theoretical (perfect correlation) Kalman filter variance is plotted for several filters. As a class, the posteriori filters provide

248

G.

V. TRUNK

better performance than the other filters. However, for very high density of false reports (480; = O.l), the a posteriori filter is 30 times worse than predicted by the standard Kalman filter approach. Thus, the standard approach should never be used in dense target (or false target) environments. Figure 25 gives the probability of making a false correlation. Again the a posteriori filters provide the best performance. Stein and Blackman (1975) have proposed a maximum likelihood approach similar to Sittler (1964) for solving the multitarget correlation problem. Their approach is unified in that they consider the total correlation-track problem, which includes track initiation, confirmation, gating, and deletion logic. They compare their results with a standard approach and show significant performance improvements. However, this author wonders how complicated the method is to implement and what their improvement is relative to a more sophisticated approach, such as that of Singer et al. (1974).

F. Radar lntegration There are many ways of integrating (combining) radar detections from multiple radars into a single-system track file. The type of radar integration that should be used is a function of the radar’s performance and its environment. While no firm rules can be generated, several methods and some general rules are: 1. Track selection: Generate a track with each radar and choose one of the tracks for the system track. The only advantage of this method is that it is the simplest method to implement. 2. Average track: Generate a track with each radar and weight the tracks to form a system track. The method can be applied when there are many radars providing unsynchronized radar data. 3. Augmented track: Generate a track with each radar, choose one as the system track, but also use selected detections from other radars to update the system track. This method should be used when one radar provides substantially better data than other radars. Detections from other radars should be used when the primary radar misses some detections or when a target maneuver is declared. 4. Average detection: Average all detections and use the average to form a system track. This method should be used when many radars are providing detections essentially at the same instant in time. 5. Merge detections: Use all detections to update the system track; tracks may or may not be initiated using all detections. Theoretically, this

RADAR SIGNAL PROCESSING

249

method provides the most information, i.e., if the detections are properly weighted, this method always provides the best performance. However, care must be taken so that bad data d o not corrupt good data. Thus, this method should be used when the radars are supplying data of comparable accuracy. The most important advantage of radar integration is that it provides the tracking information in one central source. Radar integration also provides improved track continuity and improved tracking performance on maneuvering targets. Little improvement is obtained in track initiation times since, in practice, one radar will almost always detect and establish a track before any other radar can provide some detections. Either the a-B filters or Kalman filter can be used when the radars are at the same location. However, Trunk and Wilson (1976) indicate that the Kalman filter must be used to provide triangulation effects when the radars are in different locations. Also, various methods for multiple site correlation are discussed by Singer and Kanyuck (1971). V. SUMMARY The problems involved in coherent processing have received the greatest attention in this survey. Presently, the trend appears to be a digtal implementation of the adaptive processing algorithms: direct open loop calculation of canceler weights and numerical inversion of the covariance matrix for adaptive arrays. It can be expected that such systems will be built and tested during the next several years. The area of noncoherent processing has been studied intensively since 1940. The emphasis in later years has been on techniques to limit the number of false alarms (while only suffering a small target-sensitivity loss) so as not to overload the tracking system. Many systems have been built and tested and it seems that there is little research to be done in this area. During the last several years, much progress has been made in trackwhile-scan systems. This work has given guidance on the important problem of track-detection correlation in a dense multitarget environment. The major problem still needing a solution is that of track initiation in a dense environment. One problem that has received little attention so far but will receive more and more is that of adaptively controlling the surveillance radar. Problems of interest are: when should frequency and/or polarization diversity be used; when and where should various radar modes be used; and how should the signal processing be reconfigured to cope with a changing environment. Future radars will have more flexibility and their control will become extremely important.

250

G. V. TRUNK

REFERENCES Section I Cook, C. C., and Bernfeld, M. (1967).“Radar Signals, An Introduction to Theory and Application.” Academic Press, New York. Rihaczek, A. W. (1969). Principles of High-Resolution Radar.” McGraw-Hill, New York. ‘I

Section I1 Applebaum, S. P. (1964). “Adaptive Arrays,” Rep. SPL-769.Syracuse Univ. Res. Corp., Syracuse, New York. Applebaum, S.P. (1976). IEEE Trans. Antennas Propag. 24, 585-598. Barton, D. K., and Shrader, W. W. (1969). Eascon Conu. Ref. pp. 294-297. Brennan, L. E., and Reed, L. S. (1973). IEEE Trans. Aerosp. Elxtron. Sysr. 9, 237-252. Brennan, L. E., and Pugh, E. L., and Reed, 1. S. (1971). I E E E Trans. Aerosp. Electron. Syst. 7 , 254-262. Brennan, L. E., Mallett, J. D., and Reed, 1. S. (1976).IEEE Trans. Antennas Propag. 24,607-615. Capon, J. (1964). IEEE Trans. Inf: Theory 10, 152-159. Compton, R. T. (1976). IEEE Trans. Atiretinas P ropag. 24,697-706. Gabriel, W. F. (1974).’* Proceedings, Adaptive Antenna Systems Workshop,” Vol. 1, NRL Rep. 7803. Nav. Res. Lab., Washington, D.C. Granatstein, V. L., Sprangle, P., Herndon, M., Parker, R. K., and Schlesinger, S. P. (1975).J . Appl. Phys. 46,3800-3805. Griffiths, L. J. (1976). IEEE Trans. Antennas Propag. 24, 707-720. Howells, P. W. (1976). IEEE Trans. Antennas Propag. 24, 575-584. Hsiao, J. K . (1975). Radio Electron. Eng. 45, 351-356. Hsiao, J. K., and Kretschmer, F. F. (1973). Radio Electron. Eng. 43, 689-693. Kretschmer, F. (1975). IEEE Int. Radar Con/: pp. 181-185. Kretschmer, F., and Lewis, B. (1976).“An Improved Algorithm for Adaptive Processing,” NRL Rep. 8084. Nav. Res. Lab., Washington, D.C. Kroszczynski, J. (1967). Radio Electron. Eng. 34, 157-159. Kroszczysnki, J. (1970). Radio Electron. Eng. 39, 172-176. Lewis, B. L., and Cantrell, B. H. (1975).” Short Pulse Non-Coherent MTL,” Patent Appl., Navy Case 60372, NRL. Nav. Res. Lab., Washington, D.C. McAulay, R. J. (1972). Tech. Note 1972-14. Lincoln Lab., MIT, Cambridge, Massachusetts. Muehe, C. E., Cartledge, L., Drury, W. H. Hofstetter, E., Labitt, M., McCorison, P. B., and Sferrino, V. J. (1974). Proc. I E E E 62, 716-723. Nathanson, F. E. (1969). Radar Design Principles.” McGraw-Hill, New York. O’Donnell, R. M., Muehe, C. E., Labitt, M., Drury, W. H., and Cartledge, L. (1974). Eascon Conv. Rec. pp. 71-75. Powell, M. J. D. (1970). SIAM (Soc. Ind. A p p l . Math.) Rev. 12, 79-97. Reed, 1. S., Mallett, J. D., and Brennan, L. E. (1974). IEEE Trans. Aerosp. Electron. Syst. 10, 853-863. Riegler, R., and Compton, R. (1973). Proc. IEEE 61, 748-758. Shapard, J. M., Edelblute, D., and Kinnison, G. (1971).Naval Undersea Research and Develop ment Center. NUC-TN-528 May, 1971. Skolnik, M. 1. (1962). Introduction to Radar Systems.” McGraw-Hill, New York. Skolnik, M. 1. (1970).‘’Radar Handbook.” McGraw-Hill, New York. ‘I

RADAR SIGNAL PROCESSING

25 1

Van Trees, H. L. (1961).’* Detection, Estimation and Modulation Theory,” Part 1. Wiley, New York. Van Trees, H. L. (1965). IEEE Trans. M i l . Electron. 9, 216-229. Ward, H. R., and Shrader, W. W. (1968). Eascon Conu. Rec. pp. 168-173. Washburn. T. W., and Sweeney, L. E. (1976). IEEE Trans. Antennas Propag. 24, 721-732. Widrow, B., and H o b M. (1960). I R E WESCON Conu. Rec. pp. 96-104. Widrow, B., Mantey, P., Friffiths, L., and Goode, B. (1967). Proc. IEEE 55, 2143-2159. Widrow, B., Glover, J., McCool, J., Kavnitz, J., Williams, C.. Hearn. R., Zeidler, J., Doug, E., and Goodlin, R. (1975). Proc. IEEE 63, 1692-1716.

Section Ill Applied Physics Laboratory (1975). Radar Processing Subsystem Evaluation,” Vol. 1, APL Rep. FP8-T-013. Johns Hopkins Univ., Baltimore, Maryland. Blake, L. V. (1953). P roc. I R E 41, 770-774. Cantrell, B. H., and Trunk, G. V. (1973). IEEE Trans. Aerosp. Electron. Syst. 9, 649-653. Dillard, G . M. (1967). I E E E Trans. In/: Theory 13, 2-5. Finn, H. M., and Johnson, R. S. (1968). R C A Rev. 29,414464. Hansen, V. G. (1970). IEEE Trans. In/: Theory 16. 309-315. Hansen, V. G . (1973). IEEE Int. Conf: Radar-Present Future pp. 325-332. Hansen, V. G. (1974). Comput. Electr. Eng. 1, 545-550. Hansen, V. G., and Olsen, B. A. (1971). I E E E Trans. Aerosp. Electron. Syst. 4, 942-950. Hansen, V. G., and Ward, H. R. (1972). IEEE Trans. Aerosp. Electron. Syst. 8, 648-652. Hillier, F. S., and Lieberman, G. J. (1967).“Introduction to Operations Research,” pp. 457459. Holden-Day, San Francisco, California. Lindgren, B. W. (1962). *‘Statistical Theory.” Macmillan, New York. Marcum, J. I. (1960). I R E Trans. In/: Theory 6, 59-268. Marcus, M. B.. and Swerling, P. S. (1962). IEEE Trans. In/: Theory 8, 237-245. Mitchell, R. L., and Walker, J. K. (1971). IEEE Trans. Aerosp. Electron. Sys. 7, 671-676. Neyman, I., and Pearson, E. S. (1928). Biometrika 2OA, 175-240, 263-294. Palmer, D. S., and Cooper, D. C. (1964). IEEE Trans. I n / : Theory 10, 296302. Schleher, D. C. (1975). IEEE I n t . Radar Con/: pp. 262-267. Schwartz, M. (1956). IEEE Trans. In/:Theory 2. 135-139. Swerling, P. (1956). Proc. I R E 44. 1146-1155. Swerling, P. (1960). I R E Trans. Inf Theory 6, 269-308. Trunk, G. V. (1972). Proc. IEEE 60,743-744. Trunk, G. V. (1976). Non-Rayleigh Sea Clutter: Properties and Detection of Targets,” NRL Rep. 7986. Nav. Res. Lab., Washington, D.C. Trunk, G. V., Cantrell, B. H., and Queen, F. D. (1974). IEEE Trans Aerosp. Electron. Syst. 10, 574-582. “

Section I V Benedict, T. R., and Bordner, G. W. (1962). I E E E Trans. Autom. Control 7, 27-32. Cantrell, B. H. (1972).“Behavior of a-B Tracker and Maneuvering Targets Under Noise, False Target, and Fade Conditions,” NLR Rep. 7434. Nav. Res. Lab., Washington, D.C. Cantrell, B. H. (1973).’‘ Description of an m - p Filter in Cartesian Coordinates,” NRL Rep. 7548. Nav. Res. Lab., Washington, D.C. Cantrell, B. H., Trunk, G. V., and Wilson, J. D. (1974). “Tracking System for Two Asynchronously Scanning Radars,” NRL Rep. 7841. Nav. Res. Lab., Washington, D.C.

252

G. V. TRUNK

Cantrell, B. H., Trunk, G. V., Queen, F. D., Wilson, J. D., and Alter, J. J. (1975). IEEE fnr. Radar ConJ pp. 391-395. Castella, F. R., and Dunnebacke, E. G . (1974). IEEE Trans. Aerosp. Electron Syst. 10,891-895. Cook, S . R. (1974). Development of IADT Tracking Algorithm,” APL Rep. F3C-1-061. Johns Hopkins Univ., Baltimore, Maryland. Kalman, R. E. (1960). J . Basic Eng. 82, 35-45. Kalman, R. E., and Bucy, R. S . (1961). J . Basic Eng. 83, 95-107. Quigley, A. L. (1973). IEEE Int. Radar ton/:: Radar-Present Future pp. 352-357. Quigley. A. L., and Holmes, J. E. (1975). “The Development of Algorithms for the Formation and Updating of Tracks,” WP-XBC-7512. Admirality Surf. Weapons Establ., Portsmouth, England. Singer, R. A. (1970). IEEE Trans. Aerosp. Electron. Syst. 6. 473-483. Singer, R. A., and Kanyuck, A. J. (1971). Autornatica 7 , 455-463. Singer, R. A,, and Sea, R. G. (1973). IEEE Trans. Autom. Control. 18, 571-582. Singer, R. A,, Sea, R. G., and Housewright, K. B. (1974). IEEE Trans. It$ Theory 20,423-432. Sittler, R. W. (1964). IEEE Trans. Mil. Electron. 8, 125-139. Sklansky, J. (1957). R C A Rev. 18, 163-185. Stein, J. J.. and Blackman, S . S. (1975). I E E E Trans. Aerosp. Electron. Syst. 11, 120771217, Trunk, G. V., and Wilson, J. D. (1976). ‘*TrackingFilters for Multiple-Platform Radar Integration,” NRL Rep. 8087. Nav. Res. Lab., Washington, D. C. Trunk, G. V., Wilson, J. D., Cantrell, B. H., Alter, J. J., and Queen, F. D. (1977).”Modifications to and Preliminary Results for the ADIT System,” NRL Rep. 8091. Nav. Res. Lab., Washington, D.C. White, D. M., and Silberman, S . R. (1975). ”Simulation of 2D Radar Automatic Detection and Tracking Systems : Baseline Program,” TSC-W8-60. Technol. Serv. Corp., Washington, D.C. Wilson, J. D., and Cantrell, B. M. (1976). “Tracking System for Asynchronously Scanning Radar with New Correlation Techniques and an Adaptive Filter,” NRL Rep. 7952. Nav. Res. Lab., Washington, D.C. ”

On the Teaching of Electronics to Scientists* H. E. BERGESON

AND

GEORGE L. CASSIDAY

Department of Physics The Unicersrty of Utah Sat: t a k e Ciry, Utah

1. Introduction.. . . . . . . . . . . . . . . . . . . . . .......... ....... .... 11. Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. First Quarter ................................................................... A. Simple Linear Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Simple Digital Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. Inductors, Capacitors, Transistors, and Other Vital Components . . . . . . . . . . . . . . . .. IV. Second Quarter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

253 258 260 260 268 272 278 A. Voltage Sources, Current Sources, and Wave-Shaping Techniques . . . . . . . . . . . .. . 278 . . 282 ........................................ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288 ............. ........ . . 290 292 E. Advanced Linear Devices and Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298 F. Digital-to- Analog and Analog-to-Digital Techniques . . . . . . . . . . . . . . . . . . . . . . . . . V. Third Quarter ................................................. A. Design of a Simple Computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302 B. The LSI- 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. l/O Operations . . . . . . . . . . . . . . . . . . . . . . . ............................... 312 D. The DRVll Parallel Line Interface . . . . VI. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ............................... 322 Appendix,. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322 . . . . . 329 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

I. INTRODUCTION The philosophy of teaching electronics to scientists has undergone a great many changes in the last few years. The continued explosive development of electronic devices will accelerate that change. In the past, electronics was regarded primarily a domain of the physicist; now it is a major engineering discipline. Physicists, chemists, and other scientists will continue to study

* Editor’s comment: A dimerent philosophy of the teaching of electronics is implied in the two-volume treatise: *‘ Electronic Methods,” Volumes 2A and B of“ Methods of Experimental Physics” (E. Bleuler and R. 0. Haxby, eds.), Academic Press, New York, 1975. 253

254

H. E. BERGESON A N D GEORGE L. CASSIDAY

materials, and the results of these studies will assuredly lead to the development of new electronics devices. However, these scientists are no longer the developers of electronics circuits and associated instrumentation, but more likely they are end users of such products which are developed nowadays almost solely be engineers. Consequently, the teaching of electronics to scientists has increasingly swung away from “the study of the fundamentals” and more and more toward the pragmatic “applications and uses” point of view. Indeed, at the University of Utah we have found it fruitful to adopt this approach insofar as possible. There have been pious expressions that if we will teach students “the fundamentals” then they will have the foundation necessary to use newer devices as they are developed. A careful examination of this idea is less than satisfying. So far as the behavior of the passive parts of an electronic circuit, the resistors, the capacitors and the inductors is concerned, this is a wellfounded notion. It will always be important to understand the effects of stray capacitances and inductances on frequency response and pulse shape. There will always be the necessity for pulse shaping and filtering. But so far as active devices are concerned, it is not clear what “the fundamentals ” are. At one time they may have consisted of the study of the thermal emission of a hot cathode, vacuum tube operation and biasing techniques, and the development of their characteristic curves. However, these notions were not very useful when discussing transistors. Indeed, such notions could even be detrimental since triodes are very high input-impedance devices and consequently are voltage controlled while transistors are relatively low inputimpedance devices where current control is more appropriate. So with the advent of transistors, a “study of the fundamentals ” meant learning about the band structure of semiconductors, “unlearning” vacuum tube biasing techniques, learning techniques more appropriate for transistor operation, and then deriving new sets of characteristic curves. Yet today we find that a knowledge of transistor fundamentals is not very helpful when learning about operational amplifiers and is even less helpful when interfacing an A/D converter to a microprocessor. It could thus be argued that a more useful approach for scientists today consists of circuit design using specification sheets of some of the myriads of system devices available on the market as the starting point. Since the integrated circuits available today tend to be concrete large scale system or function oriented rather than simple abstractions, such an approach has the benefit of a readily visualized and realizable result. Consequently, it is much easier to maintain one’s enthusiasm for the subject. After all, it is easy to see the virtue of an A/D converter-the name itself conveys the meaning of its intended function-but what is the virtue of an isolated transistor? Perhaps then the most important consideration in introducing the

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

255

potential scientist to electronic circuits and devices is to ask how he will use this knowledge. A11 too often cources given to such students are simply copies, perhaps abbreviated, of those given to electrical engineering students. Such courses typically place a premium upon achieving a detailed understanding of the characteristics of electronic devices with little attention given to their application. Ultimately, the study of the subject becomes an end in itself rather than a means to an end. This makes sense if one is interested primarily in pursuing increased performance possibilities or in minimizing costs. After all, these are the goals of companies involved in a competitive market. However, a typical scientist has quite a different problem. He can buy most of the electronic circuits and systems that he needs. It will be useful for him to understand their behavior, or least t o a degree. However, his most serious challenge will involve electronics applications; for example, if the output of an amplifier does not match the impedance or cover the range of the input to a recording device, what should he do? For such a person it is not useful to spend a great deal of time optimizing the behavior of the circuit or to seek the lowest cost components. It is far more important that he be able to accomplish the necessary task quickly so that he can get on with what is t o him far more important, the continuation of his research. The need for a different approach to the education of a scientist from that of an electrical engineer is underscored further by the fact that the electrical engineering student can spend two or more years studying electronics almost exclusively, whereas the student of the sciences can perhaps take a single year-long course in the subject. In this article we will argue that the most appropriate education in electronics for a scientist involves exposure to a large number of high-level devices, the design of function-oriented circuits using the specification sheets of those devices, and acceptance of working circuits which satisfy the original need. Questions of elegance should be irrelevant as should suggestions of minimization of device count, and more than a cursory examination of the physics of the devices should be considered a waste of time. On the other hand, attention must be paid to input and output characteristics, for example, when matching the output of one device to the input of another. Moreover, we feel that it is necessary to gain an understanding of computer structures and to be conversant with the interfacing of various linear and digital devices to the input-output ports of mini and microcomputers. A scientist should know how to program the mini or microcomputer in order to control data logging devices, experimental and instrumental controls, and perhaps communications systems. We do not pretend that the education given a student today will be adequate in five or ten years, only that the experience and confidence gained in learning how to use the devices of today will be useful when learning how to use those of tomorrow.

256

H. E. BERGESON AND GEORGE L. CASSIDAY

The question of order of presentation has considerable importance. The classical approach starts with direct current (dc) circuits and deals with such things as Ohm’s law, Wheatstone bridges, and Thevenin’s theorem. One moves on to a consideration of alternating currents in circuits involving capacitors and inductors, ac circuit analysis, rectifier circuits, and then the physics of active devices such as semiconductor diodes, vacuum tubes, and various types of transistors. These fundamentals” are then applied in simple vacuum tube and transistor amplifiers; subsequently negative feedback is discussed. Operational amplifiers may be introduced (in particular, concern about their inner mechanism typically dominates discussion); various types of oscillators are presented along with techniques of analog measurements, pulse circuits, and finally, digital electronics. While this approach offers a great deal of logical appeal to those who already understand electronics, we argue that it fails to motivate the neophyte and is pedagogically poor. If one accepts the premise that electronics for the science student is a tool, then one must conclude that intimate details concerning the structure of the tool are irrelevant. The many creative things that one can do with the tool move to the forefront of our perspective and thus the electronics educator should address his attention to the discussion of the tool’s capabilities. Thus, the first half of a classical approach to electronics has no obvious value to a science student. Hence we throw it out and develop an entirely new approach, keeping in mind our fundamental premise, namely, electronics is a creative tool. Let us accept the existence of a device, forget about what makes it tick,” and explore its possible consequences. At this point then we will describe how we tackle the problem here at Utah. We offer a three-quarter sequence course, outlined in detail in the Appendix. The trick is to introduce students as quickly as possible to a large variety of electronics devices. Boundary conditions are established with a precursory examination of passive circuit elements and the development of rather intuitive rules of thumb for design purposes. Within a week students are introduced to operational amplifiers and other integrated circuits permitting the design, construction, and debugging of manifestly useful circuits. A quick and dirty approach is taken, and design problems are chosen to minimize the distractions of biasing, offset voltages and currents, input and output impedances, loading, etc. In other words, emphasis is placed upon the implementation of some function which can be readily achieved with today’s integrated circuits. As the course proceeds, more difficult problems can be given which introduce the inevitable difficulties in a controlled way. At each point, a student can deal successfully with such problems, thus increasing his skill and confidence. By the end of the first quarter each student will have designed, built, and debugged more than 16 circuits, both analog and digital, “

“

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

257

most of which have obvious applications. It has been our experience that by this time a student will have gained a large measure of pragmatic experience as well as insight into the large variety of capabilities of electronics circuitry. Indeed, many students, such as those majoring in psychology or biology, terminate the course sequence here and return to their respective disciplines with an enhanced knowledge of the electronics instruments they use in their own research activity. The second quarter of the course involves more or less a rehash ” of the first quarter. Emphasis is still placed upon applications but greater attention is paid to detail and performance optimization. Rather sophisticated circuits are designed and a greater number of problems is encountered by the student when debugging those designs. Achieving working circuits thus requires increasingly higher levels of electronics knowledge. Most important, we find that students already armed with a rather large arsenal of circuit devices and techniques provided by their “ quick-and-dirty ’* first-quarter introduction are now “chomping at the bit” in their desire and eagerness to pursue more subtle problems as how stray capacitance degrades pulse rise times or how to stabilize a superfast high-gain amplifier. It is rare that a beginning student with little or no previous lab experience appreciates elegant solutions to subtle problems which are to him virtually nonexistent. It is only through experience that a student gains an appreciation for such problems. Thus, the second quarter “wraps around ” upon the first. Many of the topics covered there are looked at again but from the vantage point of experience. Additional devices are covered which serve to introduce the student to new techniques. For example, the RCA 3080 OTA (operational transconductance amplifier) is used to introduce the student to the ideas of (1) sample and hold, (2) amplitude modulation, and (3) analog switching and multiplexing. Finally, the second quarter ends with a discussion of analog-to-digital conversion techniques and digital computational circuits with a view toward introducing the student to the third quarter’s work: microprocessor and minicomputer operation. Throughout the third quarter, as in the first, our intent is not to compete with scientists in other disciplines, in this case the computer scientist. We d o not wish to teach computers per se and then turn loose upon the world scientists whose sole desire is to make more efficient computers or to pack them into your hip pocket. We view the computer as a tool which can be exploited by the intelligent but busy scientist in his own area of research. We try to avoid spending much initial time upon the details of computer architecture and operation, but instead we quickly introduce the scientist to computer input/output applications insofar as this is possible. And we really emphasize input/output operations, for therein lies the real power of the computer to the experimental scientist. It has been our experience that we “

258

H. E. BERGESON A N D GEORGE L. CASSIDAY

can take a student with a knowledge of electronics but absolutely no knowledge of computer software or hardware, and such a student will have hooked up and made to work a computer-controlled electronics system by the end of the third quarter. The core of the work in the third quarter is just such an electronics project which the student must implement successfully in order to achieve an “A” for the course. One final point should be made about this particular philosophy: we do not fed that the particulars of the topics included in the course outline of the Appendix are sacrosanct. To a large extent, the teaching of electronics is invariant under exchange of subject matter. We could easily see that certain op amp circuits, the OTA or the phase-lock loop could be eliminated and replaced with some other device of the instructor’s personal preference. The course is designed with this in mind. Our goal was to design the teaching of electronics around the manufacturer’s specification sheets. This makes for easy substitution which is not only an option for the instructor but a must. The future shock ” in electronics development is perhaps more severe than in any other technology. Today’s elegant electronics device will most assuredly be tomorrow’s obsolete Edsel. (If a young reader does not know what an Edsel was-well, that just illustrates the point.) The point is this: What the student designs and builds is not so important as the fact that he design and build and get something to work-with emphasis on the last step. There is no substitute for the confidence gained by achieving a successfully operating circuit. ‘I

11. ORGANIZATION

During the first quarter the student attends two 50-minute lectures each week: there is also an optional weekly problem and discussion section. The real heart of the course is a two-hour laboratory session held twice weekly. The student must attend the first part of the scheduled laboratory sessions to get instructions and he must demonstrate working circuits during the scheduled laboratory sessions; however, by means of an electronic combination lock, he can get into the laboratory any time the building is open to build and test his circuits. (Bench power comes through a one-hour timer to avoid problems with equipment left on after students leave.) For most laboratory assignments, the student is asked to develop a circuit with performance between certain limits (e.g., amplification of 50 k 15). No credit is given for performance outside those limits; full credit is given for performance within the limits. Other articles (English and Lind, 1973: Babcock and Vignos, 1973) have emphasized the importance of easily used breadboarding procedures. Since we usually ask the students to build two circuits per week and since the

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

259

circuits must be designed and debugged by the student, it is absolutely essential in this course that the actual connection of parts occupy a small fraction of the laboratory time. We have used a no-solder breadboarding plug-in card* Power and signals are distributed to several stations via a parallel system of 22-pin card connectors. With this system, only one set of power supplies and signal generators is required. During the second quarter, the student attends three 50-minute lectures each week. The lecture material was purposely increased in order to accommodate more time spent on understanding circuit devices and behavior. However, the lab portion of the course is still just as intensive as in the first quarter and still emphasizes design. We feel that it is appropriate to increase the workload for the student at this point in time since primarily physics majors and other hard-core would-be scientists continue with the last two quarters of the sequence. The psychologists, biologists, and premeds have pretty much vacated the premises by now. We d o allow these remaining students greater flexibility insofar as their lab schedule is concerned. They still have two circuits per week to design, build, and debug, but they can work more or less as their own schedule dictates. Every so often, an unattended student might catastrophically destroy a power supply or a signal generator, but such incidents are less frequent than one might imagine. By the second quarter, the students are experienced with the lab and its equip ment and most of the supplies and signal generators have been reasonably well idiot-proofed.’’ So the open-door ’’ policy works well. Each student still must come in at certain specified times, if only for a few minutes, each week to demonstrate working circuits to his lab instructor. As in the first quarter, no credit is given for performance outside specified tolerance limits, while full credit is given for circuits which perform within those limits. Credit obtained for achieving working circuits comprises the major component of each student’s grade. (Grading details are covered in the Appendix.) The third quarter of the course which deals with computer interfacing is arranged somewhat differently. The student again attends two 50-minute lectures each week but the laboratory is now totally open and the instructor and/or his lab assistant is formally on call for a total of two hours/day, 5 days/week. (In practice, the students may “nail” the instructor almost anytime.) By now there are no more than 10-20 students in the course (although this number seems to be on the increase!), and the electronics hardware which the student must put together in the lab has become much more complex. Furthermore, the student now must spend many sessions at a “

“

* A. P. Incorporated, 72 Corwin Drive, Painesville, Ohio 44077. Catalog number 923770, $28.90 in quantities of ten or more (1973).

260

H. E. BERGESON AND GEORGE L. CASSIDAY

minicomputer in order to learn software at the machine and assembly language level. These two requirements dictate an “open lab” policy since the lab can now only accomodate several students at a time. Many concrete examples of both computer hardware and software are presented in the lecture which must be implemented as laboratory assignments. Again no credit is given until the student demonstrates a successfully working hardware circuit or software program. The crux of this quarter is a hardware electronics project of the student’s own choosing which must be hooked up to and controlled by a minicomputer. Success at this endeavor yields the student an “A” for the course regardless of intermediate homework, lab, or exam results ! The rest ofthis article will be devoted to a discussion of the subject matter of the three quarters of this course along with that quarter’s grading procedures. In many instances, the description will be brief. Where the treatment is not standard or where it might be perhaps unfamiliar to the scientist involved in teaching electronics to scientists, more detail will be presented. 111. FIRST QUARTER

A. Simple Linear Circuits

The first two lectures deal with a simple description of passive circuits using Ohm’s law and Thevenin’s theorem. The notions of ideal voltage sources, current sources, input and output resistance are developed and emphasized. These notions crop up over and over again when discussing operational amplifiers and other integrated circuits from a black box device” point of view. One of the most common traps for each student involves loading down a voltage source with low input resistance devices or loading down a current source with a high input resistance device. Emphasis is placed upon engendering device familiarity and intuitive feelings for their behavior. Quick, mental calculations using approximations are encouraged in order to estimate such circuit characteristics as quiescent voltage levels, current flow, and voltage gain. Indeed, since standard resistors and capacitors typically have tolerances of +lo%, we react vigorously upon seeing students attempting to calculate circuit behavior to &digit accuracy from complicated equations derived in full detail. “

1. Virtual Equality of Nonsaturated Operational Amplijer Inputs

The next three lectures introduce the student to the operational amplifier (op amp) as a circuit building block. The device is initially treated as ideal, i.e., a device with infinite gain and input impedance and zero output

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

26 1

impedance. No attempt is made to look at the “innards” of an op amp. Instead time is spent upon introducing the student to the power and usefulness of the op amp. We have found this approach to be extremely powerful in terms of motivation and confidence building. In just under two weeks each student is assembling obviously useful circuits. The student has not bogged down in a morass of h-parameter analysis or complex-pole-zero mathematics which has little to d o with the world as it appears to him at this point in time. The concept of virtual equality proves particularly useful and is easily mastered. Once mastered and put to practice in a few working examples, a vista of ideas opens before each student. We develop this concept based on

FIG. 1. Schematic representation of an ‘*ideal”operational amplifier.

the “ideal op amp” approximations shown in Fig. 1. Such approximations allow the student to quickly design rather sophisticated circuits such as algebraic equation solvers. From Fig. 1, we see that V,,,

=

A(V+

-

V-)

V,,, is the output voltage, V+ is the voltage applied to the noninverting input, V- is the voltage applied to the inverting input, and A is the open-loop differential voltage amplification. Clearly, if the output voltage is to remain finite, the input voltage difference (V+ - V - ) must approach zero if the open-loop gain approaches oc). Hence, the virtual equality of the two inputs if the op amp remains out of saturation. The usefulness of this concept is demonstrated in the derivation of the voltage gain of the three common circuits shown in Fig. 2. The trick with the operational amplifier is to use feedback. By connecting the output back to the appropriate input via some passive circuit element one can argue that the op amp will d o whatever it must to keep V+ - V- 5 0, i.e., maintain the virtual equality of its two inputs. Only when the op amp saturates itself while trying to maintain this equality will the input voltages unlatch themselves and develop a sizeable difference. Assuming the op amp stays out of saturation, we then have for Fig. 2a,

. v ,

/=-

R2

Since the op amp has infinite input resistance R i , the current into the op amp is negligible, i must flow through R,. Thus current must be “sunk” by

262

H. E. BERGESON A N D GEORGE L. CASSIDAY

(a)

FIG.2. Amplifying configurations for operational amplifiers: (a) an inverting configuration, (b) a noninverting configuration, and (c) a differential configuration.

the op amp’s output and since the op amp has zero output impedance, it is not loaded down ” by R Thus, “

V,,, =

-

-

iR,

(31

(4) R, Note that the real input resistance to the circuit is trivially seen to be R, . A rather nice geometrical picture describing the gain derivation of this op-amp circuit can be drawn (Smith, 1971, Chs. 1 and 2), as in Fig. 3. To interpret this voltage-resistance graph, imagine that one starts at the source with a voltage meter and walks along the R2-R, resistance bridge of Fig. 2a towards the output measuring voltages at each position along the way. The input is held rigid at V , by the input voltage source. The op amp nails down the voltage at the inverting input to V- 2 0 as long as the op amp does not saturate itself trying to sink too much current in order to accomplish this task. Consequently, due to linearity and Ohm’s law, if we continue to the output a distance R,, we obtain the output voltage V,,, by a simple linear extrapolation of the straight line connecting V, at the input and 0 at the inverting input of the op amp. Hence, by similar triangles, - V,,,/R, = V , / R , or V,,, = - V,R, / R , . Similar plots for the circuits of vo,,

=

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

locus

IDCUS

263

locus (C)

I

Inbut locus

& locus

outpui locus

FIG.3. Voltage-resistance plot for inverting circuit of Fig. 2, diagrams (a), (b), and (c).

Fig. 2b and c are shown in Fig. 3b and c. The appropriate geometrical relations describing the gain derivations are for circuit b (Figs. 2b and 3b)

or vou,=

(1

+2)K

and for circuit c

or

The standard algebraic derivations are of similar simplicity. The ease of these derivations stems from the idea of " virtual equality." Experience has shown us that students must be continually reminded that virtual equality is a dynamic effect which is derived from the large open-loop gain of the operational amplifier.

264

H. E. BERGESON AND GEORGE L. CASSIDAY

(b) VO

FIG.4. Computing configurations for operational amplifiers: (a) an adding configuration and (b) an integrating configuration.

The behavior of computing circuit configurations is equally easy to understand with the concept of virtual equality. In Fig. 4a is an adding configuration. As before, since V- 2 0,

[One interesting student project using the adding configuration is to produce a digital-to-analog converter. If, say, binary ones are a 5-V signal and binary zeros are a 0-V signal, he need only set R , = $ R a , R , = $ R b , R , = &R,, and R , small enough that the output does not saturate to get an output voltage proportional to the binary value.] Similarly, with the integrator of Fig. 4b, i = - V,

R

Q

1/out = - -

C

Vou, =

1

' 1 V, dt RC .

--

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

265

The use of such circuits by students in the analog solution of differential equations is described well in Babcock and Vignos (1973). It is even easier for the student to set up a system to solve systems of first-order algebraic equations with several variables. 2. First-Order Perturbations The operational amplifier we have used in the course is the 741,* available at a cost of about $0.50. It has an input resistance of typically 2 MR, an output resistance of 70 R and an open-loop voltage amplification of 200,000. These characteristics are close enough to ideal for most applications. The most common difficulties encountered with this op amp relate to input and output currents. The 741 can supply only a few milliamps. Thus, the use of feedback resistors of 1 kR or less will create difficulties. Similarly, with input or feedback resistors on the order of 100 kR or so, the student will encounter difficulties due to the significance which input bias currents now assume. These limitations should be mentioned in lectures and we do so. However, the words fall upon deaf ears. These limitations represent just the sort of subtleties alluded to previously in the introduction. Such subtleties lie in wait to entrap the unsuspecting student in the lab and, indeed, this is precisely what happens when the student designs and tests circuits with 560 kR or 27 R resistors! The student is not likely to forget the experience, particular when elaborated upon by the overseeing lab instructor. These two perturbing effects can be largely eliminated by switching from the 741 to the RCA 3140,t a rather beautiful op amp which is inexpensive ( :$1.00) and pin for pin compatible with the 741. It has an input resistance of 1.5 TR, input bias currents of about 10 PA, an open-loop voltage gain of 100,000and output-current capabilities of more than 20 mA making it much more nearly an ideal device. One might argue, however, for the virtue of the perversity introduced by the less than perfect characteristics of the 741. 3. Second-Order Perturbations By requiring the student to amplify millivolt-sized signals, one can force the student to an even higher level of sophistication. When corrected to second order, Eq. (1) becomes V,", = A(V+ - V- + KO) (16) * The 741 operational amplifier is produced by a number of semiconductor manufacturers whose devices all meet essentially the same specifications. Specification sheets are provided by Fairchild. Motorola, National Semiconductor. Signetics. Texas Instruments, and others. t RCA Specification Sheet and Application Note 2M144, available from RCA Solid State, Box 3200. Route 202, Somerville, New Jersey. 08876.

266

H. E. BERGESON A N D GEORGE L. CASSIDAY

FIG.5. Circuits for correcting input errors: (a) the input offset-voltage nulling circuit for the 741. The 10-kR potentiometer is adjusted until Vo reads zero. The potentiometer is retained with the operational amplifier in whatever circuit configuration it is placed. In (b), the first operational amplifier is used to buffer a high output-impedance signal source from the inverting amplifier.

where 4,is the input offset uoltage. The value of &, may range from microvolts for precision op amps to almost 100 mV for some “ grubby” ones. It is a few millivolts for the 741. The input offset voltage is intrinsic to any particular operational amplifier, but it can be reduced considerably for most of them with an external input offset nulling circuit. The nulling circuit for the 741 is shown in Fig. 5a. The 10 kR potentiometer is adjusted until the output voltage V, equals zero. Then the 100 R and 100 kR resistors are removed and the integrated circuit with its 10 kR potentiometer are placed in whatever circuit configuration is being used. The nulling mechanisms vary from one type to another; most manufacturers’ specification sheets show how it is done for the device in question. To amplify a millivolt signal successfully, usually the student must null the input offset voltage. For signals from high impedance sources, the student will find he must pay attention to input currents. A first correction can be made by forcing input currents through equivalent resistances t o both inverting and noninverting inputs. Then, if the input currents are equal, the voltage drop to both inputs is equal and has little effect on the output. However, there is an input offset current, an intrinsic difference between the two input currents, which might still be important. The circuit designer can often use more than one operational amplifier to solve input offset current problems. Suppose, for example, he has a high output impedance source driving an operational amplifier in the inverting configuration of Fig. 2a. In this case R2 must be large to avoid loss of signal due to mismatch of input and output im-

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

267

pedances. At the same time R , must not be so large that the input offset current creates problems. Using an extra operational amplifier as a voltage follower (i.e., a noninverting amplifier with a gain of l), the signal then goes into the inverting amplifier from a low impedance source. (See Fig. 5b.) If, at this point, we still have problems, the best solution may be to obtain a different operational amplifier with a smaller input offset current such as the RCA 3 140. Indeed, use of the RCA 3 140 eliminates almost entirely these second-order input current effects. Its input offset current is only 0.5 PA, an entirely inconsequential value. However, its input offset voltage is 2 mV typically, so nulling is still required if one is to examine signal voltages of this order of magnitude. 4. Saturated Output Circuits

Another set of useful circuits involves finite differences between V+ and V- with a resulting saturation of the output voltage. Consider Fig. 6a, the voltage comparator: V- is an adjustable reference voltage. Ignoring the input offset voltage, if V , is less than V - , the output will saturate negatively.

FIG.6. Saturated output configurations: (a) a voltage comparator, (b) a Schmidt trigger, and (c) a square-wave oscillator.

268

H. E. BERGESON AND GEORGE L. CASSIDAY

Typically, operational amplifiers will saturate within about one volt of the positive or negative supply voltage, which in turn are typically k 15 V. Obviously, if one desires an inverted output, one need only reverse V+ and V- .

The Schmidt trigger of Fig. 6b is similar to the voltage comparator expect that positive feedback produces two triggering levels depending on the state of the output, and the input does to the inverting rather than the non-inverting input. To a fairly good approximation, we can regard the saturation levels to be equal to the supply voltages. With this approximation, the triggering levels are determined by a voltage divider consisting of R , , R , , and R The upper triggering level has R in parallel with R , while the lower level has R in parallel with R , . The oscillator of Fig. 6c is also simple. That it will oscillate when an ideal operational amplifier is used is obvious qualitatively. V+ is at some fraction of the output saturation voltage, while I/_ approaches the output saturation voltage as C charges through R , . When V- reaches V+ the output reverses to the opposite saturation voltage. The sign of V+ reverses and V- now approaches the opposite saturation voltage. Obviously, there is no stable state. The derivation of the oscillation frequency is most conveniently reserved for a later lecture when rise times are discussed. (For real, nonideal operational amplifiers, one must be sure that as the output reverses, V+ will change faster than V- . Since the slew rate is the rate at which a voltage change can occur on the output, one must have R,C greater than the saturation voltage divided by the slew rate.) These saturated output circuits are easy to understand. Since they have continuous inputs and bistable outputs, they make a natural transition to digital circuits which we describe next. Students are able to master this initial operational amplifier material along with material concerning resistive networks and input and output resistances in five lectures and six laboratory sessions. With our schedule of two one-hour lectures and two two-hour laboratory sessions per week, the student is able to design useful low-frequency circuits in a three week period. He is also able to make allowances for input currents and input offset voltages and to be aware of impedance matching problems. Inverting and noninverting amplifiers of arbitrary gain, differential amplifiers, digital-toanalog converters, voltage comparators, integraters, Schmidt triggers, and, to some extent, analog computation circuits are part of his repertoire. B. Simple Digital Circuits Learning to use logic and counting circuits is as easy as learning to use operational amplifiers. It is also very inexpensive. One can obtain logic circuits for about $0.25 and counting circuits for less than $1.00. Light-

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

269

emitting diodes can be used as indicators and can be obtained for under $0.25 each in lots of 100. Medium-scale integration decoding, driving, and counting circuits and seven-segment digital displays are available for $1.00 to $3.50 each. 1. Logic Families

Logic and counting circuits come in a variety of families. Within a family, inputs and outputs are compatible. That is, the output of one device will provide voltages and currents that will be recognized unambiguously as either a “ 1 ” or a “0” at inputs to other devices in its family. Typically, one output can drive ten inputs with only a direct connection between them. The only real problems are associated with initial input and final output. Two once popular families are now obsolete and should be avoided. They are resistor-transistor logic (RTL) and diode-transistor logic (DTL). Not only are they inferior to other families, but they are more expensive and less readily available. There are three popular families to choose from : emitter-coupled logic (ECL); transistor-transistor (TTL); and complementary metal oxide surface logic (CMOS). ECL has speed as its only advantage. For an introductory laboratory the few nanoseconds speed advantage is not worth the additional cost and complexity. We have chosen the 7400 TTL series because it is very inexpensive, a great variety of device types is readily available, the logic levels are convenient, and the speed (> 10 MHz) is more than adequate. Because it uses less lower and is less sensitive to power supply variations, CMOS may be a better choice now that its price has dropped. However, CMOS is slower than TTL. A CMOS exists which is interchangeable pinfor-pin with popular TTL devices. Students who learn with TTL will be able to make an easy transition to CMOS. In a sense, because of the compatibility of input and output within a logic family, the primary input and the final output represent the major design; for the central portion of the circuit one need only connect together suitable logic units. However, the use of logic circuits is new to most students, so they need practice with interconnection. With standard TTL logic (TheTTL Data Book, 1973), a logical “0” is a voltage between 0 and 0.8 V; a longical 1 is a voltage between 2.0 V and the supply voltage, nominally 5.0 V. TTL is known as current-sinking logic; i.e., a circuit presenting a logical “0” to an input must accept up to 1.6 mA from that input. A circuit presenting a logical 1 at, say, 2.4 V, must provide about 40 pA to the input. “

”

2. Simple Logic Circuits First, the student must learn about simple logic functions. Probably the easiest introduction is to examinethe truth tables. Figure 7 shows truth tables

2 70

H. E. BERGESON AND GEORGE L. CASSIDAY

for AND, OR, NAND, NOR, and NOT logic circuits. NAND is, of course, AND with an inverted output and NOR is OR with an inverted output. The student may be puzzled by NAND and NOR circuits, but they are, in fact, more generally useful than AND and OR circuits. It can be proven that any static truth table can be realized with only NOR or NAND circuits. With standard ‘ITL a logical zero can be provided by simply grounding an input,

*fify6 NAND

OR

AND

-

0

KR-g NOR

NOT

K B

ii

FIG. 7. Logic notation. Truth tables are given for AND, OR, NAND, NOR, and NOT logic. The appropriate logic symbols are placed above the tables, and the algebraicstatement of the logic is shown above the symbol.

while a single 1-kil resistor connected to the 5-Vpower supply can provide up to 27 logical 1 inputs. The student is first asked to demonstrate that the truth tables d o indeed describe the behavior of the simple integrated logic circuits. The student is then asked to construct the R-S flip-flop shown in Fig. 8a. (In it a 1 at input S , with R in the 0 state, forces Q to “O”, and then Q to 1. Then even if S drops to 0, the 1 at Q locks the state at Q = 0, Q = 1. By symmetry, a subsequent 1 to R reverses Q to 0 and Q to 1, again a stable state. That is, the circuit is bistable, remembering which input was last a 1.) Next, the student is given his first design exercise with logic circuits: he must design an EXCLUSIVE OR circuit choosing from two-input NOR, NAND, and AND, and one-input inverting circuits. (The output is a 1 if one and only one input is a 1.) One of the many possible realizations is shown in Fig. 8b.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

27 1

FIG.8. Configurations of logic elements: (a) a R-S flip-flop in which the output depends upon which input was last a logical 1; (b) one realization of an exclusive OR circuit, where the output is a logical 1 i f one. and only one, input is a logical 1.

3. M o r e Advanced Logic Circuits

After the student is familiar with simple logic circuits, he can progress to monostable multivibrators and scaling circuits. The 74121 monostable multivibrator is a versatile and dependable example. It triggers on a transition at one or more of its three inputs. The two A inputs require a 1 to 0 transition (the E input must be a 1 and the other A must not already be a 0); the E input requires a 0-to-1 transition (at least one A input must be a 0). The A inputs require a transition rate of at least 1 V/pec, whereas the E input has an internal Schmidt trigger and will trigger on a pulse rising as slowly as 1 V/sec. The pulse width can be determined by an external resistor and capacitor; the width is RC In 2. The student can become familiar with the monostable multivibrator by constructing a delay generator, a circuit commonly used with coincidence and anticoincidence devices. An input pulse triggers a monostable multivibrator whose width is the desired delay. A second monostable multivibrator is triggered on the trailing edge of the first. (The student also learns some of the niceties of triggering an oscilloscope when he is required to prove that the second pulse occurs after the specific delay.) Finally, he can construct an oscillator by connecting the monostables together to that each triggers on the trailing edge of the other. The laboratory assignment that has won most enthusiastic student approval involves counting circuits and numerical display. One simple

272

H. E. BERGESON A N D GEORGE L. CASSIDAY

assignment requires the use of a number of integrated circuits and techniques. We provide a S-V peak-to-peak 60 Hz sine wave and ask the student to count seconds and give a two-digit display of the number of‘ seconds. Since a 60-Hz sine wave is too slow to drive a TTL counting circuit, the student must first convert to a fast rise time square wave. Some students do it with the 710 voltage comparator which has a TTL compatible output; others do it by utilizing the Schmidt input to a TTL one-shot multivibrator. (Those who use the 710 have two additional problems. Because of the noise they may get a hundred or so pulses each time the slow input wave crosses the comparison voltage; this can be corrected by applying some positive feedback to make a Schmidt trigger from the comparator. The second problem is that the drive capacity is inadequate for TTL counting circuits; this can be corrected by buffering the 710 output with a TTL inverter or gate.) To divide the 60-Hz wave to a 1-sec square wave, students use parts of the 7490 decade scaler (actually a binary stage along with a divide-by-five counter) and the 7492 divide-by-twelvecounter (actually a binary counter along with a divide-by-sixcounter). The next two stages use the 7490 decade scaler. The 7490 decade scaler has four outputs whose value in binary ranges successively from 0 to 9. Such an output is called ‘‘ binary-coded decimal” or BCD. A

FIGIB D FIG.9. Seven-segment digital display. Any digit can be formed by the proper choice of segments. For example, an eight uses all segments while a seven uses segments a, b, and c.

The display is the seven-segment display shown in Fig. 9. Any digit can be formed with the proper choice of some of the seven segments. The conversion from BCD to the appropriate segments is done with a 7447 decoder. The output of the 7447 can run a light-emitting-diode seven-segment display if current-limiting resistors are placed in series with the display segments. Alternately there are available inexpensive seven-segment displays using tungsten filaments for the segments; when lit, each segment draws 8 mA, well within the capabilities of the 7447 decoder. C. Inductors, Capacitors, Transistors, and Other Vital Components After four weeks of lectures and five weeks of laboratory work the student will have designed and debugged a number of linear and digital circuits. The only explicit physics involved has been Ohm’s law. While the purist will

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

273

probably be horrified, the student is almost always delighted at the power he has to create useful circuits. Nevertheless, students must learn about inductors, capacitors, diodes, transistors, and other more exotic devices. The rest of the quarter is devoted to this end. One lecture is devoted to differential equations describing simple linear circuits with inductors and capacitors. It includes a standard treatment of rise times. Then these principles are used to derive the response of the very simple oscillator shown in Fig. 6c. If the magnitudes of the two saturation voltages are assumed to be equal, and if one ignores the transition time as the output state reverses, one can easily show that the period of oscillation (Graeme et al., 1971, Ch. 10) will be:

T=

2R,C h ( R , R2

+ 2R3)

(17)

An uncompensated operational amplifier like the 748 is best for this circuit because it will have a much shorter transition time than a compensated operational amplifier, such as the 741. (To avoid oscillations in linear amplifier circuits, the operational amplifier gain is made to fall as lfw at high frequencies. In the 741 this is accomplished with an internal compensation capacitor.) 1. Diodes

A standard treatment of diodes is given with application to clipping, clamping, and rectification. Brief mention is made of voltage regulators, but current technology makes detailed consideration obsolete. For as little as $2.50 one can now buy integrated voltage regulators which give 1-A outputs at a variety of output voltages. With only three terminals (input, output, and ground), they are very simple to use. It is important that the student be introduced to a reasonably correct description of the voltage-current characteristics of diodes. After giving such an introduction we find it useful to introduce " precise diode" circuitry (Smith, 1971,Ch. 6). Consider, for example, the circuit of Fig. 10. The inputoutput characteristics are those of an ideal diode: for positive inputs, the output is almost exactly equal to the input; for negative inputs the output is zero. The first operational amplifier compensates for the nonideal characteristics of real diodes. When the input V , is positive, point A must be at - V , for the inverting input to be at ground potential. The operational amplifier output will be more negative by the drop across diode D 2 . D1 is reversebiased and does not conduct. (The second operational amplifier simply inverts the output.) When V, is negative, the output of the first operational amplifier rises enough to cause Dl to conduct, bringing the inverting input

274

H. E. BERGESON AND GEORGE L. CASSIDAY

FIG. 10. Ideal diode circuit. The input-output characteristics are shown below the circuit diagram.

up to ground. However, Dz is reverse-biased, and the output remains at ground. By adding other current sources to the inverting input of either operational amplifier, the knee of the input-output curve can be moved arbitrarily along the horizontal axis. By combining such units, it is rather easy to produce limiters and absolute value circuits (Smith, 1971, Ch. 6).

2. Active Filters Following two standard lectures on complex impedances, Fourier analysis, and sine wave oscillators, one can introduce the very valuable topic of active filters (Graeme et al., 1971, Ch. 8), where one can filter even low frequency signals without inductors and with relatively small capacitors. Figure l l a shows one type of active filter, the multiple feedback network. With the virtual equality concept, one assumes point A is at ground potential. - V B = ilz,

V, = (il

+ i, + i3)Z2

Vo - VB = izZ4 Vo = i3Z, VB= -i3Z3

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

275

(CI VS O-

FIG. 11. Active filters: (a) shows the general form of the multiple-feedback network, (b) shows the low-pass version of the multiple-feedback network, while (c) is the high-pass version.

A low-pass filter is realized in Fig. l l b where Z, = R,, Z , = l/joC,, Z , = R 3 , 2, = R,, and Z, = l/jwC, . That this is a low frequency filter is obvious qualitatively. At low frequencies, the capacitors represent infinite impedances and the circuit is the standard inverting circuit with A = - R4/ R , . At high frequencies, the capacitors become short circuits. The input signal is attenuated as l/w at point B, and because the output is almost short-circuited to a virtual ground through C , ,the overall amplification will be proportional to l/oz.If one sets

216

H. E. BERGESON AND GEORGE L. CASSIDAY

and

and inserts these expressions into Eq. (19), he gets

5V,

H O l ~ O l

+ jcqou,, + o&

-0’

Equation (21) represents a standard second-order low-pass filter function known as the complex-conjugate, pole-pair, low-pass transfer ,function.Note that the asymptotic response is as predicted. One can also obtain band-pass and high-pass filters from the general configuration of Fig. lla. The high-pass realization is shown in Fig. llc. [Once again the asymptotic performance can be obtained by inspection. At high frequencies one can ignore the resistances, and the amplification is

At low frequencies, the signal at point B is proportional to o and the output is connected to a virtual ground through R5 giving a net gain proportional to o’.]Inserting the appropriate impedances into Eq. (19) and setting

which again has the expected asymptotic behavior. This is a standard second-order high-pass filter, the complex-conjugate, pole-pair, high-pass transfer function. The next lecture introduces the transistor in the normal way. We demonstrate its use as a switch in the circuit of Fig. 12, a sawtooth oscillator. The operational amplifier is simply an integrator. When Tl is off, current through R , flows into C1 producing a linear, negative ramp. When a given lower level is reached, the Schmidt trigger reverses, turning on Tl. If R, is very much smaller than R , , the current through R , produces a rapidly rising output. When the upper level of the Schmidt trigger is reached, it again

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

T

277

1

FIG. 12. Negative sawtooth generator. The operational amplifier is used as an integrator. When the bottom of the sawtooth is reached, the Schmidt trigger reverses, turning TI on and causing a rapid rise in the sawtooth until the upper Schmidt level is reached. It is assumed that R, 9R,.

reverses, turning Tl off and the cycle repeats. The slope of the falling ramp is - Vl /Rl C. As long as we can ignore the rising time, then the sawtooth generator can also be used as a voltage-controlled oscillator since the frequency will be proportional to the slope of the ramp which is in turn proportional to V,. The voltage-controlled oscillator can, in turn, be used for analog-to-digital conversion; if the oscillator pulses are counted for some h e d period, the count is proportional to the input voltage V,. The student is then introduced to various transistor amplifier configurations along with the h-parameters in the usual way. Field-effect transistors and the y-parameters are also given in the usual way. However, special emphasis is given to the field effect transistor as a switch and as a variable resistor. Both cases depend on the fact that for small drain-tosource voltages, the drain-source represents a relatively small or a very large resistance depending on the gate-to-source voltage. Figure 13a shows a field effect transistor in a sample-and-hold circuit. When V, is at + 15, diode D and resistor R allow the FET gate to be at the same potential as the FET source, thus keeping the drain-to-source resistance small and allowing the output to follow the input. When the FET gate is clamped at - 20 V at time thold, the output is frozen. A more sophisticated use is in the sine wave oscillator shown in Fig. 13b. ,This is a standard Wein-bridge oscillator except for the FET in the negative feedback network. As is well known, such an oscillator would work best with an ideal voltage amplification of exactly 3. With a gain of less than 3 it won’t oscillate; with a gain of more than 3 it saturates the output, generating harmonics of the fundamental frequency. With R 3 somewhat greater than twice R,, the amplification would be greater than three without the FET. As the output amplitude increases, the FET gate voltage is reduced, thus increasing the drain-to-source resistance. This lowers the overall gain. With properly chosen component values, the

278

H. E. BERGESON AND GEORGE L. CASSIDAY

p FET -INPUT v

o

OP AMP

+- -

- -20 v

I

THOLO

FIG.13. Circuits utilizing the variable resistance characteristia of field-effect transistors: in (a) the FET is used as a switch in a sample-and-hold circuit. In (b), the FET is used in a Wein-bridge oscillator. The FET resistance varies in such a way as to stabilize the gain appropriately for a distortionless sine-wave oscillator.

amplification will drop to less than 3 at output amplitudes where saturation occurs. The result is an almost distortion-free sine-wave oscillator. The course is completed with an introduction to thermistors, thyristors, phototransistors, and photon-coupled amplifiers. IV. SECONDQUARTER

We now return to a more detailed discussion of the many features of electronics that were simply glossed over during the first quarter. The first three lectures are spent on developing more precisely the notion of current and voltage sources and wave-shaping techniques using passive circuits. The development proceeds in the same fashion as presented by Professor Raphael Littauer (1965) in the first two chapters of his excellent text, Pulse Electronics. A. Voltage Sources, Current Sources, and Wave-Shaping Techniques

The idea of a voltage source or current source can be most readily visualized by asking what happens to the output of a source when connected to a load which is allowed to change. This situation is shown in Fig. 14. Most students can visualize what an ideal voltage source should do; it should preserve its output voltage V, no matter how much output current is

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

279

FIG.14. Output current vs. output voltage as the load R,, is varied.

demanded from it. In other words, the load line for a good voltage source should look like Fig. 15a. Once focused upon this aspect of its behavior, the idea of the less familiar current source is made palatable. Here we have the converse situation of a device which persists in pumping out (or extracting in) a current I,, no matter how much voltage it develops across the variable

FIG.15. Load lines for "near perfect" (a) voltage source and (b) current source.

load resistor. (A reasonably good example of a current source is a photomultiplier tube or a battery in series with a very large resistor.) The student can now readily see that ideal voltage sources have zero output impedance (R,), whereas ideal current sources have infinite output impedance. The slope of the load line thus represents R ; The student is now led quite naturally into Thevenin's theorem and Norton's theorem which can be used to simply represent all intermediate cases (sources with finite, nonzero output impedance) as shown in Fig. 16. Students are more familiar with voltage sources than with current sources, and usually devices are thought of in that light. Such thought is permissible only when the input impedance to the following stage is much larger than R , , the device's output impedance. However, there are times when one should think of a device as a current source. This is seen to be the case when the input impedance of the following stage is small compared to R o ; hence, the utility of Norton's theorem, a mode of thought that proves extremely useful when discussing the biasing of diode or transistor circuits.

(0

1

(b)

(C)

FIG. 16. Two-terminal network equivalents.

280

H. E. BERGESON A N D GEORGE L. CASSIDAY

The standard approach to the subject of waveshaping with passive networks usually consists of using differential equations to calculate output responses for various input conditions. This mathematical approach is excellent for describing the behavior of a circuit network already set up for the student but is of little value when trying to configure a new network design out of the millions of available possibilities. What is needed is a good intuitive approach to network design, and such intuition requires a “feeling” for how capacitors and inductors behave. An intuitive understanding of the behavior of capacitors and inductors follows quite naturally from Ohm’s law and the theorems of Thevenin and Norton. Ohm’s law for resistors, capacitors, and inductors is

di v2=Ldt or, alternatively, using v rather than i as the independent variable, we have

. lR

I =

Tiu

dv ic=C -

dt

iL=L

1 .

1 vdt

(25)

In particular, let us focus attention on vc and i L ; they must be continuous as long as i and v remain finite. A continuous function is essentially a constant if viewed upon a very short timescale. In other words, the voltage across a capacitor or the current through an inductor cannot change instantaneously; thus, for very short times, a capacitor may be viewed as a voltage source and an inductor as a current source. From Thevenin’s and Norton’s theorems, their output impedances are zero and infinite respectively. However, when viewed on a very long time scale, the situation reverses: a capacitor is an open circuit through which no dc current may pass while an inductor is a short circuit across which no voltage may develop. Short and long times are characterized by the only time scale a network has as its reference, namely, its own RC(L/R)internal time constant. Times much less than RC(L/R)are short; times greater than RC(L/R)are long.

ON T H E TEACHING OF ELECTRONICS TO SCIENTISTS

28 1

(b)

FIG. 17. (a) R-C low-pass and (b) R - L high-pass network.

These ideas are fully exploited in many network examples, two of which are presented in Fig. 17. In Fig. 17a the capacitor initially has u2 = 0 and since it is a short-term voltage source u2 cannot change instantaneously (if driven by a source with Jinite output impedance). Ultimately, u2 must be equal to V since the capacitor is a long-term open circuit. Thus, it charges from u2 = 0 to 0, = V on a time scale characterized by t = RC. In case (b) the current through the inductor is initially i, = 0 and since it is a short-term current source, it will preserve that value; thus u2 must suddenly jump to V (since no current can flow through R which is in series with L). Finally, u2 must equal zero since the inductor can have no dc voltage across its terminals. Thus, the current through it will be limited by R (to V / R )and u2 will go to zero exponentially with the time constant 4 R . The ideas presented here can easily be demonstrated mathematically to the demanding student, but more importantly they can be concisely tabulated into a set of simple rules which allow the student to quickly visualize how a network can be used to operate upon one wave form and change it into another. For example, on the short time scale, the low-pass RC network of Fig. 17a has operated upon an input step function to produce an output ramp, an elementary pulse wave form one order down in hierarchy from the step. On the long time scale no operation has occurred. Hence, this network is characterized as a transient-stop, dc-pass network. The L-R network of Fig. 17b is easily seen to be a transient-pass, dc-stop. In each of these cases, the resultant operation stems from the insertion of short-term short or open circuits as network shunt elements. Similar operations can be attained by the insertion of appropriate network elements in a series configuration. Obviously, these notions have their counterparts in the frequency domain. For example, many have been taught that a capacitor passes ac but blocks dc-a rather easily remembered rule. However, we find that students can visualize how things work in the time domain much more easily than in the frequency domain. Once his intuition is developed in that domain, the switch to frequency space is a somewhat simpler task. Indeed, the switch is

282

H. E. BERGESON A N D GEORGE L. CASSIDAY

crucial if a student is to understand linear amplifiers, gain vs. frequency behavior, and stabilization techniques when using feedback. Finally, to complete this section, we introduce the delay line and discuss the subject of signal transmission, cable termination, clipping, and pulse generation techniques. Again, an intuitive approach as outlined by Littauer (1965, Chs. 1 and 2) is followed. B. Transistors Lectures 4-8 are devoted to a discussion of the transistor. Many excellent texts too numerous to mention have been written on this subject. We continue to follow Littauer (1965, Chs. 5 and 6).Our intent is to establish an intuitive understanding of how transistors work in order that a student might easily utilize them in design applications. Our discussion is divided into three parts: (1) concepts of transistor operation and 1 / 0 characteristics, (2) biasing techniques, and (3) applications. 1. Concepts of Transistor Operation and 110 Characteristics

Designing transistor circuits is much more difficult for the beginning electronics student than designing op amp circuits, even though op amps are made of transistors. Circuit elements which live in the emitter region of the transistor are not isolated from circuit elements which live in the base region and each of those regions is a relatively low impedance environment. This statement is an upshot of the simple fact that the base-emitter junction of a transistor is a forward-biased diode which makes it a relatively low input impedance device. This is a point which must be emphasized repeatedly when discussing almost any transistor circuit and it serves as a good beginning point for discussing basic transistor operation. Shown in Fig. 18 is a schematic of the basic transistor along with a set of arrows denoting the direction of normal current flow. The base-emitter junction is a forward-biased diode while the base-collector junction is a reverse-biased diode. This statement establishes the logic behind the configuration depicted in Fig. 18. In either case (npn or p n p ) the terminals

FIG.18. Schematic view of basic (a) npn and (b) pnp transistor and direction of normal current flow.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

283

located higher in the figure must have higher polarities than terminals located at lower positions in order to forward-bias the base-emitter region and reverse-bias the base-collector region. Thus, current flow will always be downhill. This is an excellent convention which makes transistor circuit operation easy to visualize. The current amplification is depicted by the relative size of the arrows representing current flow. Typically, potential differences for the forward-biased base-emitter junction are of the order of several hundred millivolts while those for the reverse-biased base-collector junction are about several volts. Thus, a strong potential gradient must exist in the base-collector region and any charge injected into the base region from the emitter has a high probability of being gobbled up by the collector and never making it out through the base. If that probability is as high as 99%, then for every unit charge which manages to travel from the emitter to the base there will be 99 such charges which get swept up by the collector. This establishes the only parameter of real interest to the transistor circuit designer, namely the current amplification factor p. Thus If E P I b . Unfortunately, p is a very mercurial parameter, a fact which must be emphasized to students. That it can vary from 10 to 500 from one transistor to another is very disconcerting to the beginning student to say the least. However, this initially traumatic realization is mitigated somewhat upon learning that transistor circuits can (and must) be made to operate more or less independently of the value of jl. The basic points of emphasis can then be summarized; looking from either the base to emitter or emitter to base we see a forward biased diode. Hence, impedances are expected to be small and biasing must be established by current control. (Trying to voltage control a diode in the forward direction is like trying to thread a needle with a 10-ft thread-by holding the far end of the thread.) In particular, the output of the emitter is likely to be an excellent voltage source. Secondly, the collector-base is a reverse-biased diode. Its biasing should be established by voltage control. Collector circuit elements are likely to be well isolated from base circuit elements and the output of the collector is likely to be an excellent current source. These remarks are summarized in Fig. 19 and lead naturally into a discussion of biasing.

FIG.19. Schematics for depicting base-emitter current control and base-collector voltage control.

284

H. E. BERGESON A N D GEORGE L. CASSIDAY

2. Biasing Techniques The simplest way to operate a transistor is as a switch between cutoff and saturation. A simple example, shown in Fig. 20, suffices to explain those two regions of operation. When the input voltage ui = 0, then I b = 0 and since I , = g l , ,then I, = 0 and uo = + 5 V. The transistor is cut off, i.e., no currents flow through it and the full collector-emitter supply voltage develops across its terminals. Establishing this condition simply requires setting I b = 0 or turning off the forward bias on the base-emitter junction. Conversely, by jamming so much current into the base input such that the collector cannot fulfill its requirement that I, = /?I,, usually because of load resistor and supply voltage limitations, a short circuit condition is established in the transistor, i.e., the transistor becomes saturated. Thus, the transistor is an effective switch between open and short circuit, activated by turning off all input current or injecting large amounts.

+ FIG.20. A transistor switch from cutoff to saturation.

Biasing transistors into saturation or cutoff is fairly straightforward. Setting them up at some desired operating point in the active region is a little more difficult. Here we emphasize to the student that a knowledge of the location of the desired operating point comes first and that location is determined by the application in mind. We emphasize two excellent biasing techniques, emitter and voltage-feedback bias; each offers excellent /I independence and long-term stability. A discussion of emitter bias is particularly fruitful at this time since it leads quite naturally to the notion of impedance reflection from the emitter to the base region and vice versa. Establishing an operating point with voltage-feedback bias can be done by an iteration technique which serves to enhance one's intuitive grasp of transistor operation. Furthermore, it hints at an important concept soon to follow-negative feedback. a. Emitter Bias. Shown in Fig. 21 is a schematic which represents the effect of placing a resistor in the emitter portion of a transistor circuit. Two crucial facts, once understood, lead one to an understanding of biasing with emitter resistors: (1) When the transistor is turned on there is about 0.6 V across the forward-biased base-emitter diode. Thus, the emitter potential

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

285

Ve.Vb- 0.6 Vl

le=BIb

I

Re

FIG.21. (a) Emitter bias along with effective (b) input as seen from the base and (c) output as seen from the emitter.

more or less follows the base potential, and that junction can be thought of as a 0.6-V voltage source. (2) The current out of the emitter is times the current into the base. Thus, the equivalent input as seen by the base is as shown in Fig. 21b while the equivalent output seen from the emitter is as shown in Fig. 21c. Impedances in the emitter as seen from the base are multiplied by fl while impedances in the base as seen from the emitter are divided by p. Furthermore, step one in setting dc quiescent current in the transistor is carried out by establishing a well-defined dc potential V, at the base. Step two involves picking an emitter resistor R,. The dc standing current will then be

The collector potential is then determined by picking a collector supply voltage V,, and load resistor R,

V, 2 V',

Z,R, (27) The resulting quiescent operating point will be seen to be virtually independent of p as long as the emitter resistor R e is large enough such that /IminR e > R,. A completed example of this technique is shown in Fig. 22. b. Voltage-Feedback Bias. Shown in Fig. 23 is an example of voltage feedback bias. The point here is that a piece of the output voltage is "fed -

FIG.22. Example using emitter bias.

286

H. E. BERGESON AND GEORGE L. CASSIDAY 12

v,p=100

v,p=ro

FIG.23. Example using voltage-feedback bias.

back” to the input via the collector-emitter resistor R, . An increased voltage at the collector increases the input base current, thus turning the transistor on even harder and consequently lowering the collector voltage. In other words, this feedback resistor “ regulates ” or stabilizes the transistor against long-term drifts. Calculating the operating point for the transistor in Fig. 23 can be done in an exact way, but the result is a rather nasty equation. A little intuition goes a long way. If p is large ( z100)then the base current will be small compared to the standing current in the R, - Rb base biasing network. Consequently, the voltage across R, should be close to the voltage across Rb ,about 6.6 V. Hence V, 2: 7.2 V. But suppose p is small or comparable to the ratio R,/R,. Then, more current must flow through R , than through R, in order to provide base current to the transistor. Clearly, V, > 7.2 V. If we totally ignore all transistor currents for the moment, then only 660 pA flows through RL and we would conclude that V, = 11.34 V. Clearly, the correct value must lie within these bounds 7.2 < V , < 11.3 V. We can ascertain this value by arbitrarily raising V,, injecting additional input base current into the transistor, thus turning it on harder, which lowers V, and moves those bounds closer together until they converge. For example, we can easily see that if p = 10, increasing the lower bound by 1 V at V, increases the base current by 100 pA and the standing collector current by 1 mA, thus dropping the upper bound by 1 V. Thus, the correct value of V , must be about 9.2 V. Thus, lowering the p of the transistor by a factor of 10 induces only about a 25 % change in the operating point. 3. Applications The three common amplifier configurations (grounded-emitter, grounded-base, and grounded-collector) are presented in some detail, with emphasis on the typical uses of each. Treatment is fairly standard. Shown in Fig. 24 is an application of the (a) grounded-base and (b) groundedcollector amplifiers. In case (a) the amplifier is designed to accept positivegoing pulses from a 5 1 12 source and produce an output with a gain of about 4.5 and limited to about 5 V. In case (b) the emitter-follower is designed to

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS ‘6

-6

I .‘-

I220

J

287

-6

-6 (a I

(bl

FIG.24. (a) Grounded-base and (b) grounded-collector pulse amplifiers

drive positive pulses through a low-impedance load (say 51 Q) with a gain of 50.9. The action of each is easy to visualize. The next circuit discussed is the emitter-coupled pair as shown in Fig. 25. This circuit is quite typically the input stage for most operational amplifiers, a point brought out in discussion. The physical reason for the existence of things glossed over in the first quarter like input offset voltages and bias currents are now clarified. We emphasize that the difference amplifier is essentially a current switching device whose output is basically proportional to the difference between the two base inputs. Biasing is established by a current source looking into the low-impedance input emitters.

FIG.25. (a) Emitter-coupled pair; (b) using an emitter supply resistor as the current source.

The gain of the difference amplifier is set by the bias current in essentially the same fashion as for a common emitter amplifier. There are, however, two important differences: (1) the bias current I is shared more or less equally between two transistors T1 and T2; hence, for a given bias current the transconductance gm of the output transistor T2 is half that of a singletransistor common emitter amplifier. (2) The input voltage u1 is shared equally between the base emitter junctions of T1 and T2. In other words, T2 “loads down by about a factor of two” the output of T1. Hence, the baseemitter juntion voltage of T1 increases by u1 /2 while that of T2 decreases by a similar value. Thus, the current switched on through T1 comes at the expense of a corresponding decrease in current through T2 and the value of

288

H. E. BERGESON AND GEORGE L. CASSIDAY

that current is only half as much as what might be expected for a given input signal ul. Therefore, for a given load resistor RL and bias current I, the gain of the difference amplifier is about four times less than that of a simple common-emitter amplifier. An actual derivation of the gain yields

where V# is the scale voltage of a silicon junction diode. At room temperature, V: % 50 mV. A detailed discussion of this amplifier proves especially useful. Not only does it enhance understanding of op amps but also it sets the stage for a discussion of the ECL integrated circuit family as well as the Operational Transconductance Amplifier (OTA) which comes a little later in the course.

C. Feedback Lectures 9-12 deal with negative feedback. We begin with a formal treatment of transfer functions, the complex frequency plane, and poles and zeros in order to better understand high-frequency, negative-feedback amplifiers and stability problems (Littauer, 1965, Chs. 10 and 11; Millman and Halkias, 1967; Angelo, 1964). As usual our ultimate goal is to engender an intuitive feeling, in this case for the use of negative feedback both in design as well as stability control situations. In particular, when designing high-gain, high-frequency amplifiers (or simply when trying to use an integrated circuit version such as the 733 video amplifier) one finds to his sorrow that they always oscillate. Usually, the judicious deployment of a capacitor can kill the undesired oscillation, unfortunately at the expense of a loss in high frequency response. Such expertise usually comes with experience. We foist that experience upon the unwitting student in lab session 9. The gain of 100 negative-feedback amplifier almost invariably oscillates Shown in Fig. 26 are two applications using negative feedback. The first amplifier is a fast amplifier designed to be driven by a current source. It makes an excellent photomultiplier preamplifier, for example. The quiescent operating point is rather easy to ascertain. The potential at the collector of T1 must be about 0.6 V less than the potential at the emitter of T2. T1 must have more standing current than T2 and it must flow through R F and Rc. If T2 were off then the dc potential at the collector of T2 would be about - 7.5 V. If the current through T2 were almost equal to that through T1 (the most it could be) then that potential would be about - 5 V. Hence, - 5 < u1 < -7.5 V, and the current through T1 and T2 should each be of the order of 1 mA. The output signal voltage o1 can easily be shown to be ui = Ii,RF assuming a large loop gain for the amplifier. Representing the

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

289

f

Lo)

Lbl

FIG.26. (a) Fast current-to-voltage negative-feedback amplifier; (b) stacked-up emitterfollower driver stage.

dynamic input resistance at the emitter of T1 by re the voltage feedback factor is seen to be

But the input voltage at the emitter of T1 is Therefore, for large loop gains RF = I i , RF re That the loop gain is large can be seen by breaking the loop at the collector of T1 and the base of T2 and injecting a unit current into the base of T2.We then go around the loop and see how much current emerges from T1.This value is the loop gain - M and is seen to be u1

Uin

= - = linre-

f

which can easily be made to be greater than 100 for large B transistors. Clearly this amplifier can be used as a voltage amplifier by using the base for the signal input and then capacitively coupling R i n to ground. We would then have a voltage amplifier with a gain of RF/Rin = 100. Circuit (b) is a stacked-up emitter follower. It is highly recommended as a driver for low-impedance loads like coaxial cables. It is better than a simple emitter follower since it will actively drive both the trailing as well as the leading edge of a pulse. In this case the feedback occurs via the capacitor CF.The loop gain is essentially /?.Hence, its output impedance is that of an emitter follower reduced by the factor /I-no more than a few ohms. We

290

H. E. BERGESON A N D GEORGE L. CASSIDAY

have not shown a biasing network at the input base of T3. One very nice application involves directly connecting the output u 1 of amplifier (a) to the input u2 of amplifier (b). Furthermore, the 4.7-kSZ feedback resistor RF can now be attached to the output of the driver, thus nullifying the small voltage loss of the output stage. When the student hooks these two circuits together, it makes a great exercise trying to find a simple way to kill the resulting oscillations. In lecture 13 we introduce positive feedback. Each student is required to build a discriminator using the 733 differential video amplifier (Linear Integrated Circuits, 1976). It has a differential input stage and complementary polarity outputs. With a small RC positive-feedback network as shown in Fig. 27, it makes an ideal fast discriminator. Biasing is established with an appropriate resistor divider at the inputs. "in

pvo"i

FIG.27. Discriminator using the pA733 differential video amplifier.

D. Digital Circuits We begin this discussion with a fairly detailed analysis of the TTL, ECL, and CMOS circuit families. The virtues and drawbacks of each one are emphasized. Essentially TTL is cheap, widespread, and versatile; ECL is superfast but has poor noise immunity and consumes power; CMOS is slow but offers good noise immunity and low power consumption. CMOS is inexpensive and certainly on the upswing in usage and more and more CMOS devices are becoming available. Unfortunately, its speed limitations may never be overcome and consequently CMOS may never penetrate the highspeed circuit market. However, as far as teaching electronics goes, it should be given serious consideration by any instructor when outfitting a new laboratory. Lectures 17 and 18 are spent on digital computation circuits. First, we discuss the exclusive OR and the half, and full-adder circuits as in Malmstadt et al. (1973).We then discuss multipliers and counters. Our purpose is to introduce the student to typical circuits used in the processors of most minicomputers. Shown in Fig. 28 is an example of a hardware multiplier (Leon and Bass, 1974). It is an excellent example, for in order to see how it works a student must multiply two binary numbers together long hand and then examine exactly what he has done to see how the hardware works.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

29 1

M'3-0

I

k

Stort

Clock Timing

Frc. 28. Parallel hardware multiplier

Furthermore, when a computer multiplies two binary numbers together by software, the algorithm necessary to carry out the operation bears a simple one-to-one correspondence to the longhand multiplication. The operation is fairly easy to visualize if one first examines the details of the actual multiplication of two numbers. Note that one simply shifts the multiplicand and then adds it or a zero to an accumulated result depending upon the status of the multiplier bit. If the multiplier bit is a 1, the shifted multiplicand is added; otherwise, add zero. The hardware works in the following way. M1: M2 :

1010 1 101

(10) (13) (Add multiplicand) (Shift; add zero)

1010 0000 01 010 101 0 (Shift; add multiplicand) 110 010 1010 (Shift; add multiplicand) 10 OOO 010 (130) The multiplicand is presented to the B input of a 74181 which will serve as our adder circuit. It will add either the multiplicand or a zero to the accumulated result (contained at the A input) depending upon the condition of the select bits S 3 - S o . This condition is properly established by examination of the appropriate multiplier bit. Here the multiplier M2 is presented to the parallel load input of a 7495 register. M2 is shifted successively to the right by the clock pulse, and thus whenever a " 1 " appears, the select bits will be set to add the multiplicand to the accumulated result; otherwise, a zero is added.

292

H. E. BERGESON AND GEORGE L. CASSIDAY

The output of the 74181 adder is split; the least significant bit goes to the serial input of the 7495 while the three most significant bits are passed to the three least significant bits of the parallel input 74195 register. Thus, the accumulated result is effectively shifted one place to the right at each clock pulse. (This is equivalent to left-shifting the multiplicand.) This shifted output of the 74195 register goes to the A input of the 74181 adder as the accumulated result. At each clock pulse a “shift and add” thus occurs. Note that the LSB of the accumulated result is shifted into the serial input of the 7495 also on each clock pulse. Thus, as the multiplier bits are shifted out, the lower four bits of the resultant product are drawn in. At the end of four clock pulses the operation is completed and the upper four bits of the result appear at the parallel output of the 74195 while the lower four bits appear at the parallel output of the 7495. This example must be implemented in lab session 11. A hidden virtue of this exercise concerns inputting the multiplicand and multiplier and outputting the products as well as setting up the start and clock pulses. The student is immediately confronted with these difficulties and is forced to solve them in his own way in order to test his circuit. It is an excellent introduction to digital systems design. E. Advanced Linear Devices and Techniques This section is devoted to advanced applications involving two linear devices (1) the RCA 3080 Operational Transconductance Amplifier (OTA)* and (2) the Signetics NE565 phase-lock-loop (PLL) (Linear, Digital, MOS, 1973). We choose these devices since they can be used in a simple way to introduce the student to such techniques as sample and hold, amplitude modulation, analog multiplexing, frequency modulation, phase detection, and voltage-controlled oscillators (VCO). It further reinforces one of the primary thrusts of any modern electronics course; namely, if one looks around hard enough, one can probably find some electronics device that will closely approximate your desired circuit. A useful rule for the student is: “ Don’t try to build your circuit from scratch unless this proves to be your only alternative.” 1. The RCA 3080 Operational Transconductance AmpliJer

Shown in Fig. 29 is an equivalent circuit and simplified diagram for the OTA. Essentially the OTA is an operational amplifier. However, its output impedance is very large (typically about 10-100 MR,depending upon bias-

* RCA Specification Sheet and Application Note ICAN-6668, available from RCA Solid State, Box 3200, Route 202, Somerville, N. J. 08876.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

293

I

Amplifier Current

‘bias

V

‘bias (01

(b)

FIG.29. (a) Equivalent circuit and (b) simplified diagram for the OTA.

ing) making it a marvelous current source. Secondly, it is much more versatile than a simple op amp due to the existence of a third control input which allows the user to establish its transconductance and, hence, its gain. The operation of the OTA can easily be understood by looking at Fig. 29b. W, X, Y, and Z are current “mirrors” namely, whatever currenl is pulled down through one leg of the mirror will be pulled down through the other. In the case of current mirror W, terminal 5 is used to establish a current down through one leg of W. W will automatically extract that same current down through its other leg, thus establishing some desired bias current down through the “tail” of the input differential amplifier stage. Hence the “gain *’ of the device (in this case its transconductance) is controlled by the input bias current in the same way as for any differential amplifier. The crucial point here, of course, is that the transconductance can be varied by changing the bias current. The output current mirrors Y and Z simply act as push-pull current sources for the load. Marvelous uses for this device immediately come to mind; for example, an input analog signal can be switched on or off simply by turning on or off the input bias current. Hence, analog multiplexing as shown in Fig. 30. Either channel 1 or channel 2 is selected, depending on whether the Q or Q output of the control F / F is high. Note that the outputs of the OTA can be tied together directly. Remember they are current sources. Amplitude modulation is also trivial to implement, as shown in Fig. 3 1. In this case the bias current is adjusted up or down by the modulating frequency input, thus modulating the OTA’s “gain ” and the amplified carrier frequency. The technique of sample and hold is also trivial to implement with the OTA. Furthermore, the OTA’s output is a good current source and the built-in control circuit at terminal 5 facilitates sample and hold control circuitry. Shown in Fig. 32 is an example of a sample and hold. The output FET source follower is in the OTA’s feedback loop, thus converting

294

H. E. BERGESON A N D GEORGE L. CASSIDAY

10K

Channel I Input

Charnel Input

1

Clock Inpul

' @o

FIG.30. Two-channel linear multiplex system using the OTA

the OTA to a simple voltage follower when gated on (0 V) at terminal 5. When gated off (- 15 V) the output voltage is held by the 300 pf storage capacitor. This capacitor cannot discharge since it looks out into two very large impedances (the input of the FET and the output of the OTA). Thus, the " held " output at the FET will persist until the control is returned to 0 V, at which time the output will again track the input.

Frequency

b'9mVX

-6V 47 K

"'

*@

Modulating Freauancv

1

FIG.31. Amplitude modulation using the OTA.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

295

-15V

FIG.32. Sample and hold using the OTA.

2. The Signerics NES65 Phase-Locked Loop The phase-locked loop (PLL) is basically a negative-feedback system and as such is characterized more or less by the same equations and analysis that govern any negative-feedback system. A very comprehensive treatment of the PLL can be found in the Signetics Linear Applications Notes (Linear, Digital, Mos, 1973) and we present here a brief version of their discussion. Shown in Fig. 33 is a block diagram of the PLL. Essentially the PLL is a device that will ‘‘ lock itself” to the frequency of some input signal if that frequency lies within the limits +Aw about some well-defined “freerunning” frequency oo. The use of such a device will become apparent in a moment. In Fig. 33 we examine what happens to some input signal V , , wi as the input frequency is slowly increased from zero. The VCO (voltage controlled oscillator) will be operating at its free-running frequency wo . The phase detector compares the phase and frequency of the VCO with the input frequency and will generate an “error voltage V, that is related to the phase and frequency difference between the two signals. This error voltage is then filtered, amplified, and applied to the control input of the VCO. The VCO frequency will then shift in a direction that minimizes the difference frequency wi - oo. When the input frequency oiis sufficiently close to wo, ”

‘I=%

w -w I

0

FIG.33. Block diagram of phase-locked loop.

296

H. E. BERGESON A N D GEORGE L. CASSIDAY

the feedback nature of the PLL will cause the VCO to lock to the incoming signal. Once in lock, the VCO frequency is identical to the input signal's frequency except there is a finite phase difference 4 between the two. This net phase difference is necessary to generate the corrective VCO control voltage V, which shifts the VCOs frequency from wo to m i . Once in lock, the PLL will thus track the input frequency as long as it remains within +Am of wo (the lock range). The phase comparator is actually a multiplier circuit that mixes the input signal mi with the VCO signal, producing the sum and difference frequencies wi k wo as shown in Fig. 33. When in lock, wi- oo= 0; hence the output of the phase comparator contains a dc component. The low-pass filter removes the sum component (mi + coo),passes and amplifies the dc component to the VCO control. Additional details of the PLL can be found in the Signetics Applications Notes. We have found it useful to delve into the inner working of the PLL in great detail, thus unveiling its "black-box" shroud. In particular, we analyze the VCO, the phase comparator, and the low pass filter, transistor by transistor (we also perform a similar analysis of the RCA 3080 OTA), and as each layer of mystery is lifted, each student begins to really appreciate, perhaps for the first time in the course, the enormity of the expertise he has now acquired. LOw-Rss Filler

O '

Amplifier

I + 5V

I-" n=2

I

1

1 ( b)

FIG.34. (a) Block diagram of frequency synthesizer; (b) implementation of frequency synthesizer using the NE565.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

297

The first exercise which our students encounter with the PLL is the frequency synthesizer, shown in Fig. 34. If the loop between the VCO and the phase comparator is broken and a e n counter inserted, the VCO frequency will equal noi when lock is achieved. Hence, a desired frequency may be externally programmed ” by externally controlling the modulus n of the t n counter. The synthesizer can easily be implemented as shown in Fig. 34b where an 8281 is used as the + n counter. In this case n can be chosen to be 2,4,8, or 16 by gating either the A, B, C, or D output back to the phase comparator input of the NE565. It is up to the student to figure out how to do the gating. It is a short step to set up such a circuit under computer control. A second example involves frequency shift keying (FSK) decoding. FSK refers to data transmission by means of a carrier which is shifted between two preset frequencies (1070 Hz and 1270 Hz for the NE565). This shift is usually accomplished by driving a VCO with the binary data signal so that the two resulting frequencies correspond to the 0 and 1 of the binary data signal. “

FIG.35. FSK decoder using the NE565.

A simple scheme to implement FSK decoding is shown in Fig. 35. The NE565 locks to each of the two input frequencies (1070 and 1270 Hz) and produces different corresponding dc outputs at terminal 7. (It is the dc output in each case which controls the PLLs internal VCO.) This output is heavily filtered and input to a 710 comparator whose output state toggles from 0 to 1 as the input frequency shifts from 1070 to 1270 Hz, respectively. The maximum allowable shift rate is 300 baud, a term which will become all too familiar to each student when we discuss data transmission between a minicomputer and some external peripheral during the third quarter. In order to get this circuit to work, not only must each student build the decoder itself, but he must also build the appropriate signal source necessary to “key in ” 1070 and 1270 Hz at 300 baud. Again, there is a little more to it than first meets the eye.

298

H. E. BERGESON A N D GEORGE L. CASSIDAY

F. Digital-to- Analog and Analog-to-Digital Techniques The last set of lectures is devoted to a discussion of digital-to-analog (D/A) and analog-to-digital (A/D) techniques. This particular choice for closing out the quarter is designed to launch students into the third quarter which deals with the use of the minicomputer in data acquisition and control. Obviously, external analog signals must be converted to digital form if they are to be processed by a minicomputer and conversely digital information contained in a minicomputer must be converted into analog form if external analog signal processing devices are to be controlled. Inexpensive D/A converters are now available as 14-or 16-pin integrated circuits. We use Motorola’s MC 1408 8-bit D/A, which sells for about $4.00 (Semiconductor Data Lbrary, 1975, Sect. 8-209). Shown in Fig. 36 is a block Digital In MSB R~~~~ cmtrol I

5

LSB

6

7

9

8

10

I1

12 1-10

Current Switches

I

Vref

R 2 R Ladder

1I !

(*I ,

Bias Circuit

GND

I Reference Currenl Amplifier

I

.

p2

E

Camwn

VEE*3

~

NPN current Source Pair

FIG.36. Block diagram of the Motorola MC 1408 D/A converter.

diagram of this D/A. Again, we analyze it transistor by transistor in order to understand the details of its operation. Essentially, an R-2R ladder network divides a known reference current into binary-weighted components which can be switched to the output by a set of current switches. Hence, the output current is determined solely by the input digital word. Shown in Fig. 37 is an example of a programmable pulse generator which each student must implement in lab session 16. Programmed output voltages are obtained by using the D/A to establish appropriate bias currents in the tail of the long-tailed pair output-difference amplifier. A TTL pulse is fed into transistor Q5,thus switching the programmed bias current through Q4 and its 50-R load. Again, it is a simple step to operate this programmable pulse generator under computer control.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

FLL

Amplitude Input

"ref

1-

299

9 I%d' d

12?I

._

9?

7? 67 5

MC1508L8 MC1408L

/-

TTL Input

0 to I Volt

+5

*I5

FIG. 37. Programmable pulse generator using the MC 1408.

There are many ways to digitize signals nowadays. One merely decides how fast a signal must be digitized and what accuracy is desired, and then one purchases the appropriate D/A converter. Many vendors exist who sell marvelous A/Ds, typically as hybrid packages although some are even available in integrated circuit form such as the Motorola MC 1405 (Semiconductor Data Library, 1975, Sect. 8-166). Even with this variety of available possibilities we still think it wise to discuss the different digitization techniques commonly utilized in commercial converters. Eventually most of our students will probably be faced with a decision concerning which and whose converter they should buy, and forearmed is forewarned. We begin our discussion with the dual-ramp technique. It is slow, inexpensive to implement, and easily capable of 12-bit accuracy. Motorola used this technique in its MC 1405. It makes an ideal converter for building a 3)- or 4gdigit digital voltmeter. Our discussion is based on that given in Malmstadt and Enke (1969). Shown in Fig. 38 is a schematic of the dual ramp technique. The input signal V, is integrated for a fixed time t ,

Clock

FIG.

38. Dual-ramp A/D converter

300

H. E. BERGESON AND GEORGE L. CASSIDAY

determined by the maximum number of counts possible in a digital counter. At the end of this time, the digital counter resets and toggles a control flip-flop thus gating a known negative reference voltage into the input integrator instead of the unknown signal. The integrator then ramps down toward zero with a fixed slope determined by the reference voltage. Clearly the time t , is takes to reach zero depends upon the peak value reached by the integrator, which in turn depends upon the value of the input voltage V, . In other words, the integrator output voltage obtained by integrating the unknown input for a time t l is dissipated by integrating a known reference voltage for a time t , . Since charge is conserved, we obtain

Moreover, t , is proportional to the modulus M of the digital counter while t2 is proportional to the final number of counts N stored in the counter. Thus

Also note that the count ratio N / M is independent of the clock frequency. Furthermore, since the measurement is based upon integrals, it is fairly insensitive to noise. One must simply take care to preserve the input signal for a time long enough to perform the digitization. A second A/D converter, somewhat faster but also somewhat more expensive and difficultto implement,is the staircase converter shown in Fig. 39. We require our students to build one of these in lab session 17. The idea behind this converter is fairly straightforward; the input signal is fed into a comparator whose other input is driven by a D / A converter. The D / A converter output is toggled upwards by a counter in a staircase fashion until it crosses

Start

Control

Counter

Coincidence Converter

-

Step height equivolent lo 18061 siqnificoni digit

FIG.39. Staircase A/D converter.

Reset

Display

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

30 1

the input voltage at which point the comparator switches and halts the process. The digitized signal voltage is then automatically found in the counter which was used to toggle the D/A converter. Additional A/D converters discussed include the successive approximation and parallel comparator input array converters. These latter techniques are each faster than the previous two. The parallel comparator array is the fastest of all but very limited in accuracy and is of real use only if digitization rates in excess of a MHz or so are required. The successive approximation technique is perhaps the best of the lot in terms of accuracy and speed but it is somewhat more difficult to implement than the others. During the third quarter, lab session 11 involves a clever implementation of this particular converter with a minicomputer. We save its discussion for that time. This completes our second quarter of the sequence. By now each student has acquired quite an arsenal of electronics knowledge. They have successfully designed about 35 circuits, some of them involving a relatively complex array of electronics devices. Furthermore, much of their expertise has been engendered with a certain malice of forethought; namely, that nowadays computers is the name of the game. During the course of their future careers most of them will be forced to interact one way or another with computers, minicomputers or microprocessors. Electronics now becomes an interesting blend of hardware and computer software. Ultimately, these students will confidently tackle problems requiring knowledge of each of these areas. This next quarter is designed to extend their knowledge in those areas arld further build the confidence they require to successfully tackle research problems involving the use of computers.

V. THIRDQUARTER In order to control an experiment or some electronics hardware operation by computer, a student first must gain at least an elementary understanding of just what a computer is and how it works. Hence, as an introduction to this subject we undertake the seemingly formidable task of designing a digital computer (albeit a very rudimentary one) from the ground floor on up-gate by gate. This proves to be much simpler than one might first imagine and is well worth the effort. Each student quickly sees that computer operation is indeed quite simple and secondly that the details of computer architecture are intimately related to the instruction set that one wishes a computer to implement. Upon completion of this exercise we delve into the workings of a very specific microcomputer, Digital Equipment Corporation’s LSI-11 (Processor Handbook, 1975). We discuss its basic instruction set and several of its available 1/0 peripherals. We present many

302

H. E. BERGESON A N D GEORGE L. CASSIDAY

illustrative programming examples primarily directed towards 1/0 operations. Finally, we present several specific 1/0 applications in computer control and data acquisition. A. Design of a Simple Computer

Shown in Fig. 40 is a simple schematic of an elementary digital computer. The box labeled memory is essentially an array of n-bit cells (for the LSI-11n = 16; for many microprocessors such as the Intel 8080, n = 8) used for storing information in the form of binary words. The central processor is a box which fetches pieces of information (words) from memory, operates upon those words in some way and then stuffs the processed information back into memory. Somewhere we must have some 1/0 ports attached in order than an observer might view the processed information. A

F ~ G40. . Basic digital computer.

computer program is nothing more than a well-defined algorithm coded in some binary form and stored in memory. The algorithm is designed to process data in some way. The function of the processor is to fetch that program (and any data) stored in memory, decode it, carry out the operations upon the data as required by the algorithm and obtain a final result. This result usually is output to an external observer via some 1/0 port attached either to some designated place in the processor or memory. Consider the following operation: 1 + 1 = 2. What must a computer d o to execute such an operation? The “Add ’’ algorithm must be properly coded and stored somewhere in memory. A flow diagram and mnemonic for this coded algorithm might look like that shown in Table I. The computer is told to start executing the program by pressing a start switch. The processor then fetches a word from the first memory location, decodes that word and if it is an instruction the processor executes it accordingly. If that word is a data word, the processor usually deposits it at some temporary location where it can conveniently operate upon it. The processor continues down through the coded program until it encounters a HALT condition, at which point it does just that. This, in general, is how a computer works. Obviously, the above is a very crude explanation, but it does provide a general overview of computer operation-even though many questions have

ON THE TEACHING 01.LLECTRONICS TO SCIENTISTS

303

TABLE I Fi o h DIAGKAM 4ND MVMONIC I O K ADD O P E K A T I O ~ Memory location

Mnemonic code

Operation

START

0

STA

1

1

2

STH

3

1

4

ADD

Store following word in accumulator Data word 1 Store following word in B register Data word 1

B. A

5

w RT

6

HALT

I Display result I

Add the contents of the B register to the contents of the accumulator and leave the result in the accumulator Display contents of accumulator

Stop execution

been conveniently swept under the rug. The trick in teaching about computers is to present a continuing barrage of “overview” each of which is designed to peel away increasingly stickier levels of difficulties. In this way the student who is building a road through a swamp and currently finds himself fighting alligators does not forget why he is there in the first place. To continue then we might request a little more detail concerning the Add operation just discussed. For example, we show in Fig. 41 a more detailed picture of memory and the processor. The best way to understand this picture is by working through the example of the ADD operation again. Let us assume we have loaded the program (binary coded algorithm for ADD as in Table I ) into the first seven memory locations. We then press the R U N button. The first word in memory is loaded into the Data Out (DO) register. If that word is an instruction as determined by a decoder located in the control unit, the word is loaded into the instruction register. In this case it is an instruction (STA: store the next word in the accumulator). The

304

H. E. BERGESON A N D GEORGE L. CASSIDAY

control unit will then decode that instruction and generate the necessary sequence of pulses to carry it out; fetch the data word “1” located in memory location ‘‘ 1,” and move it into the accumulator contained in the arithmetic unit. What happens next? It is the task of a special register called the program counter (PC) to keep track of where we are in memory. In this example, after executing the first instruction, the contents of the program counter has been incremented by 2. Since, at the start of program execution, the PC was set to 0, its new value is 2, which is where the next instruction resides in memory (the STB instruction) which is to be executed next. Hence, the control unit will continue to fetch word after word from locations stored in the PC, appropriately executing an instruction or passing data to the arithmetic unit and then incrementing the PC. The process is continued until encountering HALT.

, ”i””’9 r‘I

b Ioprotion

Control

Memory Control

1

&Jump

Switches

PC Control

I

, I

Address I

I

Arith Unit

Ma nstruction

FIG.41. Basic organization of memory and control.

We have probed computer operation a little more deeply than before. This process is continued until the student understands intimately the details of the arithmetic unit, the control unit, memory and memory control, and 1/0 access. At each step along the way we ask that each student fill in many of the blanks, that is, we leave much of the hardware design to be completed by the student. For example, lab exercises 1-3 require each student to design, build and debug (1) an op code decoder necessary to decode a particular set of instructions, (2) pieces of an arithmetic unit, and (3) some 1/0 hardware necessary to display results contained in the accumulator.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

305

B. The LSI-11

Here we discuss a particular processor. We have chosen Digital Equipment Corporation’s (DEC) LSI-11 for reasons primarily of past familiarity and current expertise. We feel that the particular choice of processor is somewhat irrelevant. It is relevant, and indeed it is crucial, that the potential instructor choose some processor from some vendor and get the students to use it. There is no substitute for hands-on experience. One can read and study about computer operation ad infinitum, but until a student is forced to actually sit down and put his programs into the processor and get them to execute and in particular to carry out 1/0 operations, said student simply does not attain a working knowledge of how to use a computer in control and data acquisition situations. Most of our discussion is based upon information found in DECs own literature (ProcessorHandbook, 1975; User’s Manual, 1975).Other manufacturers put out similar literature which the potential instructor would have to use in some modified way. Such literature typically includes the basic architecture of the processor in question, its basic instruction set and adressing modes, 1 / 0 peripheral interfaces, and 1/0 programming techniques in general. We urge that any instructor augment this literature (which is usually purely factual and quite jargon-laden) with a rather liberal dosage of specific examples. For example, it is virtually worthless to invest a large amount of time explaining the details of each individual computer instruction. An example containing many instructions usually will accomplish that goal and much more besides. The student will see the context for which instructions are typically used. We find such a procedure to be a more valuable learning experience. Most students have an easier time creating their own computer code by generalizing from a specific example than by starting from scratch and attempting a grand synthesis from vaguely defined building blocks they have never seen used before. Indeed, it sometimes proves difficult to visualize the use of a specific instruction by itself, but seen within the fabric of an example, its operation is made evident by the algorithm being coded. An example of this sort of thing is easily demonstrated by following through one of our own examples. Consider DEC‘s SOB (Subtract one and branch) instruction which many readers have probably never seen. It is much simpler to explain the instruction by example instead of explaining it per se. The example is the following: Write a program to multiply two numbers (m and n ) together. Do not simply add the m numbers together n times. Assume that m, n I2’ - 1. Let m reside in the low order byte (least

306

H. E. BERGESON AND GEORGE L. CASSIDAY

significant 8 bits of the 16-bit word) of register one ( R l ) and n reside in the low order byte of R2. The resultant product (16-bits) should end up in RO. The solution is shown in Table 11. A word of explanation is in order here. We left shift the multiplicand and add to the accumulated result if the current multiplier bit under examination is a one. If that bit is a zero, we d o not add anything to the accumulated result. This operation must be performed los (eight in octal notation) times since rn and n each are 8 bits long. The multiplier “bit examination” is performed by means of the RORB (rotate right byte) and BCC (branch if C bit clear) instructions. The RORB instruction shifts all multiplier bits one place to the right, loads the LSB into a special bit cell called the C bit, and loads the previous C bit into the MSB multiplier. The BCC instruction examines the C bit. If it is a “1,” the multiplicand is added to the accumulated results; otherwise the ADD instruction is skipped. R3 initially contains the number of desired operations ( los). The SOB instruction decrements R3 and tests for a resultant zero. If no zero is encountered, the algorithm is repeated. A zero in R3 indicates that the shift-add operation has been executed eight times and the multiplication is now complete. Thus, the HALT instruction (following the SOB) is finally encountered. Clearly, the SOB instruction is useful for looping-type operations. C . I/O Operations

Each student, once familiar with basic computer operation and elementary programming techniques, is then introduced to the essentials of I/O operations. Here we discuss the concepts of (1) stacks, (2) subroutines, and (3) interrupts, features common to. many small computers. 1. The Stack A stack is just what its name implies; more specifically it is a linear arrangement of words in memory with the characteristic of last in, first out” or LIFO. The last entry placed on the stack is the first one taken off-just like a stack of cafeteria plates. This type of structure makes a very convenient temporary data storage area. The location of the last entry on the stack is contained in register R6 in the LSI-11, the stack pointer register. Typically, one decides a priori upon a location in memory as the beginning of the stack and loads R6 with that value. As words are “pushed” onto the stack, R6 is modified so that it always “points” at the address of the last entry on the stack. One can easily imagine the utility of such a structure. As external data words are read in from some 1 / 0 port, they are simply pushed onto the stack as rapidly as possible where they may be retrieved later for “

TABLE I1 SOFTWARE CODE FOR MULTIPLYING TWO B I N A R Y NUMBERS The algorithm we need to implement here by software was first alluded to during our discussion of the hardware multiplier in Section IV,D. Simply left shift m, the multiplicand, one place at a time and add the resultant number to the accumulated result if the number !I, the multiplier, has binary bit value of 1 at the position of the current shift. Thus, we have Data

RO/O R l/m R2/n R3/10

location of accumulated result multiplicand multiplier 10, times, the number of shifts required for an 8-bit word Flow Chart for Algorithm Examine LSB

I+?

NO

Add m, RO t L Shift rn

I

Program

1000/106002 1002/103001 1006/006301 1010/077305 1012J000000

Start:

RORB BCC ADD ASL SOB HALT

R2 ; Rotate right multiplier. [ + 11 ; If C is 0, skip next instruction. R1, RO ; C bit was 1 so we’ll add the shifted n. R1 ; Shift the multiplier n one place left. R4, Start; Done yet?

308

H. E. BERGESON A N D GEORGE L. CASSIDAY

analysis. By convention, stacks on DEC computers grow downward in memory. An example of a stack “push” operation is given in the following: MOV # 5 , -(R6) Before the instruction

After the instruction

R6

pb

R6

776

Values already

1;:

present on the

770

1’

[xk

776

stack +

766

770 766

In order to retrieve a word from the stack, one executes a stack “pop.” In this case, the word is “popped” from the stack into register RO.

MOV (R6)+, RO Before the instruction

After the instruction

2. Subroutines A knowledge of subroutines is not necessary in order to d o 1 / 0 programming. However, there are many similarities between subroutines and interrupts, and students usually have used subroutines in high-level programming languages like FORTRAN. Because of this familiarity we discuss them first.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

309

A subroutine is a section of computer code that is typically executed several times during the course of executing a main program. The subroutine is invoked by a call at some point in the main program. Sometimes, parameters the subroutine needs for calculational purposes must be passed to it by the main program. At all times, the subroutine has to know where to “return” to the main program in order that execution be properly continued. A schematic of a subroutine “call” is shown in Fig. 42. In the situation represented here, how does the subroutine know how to return ” to the main program to location N l ? The trick is to somehow temporarily store the return location N + 1 so that the subroutine will know where to go when it finishes. As you might guess, the stack proves especially useful for accomplishing this task. “

+

Address

Time Flow

call 1 Subroutine (JSR PC. SUB) N Main Program Executing __f

7

N.1 ‘Main Program continues at N.1

RTS PC

FIG.42. Memory map and time sequence of subroutine call.

The JSR PC, SUB is the most commonly used subroutine “call” instruction. It means, jump to the subroutine labeled SUB via the PC (program counter) as the linkage register. What does this mean? Upon executing such an instruction when encountered in the main program, the location of the next instruction at line N + 1 (which is the value contained in the PC) will get pushed onto the stack and a “jump” or branch to the memory location of the subroutine SUB will occur. Since N + 1 is the location of the next instruction which would have been executed had we not encountered the JSR PC, SUB instruction, it must be the ’* return ” address. Upon encountering the RTS PC instruction (the last one in the subroutine), the processor

3 10

H. E. BERGESON A N D GEORGE L. CASSIDAY

simply ‘‘pops’’ the return address N + 1 from the stack and executes a “jump” back to that address, thus returning control to the processor at its proper position in the main program. 3. Interrupts Suppose a minicomputer is being used to monitor some external situation which under normal circumstances does not require attention. Its attention is only required when some significant event occurs. For example, several burglar alarms in a house might be attached to a minicomputer whose job is to monitor those alarms. There is no reason for the minicomputer to lie dormant, waiting for a burglar. It might as well perform other chores such as control the heating in your house. However, it had better respond when the burglar alarm goes off.Its response might take the form of notifying the local gendarmes, for example. This is the function of the interrupt. Its action is quite similar to that of a subroutine except for one crucial difference; a subroutine is invoked at a known time and place while executing a main program, but an interrupt (which activates an interrupt service routine) occurs at some random time during the execution of a main program. This situation is schematically shown in Fig. 43; HEAT is the ,

Memory Mop

j

1

Slorl End

Start End

-

-.

’

Heat

!

1 - ! ;

Time Flow

*

Interrupt Request Occurs

I

1

;. , b r d e ;

c

;4 RTI

FIG.43. Time sequence of interrupt request.

main program which is busy controlling the heating of a house when the interrupt occurs. BURGLE is the interrupt service routine which is processed as a result of the occurrence of the interrupt. Note the similarity to the subroutine schematic of Fig. 42. What information is necessary in order that an LSI-11 minicomputer properly handle an interrupt request? Somehow the processor has to know where the interrupt service routine resides in memory, i.e., its starting address. Secondly, after executing the interrupt service routine BURGLE, the LSI-11 has to be able to restore itself to its state of affairsjust prior t o the interrupt. How can it do this? Information pertaining to the status of the main program currently executing is stored in the Processor Status word (PS) while the program Counter (PC) has the current address. When

311

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

the interrupt occurs, the PS is pushed onto the stack and then the PC. This action stores the status and return address information so that the processor can resume executing the main program after servicing the interrupt. Next, a new address is loaded into the PC and a new status word is loaded into the PS from two preassigned consecutive memory locations called the interrupt vector.” For example, let us suppose that the interrupt vector for the peripheral interface connected to the burglar alarms has been assigned the value 300. Then address 300 will contain the address of the location of BURGLE, the interrupt service routine. Suppose BURGLES starting address is 1OOO. Then vector address 300 will contain the value 1OOO. The next consecutive address after 300 is 302 (DEC’s 16-bit words are stored consecutively in even memory locations). Location 302 will contain the status word appropriate for BURGLE. Hence, at the occurrence of the interrupt, the current status and address of the main program are pushed onto the stack and the processor “vectors” to the address of the interrupt service routine and its status is set in the processor status word. Upon completion of the interrupt service routine an RTI (return from interrupt) instruction is executed whereby the return address to the main program is popped from the stack into the PC and then the processor status word back into the PS. The main program resumes execution. There are some additional fine points about interrupts. When an interrupt occurs, it is quite likely that the main program is right in the midst of some operation. Hence, the general purpose registers of the LSI- 11 processor (RO-R5) are likely to be filled with words necessary to properly continue program execution. It is thus good practice to have the interrupt service routine, as its first duty, save the contents of those registers on the system stack. These registers may then be freely used by the interrupt service routine. Finally, as its last task the service routine should restore those registers by executing six stack pops. Then, after executing an RTI, the main program is ready to resume action with its original status and processor integrity fully restored. Upon completing this discussion of stacks, subroutines and interrupts in general we present several specific 1 / 0 examples which use this knowledge. In particular, we discuss DEC‘s (1) serial line interface, the DLVll (commonly used for teletypes), (2) parallel line interface, DRVll (16-bit word in and out; commonly used in fairly high speed data-transfer situations), and (3) direct memory access interface (extremely high speed rates of data transfer such as required by refreshable video screen graphics). The DLV11 serial line interface is by far the most commonly used of these devices simply because we all talk to computers via teletype. However, it is the interface which the student is least likely ever to have to use to control an experiment or data acquisition package. Usually, he simply attaches “

“

”

3 12

H. E. BERGESON A N D GEORGE L. CASSIDAY

teletypes to the computer via this interface and then forgets about it. We forego its discussion here. The interested reader is referred to User’s Manual (1975). We will also forego a discussion of direct memory access (DMA). It is only needed when extremely high data-transfer rates are required. The interested reader is referred to Processor Handbook (1975) and User’s Manual (1975).

D . The D R V l I Parallel Line Interface One of the DEC’s most versatile interfaces is the DRV11. It is very easy to connect external hardware to it and very easy to program. We find it an excellent device to use in order to introduce beginning students to 1/0 techniques. Information concerning this interface can be found in DEC’s LSI-11 User’s Manual (EK-LSI 11-TM-002)(User’s Manual, 1975).We now present a brief description of this unit before proceeding to the applications.

To/From User

Device

FIG.44. DRVl 1 parallel-line interface.

Essentially, the DRV11 is a general purpose parallel interface that connects parallel 1/0 devices to the LSI-11 Bus as shown in Fig. 44. There are three registers on the DRVll which have three corresponding memory addresses. These addresses are set by the user via “jumpers” located on the interface board. Shown in Fig. 45 is the word format for the DRVll registers and their corresponding word formats. We have shown listed those addresses which by DEC convention are the addresses for the first DRVll interface. The function of the input and output data buffer registers should be obvious: the input data buffer receives data from the external world and inputs the data to the LSI; the output data buffer takes data from the LSI and outputs it to the external world. Interrupt capability and additional 1/0 control is provided by the control/status register DRCSR. In particular, it is

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS DRVll lnterrupls

a

Interface Format Vectors

300 304

E Label DRCSR

Address 167770

76 5

1

Reqt B

Inf EnbA

Droulbuf 167772

3 13

Cskl’ CSR0

l5xLxLd Data Out

Ikinbuf 167774

lkLIxzd Data Out (Read Onlyl

FIG.45. DRVll word formats.

capable of handling two external devices requiring interrupt service via the request and interrupt enable A and B bit pairs. Furthermore, there are two additional bits, CSRO and CSR1,which can be used for external control or monitoring purposes. Interrupt requests are generated by the DRV in the following way: Suppose the interrupt enable B bit has been set (either by the program or by the external device). When the external device sets the request B bit (bit 15) an interrupt request is generated and when the LSI acknowledges this request it “vectors” to the address contained in memory location 304 (with status PS contained in location 306). A similar interrupt can be generated via the A bits through vectors 300 and 302. This interface is extremely versatile and as a student uses it he discovers many different ways of utilizing these bits for 1/0 control. 1. Successive-Approximation AID Converter (Aldridge, 1975)

If you happen to have an LSI (or any microprocessor, for that matter) at your disposal and you are using it to hook up to the external world, one of the first things you will conclude that you need is an A/D converter. The external world is made essentially of analog signals and, inevitably, it is your task to get information about some signal into the computer for processing. The signal could represent temperature, for example. Hence, your first task will be to digitize that signal in order to get it into the LSI. Now, if you happen to know a Little about how A/D converters work, you will quickly realize that you already have a large part of the hardware necessary for an A/D already at your fingertips in the LSI itself. So why not use it?

3 14

H. E. BERGESON AND GEORGE L. CASSIDAY

Register

"in

strobe I crnp bit 2 strobe 2 cmp bit I -~ strobe 3 cmp bit 0 strobe 4

FIG.46. Four-bit successive-approximation A/D converter.

Shown in Fig. 46 is a block diagram of a 4-bit successive approximation A/D converter. First, we will review its operation and then design an 8-bit converter using the LSI. The basic elements of the converter are a clock, D/A converter, comparator, and shift register. It works in the following way: A signal of interest (here K,,) is to be digitized. It is presented to one input of the comparator. The other input is attached to the D/A output such that the comparator output is a logical 0 when K,, > D/A,,, . As soon as F,,is presented to the comparator input, a start convert command is generated, and the most significant bit of the 4-bit register is set. Thus, the D/A output jumps to one-half its maximum possible output value. The output of the comparator is sensed (i.e., is it a 1 or O?) and the most significant bit is either kept (if D/A,,, < or cleared if D/A,,, Kn). By repeating this process, the output of the D/A finally converges to within k* LSB of the input voltage. Let us follow through this process once more with a concrete numerical example. Let the input signal yn= 9.7 V; let D/AOUtmax = 16 V; hence, a digital 1000 (MSB set; others clear) presented to the D/A will produce D/AoU,= 8 V, and a 000 1 (LSB set) yields 1 V.Thus, we should be able to digitize signals to about f0.5 V. At the start of the digitization, MSB gets set and D/A,,, = 8 V which is less than K, = 9.7 V. Thus, the comparator output stays 0 and strobe 1 sets the MSB in the register. Now, at the beginning of the next cycle, bit 2 is set and a 1 100 is presented to the D/A.

=-

v,,

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

315

Its output is now D/A,,, = 12 V which is greater than I,= 9.7 V. Hence, the comparator output jumps to a 1 and strobe 2 clears bit 2, leaving a 1000 in the register and a D/A output value of 8 V. By continuing this process for the next two bits, you can easily see that we end up with a 1001 in the register which represents 9 V. However, note that this process will always yield a voltage less than the input voltage y, . Hence, our best guess for the actual value of the input voltage is obtained by adding 0.5 volts to our final estimate based on the value stored in the register. Hence, our guess would be 9.5 V f 0.5 V. Obviously, we have left out a lot of electronics in our block diagram. The point of the example is, however, that you need only an inexpensive D/A and comparator to implement this A/D with the LSI. All the additional hardware required for its implementation is in the LSI and the above control and decision-making processes can be implemented by software. Let us continue the example in more detail. First, assume that we have some sort of external status flip-flop which gets set when a signal voltage is present and ready for digitization. For example, we could have a sampleand-hold circuit which is holding the analog signal of interest and a status flip-flop which has been set just after the sample-and-hold output has settled to its quiescent value. How do we use the LSI (via the DRV11) to digitize this signal? One possible way is indicated in Fig. 47. The signal is presented to a comparator and the comparator’s output is attached to REQA (bit 7) of DRCSR. This bit will be examined in order to decide whether the D/A output during any portion of the digitization process exceeds the input signal. CSRO (bit 0) will be used to clear the external status flip-flop and thus get things ready for the next digitization. Request B (bit 15) will be used to tell the LSI that a signal is present and to start the digitization. The lower byte (bits 0-7) of DROUTBUF will be presented to our 8-bit D/A converter. When the digitization is finished, I

8 Hold

L

FIG.47. Eight-bit A I D using the DRVl I .

3 16

H. E. BERGESON A N D GEORGE L. CASSIDAY

TABLE 111 FLOWCHART: BIT SUCCESSIVE APPROXIMATION A/D CONVERSION ALGORITHM

P START

Y Examine

a Bit 15 DRCSR

Is a signal present?

Current Bit

1

NO

Clear Current Bit

(DIA < V,")

RO

(next bit)

1

=

1

O'!

(???)

Clears R E Q B (via status F/F)

Reset CSRO

ON THE TEACHING OF ELECTKONICS TO SCIENTISTS

317

DROUTBUF will contain the digitized word. It should be easy for you to visualize the digitization process about to be described: d o a BIT SET from the MSB down to the LSB in DROUTBUF, examine REQA after each BIT SET and d o a BIT CLEAR whenever REQA = 1 (indicating that D/A,,, > V,"). A flow diagram of the algorithm to be implemented is shown in Table 111. Students are asked to examine flowcharts like this carefully. Several questions come to mind: (1) What is the purpose of initially storing a 200 in RO? (2) How is the current bit set as indicated in the third step of the algorithm? (3) What is accomplished by right shifting RO in step 5? (What kind of right shift should be performed?) (4)What happens when RO = O? ( 5 ) How can the digitized word be stored? When the student is convinced that he understands the flowchart he is then asked in lab exercise 11 to actually write the 1/0 program which implements the 8-bit A/D conversion algorithm. Shown in Table IV is an example of a program which will accomplish this task. TABLE IV

CI

1000/012700 loo2/000 loo 1004/005737 10061167770 10101100375 1012/052737 10141167772 1016/105737 1020J167770 1022/000240 1024/1oooO2 1026/04OO3 7 10301167772 1032~006200 1034/005700 1036/OO 1365 1040/013746 10421167772 1044105273 7 1046/000001 1050/167770 10521042737 1054j000001 10561167770 1060/00013 7

i -

110 PROGRAM FOR %BIT SUCCESSIVE APPROXIMATION ALGORITHM MOV #loo. RO

Load RO with MSB

TST 6 # 167770

Look for REQ B (Bit 15)

BPL [ - 31

BIS RO, @ #I67772

Set bit in DROUTBUF

TSTB (u X167770

Look for 1 in REQ A

NOP BPL [+2] BIC RO, Ta X167772

Keep bit if REQ A = 0 Clear bit if REQ A = I

ASR RO TST RO BNE [ - 131 MOV ( 4 # 167772, - (SP)

Go to next significant bit Have all bits been set? If not, get next bit Push digitized word

BIS # 1, Q # 167770

Set CSRO; clears REQ B

BIC # 1, @ # 167770

Clear CSRO, restore status

IMP (@ X~OOO

G o back for next word

3 18

H. E. BERGESON AND GEORGE L. CASSIDAY

Note that REQ B of DRCSR has been used t o tell the LSI that a signal is present but since bit 6 (interrupt enable B) has not been set, no interrupt request is generated. Thus, the above program must be resident and operative. This is fine if all you are doing is digitizing words. But, if the LSI is to be mainly doing something else (as should be the case) the above routine should probably be set up as an interrupt-driven routine. Each student is asked how this could be accomplished.

2. Sixteen-Channel Data-Acquisition System One of the prime virtues of a computer is its ability to perform many tasks “simultaneously.” A typical situation concerns the processing of signals produced by a large number of transducers. If the number is too large or if the transducers are widely separated in space, the use of a computer to acquire and process the signals becomes a necessity rather than a “mere” convenience. Here, we will describe a relatively simple 16-channel data acquisition package which uses the LSI-11 in such a fashion. Shown in Fig. 48 is a block diagram of such a package. There are sixteen input channels which come typically from external transducers. In our case the transducers generate pulses which are to be integrated (if above some threshold value) and then digitized. These pulses are fed into a 16-channel analog multiplexer (MUX). By supplying a 4-bit address to the address inputs of the MUX one of the sixteen channels will be selected. As soon as the ENABLE

Dolo Pras Gate Off

FIG.48. Sixteen-channel data acquisition system.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

319

input of MUX is asserted the selected channel will be routed directly to the output of MUX. This selection and signal routing process is handled by the LSI via a D-type flip/flop (F/F). CSRO bit of DRCSR is connected to the D (data) input and CSRl bit of DRCSR is connected to the CLR (clear) input. Also, the NEW DATA READY line is connected from the DRVll to the CLK (clock) input. Routing a particular channel through MUX to the input of the external electronics signal processing package requires the following sequence of events. (1) Set bit CSRO (this loads a 1 into the F/F D input). (2) Address MUX via DROUTBUF (i.e., suppose we wish to address channel 8; simply load a 10 into DROUTBUF). (3) As soon as a word has been loaded into DROUTBUF, the DRV11 will generate for you a 300-nsec positive going pulse called NEW DATA READY. This pulse is fed to the CLK input of F/F. At that moment, the logic level present at the D input will be “clocked” to the Q output of F/F. In this case it is a 1. Consequently, the ENABLE input of MUX will now be asserted and the addressed data channel (channel 8, here) will be routed through MUX to the signal-processing electronics. The signal-processing electronics examines a pulse from the selected channel. If the pulse is large enough, it is integrated and the pulse integral is then “held” in analog form and presented to the input of a 12-bit A/D converter. After this analog value has settled down, a “start convert *’ pulse is generated by the signal-processingelectronics and presented to the A/D. The A/D digitizes the signal where it then appears as a 12-bit binary number at its output. Upon completion of the digitization an “end convert” pulse is sent back to the external electronics package. This package then generates a data-present-gate-on signal and sends it to REQ B of DRCSR. This “tells” the LSI that the A/D is holding a digitized word which it is presenting to DRINBUF. The LSI can now read DRINBUF and store the digitized signal from the selected data channel. As soon as the ESI has “read” DRINBUF, the DRVl 1 interface generates a 300-nsec positive going pulse called DATA TRANSMITTED. This pulse is fed back to the signal processing electronics where it clears everything, i.e., it destroys the analog integral still being held (and presented to the input of the A/D) and it turns off for the data-presentgate-on signal which releases REQ B. The LSI can then look for another pulse from the same data channel; it can select another data channel or it can simply stop and process data already received. There is one important thing to note about this application. The data acquisition is under control of the LSI. The LSI is selecting data channels of interest. Hence, it has to know a priori that signals are there exactly where it is looking. This implies two things: (1) you, yourself, have already put a signal source near the transducer of interest and have told the LSI that one is there or (2) the LSI is being used, somehow, to activate a signal source and

320

H. E. BERGESON AND GEORGE L. CASSIDAY

hence it knows where to look for a signal. The point is that the signal source does not simply appear at random, and, hence, this application does not require the use of interrupts. Again, each student is required to suggest modifications required to turn this into an interrupt driven process which would be necessary if the transducer signals were generated at random. One particular exercise that we have performed with the system described above is the following: We mounted a cluster of 16 photomultiplier tubes (PMT) inside a light-tight box. We also mounted a LED (light emitting diode) inside the box and hooked it to a pulser. Every pulse from the pulser would then generate a flash of light of roughly psec duration. The 16 PMTs would be exposed to this light source simultaneously. (Since each PMT in this situation “saw” the same light source, their electronic gains ” could be adjusted so that they each produced the same size electrical pulse.) Each student is then required in lab exercise 13 to write a program to get the LSI-11 to read in and store 100, light pulses from each of the 16 PMTs. The appropriate DRV11 1/0 register addresses were 17oooO (DRCSR), 170002 (DROUTBUF), and 170004 (DRINBUF). “

TABLE V 1/0 PROGRAM TO READ IN 100, PULSE HEIGHTSFROM 16 CHANNELS DATA:

RO/OOOlOa Rl/OOOlOa R2/OWO20 R3/170000 R4/170002 R5/170004 R6/002400

X words1PMT (octal)

BIS X2, (R3)

Clear F/F. disable MUX

BIS X 1, (R3)

Load enable bit at D input

r

DEC R2 MOV R2, (R4) TST (R3) BPL [ - 21 MOV (RS), -(SP) SOB RO, B MOV R1, RO BIS X2, (R3)

Set MUX address lo select PMT Address MUX, enable data channel Look for REP B (Data present)

-

TST R2 BNE [- 151

PROGRAM: 2400/0527 13 2402/000002 . t 24041052713 2406/000001 24 10/005302 2412/010214 24 14/OO57 13 2416/100376 2420/011546 2422/077004 2424/010100 242610527 13 2430/000002 2432/005702 2434/001363 24361000000

X PMT (octal)

DRCSR DROUTBUF DRINBUF Top of stack

Push word onto stack Loop until word count done Reset word count Clear F/F, disable MUX Have all tubes been read? If not, read next tube. otherwise HALT

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

32 1

Shown in Table V is our solution to the problem. Most of the instructions in the solution can be easily understood with a little thought. The first two instructions simply get the multiplexer ready for selection of a data channel. Note that R2 contains the number of channels to be examined. In this example there are 16 (20,). The addresses for the multiplexer are 0 = channel 1, 1 = channel 2, . . ., 7 = channel 8, 10 = channel 9, ..., 17 = channel 16. Hence, the DEC R2 instruction sets up the proper address in R2 for the selection of channel 16. The MOV R2, (R4) instruction addresses the MUX and thus enables channel 16. The next four instructions simply read in 100, data words and push them onto the stack. At the completion of input from a given channel, the word count is reset, the F/F is cleared thus disabling the MUX.We then examine R2, looking for a 0. If R2 contains a zero, it means that we have gone down through all data channels (16,11, . . ., 1). If R2 does not yet contain a zero, we branch back to A where the MUX is again made ready for the next channel selection. Note that data is pushed onto the stack first channel 16, ..., finally channel 1. Hence, data is popped first channel 1, . . , , finally channel 16. Also, note that the stack starts at address 2400 and goes down to address 400. Below 400, of course, resides the memory area reserved for vectors, traps, etc. Since the program starts at location 2400, machine memory is optimized. Another 1 / 0 programming exercise (lab exercise 14) that we have used and which is fairly typical involves the immersion of a radioactive source inside a sodium iodide (NaI) well-shaped crystal. This particular radioactive source emits positrons (positively charged electrons). These particles generate a flash of light in the clear NaI crystal. If the crystal is physically placed above a PMT, that PMT will produce an electrical pulse proportional to the flash of light it “sees” emanating from the crystal. Each student is required to produce a spectrum of the light pulses generated by the radioactive source, i.e., a plot of the number of pulses as a function of pulse integral. Since the signal processing electronics measures pulse integrals, and since each pulse integral is digitized and presented to the LSI as it is generated, it would be very nice if somehow the LSI could generate “on the fly” a pulse-integral distribution of the input data. In other words, we would like the student to use the LSI as a multichannel analyzer. If you think about it a little bit, the algorithm necessary to implement this function is fairly obvious. Note that out A/D converter is a 12-bit converter. Hence, each data word presented to the LSI (via DRINBUF) is a number ranging from 0 to 409610 (0 to 1oo00,). Now, let some number representing a pulse integral value be presented to the LSI, say 2002. Suppose that we simply increment the contents of that address every time it is presented to the LSI as an input data word. Stored in memory then from 0 to 1oo00, would simply be the number of times those addresses were refer-

322

H. E. BERGESON A N D GEORGE L. CASSIDAY

ences, i.e., the pulse-integral distribution desired. (It would be a simple matter to add a base address, say 3000, to each address referenced so that data would appear only in an allowable address field somewhere above the data acquisition program.) Thus, generating this distribution in the LSI should be a fairly trivial extension of our previous example. VI. CONCLUSIONS

There are many additional 1/0 exercises which each student is asked to carry out. The initial exercises have involved programming existing I/O electronics systems. Other exercises require the student to design, build and debug his own minicomputer-controlled electronics system. We find that most of our students are capable of carrying out this task if they have worked diligently through our lab exercises. By the end of this course each student has completed more than 50 exercises emphasizingdesign, as well as successful implementation. Perhaps this philosophy has sacrificed detailed understanding at many instances along the way in favor of the “quick-anddirty” approach. On the other hand, we firmly believe that the expertise and confidence each student has acquired with a wide variety of electronics devices and techniques will prove to be a far more valuable asset than that laid down by the more formal and conventional electronics teaching approach.

ACKNOWLEDGMENTS We would like to express our thanks and appreciation to (1)the National Science Foundation for its continued research support which enabled us to gain much of the expertise necessary to develop our approach to electronics, (2) Joe West for assisting with much of the hardware circuit development presented here, (3) Dave Steck for many of the programming examples and software routines we have used,, (4) Digital Equipment Corporation for allowing us to present instructional material concerning its LSI-11, ( 5 ) The American Journal of Physics for allowing us to reprint some of the ‘‘first quarter” material, (6) Betty Keuffel for preparing our figures, and finally (7) Lynette Blue for preparing the manuscript.

APPENDIX A . Lecture Topics and Laboratory Assignments

First Quarter-Lecture Topics 1. Resistors, inductors, and capacitors; Ohm’s law; voltage divider. 2. Differential circuit parameters (small-signal parameters), input and output resistance, Thevenin’s theorem. 3. Linear differential operational amplifiers.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

323

4. Virtual equality and operational amplifier circuits (amplifiers, adders, and integrators); nonideal characteristics of real operational amplifiers. 5. Analog computation, voltage comparator, Schmidt trigger. 6. Digtal logic, Boolean algebra, TTL circuitry. 7. Monostable multivibrators and counting circuits. 8. Digital display, medium-scale integration, timers, TTL compatible voltage comparators. 9. Inductors and capacitors (differential equations for linear circuits), rise time, astable multivibrator. 10. Midterm examination. 11. Semiconductor diodes, power supplies, clipping and clamping, precide diode circuitry. 12. Complex impedances. 13. Sine-wave power; Fourier analysis; sine-wave generators. 14. Active filters. 15. Bipolar transistors. 16. Bipolar transistors continued; zeroth-order performance of various transistor configurations, h-parameters, complementary transistor circuits, Darlington transistors. 17. Field effect transistors, y-parameters; switches; voltage-controlled resistance. 18. Thermistors, thyristors, phototransistors, photon-coupled devices.

First Quarter- Laboratory Assignments 1. Determine resistance networks in four-terminal black boxes. 2. Design, build, and test voltage-dividing networks with division ratios and input resistances specified. 3. Design, build and test inverting and noninverting amplifier circuits of specified gains (74 1 operational amplifier available). 4. Design, build, and test a four-bit digital-to-analog converter. 5. Design, build, and test amplifier for millivolt signals. 6. Design, build, and test one of the following: (a) Schmidt trigger with specified turn-on and turn-off voltage; (b) computer to give analog solution to two linear algebraic equations with two unknowns. 7. Empirically determine truth tables of NAND, AND, and NOR TTL circuits; build R-S flip-flop from two-input NOR circuits; design, build, and test “EXCLUSIVE OR” circuit composed of NOR, NAND, and AND devices. 8. Design, build, and test a delay generator with pulse length between specified limits (74121 monostable multivibrators are available); construct an astable multivibrator from two monostable multivibrators with the period between specified limits.

324

H. E. BERGESON AND GEORGE L. CASSIDAY

9. Design, build and test a system which uses a 5-V peak-to-peak 60-Hz sine wave to generate a digital system which will display the number of seconds from 0 to 99 using seven-segment displays. 10. Design, build, and test an operating system which will take a 1-sec period pulse train and light one light-emitting diode for three pulses, light a second light-emitting diode for three pulses, and repeat. 11. Design, build and test an astable multivibrator with a frequency of 10 (f2.5) kHz using the 741 operational amplifier. 12. Construct a square-wave generator and clip, clamp, and rectify the capacitively-coupled output. 13. Design, construct, and test an amplifier with a voltage gain of 5 k 1.5 at 100 Hz which is greater by a factor of 6 f 2 at lo00 Hz. 14. Design, build, and test an absolute value generator for use on a 100-Hz sine wave. Save for the next assignment. 15. Construct a frequency doubler by filtering out the harmonics of the output of absolute value generator constructed last time (the filter should be a second-order low-pass active filter). 16. Design, build, and test a sawtooth generator with a specified period. 17. This assignment involves some standard transistor biasing. 18 and 19. Design, build, and test any circuit where the desired characteristics are suggested in advance by the student and approved by the instructor. Second Quarter--Lecture

Topics

1. Elementary pulse signals; ideal” voltage and current sources; load lines; Thevenin and Norton theorems. 2. Response of R , J!, C networks to elementary pulse waveforms in the time domain. 3. Delay cables; transmission and wave shaping. 4. Basic transistor operation; cutoff, saturation, active region, notion of current control. 5. Transistor biasing; emitter-, self-, fixed-, and voltage-feedback bias. The simple grounded-emitter amplifier. 6. The grounded-base amplifier, current-to-voltage converter. 7. The grounded-collector amplifier, emitter followers; the cascode. 8. Emitter-coupled pairs; the difference amplifier, biasing. 9. Passive networks again, the complex frequency plane; LaPlace transform transfer functions, poles and zeros. 10. Negative feedback, bandwidth, gain stabilization, input- and outputimpedance improvements. 11. ‘Stability, methods for stability control. “

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

325

12. Emitter-coupled input stages, the 733 video amplifier, the stacked-up emitter follower. 13. Positive feedback, Schmidt trigger, oscillators. 14. Midterm Examination. 15. TTL and ECL gates; virtues and drawbacks. 16. MOS gates; virtues and drawbacks. 17. Computational circuits; exclusive-OR, half and full adders. 18. Multipliers, serial and parallel, look-ahead counters. 19. The Operational Transconductance Amplifier (OTA, RCA 3080); current mirrors; analog multiplexing. 20. The OTA; amplitude modulation, multiplication, and sample and hold. 21. The phase lock loop (NE 565); voltage-controlled oscillator, phase detection. 22. NE 565 frequency synthesizer and FSK decoder. 23. D/A techniques (Motorola 1406, 1408); R-2R ladders; programmable pulse generator. 24. A/D techniques (Motorola 1405L), dual ramp and staircase. 25. A/D techniques; successive approximation, parallel; output coding. 26. Questions and makeup. 27. Final examination.

Second Quarter-Laboratory

Assignments

1. Assemble several RC high- and low-pass networks and measure their response to step, impulse, ramp, and sine-wave inputs. Compare rise and tilt times obtained with step inputs to corner frequencies obtained with sinewave inputs. 2. Design and build two delay line clipping circuits using a mercurywetted reed relay switch toggled at 60 Hz and a 12-V power supply. The output pulse should be of 50-nsec duration and its voltage should be (a) - 1.5 V (b) 6.0 V. In each case the pulse is to be delivered to 51-0 load via an RG 58 delay line. 3. Design, build, and test a simple transistor inverting switch. 4. Design, build, and test a gain of (10 f.2) inverting transistor amplifier using emitter bias, capable of passing 10 psec pulses with less than 5 % tilt. Design and build a gain of (50 f 10) inverting transistor amplifier using voltage-feedback bias. 5. Design, build, and test a gain of (5 f 1) grounded-base amplifier. Input to the amplifier is from a 5 1 4 delay line which should be correctly terminated.

+

326

H. E. BERGESON AND GEORGE L. CASSIDAY

6. Design, build and test an emitter follower capable of driving a capacitively coupled 5142 load with a voltage gain of not less than 0.9. Assume positive-going pulses, 100-nsec pulses. Polarities should be chosen to optimize rise time. 7. Design, build, and test an emitter-coupled amplifier using two pnp transistors. Available supplies + 6 V; biasing, IC1= 1 mA and Zc2 = 9 mA, V,, = 0 V, V,, = V,, = -2 V. Coupling and bypassing should be effective at frequencies down to 10 kHz. 8. Design, build, and test a stacked-up emitter follower capable of driving pulses of either polarity with voltage gains not less than 0.95. Available supplies f 15 V. Biasing should allow for positive voltage swings of + 18 V. 9. Design, build, and test a gain of (100 k 20) transistor negativefeedback amplifier. Feedback should be to the emitter of the input stage. Use the 733 video amplifier and design and build a discriminator which delivers a 3-V minimum width 100-nsec pulse when input pulses exceed 100 mV. 10. Design, build, and test a 4-bit ripple-carry adder using the 7480 full adder and the 74195 shift register. 11. Design, build, and test a 4-bit hardware multiplier using a 74181 arithmetic logic unit, a 7495 shift register and a 74195 shift register. 12. Design, build, and test a two-channel analog multiplexer using the OTA 3080. 13. Design, build, and test a sample-and-hold circuit using the OTA 3080. 14. Design, build, and test a frequency multiplier using the NE 565 phase lock loop and a 74161 4-bit counter. Selectable outputs should multiply the input frequency by factors of 2, 4, 8, and 16. 15. Design, build, and test a FSK decoder using the NE 565 phase lock loop. 16. Design, build, and test a programmable pulse generator, using the Motorola 1408 D/A. Outputs should be 0-5 V in steps of 20 mV. 17. Design, build, and test a staircase A/D converter using the &bit Motorola 1408 D/A. 18. Finish all labs.

Third Quarter-Lecture Topics 1. Introduction; basic computer architecture; central processor and memory. 2. Building a simple computer, ECP-1 (an Educational Computer for Pedestrians); ECP-1 instruction set and architecture; two’s complement arithmetic. 3. Operation code for the basic instruction set; central processor control hardware.

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

327

4. Computer hardware; op-code decoding, program counter; memory; arithmetic unit; input/output hardware. 5. Programming Digital Equipment Company’s LSI- 11 minicomputer; LSI-11 architecture. 6. LSI-11 addressing modes; examples. 7. LSI-11 instruction set; programming examples. 8. Programming examples. 9. Input/Output; the LSI-11 stack; subroutines. 10. Programming I/O peripherals; interrupts; examples. 11. The DLV 11 serial line interface; teletype 1/0 examples. 12. Interrupts via the DLV11; examples. 13. The DRV11 parallel line interface; applications; a hardware/software successive approximation A/D converter. 14. Applications; downstream loading. 15. Applications; a 16-channel data acquisition system. 16. Applications; a light display board for visual pattern recognition. 17. Applications; a programmable pulse generator. 18. Direct Memory Access (DMA) applications; a DMA graphics generator.

Third Quarter-Laboratory Assignments 1. Design, build, and test an op-code decoder for the ECP-1 instruction set. 2. Finish the design and then build and test the arithmetic unit for the ECP- 1 minicomputer. 3. Design, build, and test the additional electronics required t o implement the WRITE instruction for ECP-1. Use three 7-segment displays for the output. 4. Write and test a program designed to check the CLR (clear) instruction of the LSI-11. 5. Write and test a program to add all the numbers stored in 1000s successive memory locations. 6. Write and test a program to convert six ASCII values which have been stored in six successive memory locations into the actual number they represent. (The ASCII values represent six decimal digits input from a teletype.) 7. Write and test a program which passes the starting location as well as size of a data array stored in memory to a subroutine. 8. (a) Write and test a program which will “ E C H O ” all characters typed on the teletype. (b) Modify the program such that the program HALTS if a rubout ” is typed. “

328

H. E. BERGESON A N D GEORGE L. CASSIDAY

9. Write and test a program to read and store in consecutive memory locations loo8 ASCII characters from the teletype. The characters should be printed out at completion of input. 10. (a) Write and test a teletype exerciser program. After it is fully debugged, install it as an interrupt service routine. (b) Install your program from lab project 9 as a main program and arm the teletype transmitter interrupt. When a character is emptied from the transmitter buffer, an interrupt should be generated and the LSI-11 should vector to the teletype exerciser routine. Disarm the transmitter interrupt after 100, characters have been transmitted and stored by the main program so they can be printed out without interruption. 11. Write and test an 1 / 0 program to execute an 8-bit successive approximation A/D conversion of a specified input voltage. 12. Write and test an 1/0 program to receive and store programs sent to the LSI-11 from a P D P 11/45 processor located downstream. 13. Write and test an 1/0 program to read in and store loo8 digitized pulse heights from each of 16 separate data channels. 14. Write and test an 1/0 program to store a 32K count (216) pulse height distribution from a single data channel. The pulse heights are generated by a radioactive Na22source mounted in a NaI crystal over a photomultiplier tube. 15. Write and test an interrupt-driven 1 / 0 program to read in a sequence of 36 addresses presented to the LSI-11 upon interrupt. Your interrupt service routine should then light up 36 external LEDs in that sequence. 16. Write and test an 1/0 program which generates a sequence of 256 pulses from the programmable pulse generator from 0-5 V in 20-mV steps. 17. Write and test an 1/0 program which draws a vector between two coordinates on the Tektronics 602 monitor. The coordinates are punched in via teletype and the vector drawing takes place via the DMA interface. Term Project: Design, build, and test some electronics project of your choice which is controlled by the LSI-11. Implementation of your project constitutes the final examination.

B. Grading First Quarter

65% lab assignment credits 25% midterm and final examinations 15 % homework assignments

ON THE TEACHING OF ELECTRONICS TO SCIENTISTS

329

Second Quarter

65 Yo lab assignment credits 257” midterm and final examinations 15”/, homework assignments Third Quarter A, successfully implemented term project B, 6 out-of-lab assignments 11-17 completed successfully C, 3-5 out-of-lab assignments 11- 17 completed successfully D, 1-2 out-of-lab assignments 11-17 completed successfully E, nothing from lab assignments 11-17 completed

Comments

Note that the grading is very achievement oriented. There is little subjective assessment by the instructor. In all cases, the student knows what he must d o to earn a high grade. He must get his circuits to perform. During the third quarter, grading is totally objective. The expressed purpose of that course is to get students to hook up some electronics to a computer and get the computer to control its operation. If the student succeeds, he gets an A. If he does not succeed he can at least d o very well by getting the computer to control existing electronics hardware. To d o this, he must understand the hardware as well as computer IjO programming at the machine language level. If he can at least get this far, then he is on the verge of being able to do it all himself. In each case, we are very pleased with the results of the grading system. Students who succeed with their lab assignments learn and understand the material much better than their comrades who d o minimal lab work-and they earn the better grade as their reward.

REFERENCES Aldridge. D. (1975). “Analog-to-Digital Conversion Techniques with the M6800 Microprocessor System,” Application Note AN-757. Motorola Semiconductor Products, Inc., Phoenix, Arizona. Angelo, E. J., Jr. (1964). ‘’ Electronic Circuits,” Ch. 19. McGraw-Hill, New York. Babcock, L. E., and Vignos, J. H. (1973). Am. J . Phys. 41, 89. English, T. C., and Lind, D. A. (1973). Am. J . Phys. 41, 81. Graeme, J. G., Tobey, G . 8.. and Huelsman, L. P., eds. (1971). “Operational Amplifiers, Design and Applications.” McGraw-Hill, New York.

330

H. E. BERGESON A N D GEORGE L. CASSIDAY

Leon, B., and Bass, S. (1974).Electron. Des. News Mar. 20, p. 55. ’’ Linear, Digital, MOS Applications Book ” (1973).Sect. 6.Signetics, Sunnyvale, California. ‘‘ Linear Integrated Circuits” (1976).Pp. 13-16. Fairchild Semiconductor, Mountain View, California. Littauer, R. (1965).‘’ Pulse Electronics.” McGraw-Hill, New York. Malmstadt, H., and Enke, C. (1969).’‘ Digital Electronics for Scientists,” Ch. 7.Benjamin, New York. Malmstadt, H., Enke, C., and Crouch, S. (1973).“Digital and Analog Data Conversions,” Module 3, Sect. 3-4.Benjamin, Menlo Park, California. Millman, J., and Halkias, C. (1967).“Electronic Devices and Circuits,” Ch. 17.McGraw-Hill, New York. “Processor Handbook, LSI-11,PDP 11/03” (1975). Digital Equipment Corp., Maynard, Massachusetts. Semiconductor Data Library ” (1975).Vol. 6. Motorola Semiconductor Products, Inc., Phoenix, Arizona. Smith, J. I. (1971).“Modern Operational Circuit Design.” Wiley (Interscience), New York. “The TTL Data Book for Design Engineers” (1973).Texas Instruments, Dallas, Texas. “User’s Manual, LSI-11,PDP 11/03”(1975).Digital Equipment Corp., Maynard Massachusetts. I’

Author Index Numbers in italics refer to the pages on which the complete references are listed.

Blank, S. L., 196(5.33),201 Bloom, L. R., 196(5.27),201 Blotekjaer, K., 20, 26, 37 Blum, J. M., 58(1.17), 70 Blum, S. E., 57(1.16), 70 Bobroff, D. L., 2, 3, 8, 37 Bok, J., 71(2.2), 73(2.2), 101 Bordner, G. W., 240,251 Born, M., 82(2.20), 101, 137 Brantley, W. A., 130(3.27), 137 Brennan, L. E., 210, 211,213, 214, 215, 216, 217,218,250 Bube. E. D., 100(2.50), 102 Bucy, R. S., 238.252 Burnham. R. D., 83(2.21), 84(2.21, 2.24). 87(2.28), 88, 101, 102 Burrus, C. A., 115(3.7), 137

A

Abram, R. A., 115, 137 Adams, M. J., 72(2.8), 101 Adams, R. F., 186(5.13).200 Aiki. K.. 85(2.26). 101 Albans. R. E.. 116(3.12), 137 Aldridge. D.. 313.329 Allen, R. W., 115, 137 Almeleh, N., 55. 70 Alter, J. J., 237. 238, 241, 243, 252 Anderson. S. J., 19, 37 Angelo, E. J., Jr., 288, 329 Apptebaum, S . P., 204,210,250 Applied Physics Laboratory, 230,251 Archer, R. J., 46. 47, 55(1.3), 69 Armstrong, J., 55(1.12), 70

C

B Babcock, L. E., 258,265, 329 Bachmann, K. J., 65(1.25), 66, 70 Bachrach, R. 2..90. 102, 148(4.2), 152(4.2), 158

Barton, D. K., 223,250 Barybin, A. A., 2, 3,4. 7, 8, 9, 10, 11, 16, 17, 18, 20.21, 22, 23.24, 25, 26, 27, 28, 29. 30. 31, 32, 33, 37 Bass, S., 290, 330 Becklund, 0. A., 140(4.1), 141(4.l), 158 Beebe, E. D., 67(1.27), 70 Benedict, T. R., 240,251 Bergh, A. A., 134(3.30), 137 Bernal, E. G . . 196(5.29).201 Bernfeld. M.. 203. 250 Bhargava, R. N., 65( 1.24). 70,96(2.38), 102 Bianco, B., 19, 37 Blackman, S. S., 248,252 Blackerall. D. M., 79(2.16), 101 Blake, L. V., 225.251

Calawa, A. R., lOl(2.53, 2.54). 102 Campbell, J. C., 40, 41,42,45,46, 48, 69 Cantrell, B. H., 223, 227, 234,235, 237, 238, 241. 243,250,251,252 Capik, R. J . , 65(1.25), 66, 70 Capon, J., 220,250 Carroll, R. C., 79(2.16), I01 Cartledge, L., 222, 250 Caruthers, J. R.. 195(5.24),201 Casey, H. C., Jr., 77, 78, 81, 84(2.22), 85, 87(2.22), 90, 101, 102 Castella, F. R., 238, 252 Caulton, M., 183(5.12),200 Chandross, E. A,. 196(5.32),201 Chang, N. S., 2, 20, 21,38 Chang, T. Y., 196(5.30),201 Chen. D.. 196(5.29),201 Chiabrera, A., 19, 37 Chin, R., 99(2.47), 102 Chinn, S. R., lOl(2.54). 102 33 1

332

AUTHOR INDEX

Chodorow. M., 2,3,37 Coleman, J . J., 54(1.7),66(1.26), 70, 98(2.45), 99, 99(2.47), 102 Compton. R. T., 206. 219, 250 Cook, C. C.. 203,250 Cook, S. R., 243, 252 Cooper, D. C., 227,251 Craford, M. G., 40, 41,42,45, 46, 48, 54(1.7), 69, 70, 95(2.37), 102, 158(5.3), 160(5.4), 200 Crawford, M. G., 49, 50, 5l(l.4), 52, 53, 54( 1.4). 55( I .4), 69 Crouch. S., 290,330

F Finn. H. M., 231, 2551 Fleming, P. L., 19,37 Fonstadt, C. G., lOl(2.52). 102 Forward, K. E., 133(3.29), 137 Fox. M. J., 95(2.37). 102 Foy. P. W., 77. 78. 101 Freire. G . F.. 2, 20. 21, 37 Friffiths. L.. 206. 208, 251

C D Dapkus, P. D., 65(1.23), 70, 136(3.33). 137 David, T., 196(5.27),201 Dawson, R. W., 115(3.7), 137 Dean, R. H., 4, 19,20,21, 28, 36.37 De Corlieu, G., 192(5.22), 201 De Winter, J. C., 67(1.27), 70, lOO(2.50). 102 Dierschke, E. G., 56, 57(1.15), 70 Dillard, G . M., 227, 251 Dixon, R. W., 100(2.50),102 Doerbeck, F. H., 79(2.16), 101 Doug. E., 204,205, 206, 207. 210, 251 Dreeben, A. B., 4, 19, 20,28, 36.37 Drury. W. H., 222, 250 Dumke, W. P.,I16(3.10), 117(3.10), 137 Dunnebacke, E. G., 238,252 Dupuis, R. D., 97(2.42), 102 Dyment, J. C., 116(3.12),137

E Edelblute, D., 217, 250 Edmonds, H. D., 11 l(3.5). 137 Edmonds, H. E., 161(5.5), 200 Engelmann, R. W. H., 20, 28, 29,37 English, T. C., 258,329 Enke, C., 290, 299,330 Enstrom, R. E., 98(2.44), 102 Ettenberg, M., 54( 1.8), 70, 78(2.15), 89(2.32), 94(2.32), 98(2.44), 102, 135(3.31), 137, 190(5.16),200

Gabriel, W. F., 219, 250 Gandhi, 0. P., 2,38 Garmire, E.. 120(3.18),137, 196(5.26, 5.28), 20 1 Garvin, H. L., 196(5.26),201 Glover, J., 204, 205, 206, 207,210, 251 Gooch, C. H., 105, 137, 158(5.1), 179(5.1), 200 Goode, B., 206,208,251 Goodfellow, R. C., 115, 137 Goodlin, R., 204, 205, 206, 207, 210, 251 Goldstein, Y., 4, 5, 10, 13, 38 Gouda, S.. 88(2.31), 97(2.40), 102 Graeme, J. G., 274,329 Granatstein, V. L., 223,250 Griffiths, L. J., 219,250 Grover, N . B., 4, 5, 10, 13, 38 Groves. S. H., lOl(2.53, 2.54), 102 Groves, W. 0..49. 50, 51( 1.4). 52. 53. 54(1.4, 1.7). 55(1.4),69, 70,95(2.37), 98(2.45), 99(2.47), 102, 158(5.3), 160(5.4), 200 Gueret, P..2, 20, 21, 37, 38 Guetin, P., 186(5.13),200

H Hackett, W. H., Jr., 55(1.12), 65(1.23), 70 Hahn, W. C.. 2, 3, 8, 38 Haisty, R. W., 56, 57( I .15), 70 Hakki, B. W., 44(1.2), 69 Halkias, C.. 288, 330 Hall, D. B., 196(5.25),201 Hall, R. N., 71, 101

AUTHOR INDEX Hansen. V. G.. 231. 232.233. 334. 235,251 Hara, T.. 58. 5Y( 1.18). 60. 70 Harman, T. C.. lOO(2.51). lOl(2.53. 2.54), I02 Hart, P. B.. 60(1.20). 61, 70 Harth. W.. 200 Hartnagel, H. L., 2. 20, 38, 129(3.22), 133(3.29). 137 Hasegdwa, H . , 133(3.29). 137 Haus. H. A., 2. 3 . 8, 37 Hawkes. T. A.. 192(5.22), 201 Hawrylo, F. 2.. 55. 70. 75, 76, 77, 101 Hayashi, I., 75. 80. 81, X4(2.17), 86(2.17), 89. SY(2.1 I). 101 Hearn, R., 204. 205, 206, 207, 510,251 Heckscher, H., 92, 93(2.35), 102 Heinen, J.. 190(5.18), 200 Heinle. W.. 2.4, 20, 21. 38 Henkel, H. J . , 80(2.18). f01. 109(3.3), 137 Henry. C. H., 136(3.33).137 Herndon, M., 223,250 Heywang. W., 71(2.5), 101 Hillier. F. S., 235, 251 Hitchens, W. R.. 98(2.45). 102 Horn. M . . 204,251 Hofmann, K. R.. 2, 19,20.21,38 Hofstetter, E., 222, 250 Holland, M. G., 126(3.19). 137 Holmes. J . E.. 238,241, 243,252 Holonyak, N . Jr.. 40, 41,42,45, 46,48, 54(1.7). 66(1.26), 69, 70, 97(2.41, 2.42). 98(2.45), 99, 99(2.47), 102 Houseweight, K. B., 245. 246. 247, 248, 252 Howells, P. W., 204, 250 Hsiao. J. K., 220. 221. 250 Hsieh, J. J., 59, 70, 92, 93(2.35). 99(2.48), 102. 119(3.15), 137 Huelsman, L. P., 274, 329 Hughes, J . J., 4, 19. 37 Hunsperger. R. G., 120(3.18), 137, 196(5.28), 20 I Hunsperger. R. H., 196(5.26).201 Hurwitz, C. E., I19(3.15), 137

333 J

.

Jaros. M 129(3.22), 137 H , 88(2 31), 97(2.40), 101 Johnson, A. D., 160(5.4), 200 Johnson, R. S., 231,251 Jjuin.

K

Kalrnan. R. E., 238,252 Kaminow, I . P., 195(5.24). 201 Karninski. J. F.. 4'. 19. 20. 28, 36. 37 Kanbe, H., 19,38 Kaner, E. A,, 2 , 38 Kanyuck, A. J.. 245, 249, 252 Kataoka. S., 19, 38 Kato, D., 117(3.14), 137 Katzir. A,. 85(2.26), I0f Kavnitr. J., 204. 205. 206. 207. 210. 'if Kawashima, M., 19.38 Keck, D. B., 197(5.34), 201 Keramidas. V. G., 130(3.27), 137 Keune, D. L.. 40,41. 42,45, 46. 48, 54(1.7), 6Y. 70. 98(2.45), 99(2.47), 102, 160(5.4), 200 Khankina. S. I . , 2; 3 , 38 Kinnison. G., 217, 250 Kino, G . S., 1, 2, 3 , 14, 20, 21, 26, 27. 28. 29, 38 Klein, E., XO(2.18). 101 Kliiver. J. W.. 2 , 3 , 8, 37 Koepke, B., 196(5.29). ,701 Korn, S. R., 151(4.4), 177(5.X). 1.58. NO Kounerth, K. L., 172(5.7), 200 Kressel, H., 5 5 , 70, 72(2.9), 73(2.9). 75. 76, 77, 78(2.15), 89(2.32), 94(2.32), 98(2.44), 101, 102, 135(3.31), 137, 190(5.16). 191(5.19), 200, 201 Kretschmer, F. F., 208, 209, 220, 221, 250 Kroemer, H., 70(2.1), I01 Kroszczynski, J., 220.250 Kuckuck, H., 80(2.18). 101 Kumabe. K . , 19,38 Kuru. I., 19, 38

I L fida, S . , I16(3.1 I), 137 fzawa, T., 196(5.31),201

Labitt, M . , 222, 250 Ladany, I., 57, 70. 191(5.19). 201

334

AUTHOR INDEX

Landsberg, P. T.,72(2.8), 101 Lee, W., 196(5.27), 201 Leon, B., 290,330 Lewis, B., 208, 250 Lewis, B. L., 223,250 Lieberman, G. J., 235,251 Lind, D. A., 258,329 Lindgren, B. W., 236,251 Littauer, R., 278,282, 288,330 Llegems, M., 85, 101 Lo, W., 101(2.55),102 Locher, R. E., 177(5.8), 200 Lockwood, H. F., 75, 76.77, 101 Loebner, E. E., 71,101 Logan, R. A., 85,86, 87(2.27), 101 Lorimor, 0. G., 65(1.23), 70, 136(3.32), 137 Ludowise, M. J ., 54( 1.7), 66( 1.26). 70, 98(2.45), 99, 99(2.47), 102

Mause, K., 19, 38 Medved, D. B.,190(5.14), 200 Mehal, E. W., 129, 137 Meixner, H., 154(4.5), 158 Melngailis, I., 100(2.51), 102 Merz, J. L., 84(2.22), 87(2.22), 101 Mettler, K., 89(2.33), 102 Metz, L. S., 2, 38 Mihara, M., 58, 59(1.18), 60,70 Mikhailovskii, A. B.,2, 3. 38 Miller, 8. I., 65(1.25), 66, 70, 119(3.16), 137 Millman, J., 288, 330 Minagawa, S., 69(I.31), 70 Mitchell, R. L., 231, 251 Muehe, C. E., 222,250 Miiran, P. C., 65( 1.24), 70 Mutter, W. E., 161(5.5), 200

N M McAulay, R. J., 222, 250 McCool, J., 204,205, 206, 207,210,251 McCorrison, P. B., 222,250 McFee, J. H., 196(5.30),201 McGee, J. C., 196(5.30),201 Macksey, H.M., 97(2.42), 102 Makita, Y.,88(2.31), 97(2.40), 102 Mallett, J. D., 213, 214, 216, 2i7, 218, 250 Malmstadt, H., 290,299,330 Mantey, P., 206, 208, 251 Many, A,, 4, 5, 10, 13, 38 Marcante, A., 2, 20, 21, 37 Marcum, J. I., 224, 225, 251 Marcus, M. B., 236,251 Marcuse, D., 84(2.23), 101, 117(3.13), 137, 198(5.37), 199, 201 Marinace, J. C., 129(3.24), 137 Martin, R. J., 196(5.33), 201 Martin, W. E., 196(5.25),201 Maruska, H.P.,68, 70 Masuda, M., 2, 20. 21, 38 Matare, H. F.,50(1.5). 51(1.5, l.6), 55(1.5, 1.6). 68(1.29), 69, 70,71(2.3,2.4), 101, 127(3.21), 137, 165(5.6), 180(5.10), 181 (5. lo), l82(5.10), I83(5. lo), 190(5.14), 194(5.23), 199(5.38),200, 201 Matarese, R. J., 4, 19, 20, 28, 36, 37 Matsuo, Y., 2, 20, 21, 38 Maurer, R. D., 197(5.34),201

Nagano, M., 190(5.17), 200 Nahory, R. E., 67(1.27). 70. lOO(2.49, 2.50). 102 Nakagome, H., 196(5.31), 201 Nakahara, S., 190(5.17), 200 Nakamura. M., 85(2.26), 101 Namikazi, H., 190(5.17), 200 Napoli, L. S., 4, 19, 37 Nathanson, F. E., 219,250 Nelson, H., 72(2.9), 73(2.9), 101 Nelson, R. J., 97(2.41), 102 Neyman, J., 225,251 Nguyen, V. T., 196(5.30),201 Novikova, S. I., 127(3.20), 137 Nuese, C. J., 54( 1.8), 62( 1.22), 63, 64, 65(1.22), 70,97(2.43), 98(2.44), 102

0 O’Donnell, R. M., 222,250 Olsen, B. A., 233,251 Olsen, G . H.,97(2.43), 102

P Palmer, D. S., 227,251 Panish, M. B., 75, 77, 78, 80, 81, 84(2.17, 2.22), 86(2.17), 87(2.17, 2.22), 89(2.11). 101

335

AUTHOR INDEX

Paoli, T. L., 81, 101 Parker, R. K., 223,250 Pashitzkii, E. A,, 2, 3, 38 Pearsall. T. P., 65(1.25), 66, 70 Pearson, E. S., 225, 251 Petroff, P. M., 130(3.27). 136(3.32), 137 Pettit, G. D., 55, 70 Pilkuhn. M. H., 74. 75(2.10), 101 Pinnow, D. A,. 197(5.35).201 Pollack, M. A,, 67(1.27), 70, lOO(2.49, 2.50). 102 Potenski, R. M., 57( I . 16). 70 Powell, M . J. D., 213, 250 Pucilowski, J., 113(3.6), 137 Pugh, E. L., 216, 250

Q Quate, C. F.. 20, 26, 37 Queen, F. D., 234, 235,237, 238, 241, 243, 251.252 Queisser, H. J., 55(1.11), 70 Quigley. A. L.. 238. 240, 241, 243, 252

R Ralson, J. M., 136(3.32), 137 Ralston. 1. M., 148(4.2), 152(4.2), 158 Read, M. H., 130(3.27), 137 Reed, 1. S.,213,214,216,217,218, 250 Reed, L. S., 210.21 I , 213, 215, 216,250 Reinhart, F. K., 85, 86, 87(2.27), 101, 119(3.16), 137, 198(5.36),201 Rich, T. C., 197(5.35),201 Richter, K., 96, I02 Ridella, S., 19, 37 Rideout, V. L.. 130, 137 Riegler. R., 206, 250 Rihaczek, A. W., 203,250 4, 19.20.21, 36,37 Robinson, B. 9.. Robinson, G. Y., 19, 37 Robson, P. N., 2, 14, 20, 38 Roccasecca, D. D., 61(1.21), 70 Rossi, J. A., 92, 93(2.35), 102, 119(3.15), 137 Rubinstein. M., 132(3.28), 137 Rupprecht, H., 55, 70, 74, 75(2.10), 101 Rutz, R. F., 68, 69. 70

S Sahm, W. H., 111, 177(5.8), 200 Saitoh, T., 69(1.31), 70 Sandberg, G. L., 190(5.14),200 Saul, R. H., 55(1.12), 6I(l.21), 70 Schlachetzki, A,, 19, 38 Schleher, D. C., 229,251 Schlesinger, S. P., 223, 250 81, 101 Schlosser, W. 0.. Schulte. H. J.. 186(5.13).200 Schultz, P. C., 197(5.34),201 Schumann, R., 113(3.6), 137 Schwartz, B., 130(3.27), 137 Schwartz, M., 229,251 Schwartz, P. M.. 19,37 Schwartzman, S., 129(3.25). 137 Scifres, D. R., 83(2.21), 84(2.21, 2.24). 87(2.28), 88,97(2.42), 101, 102 Sea, R. G., 245. 246, 247, 248,252 Seaman, J., 177(5.9), 200 Sferrino, V. J., 222, 250 Shah, 9.R., 172(5.7),200 Shank, C. V., 85, 86, 87(2.27), I01 Shapard, J. M., 217,250 Shih, K. K., 58(1.17), 70 Shimizu, N., 19.38 Shrader, W. W.. 219, 223,250,251 Silberman, S. R., 236, 252 Singer, R. A,, 242, 245, 246, 247, 248, 249,252 Singh, R.. 133(3.29), 137 Sittler. R. W., 248, 252 Sklansky, J., 238. 252 Skolnik, M . I . , 219,223, 250 Smith, J. I., 262, 273, 274, 330 Somekh, S., 85, 101, 196(5.26),201 Soshea, R. W., lll(3.4). 137 Speer, R. S., 115(3.9), 137, 191(5.20),201 Spangle, P., 223, 250 Stein, J . J., 248, 252 Stevenson, D. A,, 68, 70 Stoll, H., 120(3.18), 137, 196(5.28),201 Stone, L. E., 56, 57( 1.15). 70 Streetman, B. G . , 97(2.41), 102 Streifer, W., 83(2.21), 84(2.21,2.24), 87(2.28), 88, 101. 102 Sumi, M., 2, 38 Sumski, S., 75, 77, 78, 89(2.1I), 101 Susskind, C., 2,3,37 Sweeney, L. E., 219,251 Swerling, P. S., 224, 226, 236, 251

336

AUTHOR INDEX

T Tajima. Y., 19, 38 Tdkata, K., 116(3.1 I), 137 Tateno, H., 19.38 Taylor, H. F., 191(5.21), 201 Tien, P. K., 196(5.33), 201 ?Lam. H., 19,20,38 Tobey, G. B., 274,329 Tomasetta, L. R., 101(2.52), 102 Tomlinson, W. J., 196(5.32), 201 Torrey, H. C., 183(5.1I), 200 Toyoda, N., 58, 59(l.l8), 60, 70 Triano, A., 4, 19, 20, 28, 36, 37 Trunk, G. V., 226, 227,229, 234,235, 237, 238, 241,243,249,251,252

U Umeda, J., 85(2.26), 101 Unger, H. G., I19(3.17), 137 Unger, R., 154(4.5), 158 Unno, Y.,Il6(3.11), 137 Ulmer, E. A,, Jr., 94(2.36), 102 Ulrich, R., 196(5.32), 201

V

Van Trees, H. L.. 21 1,212, 251 Varnerin, L. J., 196(5.33), 201 Velasquez, J., 113(3.6), 137 Vignos, J. H., 258, 265, 329

W

Walker, J. K., 231, 251 Walpole, J . N., lOl(2.53, 2.54), 102

Ward, H. R.,219,232, 251 Washburn, T. W., 219, 251 Weber, H. W., 196(5.32), 201 Wei, Y., 196(5.27), 201 Wemple, S. H., 196(5.33), 201 Westermeier, H., 190(5.18), 200 Weyrich, C., 96, 102 White, D. M., 236, 252 Whitmer, C. A,, 183(5.11), 200 Widrow, B., 204,205,206,207, 208, 210, 25 1 Williams, C., 204, 205, 206, 207, 210, 251 Williams, C. S., 140(4.l), l41(4.1), 158 Winstel, G., 89(2.33), 102 Wilson, J. D., 237,238, 241, 249, 251,252 Wittke, J. P., 89(2.32), 94(2.32), 102, 190(5.16), 191(5.19), 200, 201 Wolf, H. D., 96, I02 W o k , C. M., 119(3.15), 137 Wolford, D. J . , 97(2.41), 102 Woodall, J. M., 55, 57(1.16), 70 Wright, P. D., 54(1.7), 70, 99(2.47), 102

Y Yakovenko. V. M., 2, 3,38 Yariv, A., 85(2.26), 101, 120(3.18), 137, 196(5.26, 5.28), 201 Yen, H. W., 85(2.26). 101

Z

Zack, G. W., 97(2.42), 102 Zeidler, J., 204, 205, 206, 207, 210. 251 Ziegler, G., 109(3.3). 137 Zielasek, G., 55( L9), 70 Zook, J. D., 196(5.29), 201 Zschauer, K. H., 190(5.18), 200

Subject Index A

Binary integrator, 229 Blackbody radiator, luminance of, 141 Boundary conditions, for carrier-stream surface models, 9-19 Bragg reflector, 85 Burglar alarm system, LEDs in, 164 Buried heterostructure, 88

ABSF, see Antenna beam shape factor Active filters, in electronics course, 274-278 A/D, see Analog-to-digital (dj.) Adaptive array gain, in radar signal processing, 215-216 Adaptive noise canceling concept, 205 Adaptive receiving array, in radar signal processing, 214 Adaptive thresholding, 23 1-233 ADT, see Automatic detection and tracking systems Amphoteric impurities, in light-emitting devices, 135 Amplifier inputs, teaching of, 260-26s Analog-to-digital converter, 300-301 successive-approximation type, 313-318 Analog-to-digital techniques, in electronics course, 298-301 AND, OR logic circuits, 270 Anisotropic film longitudinal propagation of waves in, 29-36 negative anisotropy in, 34-36 positive anisotropy in, 32-34 with zero diffusion, 28 Anisotropy, positive and negative, 32-36 Antenna beam shape factor, 225 Automatic detection and tracking registers, 224 Avalanche photodiode, 124

C

Capacitors, in electronics course, 272 Carrier densities, degeneration limit and, 89 Carrier stream boundary effective, 6 with no real surface sources and without surface recombination, 15-17 with real surface sources and surface recombination, 17-19 Carrier-stream surfaces boundary conditions for particular models Of, 9-19 in nondegenerate semiconductor plasmas, 1-19 Cathode ray tube displays, 159 Chemical vapor deposition, 49-50, 64 growth of junctions and heterojunctions using, 119 CFAR, see Constant false-alarm rate CMOS (complementary metal oxide surface), 269, 290 Constant false-alarm rate, 224, 234 front-end cell-averaging, 232 loss in, 231 in radar signal processing, 204 Contact geometry, 130 Correlation logic, in radar signal tracking systems, 243-248 Crystal boundary, real, see Real crystal boundary CVD, see Chemical vapor deposition

B Band-structure enhancement, 40 Batch processor, cell-averaging CFAR and, 23 1 Batch processor integrator, 230 Biasing techniques, in electronics course, 284-286 337

SUBJECT INDEX

338 D

D/A, see Digital-to-analog techniques Data-acquisition system, sixteen-channel, 318-322 DEC, see Digital Equipment Corp. LSI-11 processor Degeneration limit, carrier densities above, 89 Delay time, vs. decay time, 91 Detectors modified generalized signal test, 234-245 nonparametric, 233-235 sequential, 235-236 Device degradation, defined, 133-134 DHL, see Double heterogeneous laser Differential quantum efficiency, 149 Digital circuits, in electronics course, 268-272, 290-292 Digital computer, design of, 302-304 Digital Equipment Corp. LSI-11 processor, 305 Digital-to-analog techniques, in electronics course, 298-301 Diodes, in electronics course, 273-274 Diode-transistor logic, 269 Dispersion equations, for quasistatic waves, 20-28 Displays and indicators, LEDs in, 158-200 Doppler processing, in radar signal processing, 222-223 Double heterojunction laser, 72, 84, 86-87, 190 chemical vapor deposition and, 119 delay time for, 91 mode control in, 80-81 p-n junction pulsing in, 91 surface contact in, 116 time resolved spectra for, 90 DRV-11 parallel line interface, in electronics course, 312-322

E ECL, see Emitter-coupled logic Effective boundary, of carrier stream, 6, 9-10 Electronics course (for scientists), 253-329 advanced linear devices and techniques in, 292-297

current sources in, 280-282 digital circuits in. 290-292 digital-to-analog and analog-to-digital techniques in, 298-301 DRV-11 parallel line interface in, 312-322 feedback in, 288-290 first quarter program in, 260-278 grading in, 328-329 input-output operations in, 306-312 lecture topics and laboratory assignments in, 322-328 LSI-11 processor in, 305-306 organization for, 258-260 second quarter program in, 278-301 simple cornputer design in, 302-304 simple linear circuits in, 260-268 sixteen-channel data-acquisition system in, 318-322 successive-approximation A/D converter in, 313-318 third quarter program in, 301-322 transistors in, 282-288 voltage sources in, 278-280 wave-shaping techniques in, 280-282 Emitter-coupled logic, 269-270, 290 Energy band diagram, for two-valley semiconductors, 43 Equivalent surface sources, 4-8

F Fabry-Perot cavity, 75 as complex structure, 80 Far-infrared lasers, 94-101 Feedback, in electronics course, 288-290 Feedback integrator detectors, 228 FET (field effect transistor), in OTA feedback loop, 293-294 Fiber optics bundle, multimode, 189 Fiber optics interconnectors, 189 Fiber optics transmission, for military aircraft, 191 Fiber optics transmitters, 118 Field effect transistor, in OTA feedback loop, 293-294 Film projector sound head, LEDs in, 168 Filters, in electronics course, 274-278 First-order perturbations, in electronics course, 265

339

SUBJECT INDEX Flow control, LEDs in, 171 Four-layer heterojunction, 190 Free-surface model, in thin film semiconductor research, 25-28 Frequency shift keying (FSK), 297

G Gallium arsenide heterojunction detector,

feedback, 228 moving window, 227-228 in noncoherent detector, 227-230 two-pole, 228 Interrupts, in electronics course, 310-312 Isoelectronic trap, effects of,40 Isotropic flow, in thin film semiconductors, 26-27

Isotropic mobility. for wave propagation in anisotropic films, 31

187

Gallium arsenide homojunction laser, 95 Gallium arsenide waveguides, 179 Gunn amplifier, traveling-wave, 186 Gunn device, in photodetector-amplifer, 122 Gunn oscillator, 120 in optical microwave circuits, 182-183 self-pumped optical detector and, 188

J

Junction current, luminous efficacy and, 48 Junction performance, progress in, 40-55

K H Heterojunction laser, 72 double, see Double heterojunction laser four-layer, 190 Heterostructure waveguides, 198 Homojunction laser, 72, 74 Hybrid microwave optical circuits, 182-184 Hyperbolic modes, dispersion equation for, 29

I Index of refraction of air and compound semiconductor, 102-103

of semispherical lens, 106 Inductors, in electronics course, 272-273 Infrared LEDs, general applications of, 163-174

Infrared parametric amplification, 171 Input offset current, 266 Input offset voltage, 266 Inspection systems, LEDs in, 165-170 Integrated optics, 189-200 light-emitting devices and, 117 Integrators batch processor, 230 binary, 229

Kalman filter, 238 Kino surface sources, 3

L Large optical cavity laser, 72, 82, 190 delay time for, 91 efficency of, 76 surface contact in, 116 Laser buried heterostructure, 88 compared with LED, 89,93-94 degradation of, 133-136 DH, see Double heterojunction laser gain efficiency in, 71-80 gallium arsenide homojunction, 95 large optical cavity, 72. 82, 190 modulation frequency for, 89-93 passivation of, 131-133 photoluminescent decay in, 96 power efficiency of, 75 quantum efficiency in, 71-80 radiation pattern of, 80-88 SCH (separate component heterostructure), 32-85 taper-coupled, 86 in ternary compound material, 97 types of, 72 visible and far-infrared, 94-101

340

SUBJECT INDEX

Laser emission, temperature dependence in, 89 Laser production, light output and, 105 Laser-structure interface, corrugated, 83 Laser structures, stripe geometry for, 115-116 Laser technology chemical vapor deposition in, 71 device lifetime in, 191 progress in, 70-101 Least mean square adaptive algorithm, 206 LEDs, see light-emitting devices LED technology, progress in, 55-69 LED watch, 162 Light-emitting crystal surface, as diffuse radiator, 138 Light-emitting devices, 39-101 applications of, 158-200 band-structure enhancement and, 40 bandwidth measurements for, 152 compounds and substrates in, 60-69 construction of, 110-1 11 contrast enhancement for, 111 cross section of, 115 defect density in, 136 degradation of, 133-136 design and technologies for, 40-69 device types and technological progress in, 102- 136 as displays and indicators, 158-200 efficent use of, 145 emitters and current sources for, 156 in flow control, 170-171 gallium arsenide in, 78-79 high-efficiency, 78 infrared, 163-174 in inspection systems, 165-170 interconnection metallurgy for, 160 junction performance and, 40-55 junction plane light from, 110 laser compared with, 89, 93-94 luminous efficacy of, 40-41 luminous efficiency of, 141 measurement techniques for, 138-158 modulation frequency for, 89 in monoblock photodetector modulator, 125 multijunction, 112 multiple layer structures in, 190 multivibrator modulator for, 164 null indicator with, 173

in optical communications and integrated optics, 189-200 optical problems in, 102-1 17 in optical radar, 177-188 in oscilloscopes, 172-174 output vs. junction current for, 48 output power and efficiency of, 142-151 output power vs. wavelength in, 154-155 passivation and degradation of, 13 1-136 in photocouples, 174-177 in photoelectric amplifier, 122 photon flux of, 143 pulsed operation of, 164 quantum efficiencies of, 47 risetime measurements for, 151-152 Schottky barrier and, 124 semispherical epitaxial-diffused, 56-57 silicon as dopant for, 79 special circuits with, 172-174 spectral intervals for, 149 spatial power distribution curve for, 112 surface morphology in, 58 technological progress in, 55-69 thermal and contact problems in, 125-131 thermal impedance of, 153-154 for watches, 162 Light guiding structures, in photoresist films, 196 Light-pumped converter, Schottky barrier as, 123 Linear circuits, in electronics course, 260-268 Linear expansion coefficients, of Group 111-V compounds, 127 Liquid phase epitaxy heterojunctions and, 100 in high-efficiency LEDs, 78 surface morphology in, 58 vs. vapor phase epitaxy, 52-54 Liquid phase epitaxy devices, efficiency of, 160 Liquid phase epitaxy growth, for double heterostructure devices, 197 LMS (least mean square) adaptive algorithm, 206 LOC, see Large optical cavity laser Logic circuits, in electronics course, 269-272 Logic families, teaching of, 269 Luminous efficacy defined, 40 junction current and, 48

34 1

SUBJECT INDEX

Luminous efficiency defined, 42 enhancement of, 42 Luminous energy, defined, 142 Luminous flux calculation of, 130 defined, 142 Luminous flux density defined, 142 per unit solid angle, 142 Luminous intensity, defined, 142 Luminous output, as function, 130-131

M Mesa-stripe geometry device, 118 MGST, see Modified generalized sign test detector Military aircraft, fiber optics transmission for, 191 Modified generalized sign test detector, 234-235 Modulation frequency, for lasers, 89-93 Moving target detector system, 204 Doppler processing in, 222-223 Moving target indicator noncoherent, 223-224 in radar signal processing, 204, 219-223 two-pulse, 220 Moving target indicator filter, frequency response curve for, 221 Moving target indicator gain, 217 Moving window detectors, 227-228 MTD, see Moving target detector system MTI, see Moving target indicator Multivibrator modulator, for LED, 164

N Narrow gap semiconductor compounds, 100 Noncoherent detection classical theory in, 225-226 false alarms and, 230-234 integrators in, 227-230 nonparametric detectors in, 233-235 in radar signal processing, 224-236 sequential detectors as. 235-236

Noncoherent moving target indicator, 223-224 Nondegenerate semiconductor plasmas, 1-19 0

Operational amplifier, in electronics course, 262-265 Operational amplifier inputs, nonsaturated, 260-264 Operational transconductance amplifier, 292-294 Optical communications, 189-200 Optical radar, LEDs in, 177-188 Optoelectronics, next-generation, 197-200 Opto-hybrid microwave circuit, 183, 185, 188 Oscillators, voltage-controlled, 292, 296 Oscilloscopes, LEDs in circuitry of, 172-174 OTA, see Operational transconductance amplifier

P Parametrically related frequencies, 182 Parametric conversion, waveguides in, 179 Passivation, of light-emitting devices, 131-133 Perturbations, first- and second-order, 265-267 Phase-lock loop, 292,295-297 Photoconductor as high-frequency modulated light detector, 120 transit time reduction factor for, 121 Photocouplers, 174- 177 Photo-Darlington coupler, 175 Photoelectric amplifier, 122 Photoluminescent decay, 96 Photoresist films, light-guiding structures in, 196 Phototransistor coupler, 176 P I N detectors, 163 PIN diode, 120-121 Planck’s radiation law, 140 Plasma media, electrodynamic boundary conditions for, 2-3 PLL, see Phase-lock loop p-n junction, puking of, 91 Pulse modulation frequency limitation, 178

342

SUBJECT INDEX

0 Quasi-direct crystal, 40 Quasistatic waves, dispersion equations for, 20-28

R Radar signal processing, 203-249 adaptive arrays and radars in, 210-219 array gain in, 214-216 coherent processing in, 204-224 Doppler processing in, 222-223 false alarms in, 230-234 moving target indicators in, 219-223 noncoherent detection in, 224-236 radar integration in, 248-249 tracking system in, 236-249 RCA 3080 Operational Transconductance Amplifier, 292-294 Real crystal boundary, 11-15 with no surface sources, 11-12 with surface sources and surface recombination, 12-15 without surface recombination, 11-12 Real surface sources, defined, 7 Refractive index, of air vs. compound semiconductor, 102-103 Resistor-transistor logic (RTL), 269

s Saturated output circuits, in electronics course, 267-268 SCH laser, see Separate component heterostructure laser Schottky barrier, back biasing of, 194 Schottky barrier diodes, 121-124 as light-pumped converter, 123 mixer for, 187 Scientists, teaching electronics to, 253-329 SCR, see Silicon-controlled rectifier Second-order perturbations, in electronics course, 265-267 Self-pumped optical detector, 187-188 Semiconductor(s) thin film, see Thin film semiconductors

traveling wave, 123 two-valley, 4345 Semiconductor compounds, narrow-gap, 100 Semiconductor laser, 70-71 see also Laser; Laser technology Semiconductor plasmas, nondegenerate, 1-19 Semiconductor surface models, 4-8 Semiconductor surfaces, electronic structure and electrical properties of, 4-8 Separate component heterostructure laser, 82-85 Sequential detectors, 235-236 Sequential likelihood ratio test, 235 SHL, see Single heterojunction laser Sidelobe cancelers, 204-210 Silicon-controlled rectifier, 174 Silicon readout amplifier, 164 Simple digital circuits, in electronics course, 268-272 Simple linear circuits, teaching of, 260-268 Single heterojunction devices, 190 Single heterojunction laser, 72-73 degradation curves for, 135 delay time for, 91 structure of, 135 SLRT, see Sequential likelihood ratio test Snell’s law, 103, 107 Stack, in electronics course, 306-308 Stripe geometry, for laser structures, 115-116 Stripe laser, 118 Subroutine “call” instruction, 309 Subroutines, in 1 / 0 programming, 308-310

T Taper-coupled laser, 86 Ternary compound material, laser operation in, 97 TFSS, see Thin-film semiconductor structures Thermal conductivity, vs. temperature, 126 Thin-film semiconductor amplifier amplification mechanism of, 34-36 positive anisotropy in, 32-34 Thin-film semiconductors and anisotropic film with zero diffusion, 28 and dispersion equations for quasistatic waves, 20-28 free-surface model and, 25-28 isotropic film in, 26-27

343

SUBJECT INDEX

longitudinal propagation of waves with zero diffusion in, 29-36 normal modes in, 19-36 wave interactions in, 1-37 Tracking system correlation logic in, 243-248 maneuver-following logic in, 240-242 radar integration in, 248-249 in radar signal processing, 236-249 system outline in, 237-238 tracking filters in, 238-240 Track-while-scan system, 236-238 Transistors applications of, 286288 in electronics course, 272-273, 282-288 field-effect, 293-294 1 / 0 characteristics of, 282-284 operation of, 282-284 Transistor-transistor logic, 269, 272, 290 Transverse electrical mode, in laser radiation pattern, 81 Traveling-wave Gunn amplifier, 186-187 Traveling-wave semiconductor amplifier, 123 Trigonometric modes, dispersion equation for, 30 TTL, see Transistor-transistor logic Two-pole filter integrator, 228 Two-valley semiconductor, energy band diagram for, 43

U UHF oscillators, 120-121

v Vapor phase epitaxy, 52 mass production and, 60 Vapor phase growth system, 63 VCO, see Voltage-controlled oscillators Vegard’s law, 66 Visible-frequency lasers, 94- 101 Voltage-controlled oscillators, 292, 296 W

Waveguides band-filter effect between, 195 coupled optimal, 197 energy coupling in, 195 in parametric conversion, 179 Wave interactions, in thin-film semiconductor structures, 1-37 Wave propagation, positive anisotropy in, 32-34 Weierstrass sphere, 106 construction of, 108 LEC with, 116 Widrow-Hoff LMS algorithm, 206

Cumulative Author Index, Volumes 1-45 A

Ables, H. D.: see Hewitt, A. V. Ables, H. D.: see Kron, G. E. Abraham, George: Multistable semiconductor devices and integrated circuits, XXXV, 269 Abraham, J. M., Wolfgang, L. T., and Inskeep, C. N.,: Application of solid-state elements to photoemissive devices, XXII B, 671 Abrahams, E.: Relaxation processes in ferromagnetism, VI, 47 Aceto, P.: see Roehrig, H. Adams, J.: X-ray detection by channel electron multipliers, XXII A, 139 Adams, K. M., Depretterne, E. F. A,, and Voorman, J. 0.:The gyrator electronic systems, XXXVII, 79 Ahmad, N., Gale, B. C., and Key, M. H.: Time resolution limitations in single-stage image converter photography, XXVII B, 999 Aihara, S.: see Shimizu, K. Aikens, R.: see Hynek, J. A. Airey, R. W.: see Delori, F. C. Airey, R. W.: see McGee, J. D. Alexander, J. W. F., and Burtt, R. B.:Bombardment-induced conductivity targets for image orthicons, XVI, 247 Allan, F. V., and Garfield, B. R. C.: The study of photocathode composition by microbalance methods, XVI, 329 Allen, J. Denton: see Malting, L. R. Allen, J. Denton: The Manner IV space-craft television system, XXII B, 849

Allen, K. G.R., Anderson, B. E., Boksenberg, A., and Oliver, M. B. Properties of the detection system for the International Ultraviolet

Explorer satellite, XL A, 223 Allen, K. G.R., Anderson, B. E., Boksenberg, A., and Ross, D. G.: Wavelength dependent resolution in the far ultraviolet for proximity focused imaging from a caesium telluride photocathode, XL A, 449 Alpern, M., Bijaoui, A,, and Duchesne, M.: Sur le gain en sensibilite, dans l'infra-rouge Proche, de la camera electronique par rapport a la photographie classique, XXII A, 5 Alsberg, H.: see Beyer, R. R. Alsberg, H., and Hartman, R. E.: High resolution electron microscope imaging with silicon diode array target vidicons, XL A, 287 Amboss, K.: The analysis of dense electron beams, XXVI, 1 Anderson, A. E.: see Wachtel, M. M. Anderson, A. E., and Schneeberger, R. J.: Limitations to resolving power in electronic imaging, XVI, 299 Anderson, B. E.: see Allen, K. G.R. Anderson, B. E.: see Sturgell, C. C. Anderson, B. E.: Properties of the detector system for the International Ultraviolet Explorer satellite, XL A, 223 Anderson, D. G.: see Flanagan, T. P. Anderton, H.: An x-ray image intensification system for use with a point projection x-ray microscope, XXII, B, 919 Anderton, H., and Beyer, R. R.:Dynamic imaging with television cameras, XXVIII A, 229 Arndt, U. W., and Gilmore, D. J.: A television x-ray diffractometer, XXVIII A, 229 Arndt, U. W., Gilmore, D. J., and Boutle, S. H.: Television recording and analysis of x-ray diffraction patterns, XXXIII 8, 1069 Asano, M.: see Hirashima, M. 344

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

Ashworth, F.: Field emission microscopy, Ill, 1 Aslam, M.: see McGee, J. D. Aukerman, L. W.: see Seib, D. H. Auman, J. R.: see Walker, G. A. H. Authinarayanan, A., and Dudding, R. W.: Changes in secondary electron yield from reduced lead glasses, XL A, 167

B Bacik, H.: see McGee, J. D. Bacik, H. Coleman, C. I., Cullum, M.J., Morgan, B. L., Ring, J. and Stephens, C. L.: The analysis of direct Spectracon exposures obtained on the Isaac Newton telescope, XXXIII 8. 747 Bailey, P. C.: see Garfield, B. R. C. Bakken, H.: see McGee, J. D. Bakos, G.: see Hynek, J. A. Bakos, J. S.: Multiphoton ionization of atoms, XXXVI, 57 Baldinger, E., and Frazen, W.: Amplitude and time measurement in nuclear physics, VIII, 255 Ball, Jack. Niklas, Wilfred F., Dolon, Paul J., and Ter-Pogossian. M.: Image intensifying chains for medical scintillation cameras, XXII B, 927 Baranne, A., and Duchesne, M.: Un montage de spectrographe specialement adapte a une camera electroniaue de type Lallemand, XL B, 641 Barford, N. C.: Electron beam scanning, XXXIII A, 535 Bargellini, P. L., and Rittner, E. S.: Advances in satellite communications, XXXI, 119 Barlow, G. E., Ovenstone, J. A., and Thonemann, F. F.: Automatic data processing in the physical sciences, XI. 185

Barnett, M. E., Bates, C. W.. Jr., and England, L.: Electron optics of a photoconductive image converter, XXVlll A, 545 Barnes, Aaron: Theoretical studies of the large-scale behaviour of the solar wind, XXXVI, 1 Barton, G.: see Hynek, J. A.

345

Barybin, A. A,: Electrodynamic concepts of wave interactions in thin-film semiconductor structures, I, XLIV, 99 Barybin, A. A,: Electrodynamic concepts of wave interactions in thin-film semiconductor structures. 11, XLV, 1 Baskett, J. R.: see Liu, J. D. Bates, C. W., Jr.: see Barnett, M. E. Bates, C. W., Jr.: Scintillation processes in thin films of CsI(Na) and CsI(TI) due to low energy x-rays, electrons and protons, XXVIII A, 451 Bates, David, J., Knight, Richard I., Spinella, Salvatore, and Silzars, A r k : Electronbombarded semiconductor devices, XLIV, 221 Batey, P. H., and Slark, N. A.: Performance of the transmission secondary-electron image intensifier, XXII A, 63 Baudrand, J., Combes, M., Felenbok, P., Fort, B., and Picat, J. P.: Development of a new kind of Lallemand camera, XXXlIl A, 7 Baum, W. A., see Frederick, L. W. Baum, W. A,: see Hall, J. S. Baum, W. A,: see McGee, J. D. Baum, W. A,: see Wilcock, W. L. Baum, W. A.: A critical comparison of image intensifiers for astronomy, XXVIII B, 753 Baum, W. A,: Laboratory evaluation of image tubes for astronomical purposes, XVI, 391 Baum, W. A,: Magnetic focusing of image tubes, XXII A, 617 Baum, W. A,: The potentialities of photoelectronic imaging devices for astronomical observations, XII, 1 Baum, W. A,, Busby, D. M., and Pettauer, T. V.: The stabilization of planetary images, XXXIII B, 781 Baumgartner, W.: A light amplifier with high light output, XXVIII A, 151 Baumgartner, W., and Gilliard, B.: Space charge in channel multipliers, XL A, 113

Baumgartner, W., and Zimmerman, U.: A high-gain channel electron multiplier (CEM) array and some of its operational characteristics, XXIll A, 125

346

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

Beauvais, Y.,and Blamoutier, M.: Integrating ultraviolet sensitive camera tube, XL A, 201 Beaver, E. A., Harms, R. J., and Schmidt, G. W.: Digicon applications in astronomy, XL B, 745 Beaver, E. A., Mcllwain, C. E., Choisser, J. P., and Wysoczanski, W.: Counting image tube photoelectrons with semiconductor diodes, XXXIlI B, 863 Beckman, J. E.: Application of information theory to the evaluation of two image intensifier tubes, XXII A, 369 Beckman, J. E., and Egan, D. W.: A search for molecular hydrogen in the interstellar medium, XXVIII B, 801 Beesley, J., and Norman, D. J.: Highresolution phosphur screens, XXII A, 55 1 Bell, A. E.: see Swanson, L. W. Bellier, Mlle M.: see Wlerick, G. Bent, G. J.: see Geneux, E. Bennet, A. W.: see Mayo, B. J. Berenyi, D.: Recent applications of electron spectroscopy, XLII, 55 Berg, A. D., Smith, R. W., and Prosser, R. D.: An electron image store and analyser, XXII B, 969 Berger, Harold, see Niklas, Wilfrid, F. Bergeson, H.E., and Cassiday, George L.: On the teaching of electronics to scientists, XLV, 253 Beurle, R. L.: see Hodgson, R. M. Beurle, R. L., and Jenkinson, G. W.: A charge image storage tube for character recognition, XXVIlI B, 1043 Beurle, R. L., and Slark, N. A.: An experimental image storage tube for the detection of weak optical images of low contrast, XII, 247 Beurle, R. L., and Wreathall, W. M.: Aberration in magnetic focus systems, XVI, 333 Beurle, R. L., Daniels, M. V., and Hills, B. L. : Image intensifier design and visual performance at low light-levels, XXVIII B, 635 Beurle, R. L., Hodgson, R. M., and Gelade, C. A.: Visual thresholds using high-grain image-intensifying systems, XXXIII B, 63 1

Beyer, R. R.: see Anderton, H. Beyer, R. R.: see Boerio, A. H. Beyer, R. R.: see Collings, P. R. Beyer, R. R.: see Fenner, E. Beyer, R. R., and Alsberg, H.: The development of an intensifier-vidicon for space applications, XXXIII B, 937 Beyer, R. R., and Goetze, G. W.: An optically scanned SEC camera tube, XXII A, 24 1 Beyer, R. R., Green, M., and Goetze, G. W.: Point-source imaging with the SEC target, XXII A, 251 Bhatia, T. B., Bride, G. K., Ghosh, C., Kelkar, G. N., Srininasan, M., Varma, B. P., and Verma, R. L.: Photoelectronic device development and related research at B.A.R.C., XL A, 409 Bied-Charreton, P., Bijaoui, A., Duchesne, M., and Le Contel, J. M.: Sur quelques progres recents apportes a la camkra electronique a focalisation electrostatique et sur son application en physique et en astronomie, XXVIII A, 27 Bijoui, A,: see Alpern, M. Bijaoui, A,: see Bied-Charreton, P. Billig, E., and Holmes, P. J.: Defects in diamond-type semiconductor crystals, X, 71 Binnie, D. M., Jane, M. R., Newth, J. A., Potter, D. C., and Walters, J.: Work at Imperial College, London, on the use of image intensifiers in nuclear physics, XVI, 501 Biondi, Manfred A,: Atomic collisions involving low energy electrons and ions, XVII, 67 Bird, P. R., Bradley, D. J., and Sibbet, W.: Photochron 11: an image tube for sub-picosecond chronography, XL A, 51 Blake, J., and Burtt, R. B.: Image orthicons with magnesium oxide targets, XVI, 213 Blamoutier, M.: Integrating ultraviolet sensitive camera tube, XL A, 201 Blamoutier, M.: Un tube de prise de vues sensible aux rayons X, XXVIII A, 273 Blewett, John P.: Recent advances in particle accelerators, XXIX, 223 Bloch, F.: see Brillouin, L.

CUMULATIVE AUTHOR INDEX, VOLUMES

Bloom. J. H.:see Shepard, F. D., Jr. Boerio, A. H.: see Goetze, G. W. Boerio, A. H., Beyer, R. R., and Goetze, G. W.: The SEC target, XXII A, 229 Bogdanov, E. V.: see Kislov, V. Ya. Boischot, A., and Denisse, J. F.: Solar radio astronomy, XX, 147 Boksenberg, A,: see Allen, K. G. R. Boksenberg, A,: see Anderson, B. E. Boksenberg, A,: see Sturgell, C. C. Boksenberg, A,, and Burgess, D. E.: An image photon counting system for optical astronomy, XXXIII B, 835 Boksenberg, A,. and Newton, A. C.: An electromechanical picture signal generating device, XXVIII A, 297 Boksenberg A., Burgess, D., Fordham, J. L. A., Shortridge, K., and Wright. S. L.: Astronomical observations with the University College London image photon counting system, X L B, 877 Bostock. D.: see McMullan, D. Boulmer. J.: see Delpech, J.-F. Boussage, C.: see Rosch, J. Boutle. S. H.: see Arndt. U. W. Boutot, J. P., Eschard, G.. Polaert, R., and Duchenois, V.: A microchannel plate with curved channels: an improvement in gain, relative variance and ion noise for channel plate tubes, XL A, 103 Bouwers, A.: Low brightness photography by image intensification, XVI, 85 Bowen. J. S.: see Dennison. E. W. Bowers, Michael T.and Su. Timothy: Thermal energy ion-molecule reactions, XXXIV, 223 Bowers, Raymond: see Frey, Jeffrey Bowhill, S. A., and Schmerling, E. R.: The distribution of electrons in the ionasphere, XV, 265 Bowles. K. L.: Radio wave scattenng in the ionosphere, XIX, 55 Boyer, L. A,: see Flory. L. E. Boyes, E. D.: see Turner, P. J. Bradley, D. J.: see Bird, P. R. Bradley, D. J., and Majumdar, S. : Application of electron-optical deflexion and storage techniques to time-resolved interference spectroscopy. XXII B, 985

Bradley, D. J., Liddy, B.. Roddie. A. G.,

145

347

Sibbett, W., and Sleat. W. E.: Picosecond chronography with image tubes, XXXIII B, 1145 Brand, P. W. J. L.: see Smyth, M. J. Brand, P.W. J. L., and Smyth, M. J.: Use of a Lenard-window image tube for astronomical spectrophotometry, XXII B, 741 Brand, P. W. J. L., and Wolstencroft, R. D.: Recent astronomical applications of a Spectracon, XXVIII B, 783 Branscomb, L. M.: Negative ions, IX, 43 Bratton, J.: see Pollehn, H. Brauer, W.: see Hachenberg, 0. Bride, G. K.: see Bhatia, T. B. Brillouin, L.: Electronic theory of the plane magnetron, 111, 85 Brillouin, L., and Bloch, F.: Electronic theory of the cylindrical magnetron, 111, 15

Broerse, P. H.: Electron bombardment induced conductivity in lead monoxide, XXII A, 305 Brooks, F. P., Jr.: Recent developments in computer organization, XVIII, 45 Brooks, H.: Theory of the electrical properties of germanium and silicon, VII, 85 Broussaud, G., and Simon, J. C.: Endfire antennae, XIX, 255 Brown, J.: Microwave optics, X, 107 Bruin, Frans: The autodyne as applied to paramagnetic resonance, XV. 327 Buchholz, V. L.: see Walker, G. A. H. Buchholz, V. L., Walker, G. A. H., Glaspey, J. W., Isherwood, B. C., and LaneWright, D.: The use of linear silicon array for astronomical spectroscopy, XL B, 879 Bulpitt, C. J.: see Delori, F. C. Burgess, D. E.: see Boksenberg, A. Burns, J., and Neumann, M. J.: The channeled image intensifier, XI], 97 Burstein, E., and Egli, P.H.: The physics of semiconductor materials, VII, 1 Burtt, R. B.: see Alexander, J. W. F. Burtt. R. B.: see Blake, J. Busby, D. M.: see Baum, W. A. Butler, D. J.: see Garfield, B. R. C. Byatt, D.: Bright displays for radar applications, XVI, 265

348

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

C

Calderwood, J. H.: see Smith, C. W. Caldwell, D. 0.:see Hill, D. 0. Caldwell, D. 0.: Scintillation chamber comparisons: fibres u. NaI and image intensifiers u. orthicons, XVI, 469 Campagna, M., Pierce, D. T., Meier, F., Sattler, K., and Siegmann, H. C.: Emission of polarized electrons from solids, XLI, 113 Capone, B. R.: see Shepherd, F. D., Jr. Carruthers, G. R.: Further developments of magnetically focused internal-optic image convertors, XXXIII B, 881 Carruthers, G. R.: Internal-grating electronographic spectrographs for the far-ultraviolet and X-ray wavelength ranges, XXXIII B, 895 Carruthers, G. R., Kervitsky, A,, and Opal, C. B.: Some applications of microchannel plates to electronic imaging devices, XL A, 91 Cartwright, P.: see Turner, P. J. Cassiday, George L.: see Bergeson, H. E. Castaing, Raymond: Electron probe microanalysis, XIII, 317 Catchpole, C. E.: see McGee, J. D. Catchpole, C. E.: Measurement of the spatial frequency response of image devices, XXII A, 425 Catchpole, C. E.: X-ray image intensification using multistage image intensifiers, XVI, 561 Chalmeton, V., and Eschard, G.: Reduction of the relative variance of the singleelectron response at the output of a microchannel plate, XXXIII A, 167 Chang, I. F.: see Kazan, B. Charles, D. R.: see Guillard, C. Charles, D. R., and Duchet, M.: Visible and x-ray image devices working on the induced conductivity principle, XXII A, 323 Charles, D. R., and Le Carvennec, F.: Infrared pick-up tube with electronic scanning and uncooled target, XXXIII A, 279 Charman, W. N.: Cosmic rays and image intensifier dark current, XXIII B, 705 Charman, W. N., and Hewitt, A. V.:The

influence of temperature on the performance of a cascade image intensifier, XXII A, 101 Charrier, Mlle, S., and WICrick, G.: Proprietes des Photocathodes Liberees dans un Vide Eleve, XVI, 5 Chatterton, P. A.: see Smith, W. A. Chenette, Eugene, R.: Noise in semiconductor devices, XXIII, 303 Chernov, Z. S.: see Kislov, V. Ya. Chevillot, A.: see Picat, J. P. and Heckathorn, H. M.: Chincarini, G., Electronography of extended objects, XL B, 791 Chodorow, M.: see Warnecke, R. R. Choisser, J. P.: see Beaver, E. A. Choisser, J. P.: Recent developments in the use of parallel and self-scanned diode arrays t o detect photoelectrons, XL, B, 735

Choudry, A.: Characteristics of an optically scanned SEC device, XL A, 253 Choudry, A., Goetze, C. W., Nudelman, S., and Shen, T. Y.: Photoelectronic image recording device optimized for high detective quantum efficiency, XXXIII B, 903

Churchill, J. L. W., and Curran, S. C., Pulse amplitude analysis, VIII, 317 Clarke, J. A,: see Yeadon, E. C. Clayton, R. H., and Gumnick, J. L.: Use of the image dissector in photocathode research, XXII A, 507 Clement, G.: An ultra-fast shutter tube for exposure times below 0.5 nanoseconds, XXXIII B, 1131 Cochrane, J. A,, and Thumwood, R. F.: The effects of high electric fields on photocathodes, XL A, 441 Cohen, M.: see Kahan, E. Cohen, M., and Kahan, E.: Linearity and optimum working density of optical and nuclear emulsions, XXXIII A, 53 Coleman, C. I.: see Bacik, H. Coleman, C. 1.: see McGee, J. D. Coleman, C . I.: The detective quantum efficiency of the Spectracon, XL B, 661 Coleman, C. I., Reay, N. K., and Worswick, S. P.: Spectracon observations of planetary nebulae, XL B, 817 Coleman, L. W.: see Thomas, S. W.

CUMULATIVE AUTHOR INDEX, VOLUMES Coles, D. K.: Microwave spectroscopy, 11, 300

Collings, P. R., Beyer, R. R.,Kalafut, J. S., and Goetze, G. W.: A family of multi-stage direct-view image intensifiers with fibreoptic coupling, XXVIII A, 105 Collings, P. R.,Healy, L. G., Laponsky, A. B., and Shaffer, R. A.: A proximity focused ultraviolet sensitive SEC camera tube, XXXIII A , 253 Combes, M.: see Baudrand. J. Combes, M.: see Picat, J. P. Combes, M., Felenbok, P., Guerin, J., and Picat. J. P.: Electronic cameras for space research, XXVIII A, 39 Conder, P. C.: see Holeman, B. R. Condon, P. E.: Image tubes in nuclear physics, XU, 123 Conger, G. B., 111: see Santilli, V. J. Conrad, A. C., Jr.: see Jordan, J. A,, Jr. Cooper, A. W.: see Oleson, N. L. Cooper, R.,and Elliot, C. T.: Pre-breakdown light emission from alkali halide crystals, XXII B, 995 Cope, A. D.: see Wronski, C. R. Cope. A. Danforth, and Leudicke, Eduard: The development of camera tubes for recording 175 Corney, A.: The measurement of lifetimes of free atoms, molecules, and ions, XXIX, 115 Corps, R. J.: see Groves, P. R. Cozens, J. R.: see von Engle, A. Cranstoun, G. K. L.: The application of high-gain image intensification and closed-circuit television to field-ion microscopy, XXVIII B, 875 Crompton, R. W.: The contribution of swarm techniques to the solution of some problems in low energy electron physics, XXVII, 1 Cromwell, R. H., and Dyvig, R. R.: Evaluation of image intensifiers for astronomy, XXXIII B, 677 Cullum, M. J.: see Bacik, H. Cullum, M. J.: see McGee, J. D. Cullum, M. J., and Stephens, C. L.: Data reduction for direct astronomical electronography, XXXIII B, 757 Culshaw, W.: Millimeter wave techniques, XV, 197

1-45

349

Curran, S. C.: see Churchill, J. L. W. Curtis, N. A,: see McMullan, D. Curzon, A. E. and Lisgarten, N. D.: The electron-beam shadow method of investigating magnetic properties of crystals, XXIV, 109 Czekalowski, G. W. A., and Hay, G. A,: A quadrature spatial-frequency Fourier analyser, XXVIII B, 653

D Danforth, W. E.: Thorium oxide and electronics, V, 169 Daniels, M. V.: see Beurle, R. L. Davies, J. G.: Radio observation of meteors, IX, 95 Davis, G . P.: Experiences with magnetically focused cascade image intensifiers, XVI, 119

Davis, Robert, J.: The use of the UriconClescope television system for ultraviolet astronomical photometry, XXII B, 875 Dawson, P. H., and Kimbell, G. H.: Chemical lasers, XXXI, 1 Dawson, P. H., and Whetten, N. R.: Mass spectroscopy using rf quadropole fields, XXVII, 59 Day. J. E.: Recent developments in the cathode-ray oscilloscope, X, 239 Dean, R. J.: see Jennings, A. E. Deasley, P. J., and Faulkner, K. R.:Electron emission from forward biased p-n junctions, XXXIII A, 459 Decker, R. W.: Decay of S.20 photocathode sensitivity due to ambient gases, XXVIII A, 357 Decker, R. W., and Mestwerdt, H.: Large-image electronographic camera, XXVlIl A, 19 DeCorpo, J. J.: see Saalfeld, F. E. de Haan, E. F.: Signal-to-noise ratio of image devices, XXI, 291 Delcroix, Jean-Loup, and Trindade, Armando Rocha: Hollow cathode arcs, XXXV, 87 Delori, F. C.: see McGee, J. D. Delori, F. C., Airey, R. W., Dollery, C. T., Kohner, E. M., and Bulpitt, C. J.: Image intensifier cineangiography, XXXIII B, 1089

350

CUMULATIVE AUTHOR INDEX, VOLUMES 1 4 5

Delori, F. C., Airey, R. W., and McGee, J. D.: Further research on the Imperial College cascade image intensifier, XXXIII A, 99 Delpech, J.-F., Boulmer, J., and Stevefelt, J.: Low-temperature rare-gas stationary afterglows, XXXIX, 121 Deltrap, J. H. M., and Hanna, A. H.: Image intensifier system using reflective photocathode, XXVIII A, 443 Denisse, J. F.: see Boischot, A. Dennison, Edwin W.: A microphotometer for use with photographic and electronographic recording image tubes, XXII A, 435 Dennison, Edwin W.: An isophote converter for use with signal-generating image tubes, XII, 307 Dennison, E. W.: An integrating television system for visual enhancement of faint stars, XXXIII B, 795 Dennison, E. W.: Memory systems for signal generating photoelectron image detectors XL, B, 729 Dennison, E. W.: The image orthicon applied to solar photometry, XVI, 447 Dennison, E. W., Schmidt, M., and Bowen, 1. S.: An image-tube spectrograph for the Hale 2M)-in. telescope, XXVIII B, 767 Deprettere, E. F. A.: see Adams, K. M. Desbois, H.: see Pauty, F. Deutscher, K.: see Kossel, D. Deutschman, W. A,: Orbital operation and calibration of SEC-vidicons in the Celescope experiment, XXXIII B, 925 De Witt, John H.,Jr.,: A report on the image orthicon using slow readout, XVI, 419 Diamant, L.: Grid shuttered image converter tube in nanosecond operating mode, XL A, 59 Dickson, J.: see McMullan, D. Doe, L. A,: see Livingston, W. C. Dolan, W. W.: see Dyke, W. P. Dolizy, P., and Legoux, R.: A new technology for transferring photocathodes, XXVIII A, 367 Dollery, C. T.: see Delori, F. C. Dolon, Paul J.: see Ball, Jack Dolon, Paul J.: Niklas, Wilfrid F.

Donal, J. S.: Modulation of continuouswave magnetrons, IV, 188 Donati, S., Gatti, E., and Svelto, V.: The statistical behaviour of the scintillation detector: theories and experiments, XXVI, 251 Donelti, G., and Paoletti, L.: Electron micrograph analysis by optical transforms, XLIII, 1 Doolittle, R. F., and Graves, C. D.: Further developments in the application of scintillation chambers to space research, XXlI B, 823 Doolittle, R. F., 11, and Graves, C. D.: The application of scintillation chambers to space research, XVI. 535 Doughty, D. D.: see Schneeberger, R. J. Doughty, D. D.: see Wachtel, M. M. Doughty, D. D.: Ultra-violet sensitive camera tubes incorporating the SEC principle, XXII A, 261 Dow, W. G.: Nonuniform D-C electron flow in magnetically focused cylindrical beams, X, 1 Dow, W. G.: The general perturbational theory of space-harmonic traveling-wave electron interaction, XVII, 1 Dracass, J.: see Flanagan, T. P. Driard, B.: see Guyot, L. F. Driard, B.: ContrBle des monocristeaux par tube intensificateur de luminance. XXVIII B, 931 Driard, B., Guyot, L. F., and Verat, M.: A 35-cm input-field image intensifier for scintillation cameras, XXXIIl 8, 1031 Driard, B., Roziere, G., Guyot, L. F., and Verat, M.: Improvements to an image intensifier for a y-ray scintillation camera, XL A, 41 Duchenois, V.: see Boutot, J. P. Duchesne, M.: see Alpern, M. Duchesne, M.: see Baranne, A. Duchesne, M.: see Bied-Charreton, P. Duchesne, M.: see Lallemand, A. Duchesne, M.: Sur la Realisation d’une Camera Electronique de Grandissement 1/7, XVI, 27 Duchesne, M.:Sur une Nouvelle Technique d’Utilisation de la Camera Electronique, XVI, 19

CUMULATIVE AUTHOR INDEX, VOLUMES

Duchesne, M., and Hezard, C.: Sur la realisation d’un objectif a immersion a lentilles cylindriques croisees en vue de son utilisation comme systeme focalisateur de la camera klectronique: resultats preliminaires, XXlI A, 609 Duchet, M.: see Charles, D. R. Duchet, M.: Time-response of photocathodes, XXII A, 499 Dudding, R. W.: see Authinarayanan, A. Dunham, Theodore, Jr.: Performance of image tubes in the coude spectrograph at Mount Stromlo observatory, XXlI B, 129

Dunlap, J.: see Hynek, J. A. Dunlap, J. R.: Astronomical photometry and other recent applications of the image orthocon, X L B, 901 Dunlap, J. R., Hynek, J. A,, and Powers, W. T.: Improvements in the application of the image orthicon to astronomy, XXXIIl B, 789 Dunlap, J. R.. Weiler, E. J.. and Hynek, J. A.: Astronomical photometry and other recent applications of the image orthocon, X L B, 901 Dupre, Mlle M.: see Wlerick, G. Du Toit, A. G.: A method for efficient numeric computation of axially symmetric electrostatic fields in image tubes, X L A, 485 Dvoiak. M.: see JareS, V. Dvoiak, M.: Some properties of the trialkali Sb-K-Rb-Cs photocathode, XXVIII A, 347 Dyke, W. P., and Dolan, W. W.: Field emission, VIII, 89 Dyvig, R. R.: see Cromwell, R. H.

E Edgecumbe, J.: see Garwin, E. L. Egan, D. W.: see Beckman, J. E. Egli, P. H.: see Burstein, E. Eichmeier, J.: see Knoll, M. Einstein, P. A.: see Haine, M. E. Eisenstein, A. S.: Oxide coated cathodes, 1, 1

Elliot, C. T.: see Cooper, R. Elliot, C. T.: see Smith, W. A.

1-45

351

Elvey, C. T.: Aurora borealis, IX, 1 Emberson, C. J.: see Wheeler, B. E Emberson, D. L.: A comparison of some properties of image intensifiers of the transmitted secondary emission multiplication type and of the cascade type, XXII A, 129 Emberson, D. L., and Holmshaw. R. T.: Some aspects of the design and performance of a small high-contrast channel image intensifier, XXXlII A, 133

Emberson, D. L., and Long, B. E.: Some aspects of the design and manufacture of a fibre-optic coupled cascade image intensifier, XXVIll A, 119 Emberson, D. L., Todkill, A,, and Wilcock, W. L.: Futher work on image intensifiers with transmitted secondary electron multiplication, XVI. 127 Emeleus, K. G.: Plasma oscillations, XX, 59 England, L.: see Barnett, M. E. Ennos, A. E.: see Haine, M. E. Erickson, William C., and Kerr, Frank J.: Technology and othervations in radio astronomy, XXXII, 1 Eschard, G.: see Boutot, J. P. Eschard, G.: see Chalmeton, V. Eschard, G., and Graf, J.: Quelques problemes concernant les multiplicateurs canalises pour intensificateur d’image, XXVIII A, 499 Eschard, G., and Polaert, R.: Tubes obturateurs pour photographie ultrarapide au temps, de pose d’une nanoseconde. XXVIII B, 989 Eschard, G.. Graf, J., and Polaert, R.: Signal to noise and collection efficiency measurements in microchannel wafer image intensifiers, X L A, 141 Essig, Sanford, E.: Field emission in image tubes, XII, 73 Evans, H. D.: see McGee, J. D. Evrard, R.: Image intensifier tubes with new very simple electron optics, X L A, 83

F Farago, P. S.: The polarization of electron beams and the measurement of the g-factor anomaly of free electrons, XXI, 1

CUMULATIVE AUTHOR INDEX, VOLUMES 1 4 5

352

Faulkner, K. R.: see Deasley, P. J. Fay, Theodore D.: see Frederick, Lawrence W. Fawcett, 1. M.: see Jensen, A. S. Feibelman, W. A.: see Schneeberger, R. J. Feibelman, W. A.: see Sturgill, C. C. Feingold, R.: see Pollehn, H. Feinstein, David L.: see Granatstein, V. L. Felenbok, P.: see Baudrand, J. Felenbok, P.: see Combes, M. Felenbok, P.: see Picat, J. P. Fenner, E., Franz, F., Gudden, F., Heinrich, H., and Hofmann, F. W.: X-ray image intensifiers: image quality and possibilities for enhancement, XXXIII B, 1049

Fenner, E., Heinrich, H.,Schweda, S., Goetze, G . W., and Beyer, R. R.: X-ray camera tube with SEC target XXXIII B, 1061 Ferguson, Eldon E.: Thermal energy ionmolecule reactions, XXIV, 1 Fiermans, L., and Vennik, J.: Electron beams as analytical tools in surface research: LEED and AES, XLIII, 139 Filby, R. S.,Mende, S.B., and Twiddy, N. D.: A television camera-tube using a low density potassium chloride target, XXII A, 273 Fisher, D., Lee, R., McCollough, V., Nudelman, S.,Tufts, D., and Wilkinson, M.:Methods for evaluating camera tubes, XXXIII B, 601 Flanagan, T. P., Anderson, D. G., Noe, E. H., and Dracass, J.: Properties and applications of glass scintillators, XVI, 547

Fleming, W. J., and Rowe, J. E.: Acoustoelectric interaction in 111-V compound semiconductors, XXXI, 161 Flinn. E. A.: see McGee, J. D. Flinn, E. A.: Progress report on a channelled image intensifier, XVI, 155 Flory, L. E., Pike, W. S.,Morgan, J. M., and Boyer, L. A.: A programmable integrating television system for use with the Stratoscope, XXII B, 885 Folkes, J. R.: see Garfield, B. R. C. Folkes, J. R.: Introduction of preformed photocathodes into vacuum systems, XVI, 325 Foote, D.P.: see Kazan, B. Ford, W. K.,Jr.: see Frederick, L. W.

Ford, W. K., Jr.: see Hall, J. S. Ford, W. Kent, Jr.: Astronomical uses of cascade intensifiers, XXII B, 697 Fordham, J. L. A.: see Boksenberg, A. Foreman, P. H., and Thumwood, R. F.: An image intensifier tube using the multipactor principle, XVI, 163 Fort, B.: see Baudrand, J. Fort, B.: see Picat, J. P. Fouassier, M.: see Graf, J. Fowler, Richard G.: Electrons as a hydrodynamicaLfluid, XX, 1 Fowler, Richard, G.: Nonlinear electron acoustic waves, part 1, XXXV, 1 Fowler, Richard G.: Nonlinear electron acoustic waves, part 11, XLI, 1 Fowweather, F., and Harbour, J.: The application of image storage tubes to the observation of optical diffraction patterns, XI1 311 Frank, K.: see Hambrecht, F. T. Franz, F.: see Fenner, E. Franz, K., Kochmann, G., and Lahmann, R.: Modulation transfer function output screens, XXXIII A, 483 Franzen, W.: see Baldinger, E. Franzen, Wolfgang, and Porter, John H.: Energy spectrum of electrons emitted by a hot cathode, XXXIX, 73 Frederick, Lawrence W., Fay, Theodore D., and Johnson, Hollis R.: Infra-red stellar spectroscopy with a mica-window tube, XXII B, 723 Frederick, L. W., Hall, J. S., Baum, W. A., and Ford, W. K., Jr.: Some astronomical uses of image intensifying tubes, XVI, 403 Freeman, K. G.: see Taylor, D. G. Frey, Jeffrey, and Bowers, Raymond: The impact of solid state microwave devices: a preliminary technology assessment, XXXVIII, 147 Frohlich, H., and Simpson, J. H.: Intrinsic dielectric breakdown in solids, 11, 185

Fromm, W. E.: The magnetic airborne detector, IV,258 Frost, M. M.: see Nudelman, S. Frost, M.M.,and Roehrig. H.:Image tubes and detective quantum efficiency, XL B, 5 19

CUMULATIVE AUTHOR INDEX, VOLUMES

G

Gale, B. C.: see Ahmad, N. Granson, A.: see McGee, J. D. Garfield, B. R. C.: see Allan, F. V. Garfield, B. R. C.: Multialkali photocathodes, XXXIII A, 339 Garfield, B. R. C., and Thumwood, R. F.: A microbalance study of the Cs-Sb and Na-K-Sb photocathodes, XXII A, 459 Garfield, B. R. C.. Bailey, P. C., and Marshall, R.: Developments in image tubes for ultra-high-speed photography, XXXIII B. 1137 Garfield, B. R. C., Folkes, J. R., and Liddy, B. T.: Improvements to photocathodes for pulse operation, XXVIII A, 375 Garfield, B. R. C., Wilson, R. J. F.. Goodson, J. H., and Butler, D. J.: Developments in proximity focused diode image intensifiers, XL A, 1 I Garlick, G. F. J.: Cathodoluminescence, 11, I52 Garlick, G. F. J.: Recent developments in solid state image amplifiers, XVI, 607 Garrett, C. G. B.: The electron as a chemical entity, XIV, 1 Garthwaite, E.: X-ray image intensifier using image orthicon tubes, XII, 379 Garwin, E. L., and Edgecumbe, J.: Response of low-density KC. foils to multi-meV electrons, XXlI A, 635 Gatti, E.: see Donati, S. Gaucher, J. C.: see Roux, G. Gebel, R. K. H.: Low-energy quanta image transducers using a controlled recombination mode, XXlI A, 189 Gebel. R. K. H.: The fundamental infra-red threshold in thermal image detection as affected by detector cooling and related problems, XXVIII B. 685 Gebel. R. K. H.: The potentialities of electronically scanned photoconductive image detectors for astronomical uses, XVI, 451 Gebel, R. K. H.,and Deval, Lee: Some early trials of astronomical photography by television methods, XII, 195 Gebel, R. K. H., Mestwerdt, H.R., Spiegel, H. J.. and Hayslett, R. R.:The limitations of optoelectronic image recording using

145

353

night-sky illumination considering optimum gain and optimized components, XXXIII B, 999 Geise, R.: see Gildemeister, 0. Gelade, G. A,: see Beurle, R. L. Geneux, E., Bene, G. J., and Perrenoud, J.: Magnetic coherence resonances and transitions at zero frequency, XXVII, 19 Geurts, A,: see Kiihl, W. Ghosh, C.: see Bhatia. T. B. Gibbons, D. J.: The tri-alkali stabilized C. P. S. Emitron: A new television camera tube of high sensitivity, XII, 203 Giese, R., Gildemeister, 0.. and Schuster, G.: Test of a high-resolution terenkov chamber with a four-stage image intensifier, XXVIII B, 919 Gilbert, G. R., Angel, J. R. P., and Grandi, S.: A digital television system for astronomy, XL B, 699 Gildemeister, 0.:see Giese, R. Gildemeister, 0..and Giese, R.: An image intensifier for track recording, XVI, 113 Gilliard, B.: see Baumgartner, W. Gilmore, D. J.: see Amdt, U. W. Gilmore, D. J.: A television X-ray diffractometer, XL B, 913 Ginzton, E. L.: see Warnecke, R. R. Gislason, G. A,: see Sackinger. W. M. Glaspey. J. W.: see Buchholz, V. L. Gorlich, P.: Problems of photoconductivity, XIV, 31 Gorlich, P.: Recent advances in photoemission, XI. 1 Goetze, G. W.: see Beyer, R. R. Goetze, G. W.: see Boerio, A. H. Goetze, G. W.: see Choudry, A. Goetze, G. W.: see Collings, P. R. Goetze, G. W.: see Fenner. E. Goetze, G. W.: Secondary electron conduction (SEC) and its application to photoelectronic image devices, XXII A. 219 Goetze, G. W.: Transmission secondary emission from low deposits of insulators, XVI, 145 Goetze, G. W., and Boerio, A. H.: SEC camera-tube performance characteristics and applications XXVIl A, 159

354

CUMULATIVE AUTHOR INDEX, VOLUMES

Goetze, G. W., and Taylor, A.: Recent applications of transmission secondary emission amplification, XVI, 557 Goldberg B. A.: see Walker, G. A. B. Goldberg, Seymour, and Rothstein, Jerome: Hydrogen thyratrons, XIV, 207 Goldstein, L.: Electrical discharge in gases and modern electronics, VII, 399 Goodson, J., Woolgar, A. J., Higgins, J., and Thumwood, R. F.: The proximity focused diode image intensifier, XXXIII A, 83 Goodson, J. H.: see Garfield, B. R. C. Gordon, A. W.: see Raffan, W. P. Goto, S.: see Sasaki, T. Cower, A. C.: see Walker, G . A. B. Graf, J.: see Eschard, G. Graf, J., Fouassier, M., Polaert, R., and Savin, G.: Characteristics and performance of a microchannel image intensifier designed for recording fast luminous events, XXXIII A, 145 Granatstein, V. L., and Feinstein, David L.: Multiple scattering and transport of microwaves in turbulent plasma, XXXII. 3 1 1 Graves, C. D.: see Doolittle, R. F. Greatorex, C. A.: Image intensification using a flying-spot x-ray tube, XII, 327 Greatorex, C. A.: Image storage techniques applied todiagnostic radiology, XVI, 593 Green, M.: see Beyer, R. R. Green, M., and Hansen, J. R.: The application of SEC camera tubes and electrostatic image intensifiers to astronomy, XXVIII B, 807 Greschat, W., Heinrich, H., and Romer, P.: Quantum yield of Cs,Sb photocathodes as a function of thickness of incidence, X L A, 391 Griboval, D.: see Griboval, P. Griboval, P.: A five centimeter magnetically focused electronographic camera: description and first tests, X L B, 613 Griboval, P., Griboval, D., Marin, M., and Martinez, J.: Properties of commercial electron-sensitive plates for astronomical electronography, XXXlII A, 67 Grivet, P.: Electron lenses, 11.48

1-45

Grivet, P. A., and Malnar, L.: Measurement of weak magnetic fields by magnetic resonance, XXIII, 39 Grosch, G. A., and Krieser, J. K.: Leistungsgrenze eines Sichtsystems mit Bildverstarker, XXVIlI B, 603 Grosse, Achilles: see Wlerick, Gerard Groves, P. R., and Corps, R. J.: Applications of the image isocon tube, XXVIII B, 827 Gudden, G.: see Fenner, E. Guenard, P. R.: see Warnecke, R. R. Guerin, J.: see Combes, M. Guest, A,: see Manley, 8. W. Guillard, C., and Charles, D. R.: On some properties of electron bombardment induced conductivity, XXII A, 315 Guillemin, E. A,: A summary of modern methods and network synthesis, 111, 261 Gumnick, J. L.: see Clayton, R. H. Gunshor, L.: Optical imaging with acoustic waves and photo-excited charge carriers, X L B, 993 Guyot, L. F.: see Driard, B. Guyot, L. F.: Derniers Developpements sur les lntensificateurs d’lmage Rayons X a Grand Gain at les Tubes Convertisseurs d’lmage, XVI, 91 Guyot, L. F., Driard, B., and Sirou, F.: Tubes intensificateurs d’image pour observation des phenomenes lumineux rapidement evolutifs, XXII B, 949

H Hachenberg, O., and Brauer, W.: Secondary electron emission from solids, XI, 413 Hagino, M., Yoshizaki, S., Kinoshita, M., and Nishida, R.: Caesium activated CsI transmission-type secondary emission dynode, XXXIII A, 469 Haine, M. E.: the electron microscope-a review, VI, 295 Haine, M. E., Ennos, A. E., and Einstein, P.A,: An image intensifier for the electron microscope, XII, 317 Hall, J. A.: Uniform layer heterojunction targets for television camera tubes. XXXIll A, 229 Hall, J. S.: see Frederick, L. W. Hall, J. S., Ford, W. K., Jr., and Baum,

CUMULATIVE AUTHOR INDEX, VOLUMES W. A.: Astronomical tests of barriermembrane image converters, XII, 21 Hallam, K. L.: see Johnson. C. 9. Hallam, K. L., and Johnson, C. 9.: Image transfer and conversion for photoelectronic imaging devices, X L 9, 601 Hambrecht. F. T., and Frank. K.: The future possibilities for neural control, XXXVIII, 55 Hanna, A. H.: see Deltrap, J. H. M. Hansen, J R.: see Green, M. Harbour, J.: see Fowweather, F. Harmer, A. L., and Wreathall, W. M.: Thermal diffusion limitations of the resolution of a pyroelectric vidicon, X L A, 313 Harms, R. J.: see Beaver, E. A. Harmuth, Henning F.: Research and development in the field of Walsh functions and sequency theory, XXXVI. 195 Harmuth, Henning F.: Generation of images by means of two-dimensional. spatial electric filters, XLI, 167 Harris, R. 9. A,: see Huston, A. E. Harth, W.: see Schaff, F. Harth, W.: see Unger, H.-G. Hartley, K. F.: see McMullan, D. Hartley, K. F.: see Pilkington, J. D. H. Hartley, K. F., and McMullan, D.: An image tube for experimental electron optics, X L A. 493 Hartman, R. E.: see Alsberg, H. Hartmann, P.: see Vernier, P. Hasegawa, S.: Effect of optical pulse height distribution on the resolving power of an image tube, XXXIII B, 617 Hasegawa, S.: Resolving power of image tubes, XXVIII 9, 553 Hasegawa, S., and Kaneko, Y.: An edge detecting system, X L B, 963 Hasker, J.: Imaging, beam-acceptance lag in camera tubes, XXXIII A, 317 Hasted, John 9.: Inelastic collisions between atomic systems, XIII, 1 Haus, Hermann, A.: see Pucel, Robert A. Hay, G. A.: see Ozekalowski, G. W. A. Hay, G. A. : The image orthicon in diagnostic radiology, XVI, 581 Hay, G. A,: X-ray image intensification using

145

355

optical television methods, XII, 363 Hayslett, R. R.: see Gebel, R. K. H. Hayward, R. W.: Beta-ray spectrometers, v, 97 Healy, L. G.: see Collings, P. R. Heckathorn, H. M.: see Chincarini, G. Heimann, W.: Experiments with a simple photo-electronic storage tube, XII, 235 Heimann, W.: Possibilities of reducing image defects in electron-optical imaging devices using electrostatic lenses, XXll A, 601 Heimann, W., and Hoene, E. L.: Improvement of signal-to-noise ratio of image converters with S.l photocathodes, XXVIIl 9, 677 Heimann, W., and Kunze, C.: Development of an infra-red vidicon-type pick-up tube with a lead sulphide target, XVI, 217 Heinrich, H.: see Fenner, E. Heinrich, H.: see Greschat, W. Heinrich, Hans: see Stahnke, Ingeborg Herbstreit, J. W.: see Rice, P. L. Herbstreit, J. W.: Cosmic radio noise, I. 347 Herrman, M., and Kunze, C.: A new multiplier system with forty separate channels, XXVIII 9, 955 Hersey, J. B.: Electronics in oceanography, IX, 239 Herstel, W.: Some experiences with x-ray image intensifiers and television channels, XVI, 610 Herstel, W.: The assessment of image quality in medical fluoroscopy, XXII A, 363 Herstel, W.: The evaluation of image quality of radio-isotope scanners and pray cameras. XXXIII 9, 1041 Herstel, W.: The observation of moving structures with x-ray image intensifiers, XXVIII B, 647 Hewitt, A. V.: see Charmann, W. N. Hewitt, A. V.: see Kron, G. E. Hewitt, A. V., Kron, G. E., and Ables, H. D.: Photometry with the electronic camera, XXXlIl B, 737 Htzard, C.: see Duchnesne, M. Higgins, J.: see Goodson, J. Hill, D. A,, and Porter, N. A,: Photography of extensive air showers in the atmosphere, XVI, 531

356

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

Hill, D. A., Caldwell, D. O., and Schulter, R. A.: Performance of an image intensifier system, XVI, 475 Hill, G. E.: Secondary electron emission and compositional studies on channel plate glass surfaces, X L A, 153 Hills, B. L.: see Beurle, R. L. Hiltner, W. A., and Niklas, W. F.: A low background image tube for electronography, XVI, 37 Hiltner, W. A., and Pesch, Peter: Image tube research at Yerkes Observatory, XII, 17 Hinder, G. W.: see Iredale, P. Hinder, G. W., and Iredale, P.: The image quality of an image intensifier expressed in terms of its equivalent quantum efficiency, XXXIII B, 639 Hirashima, M.: Optimum conditions for activating silver-magnesium alloy dynodes in water vapour, XXII A, 661 Hirashima, M., and Asano, M.: Effects of caesium vapour upon target glass of image orthicon, XXII A, 651 Hirashima, M., and Asano, M.: Reaction of caesium vapour with gold, XXII A, 643 Hirashima, M., and Asano, M.: Some better materials for caesium vapour, XXVIII A, 381 Hirashima, M., Sano, T., and Asano, M.: A method of testing the effectiveness of a protective coating layer on any glass surface against alkali vapours, XXXIII A, 381 Hirayama, T.: see Kajiyama, Y. Hirsch, C. J.: A review of recent work in color television, V, 291 Hirschberg, K.: see Kossel, D. Hiruma, T., Suzuki, Y.,and Kurasawa, K.: Pick-up storage having an electronic shutter, automatic exposure control, wobbling correction, and slow scanning, XXXlll A, 263 Hobson, J. P.: see Redhead, P. A. Hoene, E. L.: see Heimann, W. Hodgson, R. M.: see Beurle, R. L. Hodgson, R. M., and Beurle, R. L.: Image intensifier noise and its effects on visual pattern detection, X L B, 539

Hoene, E. L.: Optical and photoelectric properties of multialkali photocathodes, XXXIII A, 369 Hofmann, F. W.: see Fenner, E. Hok, G.: The microwave magnetron, 11, 220 Holeman, B. R., Conder, P. C., and Skingsley, J. D.: Proximity focused image intensifier with GaAs photocathode, X L A, 1 Holmes, P. J.: see Billig, E. Holmshaw, R. T.: see Emberson, D. L. Holmshaw, R. T.: see Manley, B. W. Holtom, R.:see Howorth, J. R. Hooper, E. B., Jr.: A review of reflex and Penning discharges, XXVII, 295 Hopmann. W.: The image orthicon in highspeed photography, XXII B, 101 1 Hopmann, W.: The influence of photocathode resistance and space charge on the resolution of magnetic focus systems, XXII A, 591 Hori, H., Tsuji, S.,and Kiuchi, Y.: An infra-red sensitive vidicon with a new type of target, XXVIII A, 253 Houston, J. M., and Webster, H. F.: Thermionic energy conversion, XVII, 125 Houston, John M.: see Vosburgh, Kirby G . Howorth, J. R., Holtom, R., Sheppard, C. J. R., and Trawny, E. W. L.: Thermionic silicon, X L A, 387 Howorth, J. R., Surridge, R. K., and Palmer, I. C.: The negative electron affinity GaAsP cold cathode silicon vidicon, X L A, 463 Hubbard, Edward L.: Linear ion accelerators,

xxv, 1

Hughes, J. S., Wilcock, W. L., and Smith, D. G.: Absolute photoelectric emission statistics of aluminium, silver and potassium chloride in the ultra-soft X-ray region, XXXIII A, 433 Huston, A. E.: Image tube highspeed cameras, XXII B, 957 Huston, A. E., and Harris, R. B. A,: Developments in image tube high-speed framing cameras, XXXlII B, 1109 Huston, A. E., and Walters, F. W.: Electron tubes for high-speed photography, XVI, 249

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

Hutter. E. C.: see Vance, A. W. Hutter, R. G. E.: The deflection of beams of charged particles, 1, 167 Hutter, R. G. E.: Travelling-wave tubes, VI, 371 Hynek, J. A,: see Dunlap, J. R. Hynek, J. A,, Bakos, G., Dunlap, J., and Powers, W.: Advances in the application of the image orthocon to astronomy, XXll B, 713 Hynek, J. A., Barton, G . ,Aikens, R., and Powers, W.: Potentialities and limitations of image scanning techniques in astronomy, XVI, 409

I lida, H.: see Kurasawa, K. II, M.: see Kurasawa, K. Inghram, M. G.: Modem mass spectroscopy, I., 219 Inskeep. C. N.: see Abraham, J. M. Iredale, P.: see Hinder, G. W. Iredale, P.. and Ryden, D. J.: On the quality of photographic images recorded with the use of image intensifiers, XXVllI B. 589 Iredale, P., Hinder, G. W., and Smout, D. W. S.: Position-sensitive photon counters, XXVlIl 9, 965 Iredale, P., Hinder, G. W., Parham, A. G., and Ryden, D. J.: The observation of cerenkov ring images with an image intensifier system of high gain, XXII B, 801 Isherwood, B. C.: see Buchholz, V. L. Isherwood. B. C.: see Walker, C. A. H. Ito, M.: see Suzuki, Y. hey, H. F.: Space charge limited currents, VI. 137

J Jackson, F. W.: see Wardley, J. Jackson, R. N., and Johnson, K. E.: Gas discharge displays: a critical review, XXXV, 191

357

Jane, M. R.: see Binnie, D. M. JareS, V.: A flat channel system for imaging purposes, XXXllI A, 117 JareS, V.: Possibilities of eliminating the circular, spurious signals in vidicons caused by secondary emission, XXXlll A, 307 JareS, V., and Dvoiak, M.: A flat channel system for imaging purposes, XXXIIl A, 117 JareS, V., and Novotny, B.: Some problems of electron optical system design using a computer XL A. 473 JareS, V., and Novotny, B.: Two methods for the determination of the imaging properties of electron-optical systems with a photocathode, XXVIII A, 523 Jaumot, Frank E., Jr.: Thermoelectricity, XVII, 207 JedliEka, M.: Research on photocathodes in Czechoslovakia, XXVIll A, 323 JedliEka. M., and Vilim, P.: Some properties of the Sb-Rb-Cs photocathode, XXII A, 449 JedliEka, M., Ladman, R., Vitovsky, Lezal, D., and Srb, I.: Some properties of evaporated and sprayed CdSe layers for heterojunction vidicon targets, XL A, 323 JefTers, S., and McGee, J. D.: On the transmission of medium energy electrons through mica, XXII A, 41 Jeffers. S., and Weller, W.: An image intensifier multichannel analyser astronomical spectroscopy, XL B, 887 Jenkinson, G . W.: see Burle, R. L. Jennings, A. E., and Dean, R. J.: Sensitization of electrostatically focused image converters, XXII A, 441 Jensen, A. S., and Fawcett, J. M.: Measurement of TV camera noise, XXVIII A, 289 Jensen, Arthur S., Reininger, Walter G., and Lamansky, Igor: The grating storage target, XXII A, 155 Johnson, C. B.: see Hallam, K. L. Johnson, C. B.: A magnetically focused image intensifier employing evaporated field electrodes, XXXIll A, 93

358

CUMULATIVE AUTHOR INDEX, VOLUMES

Johnson, C. B.: Classification of electronoptical device modulation transfer functions, XXXIII B, 579 Johnson, C. B., and Hallam, K. L.: The oblique image counter, XL A, 69 Johnson, Hollis R.: see Frederick, Lawrence W. Johnson, J. M.: see Sackinger, W. M. Johnson, K. E.: see Jackson, R. N. Jones, G. R.: see Watton, R. Jones, Lawrence W.: see Perl, Martin L. Jones, Lawrence W., and Loo, Billy W.: The use of image intensifiers with streamer chambers, XXII B, 813 Jones, L. W., and Perl, Martin L.: Two high-energy physics experiments using the luminescent chamber, XVI, 513 Jones, R. Clark: Performance of detectors for visible and infrared radiation, V, 1 Jones, R. Clark: Quantum efficiency of detectors for visible and infrared radiation, XI, 87 Jones, T. J. L.: see Sturgell. C. C. Jordan, J. A., Jr., Bakken, G. S., and Conrad, A. C., Jr.: A cascade image intensifier camera for beam-foil spectroscopy, X X V l l l B, 907

K Kahan, E.: see Cohen, M. Kahan, E., and Cohen, M.: Comparison of the eficiency of image recording with a Spectracon and with Kodak IIa-0 emulsion, XXVIII B, 725 Kajiyama, Y., Kawahara, T.,and Hirayama, T.: Newly developed image orthicon tube with a MgO target, XXVIII A, 189 Kalafut, J. S.: see Collings, P.R. Kan, S. K.: see Sauzade, M.D. Kaneko, Y.:see Hasegawa, S. Kansky, E.: Some physico-chemical aspects of the synthesis of antimonide photccathodes, XXXIIl A, 357 Kao, K. C.: see Smith, C. W. Karady, George: High-power electronic devices, XLI, 3 11 Karmohapatro, S. B.: Laboratory isotope separators and their applications, XLII, 113

145

Kaufman, Harold, R.: Technology of electron-bombardment ion thrusters, XXXVI, 265 Kaw, Predhiman Krishan: see Sodha, Mahendra Singh Kawahara, T.: see Kajiyama, Y. Kawakami, H.: see Uno, Y. Kay, Eric: Impact evaporation and thin film growth in a glow discharge, XVII, 245

Kazan, B.: see Knoll, J. Kazan, B., and Chang, I. F.: Optical writing and erasing with bistable-phosphor storage tubes, XXXIII A, 331 Kazan, B., and Foote, D. P.: Recent developments in field-effect image storage panels, XXVIII B, 1059 Keen, Ralph S.: see Schnable, George L. Kelkar, G. N. : see Bhatia, T. B. Kelly, John: Recent advances in electron beam addressed memories, XLIII, 43 Kennedy, David P.: Semi-conductor device evaluation, XVIII, 167 Kennedy, S. W.: see Weingartner, H. C. Kerr, Frank J.: see Erickson, William C. Kervitsky, J.: see Carruthers, G. R. Kerwin, L.: Mass spectroscopy, VIII, 187 Key, M. H.: see Ahmad, N. Khogali, A,: see McGee, J. D. Kidger, M. J.: see Wynne, C. G. Kimbell, G. H.: see Dawson, P. H. King, J. G., and Zacharias, J. R.: Some new applications and techniques of molecular beams, VIII, 1 Kinoshita, M.: see Hagino, M. Kislov, V. Ya., Bogdanov, E. V.,and Chemov, Z. S.: Physical foundations of plasma applications for generation and amplification of microwaves, XXI, 287 Kissell, K. E.: see Rork, E. W. Kissell, K. E.: Quantitative performance of single- and twc-stage image tubes in spectroscopy, XXXIII B, 653 Kistemaker, J: see Snoek, C. Kiuchi, Y.:see Hori, H. Kiuchi, Y.: see Shimizu, K. Klein. N.: Electrical breakdown in solids, XXVI, 309 Knight, R.: see Walker, G. A. H. Knight, Richard I.: see Bates, David J.

CUMULATIVE AUTHOR INDEX, VOLUMES

Knoll, M., and Kazan, B. : Viewing storage tubes, VIII, 447 Knoll, M., Eichmeier, J., and Schon, R. W.: Properties, measurement, and bioclimatic action of "small" multimolecular atmospheric ions, XIX, 178 Kochmann, G.: see Franz, K. Kohashi. T., Nakamura, T., Maeda, H., and Miyaji, K.: A fast-response sotid-state image converter, XXII B, 683 Kohashi. T., Nakamura, T., Nakamura. S., and Miyaji, K.: Recent developments in solid-state infra-red image converters, XXVlIl B, 1073 Kohn, E. S.: see Williams. B. F. Kohner, E. M.: see Delori, F. C. Komrska, Jiii: Scalar diffraction theory in electron optics, XXX, 139 Konigsberg, R. L.: Operational amplifiers, XI, 225 Konrad, G. T., and Rowe, J. E : Harmonic generation and multisignal effects in nonlinear beam plasma systems, XXIX, 1 Kornelsen, E. V.: see Redhead, P. A. Kossel, D., Deutscher, K., and Hirschberg, K.: Interference photocathodes, XXVIII A, 419 Krieser, J. K.: see Grosch, G. A. Kron, G. E.: see Hewitt, A. V. Kron, G. E.: Advantages of a bakeable electronographic plate. XVI, 35 Kron, G. E.: A modified Lallernand image tube. XVI, 25 Kron, G. E., Ables, H. D., and Hewitt, A. V.: A technical description of the construction, function, and application of the U.S.Navy electronic camera, XXVIII A, 1 Kron, Gerald E., and Papiashvili, I. I.: Progress in the development of the Lick-Stromlo electronic camera, XXII A, 59 Kiihl, W.. Geurts, A,, and v. Overhagen, J.: Information transfer with high-gain image intensifiers, XXVlIl 8, 615 Kunze, C.: see Heimann. W. Kunze, C.: see Herrmann, M. Kunze, W , Meyerhoff, K., and Retzlaff, G.: The useful luminance gain of image

1-45

359

intensifier systems with respect to noise limitations, XXVlIl B, 629 Kurasawa, K.: see Hiruma, T. Kurasawa, K., 11, M. Iida. H., and Suzuki, Y.: A tracking television system for medical applications, XL B, 951 Kusak, Lloyd: Basic concepts of minicomputers, XLIV, 283

L Labeyrie, A.: An image-tube Fourier spectrograph, XXVIII B, 899 Ladman, R.: see Jedlkh, M. Lahmann, R.: see Franz, K. Line, D. C.: Advances in molecular beam masers, XXXIX, 183 Lallemand, A.: Perfectionnement de la camera electronique-application a I'infra-rouge, XXII A, 1 Lallemand, A.: Quelques reflexions sur la camera electronique, XVI, 1 Lallemand, A., Duchnesne, M., and Wlerick, G.: La photographie electronique, XII, 5 Lamport, D. L.: see Millar, 1. C. P. Lamport, D. L.: see Stark, A. M. Lane-Wright, D.: see Buchholz, V. L. Lange, F. W., and Schweda, S.: Asymmetrical astigmatism of X-ray image intensifiers, XL A, 507 Lansiart, A,: see Roux, G. Lansiart, A., and Roux, G.: Spark chambers and image intensifiers used in the scanning of radioactive objects, XXII B, 94 1 Lapnosky, A. B.: see Collings, P. R. Laques, P.: Photographie des etioles doubles au moyen de la camera electronique Lallemand, XXII B, 755 Lashinsky, Herbert: Cerenkov radiation at microwave frequencies, XIV, 265 Laverick, Charles: Superconducting magnet technology, XXIII, 385 Laviron, E.: see Louis-Jacquet, M. Lawless, W. L.: Developments, in computer logical organization, X, 153 Le Carvennec, F.: see Charles, D. R. Le Carvennec, F.: Recherche d'un dispositif nouveau de television thermique, XXVIII A, 265

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

360

Le Contel, J. M.: see Bied-Charreton, P. Leder, L. 8.: see Marton, L. Lee, R.: see Fisher, D. Legoux, R.: see Dolizy, P. Lechmann, J.: see Vance, A. W. Leifer, M., and Schreiber, W. K.: Communication theory, 111, 306 Lelieure, G.: Photometrie de radiogalaxies de type N par electronographie, XL B, 867

Lelievre, G., and Wlerick, G.: Etude d’astres faibles en lumiere totale avec la camera electronique, XXXIII B, 719 Lengyel, Nardone, S., and Pommerrenig, D.: Measurement of electron diffusion length by photo-luminescence in p-doped GaAs substrates and p-doped epitaxially grown GaAs photocathodes, XXXIII A, 389 Lenz, F.: see Mollenstedt, G. Lequais, J.: see Roux, G. Leial, D.: see JedliEka, M. Liddy, B.: see Bradley, D. J. Liddy, B. T.: see Garfield, B. R. C. Liebmann, G.: Field plotting and ray tracing in electron optics: A review of numerical methods, 11, 102 Limansky, Igor: see Jensen, Arthur S. Linden, B. R.: A survey of work at CBS laboratories on photoelectronic image devices, XVI, 31 1 Lindsay, P. A.: Velocity distribution in electron streams, XII, 181 Lisgarten, N. D.: see Curzon, A. E. Liu, I. D., and Baskett, J. R.: A high-gain time-resolving spectrograph for diagnostics of laboratory simulated re-entry objects, XXVIII B, 1021 Livingston, M. S.: Particle accelerators, I, 269

Livingston, W. C.: Properties and limitations of image intensifiers used in astronomy, XXIII, 347 Livingston, W. C.: Stellar photometry with an image orthicon, XVI, 43 1 Livingston, W. C., Lynds, C. R., and Doe, L. A.: Recent astronomical research utilizing a high gain image intensifier tube, XXII 9, 705 Long, B. E.:see Emberson, D. L.

Loo, Billy W.: see Jones, Lawrence W. Louis-Jacquet, M. and Laviron, E.: Camera electronographique avec dispositif d’obturation et de dtflexion pour cinematographie ultra-rapide, XXXIII B, 1101 Low, W.: Electron spin resonance-a tool in mineralogy and geology, XXIV, 51 Lowrance, J. L.: see Zucchino, P. M. Lowrance, J. L., and Zucchino, P. M.: Integrating television sensors for space astronomy, XXVIII B, 851 Lowrance, J. L.,Renda, G., and Zucchino, P.: The I-SIT isocon photon sounting TV system, XL 9, 71 1 Ludington, C. E.: see Shepherd, F. D., Jr. Luedicke, Eduard, see Cope, A. Danforth Lynds, C. R.: see Livingston, W. C. Lynds, R.: see Powell, J. R. Lynton, E. A., and McLean, W. L.: Type I1 superconductors, XXIII, 1

M McCollough, V.: see Fisher, D. McCollough, W. V., and Tufts, D. W.: Measurements of noise in camera tubes, XL B, 585 McCombe, Bruce D., and Wagner, Robert J.: Intraband magneto-optical semiconductors in the far infra-red, I. XXXVII, 1 McCombe, Bruce, D., and Wagner, Robert J.: Intraband magneto-optical studies of semiconductors in the far infra-red, 11. XXVIII, 1 McCullough, W. V.: Measurements of noise in camera tubes, XL B, 585 McGee, J. D.: see Delori, F. C. McGee, J. D.: see Jeffers, S. McGee, J. D.: see Smith, C. W. McGee, J. D., and Wheeler, B. E.: An image tube with Lenard window, XVI, 47 McGee, J. D., Airey, R. W., and Aslam, M.: High quality phosphor screens for cascade image intensifiers, XXII A, 571 McGee, J. D., Airey, R. W., and Varma, E.P.: Cascade image intensifier developments, XXVIII A, 89

CUMULATIVE AUTHOR INDEX, VOLUMES

McGee, J. D., Airey, R. W., and Wheeler, B. E.: Thin-window image intensifier with phosphor output, XVI, 61 McGee, J. D.. Airey. R. W., Aslam, M., Powell, J. R., and Catchpole. XXII A, 113

McGee, J. D., Aslam, M., and Airey, R. W.: The evaluation of cascade phosphorphotocathode screens, XXlI A, 403 McGee, J. D., Bacik, H., Coleman, C. I., and Morgan, B. L.: Extended field Spectracon, XXXIII A, 13 McGee, J. D., Flinn,E. A,, and Evans, H. D.: An electron image multiplier, XII, 87 McGee, J. D., Khogali, A., and Ganson, A.: Electron transmission through mica and the recording efficiency of the spectracon, XXII A, 31 McGee, J. D., Khogali, A., Ganson, A, and Baum, W. A.: The spectracon-an electronographic image recording tube, XXII A, 11 McGee, J. D., McMullan, D., Bacik, H., and Oliver, M.: Further Developments of the spectracon, XXVIII A, 61 McGee, J. D., Morgan, B. L.. Delori, F. C., Airey, R. W., Cullum, M. J., and Stephens, C. L.: A photon-counting detector for stellar spectrophotometry, XXXIII B, 851 McIlvaine, P. M.: see Roehrig, H. Mcllwain, C. E.: see Beaver, E. A. Mackay, C. D.: Photometry of galaxies with a Spectracon, XL B, 847 McKay. K. G.: Secondary electron emission, 1. 66 MacKrell, G. E.: see Sturgell, C. C. McLane, C. K.: Experimental plasma turbulence. XXX. 1 McLean, W. L.: see Lynton, E. A. McMullan, D.: see Hartley, K. F. McMullan, D.: see McGee, J. D. McMullan, D.: A silicon diode array image tube with aerial read out, XL 8, 777 McMullan, D., and Powell. J. R.: Residual gases and the stability of photocathodes, XL A, 427 McMullan, D., and Towler, G. 0.:Some properties of SEC targets, XXVIII A, 173

145

361

McMull;in, D., Powell, J. R., and Curtis, N. A.: Progress towards an 8 cm electronographic image tube, XL B, 627 McMullan, D., Powell, J. R., and Curtis, N. A,: Electronographic image tube development at the Royal Greenwich Observatory, XXXIII A, 37 McMullan, D., Wellgate, G. B., Hartley, K. F., Dickson, J., and Bostock, D.: A silicon diode array image tube with serial read out. XL B, 777 McMullan, D., Wellgate, G. B., and Ormerod, J.: Serial read-out from image tubes incorporating silicon diode arrays, XXXIII B, 873 McNish. A. G.: Ionospheric research, 1, 317 Maeda, H.: see Kohashi, T. Maeda, H.: see Miyazaki, E. Maeda, H.: see Oba, K. Maeda, H.: see Uno, Y. Majumdar, S.: see Bradley, D. J. Malherbe, A., Tessier, M., and Veron, S.: Spectral response of S.1 photocathodes in the near infra-red, XXII A, 493 Malling. L. R., and Allen, J. Denton: The slow-scan vidicon as an interplanetary imaging device, XXII B, 835 Malnar, L.: see Grivet, P. A. Manley, B. W., and Schagen, P.: tenicon: A high resolution information storage tube, XVI, 287 Manley, B. W., Guest, A., and Holmshaw, T. R.: Channel multiplier plates for imaging applications, XXVIII A, 471 Manly, P. L.: see Rork, E. W. Manson, Steven T.: Atomic photoelectron spectroscopy, part I, XLI, 73 Manson, Steven T.: Atomic photoelectron spectroscopy. 11, XLIV, 1 Mardix, S.: see Roehrig, H. Mardix, S., and Sadasiv, G.: Photoresponse beyond the absorption edge in silicon p-n junctions, XXXIII A, 409 Marin, M.: see Griboval, P. Marshall, F. B., and Roane, G. D.: Performance comparison of the SEC camera tube and the image orthicon, XXII A, 291 Marshall, R.: see Garfield, B. R. C. Martin, R.: see Wise, H. S.

CUMULATIVE AUTHOR INDEX, VOLUMES 1 4 5

362

Martinelli, R. U.: see Williams, B. F. Martinez, J.: see Griboval, P. Marton, L., Leder, L. B., and Mendlowitz, H.: Characteristic energy losses of electrons in solids, VII, 183 Massey, H. S. W.: Electron scattering in solids, IV, 2 Masuda, Kohzoh: see Namba, Susumu Matare, Herbert F.: Light-emitting devices, part I: methods, XLII, 179 Matare, Herbert F.: Light-emitting devices, part 11: device design and applications, XLV, 39 Mathey, L., Plociennik, J. M., and Vernier, P.: Utilization d'une camera electronique pour etudier I'evolution de seuil photoelectrique de couches minces d'or deposees sous ultra-vide, XXXIII A, 423

Mayer, H. F.: Principles pulse code modulation, 111, 221 Mayo, B. J., and Bennett, A. W.: The use of meshes to reduce the effect of errors in certain types of electron tube, XXXIII A, 571 Medved, David, B., and Strausser, Y. E.: Kinetic ejection of electrons from solids, XXI, 101 Meier, F. : see Campagna, M. Melton, B. S.: Contributions of electronics to seismology and geomagnetism, IX, 297

Mende, S. B.: see Filby, R. S. Mende, S. B., and Shelley, E. G.: Single electron recording by self-scanned diode arrays, XL B, 119 Mendlowitz, H.: see Marton, L. Mestwerdt, H.: see Decker, R. W. Mestwerdt, H. R.: see Gebel, R. K. H. Metson, G. H.: On the electrical life of an oxide-cathode receiving tube, VIII, 403 Meyerhoff. K.: see Kunze, W. Midgley, D.: Recent advances in the Hall effect: research and application, XXXVI, 153

Milch, J. R.: see Reynolds, G. T. Millar, I. C. P., Washington, D.. and Lamport, D. L.: Channel electron multiplier plates in X-ray image inten&kation, XXXIII A, 153-

Miller, D. E.: see Wilcock, W. L. Miller, J. S., Robinson, L. B., and Wampler, E. J.: The present status of the Lick Observatory image tube scanner, XL B, 693

Milsom, A. S.: see Palmer, D. R. Misell, D. L.: Image formation in the electron microscope with particular reference to the defects in electronoptical image, XXXII, 63 Miyaji, K.: see Kohashi, T. Miyaji, K.: see Miyazaki, E. Miyashiro, S. and Makayama, Y.:Electronic zooming with the image orthicon television pick-up tube, XVI, 195 Miyashiro, S., and Nakayama, Y.:Some methods of minimizing the black-border effect in the image orthicon television pick-up tube, XVI, 171 Miyashiro, S., and Shirouzu, S.: A supersensitive camera tube incorporating a silicon electronmultiplication target, XXXlll A, 207 Miyashiro, S., and Shirouzo, S.: Electrostatically scanned image orthicon, XXVIlI A, 191 Miyazaki, E.: see Uno, Y. Miyazaki, E., Maeda, H., and Miyaji, K.: The evoscope-a fixed-pattern generator using a Au-Si diode; XXIl A, 33 1 Mladjenovic, Milorad S.: Recent advances in design of magnetic beta-ray spectrometers, XXX, 43 Mockler, Richard C.: Atomic beam frequency standards, XV, 1 Mollenstedt, G., and Lenz, F.: Electron emission microscopy, XVIII, 251 Moreno, T.: High-power axial-beam tubes, XIV, 299 Morgan, B. L.: see Bacik, H. Morgan, B. L.: see McGee, J. D. Morgan, B. L.: A photoelectron counter using Spectracon and diode array, XL B, 765

Morgan, B. L., and Ring, J.: Applications of the Spectracon in astronomy, XL B, 803 Morgan. - B. L., Smith, R. W., and Wilson, G. A,: A storage image tube for optoelectronic computing, XXVIlI B, 1051

CUMULATIVE AUTHOR INDEX, VOLUMES

Morgan, J. M.: see Flory, L. E. Morton, G. A,: The scintillation counter, IV, 69 Morton, G. A,, and Ruedy, J. E.: The low light level performance of the intensifier orthicon, XII, 183 Moss, H.: Cathode ray tube progress in the past decade with special reference to manufacture and design. 11, 2 Motz, H.. and Watson. C. J.: The radiofrequency confinement and acceleration of plasmas, XXIII, 153 Mulder, H.: A colour image intensifier system for night vision, X L A, 33 Mulder. H.: Electron-optical transfer functions of image intensifiers. XXXIIl A, 563 Miiller, Erwin W.: Field ionization and field ion microscopy, XIII, 83

N Nakamura, S.: see Kohashi, T: Nakamura, T.: see Kohashi, T. Nakamura, T.: see Sasaki, T. Nakayama, Y.:see Miyashiro, S. Namba. Susumu, and Masuda, Kohzoh: Ion implantation in semiconductors, XXXVII, 263 Narcisi, Rocco S., and Roth, Walter: The formation of cluster ions in laboratory sources and in the ionosphere, XXIX, 79 Nardone, S.: see Lengyel, G. Nassenstein, H.: The boundary layer image converter, XVI, 633 Needham, M. J.. and Thumwood, R. F.: A proximity-focused image tube, XXVIII A, 129 Nelson, P. D.: The development of image isocons for low-light applications, XXVIII A. 209 Neumann, M. J.: see Burns, J. Newth, J. A,: see Binnie, D. M. Newton, A. C.: see Boksenberg, A. Nielsen, R. Florentin: Photoelectron energy spectrophotometry. XL B, 973 Niklas, Wilfrid, F.: see Ball, Jack Niklas, Wilfrid F., Dolon. Paul J., and Berger, Harold: A thermal-neutron

1-45

363

image intensifier, XXlI 9, 781 Niklas, W. F.: see Hiltner, W. A. Ninomiya, T., Taketoshi, K., and Tachiya, H.: Crystal structure of multialkali photocathodes, XXVIIl A, 337 Niquet, G.: see Vernier, P. Nishida, R.: see Hagino, M. Nishida, R.: see Nogami, M. Nixon, W. C.: see Oatley, C. W. Noe, E. H.: see Flanagan, T. P. Nogami, M., Okamoto, S., and Nishida, R.: A uniform CdS-CdTe-As,& heterojunction target for TV camera tubes, X L A, 335 Norman, D. J.: see Beesley, J. Norton, K. A.: Propagation in the FM broadcast band, I. 381 Novice, M.: see Szepesi, Z. Novotny, B.: see JareS, V. Nozawa, Y.: A digital television system for a satellite-borne ultra-violet photometer, XXII B, 865 Nozawa, Y.:Characteristics of a television photometer. XXVIII B, 891 Nudelman, S.: see Choudry, A. Nudelman, S.: see Fisher, D. Nudelman, S.: see Roehrig, H. Nudelman, S.: Image tubes and detective quantum efficiency, X L B, 539 Nudelman, S.: Intensifiers: detective quantum efficiency, efficiency contrast transfer function and the signal-to-noise ratio, XXVIII B, 577

0 Oatley, C. W., Nixon, W. C.. and Pease, R. F. W.: Scanning electron microscopy, XXI, 181

Oba, K., and Maeda, H.: An analysis of the direct current operation of channel electron multipliers, XXXIlI A, 183 Oba, K.. and Maeda, H.: Impulse and frequency response of channel electron multipliers, X L A. 123 Okamoto, S.: see Nogami, M. O'Keefe, T. W., and Vine, J.: A highresolution image tube for integrated circuit fabrication, XXVIII A, 47

364

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

Okress, E. C.: Magnetron mode transitions, VIII, 503 Oleson, N. L., and Cooper, A. W.: Moving striations. XXIV, 155 Oliver, M.: see McGee, J. D. Oliver, M.: Sources of spurious background in the Spectracon, XXXIIl A, 27 Oliver, M. B.: see Allen, K. G . R. Oman, R. M.: Electron mirror microscopy XXVI, 217 Opal, C. B.: see Carruthers, G. R. Ormerod, J.: see McMullan, D. Ovenstone, J. A.: see Barlow, G. E.

P Palmer, D. R., and Milsom, A. S.: Problems in the use of image intensifiers in astronomical Cassegrain spectrographs, XXXIII B, 169 Paoletti, L.: see Donelli, G. Papiashvili, I. 1.: see Kron, Gerald E. Parham, A. G.: see Iredale, P. Pauty, F., Desbois, H., and Vernier, P.: Etude d’un spectrographe fournissant les distributions angufaire et energetique des photoelectrons, XXXIII A, 415 Pawley, M. G., and Triest, W. E.: Multichannel radio telemetering, IV, 301 Pease, R. F. W.: see Oatley, C. W. Pennings, J. C.: see van Leunen, I. A. J. Perl, Martin, L.: see Jones, L. W. Perl, Martin L., and Jones, Lawrence W.: The regenerative image intensifier and its application to the luminescent chamber, XII, 153 Perrenoud, J.: see Geneux, E. Pesch, Peter, see Hiltner, W. A. Petley, C. H.: see Taylor, D. G . Pettauer, T. V.: see Baum, W. A. Piaget, C., Polaert, R., and Richard, J. C.: Gallium indium arsenate photocathodes, XL A, 311 Picat, J. P.: see Baudrand, J. Picat, J. P.: see Combes, M. Picat, J. P., Chevillot, A., Combes, M., Felenbok, P., and Fort, B.: A Lallemand electronic camera focused by a superconducting magnetic coil, XXXIII A, 1

Picat, J. P., Combes, M., Felenbok, P., and Fort, B.: The electronic camera used in a reflection mode, XXXIII A, 557 Pickels, J. C.: see Theodorou, D. G. Pierce, D. T.: see Campagna, M. Pierce, J. A,: Electronic aids to navigation I, 425 Pike, W. S.: see Flory, L. E. Pilkington, J. D. H., and Hartley, K. F.: Distortion of electron images focused by almost uniform electric and magnetic fields, XXXIII A, 545 Pinsker, Z. G.: Electron diffraction structure analysis and the investigation of semiconducting materials, XI, 3 5 5 Pippard, A. B.: Metallic conduction at high frequencies and low temperatures, V1, 1 Plociennik, J. M.: see Mathey, L. Polaert, R.: see Boutot, J. P. Polaert, R.: see Eschard, G. Polaert, R.: see Graf, J. Polaert, R.: see Piaget, C. Pollehn, H., Bratton, J., and Feingold, R.: Low noise proximity focused image intensifiers, XL A, 21 Pommerrenig, D.: see Lengyel, G. Porter, John H.: see Franzen, Wolfgang Porter, N. A.: see Hill, D. A. Porti, A., and Sashin, D.: Electronic imaging techniques for improved diagnostic radiology, XL B, 945 Potter, D. C.: see Binnie, D. M. Poultney, Sherman K.: Single photon detection and timing: experiments and techniques, XXXI, 39 Powell, J. R.: see McGee, J. D. Powell, J. R.: see McMullan, D. Powell. J. R., and Lynds, R.: Methods of increasing the storage capacity of highgain image intensifier systems, XXVlll B, 745 Powers, W.: see Hynek, J. A. Powers, W. T.: see Dunlap, J. R. Prosser, R. D.: see Berg, A. D. Pucel, Robert, A., Haus, Hermann A., and Statz, Hermann: Signal and noise properties of gallium arsenide microwave field-effect transistors, XXXVIII, 195 Pulfrey, D. L.: see Smith, W. A.

CUMULATIVE AUTHOR INDEX, VOLUMES

Putley, E. H., Watton, R., Wreathall, W. M., and Savage, S. D.: Thermal imaging with pyro-electric television tubes, XXXIII. 285

R Rado. G. T.: Ferromagnetic phenomena at microwave frequencies, 11, 251 Raffan, W. P., and Gordon, A. W.: The development and application of interference photocathodes for image tubes, XXVIII A, 433 Randall, R. P.: Charge integration experiments with a C.P.S. Emitron. XII, 219 Randall, R. P.: Dark current scintillations of cascade image intensifiers, XXVIII B, 713 Randall, R. P.: Operating characteristics of a four-stage cascade image intensifier, XXII A, 87 Reay, N. K.: see Coleman, C. 1. Redhead, P. A., Hobson, J. P., and Kornelson, E. V.: Ultra-high vacuum, XVII, 323 Reininger, Walter, G.: see Jensen, Arthur S. Renda, G.: see Lowrance, J. L. Retzlaff, G.: see Kunze, W. Reynolds, Geo. T.: Sensitivity of image intensifier-film systems for observing weak light sources, XXII A, 381 Reynolds, Geo. T.: The distribution of single electron pulse sizes from multi-dynode electron multipliers, and single electron detection, XXII A, 71 Reynolds, G.T.: Photon interference experiments utilizing photoelectronic devices, XXVIII B, 939 Reynolds, G. T., and Milch, J. R.: Image tube characteristics of importance in X-ray diffraction studies, XL B, 923 Reynolds, T. T., Scarl, D. B., Swanson, R. A., Waters, J. R., and Zdanis, R. A.: Filament scintillation chamber experiments at Princeton University, XVI, 487 Riblet, Henry B.: Radio telemetering, XI, 287

145

365

Rice, P. L., and Herbstreit, J. W.: Tropospheric propagation, XX, 199 Richard, J. C.: see Piaget, C. Richards, E. A.: Contrast-enhancement in imaging devices by selection of input photosurface spectral response, XXVIII B, 661 Richards, E. W. T.: see Wise, H. S. Richmond, J. C.: Image quality of photoelectronic imaging systems and its evaluation, XL B, 519 Rindfleisch, T., and Willingham, D.: A figure of merit measuring picture resolution, XXII A, 341 Ring, J.: see Bacik, H. Ring, J.: see Morgan, B. L. Ring, J., and Worswick, P.: Photometric accuracy in electronography, XL B, 679 Rittner, E. S.: see Bargellini, P. L. Rittner, E. S.: Recent advances in silicon solar cells for space use, XLII, 41 Roach, F. E.: The nightglow, XVIII, 1 Roane, G. D.: see Marshall, F. B. Roberts, Arthur: Amplification of transient images in high-gain photocathcdephosphur image intensifier systems, XII, 135 Robinson, L. B.: see Miller, J. S. Robinson, L. C. : Generation of far-infrared radiation, XXVI, 171 Roddie, A. G.: see Bradley, D. J. Roehrig, H.: see Frost, M. M. Roehrig, H.: see Nudelman, S. Roehrig, H., Aceto, P., Mardix, S., McIlvaine, P. M., and Nudelman, S.: Near infrared camera tube studies with an Ag2S target, XL A, 365 Romer, P.: see Creschat, W. Roosild, S. A.: see Shepherd, F. D., Jr. Rork, E. W., St. John, M. R., Manly, P.L., and Kissell, K. E.:Measurement of a point source sensitivity of three high gain camera tubes, XL A, 263 Rosch, J.: see Wlerick, G. Rosch, J.: Le gain possible de rbolution dans I’observation astronomique par l’emploi de la camera electronique de Lallemand, XII, 113 Rosch, J., Wltrick. G., and Boussuge, C.: Photographie des Etoiles doubles au

366

CUMULATIVE AUTHOR INDEX. VOLUMES 1-45 Moyen de la camera electronique, XVI, 357

Rose, A.: Television pick-up tubes and the problem of vision, I, 131 Rose, D. C.: Intensity variations in cosmic rays, IX, 129 Ross, D. G.: see Allen, K. G. R. Roth. Walter, see Narcisi, Rocco S. Rothstein, Jerome: see Goldberg, Seymour Rougeot, H.: TV camera tube with a gallium arsenide target, XL A, 185 Roux, G.: see Lansiart, A. Roux, G. Gaucher, J. C., Lansiart, A., and Lequais, J.: Detecteur photoelectronique analogue de la position de scintillations faiblement lumineuses, XXXIII B, 1017 Rowe, E. G.: On some aspects of tube reliability, X, 185 Rowe, J. E.: see Flemming, W. J. Rowe, J. E.: see Konrad, G. T. Rowlands, Richard 0.:Electronic engineering in river and ocean technology, XXXI, 267 Roziere, G.: see Driard, 8. Ruedy, J. E.: see Morton, G. A. Russell, L. A,: High-speed magnetic-core memory technology, XXI, 249 Rutherford, R. E., Jr.: An electron beam readout technique, XL A, 279 Ryden, D. J.: see Iredale, P. S

Saalfeld, F. E., DeCorpo, J. J., and Wyatt, J. R.: Mass spectroscopy, XLII, 1 Sackinger, W. M., and Gislason, G. A.: Ion feedback noise in channel multipliers. XXXllI A, 175 Sackinger, W. M., and Johnson, J. M.: An analysis of the low-level performance of channel multiplier arrays, XXVIII A, 487

Sackinger, W. M., and Johnson, J. M.: Effects of vacuum space charge in channel multipliers, XXVlIl A, 507 Sadasiv, G.: see Mardix, S. St. John, M. R.: see Rork, E. W. Sano, T.: see Hirashima, M. Santilli, V. J., and Conger, G. B., Ill: TV camera tubes with large silicon diode

array targets operating in the electroo bombarded mode, XXXIII A, 219 Sasaki, T., Nakamura, T., and Goto, S.: Experiments on a wire-electrode type image intensifier using electroluminescence, XVI, 621 Sashin, D.: see Porti, A. Sato, K., and Takahashi, M.: A magnetically focused SEC camera tube, XXXIII A, 241 Sattler, K.: see Campagna, M. Sauzade, M. D., and Kan, S. K.: High resolution nuclear magnetic resonance in high magnetic fields, XXXIV, I Savage, S. D.: see Putley, E. H. Savin, G.: see Graf, J. Scarl, D. B.: see Reynolds, G. T. SchaK, F., and Harth, W.: Computation of imaging properties of image tubes from an analytic potential representation, XXVIIl A, 535 Schaffner, J.: Junction transistor applications, V, 367 Schagen, P.: see Manley, B. W. Schagen, P.: see Taylor, D. G. Schagen, P.: see Woodhead, A. W. Schagen, P.: An image intensifier system for direct observation at very low light IeveIs, XVI, 75 Schagen, P., and Turnbull, A. A.: New approaches to photo-emission at long wavelengths, XXVlIl A, 393 Scharfe, M. E., and Schmidlin, F. W.: Charged pigment xerography, XXXVIII, 83

Schluter, R. A.: see Hill, D. A. Schmerling, E. R.: see Hill, D. A. Schmidlin, F. W.: see Scharfe, M. E. Schmidt, G. W.: see Beaver, E. A. Schmidt, M.:see Dennison, E. W. Schnable, George, L., and Keen, Ralph, S.: On failure mechanisms in large-scale integrated circuits, XXX, 79 Schneeberger, R. J.: see Anderson, A. E. Schneeberger, R. J., Skorinko, G., Doughty, D. D., and Feibelman, W. A.: Electron bombardment induced conductivity including its application to ultra-violet imaging in the Schuman region, XVI, 235

CUMULATIVEAUTHOR INDEX, VOLUMES

Schon, R. W.: see Knoll, M. Schooley. Allen, H.: Electronic instrumentation for oceanography XIX, 1 Schuster, G.: see Giese, R. Schut, Th. G.: Beam-current-induced dark current in Plumbicons, XXXIII A, 319 Schweda. S.: see Fenner. E. Schweda, S.: see Lange, F. W. Seib. D. H., and Aukerman, L. W.: Photodetectors for the 0.1 pm spectral region XXXIV, 95 Septier, Albert: Strong-focusing lenses, XIV. 85

Shaffer, R. A.: see Collings, P. R. Shapiro, G.: Subminiaturization techniques. 111. 195 Shelley. E. G.: see Mende, S. B. Shen, T. Y.: see Choudry, A. Shepherd, F. D., Jr., Yang, A. C., Roosild, S. A., Bloom, J. H., Capone, B. R., Ludington, C. E., and Taylor, R. W.: Silicon Schottky barrier monolithic IRTV focal planes, X L B, 981 Sheppard, C. J. R.: see Howorth, J. R. Shimizu, K., Yoshida, 0.. Aihara, S., and Kiuchi, Y.: Characteristics of a new camera tube with a CdSe Photoconductive target, XXXIII A, 293

Shirouzo, S.: see Miyashiro, S. Shortridge. K.: see Boksenberg, A. Shrager. Peter G.. and Susskind, Charles: Electronics and the blind, XX, 261 Sibbet, W.: see Bird, P. R. Sibbett, W.: see Bradley, D. J. Siegmann, H. C.: see Campagna, M. Silzars, Aris: see Bates, David J. Simon, J. C.: see Broussaud, G. Simon, J. H.: see Frohlich, H. Singer, J. R.: Masers and other quantum mechanical amplifiers, XV, 73 Sirou, F.: see Guyot, L. F. Skingsley, J. D.: see Holeman, B. R. Skonnko, G.: see Schneeberger, R. J. Slark, N. A.: see Batey, P. H. Slark, N. A.: see Beurle, R. L. Slark, N. A., and Woolgar, A. J.: A transmission secondary emission image intensifier, XVI, 141 Sleat, W. E.: see Bradley, D. J.

145

367

Smit. J., and Wijn, H. P. J.: Physical properties of ferrites, VI, 69 Smith, C.: see Watton, R. Smith, C. V. L.: Electronic digital computers, IV, 157 Smith, C. W.: An x-ray sensitive photoconductive pick-up tube, XII, 345 Smith, C. W., Kao. K. C., Calderwood, J. H., and McGee, J. D.: A study of prebreakdown phenomena in n-hexane using an image intensifier tube, XXII B, 1003

Smith, D. G.: see Hughes, J. S. Smith, R. W.: see Berg, A. D. Smith R. W.: see Morgan, B. L. Smith, R. W.: The application of the electron image store and analyser to high-speed photography, XXVIII B, 1011 Smith, W. A,, Chatterton, P. A., Elliot, C. T., and Pulfrey, D. L.: A high speed photographic study of the electrical breakdown of small gaps in vacuum, XXVIII B, 1041 Smith-Rose, R. L.: Radiowave propagation: A review, IX, 187 Smout, D. W. S.: see Iredale. P. Smyth, M. J.: see Brand, P.W. J. L. Smyth, M. J., and Brand, P. W. J. L.: Linearity of electronographic emulsions, XXVIII B, 737 Snoek, C., and Kistemaker, J.: Fast ion scattering against metal surfaces, XXI. 67 Sodha, Mahendra Singh, and Kaw, Prehiman Krishan: Theory of the generation of harmonics and combination frequencies in a plasma. XXVII, 187 Somers, L. E.: The photoemitter-membrane light modulator image transducer, XXXIII A, 493 Southon, M. J.: see Turner, P.J. Southon, M. J.: see Whitmell, D. S. Spiegel, H. J.: see Gebel. R. K. H. Spinella, Salvatore: see Bates, David J. Srb, I: see JedliEka, M. Srininasan, M.: see Bhatia, T. B. Stahnke. Ingeborg, and Heinrich, Hans: Special problems in measuring the modulation transfer function of x-ray image intensifiers, XXII A, 355

3 68

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

Stark, A. M., Lamport, D. L., and Woodhead, A. W.: Calculation of the modulation transfer function of an image tube, XXVIII B, 567 Statz, Hermann: see Pucel, Robert, A. Stephens, C. L.: see Bacik, H. Stephens, C. L.: see Cullum, M. J. Stephens, C. L.: see McGee, J. D. Sternheimer, R. M.: Parity nonconservation in weak interactions, XI, 31 Stevefelt, J.: see Delpeck, J.-F. Stone, H. D. : Preparation of high-resolution phosphor screens, XXII A, 565 Stoudenheimer, R. G.: Image intensifier developments in the RCA electron tube division, XII, 41 Strausser, Y.E.: see Medved, David B. Stricker, S.: The Hall effect and its applications, XXV, 97 Sturgell, C. C., Williams, J. T., Feibelman, W. A., Boksenberg, A., Anderson, B. E., MacKrell, G . E., and Jones, T. J. L.: Application of new ultraviolet television detectors in an astronomical satellite, XXXIII B, 91 1 Sturimer, W.: Some applications of solid state image converters (SSIC), XVI, 613 Su, Timothy,: see Bowers, Michael T. Susskind, Charles, see Shrager, Peter G. Susskind, C.: Electron guns and focusing for high-density electron beams, VIII, 363

Suzuki, Y.:see Hiruma, T. Suzuki, Y.:see Kurasawa, K. Suzuki, Y.,Uchiyama, K., and Ito, M.: A large diameter X-ray sensitive vidicon with beryllium window, XL A, 209 Svelto, V.: see Donati, S. Swainston, P.: see Towler, G. 0. Swank, Robert K.: see Vosburgh, Kirby G. Swanson, L. W., and Bell, A. E.: Recent advances in field eiectron microscopy of metals, XXXII, 193 Swanson, R. A.: see Reynolds, G. T. Swart, L. M., and van Rooy, H. J.: Combined magnetic deflection and focusing in a pick-up tube with the scanning focus coil, XXXIII A, 527 Svms. C. H.A,: Gallium arsenide thin-film photocathodes, XXVIII A, 399 r

.

Szepesi, Z., and Novice, M.: Solid-state radiographic amplifiers and infra-red converters, XXVIII B, 1087

T

Tachiya, H.: see Ninomiya, T. Takahashi, M.:see Sato, K. Taketoshi, K.: see Ninomiya, T. Taylor, A.: see Goetze, G . W. Taylor, D. G.: see Schagen, P. Taylor, D. G.: see Woodhead, A. W. Taylor, D. G.: The measurement of the modulation transfer functions of fluorescent screens, XXII A, 395 Taylor, D. G., and Schagen, P.: The application of channel image intensifiers to low light-level television, XXXIII B, 945

Taylor, D. G., Petley, C. H., and Freeman K. G.: Television at low light-levels by coupling an image intensifier to a Plumbicon, XXVIII B, 837 Taylor, R. W.: see Shepherd, F. D., Jr. Taylor, S.: An infra-red-sensitive television camera tube, XII, 263 Tepinier, M.: see Vernier, P. Ter-Pogossian, M.: see Ball, Jack Tessier, M.:see Malherbe, A. Teszner, J. L.: see Teszner, S. Teszner, S.,and Teszner, J. L. : Microwave power semiconductor devices, I, XXXIX, 291 Teszner, S., and Teszner, J. L.: Microwave power semiconductor devices. 11. critical review, XLIV, 141 te Winkel, J.: Past and present of the charge-control concept in the characterization of the bipolar transistor, XXXIX, 253 Theile, R.: On the signal-to-noise ratio in television storage tubes, XII, 277 Theodorou, D. G.: Research on photocathode surfaces at the Bendix Corporation Research Laboratories Division, XXII A, 477 Theodorou, D. G., and Pickels, J. C.: Salient sensor characteristics for low light-level TV systems, XXXIII, B, 979

CUMULATIVE AUTHOR INDEX, VOLUMES Thomas, B. R.: A high-resolution diode for high-speed photography, XXXIII B, 1119

Thomas, S. W., Tripp. G. R., and Coleman, L. W.: S.1 photocathode response linearity and dynamic range with picosecond 1.06 pm laser pulses, XL A,

1-45

369

Uno, Y., Kawakami. H., Maeda, H., and Miyazaki, E.: Cathode-ray tube with thin electron-permeable window, XXVIII A, 81

V

451

Thonemann, F. F.: see Barlow, G . E. Thumwood, R. F.: see Cochrane, J. A. Thumwood, R. F.: see Foreman, P. H. Thumwood, R. F.: see Garfield, B. R. C. Thumwood, R. F.: see Goodson, J. Thumwood, R. F.: see Needham, M. J. Todkill, A.: see Emberson, D. L. Tomovik, R.: Systems approach to skeletal control: concept of the system, XXX, 2 73

Towler, G. 0.:see McMullan, D. Towler, G. O., and Swainston, P.: Assessing the performance of low light-level camera tubes, XXXlll B, 961 Trawny, E. W. L.: see Howorth. J. R. Triest, W. E.: see Pawley. M. G. Trindade. Armando Rocha, sec Delcroix, Jean-Loup Tripp. G . R.: see Thomas, S. W. Trolander, Hardy W., and Veghte, James H.: Recent advances in biological temperature measurements, XXX, 235 Trunk, G . V.: Radar signal processing, XLV, 203 Tsuji, S.: see Hori. H. Tufts, D.: see Fisher, D. Tufts, D. W.: see McCullough. W. V. Turnbull, A. A,: see Schagen, P. Turner, P. J., Cartwright, P., Boyes, E. D., and Southon, M. J.: Use of channelplate intensifiers in the field-ion microscope, XXXIII B, 1077 Twiddy, N. D.: see Filby, R. S. Twiss. R. Q.: On the steady state theory of the magnetron, V, 247

u Uchiyama, K.: see Suzuki, Y. Unger, H.-G. and Harth, W.: Physics and applications of MIS varactors, XXXIV. 281

van Alphen, W. M.: A small high-precision electrostatic pick-up tube, XL A, 183 van Alphen, W. M.: Combined electrostatic focussing and deflection, XXXlII A, 51 1 Vance, A. W., Hutter, E. C., Lehmann, J., and Wadlin, M. L.: Analog computers, VII, 363 Van de Handel, J.: Paramagnetism, VI, 463 Van der Ziel, A.: Fluctuation phenomena, IV, 110 van Huyssteen, C. F.: A simple photocathode transfer system, XL A, 419 van Leunen, J. A. J.: The multiplication rule in the O.T.F. concept, XXXlII B, 585 van Leunen, J. A. J., and Pennings, J. C.: A simple instrument for measuring image sharpness and noise characteristics of image intensifiers, XL 9, 577 van Roosmalen, J. H. T.: Adjustable saturation in a pick-up tube with linear light transfer characteristic, XXVIII A, 28 1 van Rooy, H. J.: see Swart, L. M. Varma, B. P.: see Bhatia, T. B. v. d. Polder, L. J.: Beam-discharge lag in a television pick-up tube, XXVIII A, 237 v. Overhagen, J.: see Kuhl, W. Varma, B. P.: see McGee, J. D. Veghte, James, H.: see Trolander, Hardy W. Vennik, 3.: see Fiermans, L. Verat, M.: see Driard, B. Verma, R. L.: see Bhatia, T. B. Vernier, P.: see Mathey, K. Vernier, P.: see Pauty, F. Vernier, P., and Hartmann. P.: Resultats obtenus a I'aide de la camera electronique Lallemand dans I'etude de l'emission photoelectrique. XXII A, 519 Vernier, P., Hartmann, P., Niquet, G., and Tepinier, M.: h u d e de I'emission photoelectrique des structures metalisolant-metal, XXVIll A, 409

3 70

CUMULATIVE AUTHOR INDEX, VOLUMES

Veron, S.: see Malherbe, A. Veron, S.: Quelques aspects der essais de de@t de photocathodes S.20 et d'ecrans fluorescents sur fibres optiques, XXVIII A, 461 Verwey, J. F.: Nonvolatile semiconductor memories, XLI, 249 Vilim, P.: see JedliEka, M. Vine, J.: see OKeefe, T. W. Vine, J.: The design of electrostatic zoom image intensifiers, XXVIII A, 537 Vitovsky, V.: see JedliEka, M. Vodovnik, L.: Functional electrical stimulation of extremities, XXX, 283 von Engel, A., and Cozens, J. R.: Flame plasmas, XX, 99 Voorman, J. 0.:see Adams, K. M. Vosburgh, Kirby G., Swank, Robert K., and Houston, John M.: X-ray image intensifiers, XLIII, 205

W Wachtel, M. M., Doughty, D. D., and Anderson, A. E.: The transmission secondary image intensifier, XII, 59 Wadlin, M. L.: see Vance, A. W. Wagner, K. H.: Application of image intensifiers and shutter tubes to the study of gas discharges, XXVIII B, 1033 Wagner, Robert, J.: see McCombe, Bruce, D. Wait, James, R.: Recent theoretical advances in the terrestrial propagation of ULF electro-magnetic waves, XXV, 145

Walker, G. A. H.: see Buchholz, V. L. Walker, G. A. H., Auman, J. R., Buchholz, V. L., Goldberg, B. A,, Gower, A. C., Isherwood, B. C., Knight, R., and Wright, D.: Application of an image isocon and computer to direct digitization of astronomical spectra, XXXIII B, 819 Walker, M. F.: Performance of the Spectracon in astronomical spectroscopy, XXVIII B, 773 Walker, M. F.: Recent astronomical observations obtained with the Lallemand electronic camera, XVI, 341 Walker, M. F.: Recent progress in the use of

1-45

the Lallemand electronic camera in astronomical spectroscopy, XXII B, 761 Walker, M. F.: Recent results in the use of the Spectracon for direct electronography, XL B, 819 Walker, M. F.: The use of electronographictype image tubes in astronomical photometry, XXXIlI B, 697 Walters, F. W.: see Huston, A. E. Walters, J.: see Binnie, D. M. Wampler, E. J.: see Miller, J. S . Ward, R.: Noise measurements on image intensifiers, XL B, 553 Wardley, J.: A high-resolution ruggedized half-inch vidicon, XXII, 211 Wardley, J.: An improved ultra-violet sensitive vidicon, XVI, 227 Wardley, J., and Jackson, F. W.: A 13-mm all-electrostatic vidicon, XXVIII A, 247 Warnecke, R. R., Chodorow, M., Guenard, P. R., and Ginzton, E. L.: Velocity modulated tubes, 111.43 Washington, D.: see Millar, 1. C. P. Waters, J. R.: see Reynolds, G. T. Watson, C. J.: see Motz, H. Watton, R.: see Putley, E. H. Watton, R., Jones, G. R., and Smith C.: Pyroelectric materials for operation in a hard vacuum, XL A, 301 Webster, H. F.: see Houston, J. M. Webster, W. M.: A comparison of analogous semiconductor and gaseous electronic devices, VI, 257 Wehner, G. K.: Sputtering by ion bombardment, VII, 239 Weiler, E. J.: see Dunlap, J. R. Weimer, Paul K.: Television camera tubes: A research review, X111, 387 Weimer, P. K.: Image sensors for solid state cameras, XXXVII, 181 Weingartner, H. C., and Kennedy, S. W.: Modern vacuum pumps in electronics manufacturing, V, 213 Weller, W.: see Jeffers, S. Wellgate, G. B.: see McMullan, D. Wendt, G.: INTIC, an image intensifying, integrating and contrast-enhancing storage tube, XXVIII A, 137 Wheeler, B. E.: see McGee, J. D. Wheeler, B. E., and Emberson, C. J.: Some measurements on the direct recording of

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

electron images using thin windows, XXII A, 51 Whelan, M. J.: Electron diffraction theory and its application to the interpretation of electron microscope images of crystalline materials, XXXIX, 1. Whetten, N. R., see Dawson, P. H. White, J. E.:Tube miniaturisation, 111, 183 Whitmell, D. S., and Southon, M. J.: Image intensification in field-ion microscopy, XXIl B, 903 Wijn, H. P.J.: see Smit, J. Wilcock, W. L.: see Emberson, D. L. Wilcock, W. L.: see Hughes, J. S. Wilcock, W. L.: Routine measurement of the responsive quantum efficiency of photoemissive cathodes, XXII A, 535 Wilcock, W. L.: Statistics of transmitted secondary electron multiplication, XXll A, 629 Wilcock, W. L., and Baum, W. A.: Astronomical tests of an imaging photomultiplier, XVI, 383 Wilcock, W. L., and Miller, D. E.: Statistics of transmitted secondary electron emission, XXVlIl A, 513 Wild, 1. P.: Observational radio astronomy, VII, 299 Wilkinson, M.: see Fisher, D. Williams, B. F., Martinelli, R. U., and Kohn, E. S.:Negative electron affinity secondary emitters and cald cathodes, XXXIII A, 447 Williams, F. E.: Sotid-state luminescence, V, 137 Williams, J. T.: see Sturgell, C. C. Willingham, D.: see Rindfleisch, T. Wilson, G. A.: see Morgan, B. L. Wilson, R. J. F.: see Garfield, B. R. C. Winkler, Gernot M. R.: Time keeping and its applications, XLIV, 33 Wise, H. S., Richards, E. W. T., and Martin, R.: Digital read-out of an image intensifier using a vidicon or a scanning spiral slit plus a digital memory oscilloscope, XXVIIl B, 981 Wlerick, G.: see Lelieve, G. Wltrick, G.: Photombtrie Bidimensionelle avec la camera electronique, XL B, 855 Wlerick, Gerard, and Grosse, Achilles: La camera electronique: un recepteur

371

d'images sans lumiere diffusee, XXII A, 465 Wlerick, G.: see Charrier, Mlle S. Wlerick, G.: see Lallernand, A. Wlerick, G.: see Rosch, J. Wlerick, G.: Etudes d'astres faibles en lumiere totale avec la camera electronique, XXVllI B, 787 Wlerick, G., Rosch. J., Dupre, Mlle M., and Bellier, Mlle M.: La photographie electronique des planetes et ses applications photometriques, XVI, 371 Wolfgang, L. G.: see Abraham, J. M. Wolstencroft, R. D.: see Brand, P. W. J. L. Woodhead, A. W.: see Schagen, P. Woodhead, A. W.: see Stark, A. M. Woodhead, A. W., Taylor, D. G., and Schagen, P.: A two-stage electrostatic image intensifier with a large photocathode area, XVI, 105 Woolgar, A. J.: see Goodson, J. Woolgar, A. J.: see Slark, N. A. Woonton, G . A.: Relaxation in diluted paramagnetic salts at very low temperatures, XV, 163 Worswick, P.: see Ring, J. Worswick, S. P.: see Coleman, C. I. Wreathall, W. M.:see Beurle, R. L. Wreathall, W. M.: see Harmer, A. L. Wreathall, W. M.: see Putley, E. H. Wreathall, W. M.: Aberrations of diode image tubes, XXII A, 583 Wright, D.: see Walker, G. A. H. Wright, S. L.: see Boksenberg, A. Wronski. C. R., and Cope, A. D.: Antimony triculfide heterojunction vidicon structures, XL A, 349 Wyatt, J. R.: see Saalfeld, F. E. Wynne, C. G., and Kidgear, M. J.: The design of optical systems for use with image tubes, XXVlII B, 759 Wysoczanski, W.: see Beaver, E. A.

Y

Yank A. C.: see Shepherd, F. D., Jr. Yang, Edward S.: Current saturation mechanisms in junction field-effect transistors, XXXI, 247

3 12

CUMULATIVE AUTHOR INDEX, VOLUMES 1-45

Yeadon, E. C., and Clarke, J. A.: Modulation transfer function measurements on channel image intensifiers, XXXIII B, 593 Yoshida, 0.:see Shimizu, K. Yoshizaki, S.: See Hagino, M.

Z

Zacharias, J. R.: see King, J. G. Zacharov, B.: Image resolution in thinwindow intensifiers using homogeneous fields, XVI, 67 Zacharov, B.: A demagnifying image tube for nuclear physics applications, XVI, 99 Zacharov, B., and Dowden, S.: An

image intensifier with a thin endwindow, XII, 31 Zalm, P.: Thermionic cathodes, XXV, 211 Manis, R. A.: see Reynolds, G. T. Zeitler, E.: Resolution in electron microscopy, XXV, 277 Zimmermann, Bodo: Broadened energy distributions in electron brams, XXIX, 257 Zimmerman, U.: see Baumgartner, W. Zucchino, P.: Photometric statistical performance of the SEC target, X L A, 239 Zucchino, P. M.: see Lowrance, J. L. Zucchino, P. M., and Lowrance, J. L.: Recent developments and applications of the SEC-vidicon for astronomy, XXXIII B, 801

Cumulative Subject Index, Volumes 1-45 A

Accelerators linear ion, XXV, 1 particle, I, 269 Acoustoelectric interactions, in 111-IV compound semiconductors, XXXI, 161 Aids to navigation, electronic, I, 425 Airborne detector, magnetic, IV, 258 Alkali halide crystals, pre-breakdown light emission from, XXII B, 995 Amplification of transient images, XII, 135 transmission secondary emission, XVI, 557 Amplifiers operational, XI, 225 quantum mechanical, XV, 73 Amplitude, pulse, analysis, VIII, 317 Amplitude measurement in nuclear physics VIII, 256 Analog computers, VII, 353 Antennae, endfire, XIX, 255 Astronometric images, camera tubes for recording, XXII A, 175 Astronomical cassegrain spectrographs, XXXlII B, 769 Astronomical electronography data reduction for, XXXllI B, 757 electron-sensitive plates XXXIII A, 67 Astronomical observations recent, obtained with Lallemand electronic camera, XVI, 341 with University College London image photon counting system, XL B, 877 Astronomical photometry electronographic-type image tubes in, XXXlIl B, 697 image orthicon and, XL B, 901 Astronomical spectra digitization of, XXXIII B, 819

Astronomical spectroscopy image intensifier multichannel analyser for, XL B, 887 linear silicon array for, XL B, 879 Astronomical tests of barrier-membrane image converters, XII, 21 of imaging photomultiplier, XVI, 383 Astronomical uses of image intensifying tubes, XVI, 403 Astronomy applications of spectracon in, XL B, 803 comparison of image intensifiers for, XXVlII B, 753 digicon applications in, XL B, 745 digital television system for, XL B. 699 image scanning techniques in, XVI, 409 improvements in image orthicon and, XXXIII B, 789 observational radio, VII, 299 SEC vidicon for, XXXIII B, 801 solar radio, XX. 147 Asymmetrical astigmatism of X-ray image intensifiers, XL A, 507 Atomic collisions, XVIII, 67 Atomic photoelectron spectroscopy, XLI, 73; XLIV, 1 Auger electron spectroscopy, in surface research, XLIII, 139 Aurora borealis, IX, 1 Automatic data processing, XI, 185 Axial-heam tubes, XIV, 299

B Barrier-membrane image converters, astronomical tests of, XII, 21 Beams deflection of, I. 671 high density electron, VIII, 363 magnetically focused cylindrical, X, 1 373

374

CUMULATIVESUBJECT INDEX. VOLUMES

molecular, new applications and techniques, VIII, 1 Beta-ray spectrometers, V, 97 Biological temperature measurements, XXX, 23 5 Bistable-phosphor storage tubes XXXIII A, 33 1

C Caesium vapor effects upon target glass, XXII A, 561; XXVIII A, 309 getter materials for, XXVIII A, 381 reaction with gold, XXII A, 643 Camera tubes beam acceptance in, XXXIII A, 317 beam-discharge lag in, XXXIII A, 317 with CdSe photoconductive target, XXXIII A, 293 evaluating, XXXIII 9, 601 high grain, point source sensitivity, of, XL A, 263 imaging in, XXXIII A, 317 noise measurements in, X L 9, 585 SEC magnetically focused, XXXIII A, 241 proximity focused, XXXIII A, 253 with Si electron multiplication target, XXXIII A, 207 TV with large Si dole array targets, XXXIII A, 219 uniform layer heterojunction targets for, XXXIII A, 229 Camera tube studies near infrared, with Ag,S target, XL A, 365 Cascade image intensifier, XXII A, 113; XXVIII A, 89 astronomical uses, XXII 9, 697 comparison with transmission secondary emission type, XXII A, 129 dark current scintillations of, XXVIII 9, 713 fibre-optic coupled, XXVIII A, 119 four-stage, characteristics of, XXII A, 87 influence of temperature on, XXII A, 101 magnetically focused, XVI, 113 Cascade image intensifier camera, for beamfoil spectroscopy, XXVIII 9, 907

145

Cascade phosphor-photocathode screens, evaluation, XXII A, 407 Cathode ray oscilloscopes, recent developments, X, 239 Cathode ray tube with electron-permeable window, XXVIII A, 81 manufacture and design, 11, 2 progress, 11, 2 Cathodes oxide coated, I, 1 thermionic, XXI, 21 1 Cathodoluminescence, 11, 152 Cerenkov chamber, with four-stage image intensifier, XXVIII B, 919 Cerenkov radiation, at microwave frequencies, XIV, 265 Channel electron multiptier(s) dc operation, analysis of, XXXIII A, 183 high, gain, characteristics of, XXXIII A, 125 ion feedback noise in, XXXIII A, 175 plates, in x-ray image intensification, XXXIII A, 153 Channel multiplier for imaging applications, XXVIII A, 471 impulse and frequency response of, XL A, 123 low-level performance of, XXVIII A, 487 problems concerning, XXVIII A, 499 space charge in, XL A, 113 vacuum space charge in, XXVIII A, 507 x-ray detection by, XXII A, 139 Channel plate glass surfaces XL A, 153 Channel plate intensifiers, in field ion microscope, XXXIII 9, 1077 Channelled image intensifier, XII, 97 progress report on, XVI, 155 Characteristic energy losses, of electrons in solids, VII, 183 Charge-control concept, and the bipolar transistor, XXXIX, 253 Charge integration experiments, XII, 219 Charge particle beams, deflection of, I, 167 Charged pigment xerography, XXXVIII, 83 Chemical lasers, XXXI, 1 Cluster ions, in laboratory and ionosphere, XXIX, 79 Cold cathode silicon vidicon, XL A, 463 Collisions, inelastic, between atomic systems, XIII, 1

CUMULATIVE SUBJECT INDEX. VOLUMES

Color television, recent work in, V, 291 Combination frequencies, in plasm& 187 Combined electrostatic focusing and deflection, XXXlll A, 5 1 1 Combined magnetic deflection and focusing, XXXllI A, 527 Communications, satellite, XXXI. 119 Communication theory, 111, 306 Computer logic organization, X, 153 Computer organization, recent developments in, XVIII, 45 Computers analog, VII, 363 electronic digital, IV, 157 Computers (mini), basic concepts of, XLIV, 283 Conduction, metallic, at high frequencies and low temperatures, VI, 1 Conductivity, electron bombardment induced, XVI, 235 Continuous-wave magnetrons, modulation of, IV, 188 Contrast-enhancement, in imaging devices XXVIIIB. 661 Cosmic radio noise, 1, 347 Cosmic rays and image intensifier dark current, XXVIII B, 705 intensity variations in, IX, 129 Coude spectrograph, performance of image tubes in, XXII B, 729 Counter, scintillation, IV, 69 Crystals electron-beam investigating of, XXIV, 109 magnetic properties of, XXIV. 109 Currents, space-charge-limited, VI, 138 Current saturation mechanisms, in junction field-effect transistors, XXI, 247 Cylindrical beams, nonuniform D-C electron flow in, X, 1 Cylindrical magnetron, electronic theory of, 11. 15

D Data processing, automatic, in physical sciences, XI, 185 D-C electron flow.nonuniform, in magnetically focused beams, X. 1

1-45

375

Defects in diamond-type semiconductor crystals, X. 71 Deflection of beams of charge particles, 1, 167 Design of cathode ray tubes, 11, 26 Detective quantum efficiency image tubes and, XL A, 539 of intensifiers, XXVIII B, 577 Detector@) magnetic airborne, IV, 258 quantum efficiency of, XI, 87 for visible and infrared radiation, V, 1 Detector system, for ultraviolet Explorer satellite, XL A, 223 Diagnostic radiology, electronic imaging techniques for, XL B, 945 Dielectric breakdown in solids, intrinsic, 11, 185

Digicon, applications in astronomy, XL B, 745 Digital computers. electronic, IV, 157 Digital memory oscilloscope, XXVIII B, 98 1 Digital television system, for astronomy, X L B, 699 Diode arrays parallel and self-scanned, to detect photoelectrons, XL B, 735 self-scanned, single electron recording, by, XL B, 779 Diode image tubes, abberations of. XXIl A. 583 Discharges, electrical, in gases, VII, 401 Distortion of electron images, XXXIII A, 545

Distribution of electrons, XV, 265

E Edge detecting system, XL B, 963 Efficiency contrast transfer function, of intensifiers, XXVllI B, 577 Ejection, kinetic, of electrons from solids, XXI, 101 Electroluminescence, XVI. 621 Electromechanical picture signal generating device, XXVIII A, 297 Electron acoustic waves, nonlinear, XXXV, 1; XLI, 1

376

CUMULATIVE SUBJECT INDEX, VOLUMES

Electron beams as analytical tools in surface research (LEED and AES), XLIII, 139 broadened energy distributions, XXIX, 257 dense, analysis of, XXVI, 1 high density, VIII, 363 polarization of, XXI, 1 Electron beam addressed memories, XLIII, 43

Electron beam readout technique, X L A, 219

Electron beam scanning, XXXIII, 535 Electron-bombarded semiconductor devices, XLIV, 221 Electron bombardment induced conductivity image devices working on, XXII A, 323 properties of, XXII A, 315 Electron diffraction structure analysis, XI, 355

Electron diffraction theory, applications, XXXIX, 1 Electron emission from forward biased p n junctions, XXXIII A, 459 secondary, I, 66; XI, 413 from solids, XLI, 113 Electron emission microscopy, XVIII, 25 1 Electron flow, in magnetically focused beams, X, 1 Electron guns, for high density electron beams, VIII, 363 Electron image, direct recording, using thin windows, XXll A, 51 Electron image multiplier, XII, 87 Electron image store and analyser, XXII, B, 969

Electron interaction, space-harmonic traveling-wave, XVII, 1 Electron lenses, 11, 48 Electron micrograph analysis by optical transforms, XLIII, 1 Electron microscope, V1, 269; XII, 317 Electron microscope imaging, with silicon diode array vidicons, X L A, 287 Electron microscopy, resolution in, XXV, 277

Electron mirror microscopy, XXVI, 217 Electron multiplication, secondary image intensifiers, XVI, 127 Electron-optical deflexion and storage techniques, XXII B, 985

1-45

Electron-optical systems imaging properties of, XXVIII A, 523 problems, using computer, X L A, 473 Electron-optical transfer functions of image intensifiers, XXXIII A, 563 Electron optics field plotting and ray tracing in, 11, 102 scalar diffraction in, XXX, 139 Electron probe microanalysis, XIII, 3 17 Electron scattering, in solids, IV, 2; VII, 183 Electron spectroscopy, recent applications of, XLII, 55 Electron spin resonance, in mineralogy and geology, XXIV, 51 Electron streams, velocity distribution in, XIII, 181 Electron tubes for high-speed photography, XVI, 249 use of meshes to reduce errors in, XXXIII A, 571 Electronic aids to navigation, I, 425 Electronic camera, XVI, I, 19 in astronomical spectroscopy, XXIl B, 761 bidimensional photometry with, X L B, OOO diffused light in, XXII A, 465 in double-star photography, XXII B, 755 electrostatically focused, in physics and astronomy, XXVIII A, 27 for enlargement 1/7, XVI, 27 focusing with cylindrical lens, XXII A, 609 for high-speed cinematography, XXXIII B, 1101 infra-red application of, XXII A, 1 new technique for utilization of, XVI, 19 photography of double stars by means of, XVI, 357 photometry with, XXXIII B, 737 relation to standard photography, xxii A, 5 for space research, XXVIII A, 39 studies of weak stars in daylight with. XXXlIl B, 119 in study of photoelectric emission, XXXIIl A, 519 study of photoelectroic threshold, XXXIII A, 423 study of weak stars, XXVIII B, 787 used in reflection mode, XXXlIl A, 557 Electronic devices gaseous, comparison with semiconductors, VI, 257

CUMULATIVE SUBJECT INDEX, VOLUMES high-power, XLI, 3 1 1 Electronic imaging limitation to resolving power in, XVI 299 techniques for improved diagnostic radiology, X L B. 945 Electronic imaging devices, micro-channel plates in, XL A, 91 Electronic photography, of planets, XVI. 371 Electronic systems, gyrator in, XXXVII, 79 Electronic theory of cylindrical magnetron, 111, 15 of plane magnetron, 111, 185 Electronic zooming, XVI, 195 Electronics and the blind, XX, 261 contributions to seismology and geomagnetism, IX, 297 modern, and electrical discharges in gases, VII, 401 modern vacuum pumps in, V, 213 in oceanography, IX, 239 thorium oxide and, V, 169 Electronographic camera five-centimeter magnetically focused, XL B, 613 large-image, XXVIII A, 19 Electronographic emulsions, linearity of, XXVIII B, 737 Electronographic image tube, progress toward 8 cm., XL B, 627 Electronographic plate, bakeable, advantages of, XVI, 35 Electronographic spectrographs, intemalgrating, XXXIII B, 895 Electronography direct, use of spectracon for, X L B, 829 of extended objects, X L B, 791 photometric accuracy of, XL B, 679 Electron(s) as a chemical entity, XIV, 1 energy spectrum of, XXXIX, 73 distribution of, in ionosphere, XV, 265 as hydrodynamical fluid, XX, 1 and ions, low energy, atomic collisions involving, XVIII, 67 Electrostatic fields, computation of axially symmetric, X L A, 485 Electrostatic image intensifiers. application to astronomy, XXVlII B, 807 Electrostatic lenses, reducing defects in imaging devices using, XXII A, 6 0 1

145

377

Electrostatic pick-up tube, XL A, 103 Emission field, 111, 1 ; VIII. 90;XII, 73 transmission secondary, XII, 59 Emitron, C. P. S., charge integration experiments with, XII, 219 Endfire antennae, XIX, 255 Energy conversion, thermionic, XVII, 125 Energy losses, of electrons in solids, VII, 183 Energy spectrum, of electrons from hot cathode, XXXIX, 73 Evaluation, semiconductor device, XVIII, 167 Evaporation, impact, in glow discharge, XVII, 245 Evoscope, h e d pattern generator, XXII A, 331

F Ferrites, physical properties of, VI, 70 Ferromagnetic phenomena at microwave frequencies, 11, 251 Ferromagnetism, relaxation processes in, VI, 47 Field-effect image storage panels, XXVIII B, 1059 Field effect transistors, GaAs microwave, properties of, XXXVIII, 195 Field electron microscopy of metals, XXXII, 193 Field emission, VIII, 90 in image tubes, XII, 73 microscopy, 111, 1 Field ionization, XIII, 83 Field ion microscopy, XIII, 83 image intensification in, XXII B, 903 Field plotting, in electron optics, 11, 102 Filament scintillation chamber, XVI, 487 Flame plasmas, XX, 99 Fluctuation phenomena, IV, 110 Fluoroscopy, medical, image quality in, XXI A, 363 FM broadcast band, propagation in, I, 381 Focused cylindrical electron beams, magnetically, X, 1 Focusing for high density electror beams, VIII, 363 Free atoms, lifetimes of, XXIX, 115 Frequencies high metallic conduction at. VI, 1 microwave, XIV, 265

378

CUMULATIVE SUBJECT INDEX, VOLUMES

Frequency standards, atomic beam, XV, 1 Functional electrical stimulation of extremities, XXX, 283

G

Gallium arsenide, thin-film photocathodes, XXVIII, 399 Gallium arsenide substrates, measurement of diffusion length in, A, 389 Gamma ray cameras, evaluation of image quality, XXXIll B, 1041 Gas discharge displays, XXXV, 191 Gaseous electronic devices, VI, 257 Gases, electrical discharges in, VII, 401 Geomagnetism, contributions of electronics to, IX, 297 Germanium, electrical properties of, VII, 87 8-factor anomaly, of free electrons, XXI, 1 Glass scintillators applications of, XVI, 547 properties of, XVI, 547 Glow discharge, impact evaporation and thin film growth in, XVII, 245 Grating storage target, XXII A, 155 Guns. electron, high density, VIII, 363

H Hale 200-in. telescope, image-tube spectrograph for, XXVIII 9, 767 Hall effect advances in, XXXVI, 153 and applications, XXV, 97 Harmonic generation in nonlinear beam plasma systems, XXIX, 1 in plasma, XXVII, 187 Heterojunction target, DdS-CdTe-As,Se, , X L A, 335 Heterojunction vidicon structures, antimony trisulfide. X L A, 349 Heterojunction vidicon targets, evaporated and sprayed CdSe layers for, X L A, 323

n-Hexane, pre-breakdown using image intensifier, XXll 9, 1003 High density electron beams, VIII, 363

1-45

High frequencies, metallic conduction at, VI, 1

High gain camera tubes, X L A, 263 High-gain image intensifier, XII, 135 and field-ion microscopy, XXVIII 8, 875 increasing storage capacity, XXVIII 9, 745 High-power electronic devices, XLI, 31 1 High-power tubes, XIV, 299 High speed framing camera(s), XXXIII 9, 1109

High-speed photography, XXXIII B, 1119 electron store and analyser application to, XXVIII 9, 1011 image orthicon in, XXII B, 1101 Hollow cathode arcs, XXV, 87 Hydrodynamical fluid, XX, 1 Hydrogen thyratrons, XIV, 207

I Image amplifiers, solid state, recent developments in, XVI, 607 Image converter(s) barrier-membrane, astronomical tests, XII, 21

boundary layer, XVI, 633 electrostatically focused, XXII A, 441 solid state, applications of, XVI, 613 solid state fast response, XXII B, 683 Image converter tube oblique, X L A, 69 grid shuttered, X L A, 59 Image detectors, photoconductive, for astronomical uses, XVI, 451 Image devices signal-to-noise ratio of, XII, 291 spatial frequency response of, XXII A, 425 Image formation, in electron microscope, XXXII, 63 Image generation by means of twodimensional, spatial electric fitters, XLI, 167 Image intensification, XII, 327 low brightness photography by, XVI, 55 x-ray, CEM plates in, XXXlll A, 153 Image intensifier(s) application to gas discharges, XXVlll 9, 1033

application to luminescent chamber, XII, 153

CUMULATIVE SUBJECT INDEX, VOLUMES 1 4 5 in astronomical cassegrain spectrographs, XXXIII B, 769 in astronomy, XXIII, 347 for astronomy, XXXIII B, 677 cascade. research on, XXXIII A, 99 channel, for low light television, XXXIII B, 945 channeled. XII. 97; XVI, 155 developments of, XII, 41 digital read-out of, XXVIII B. 981 for electron microscope, XII, 317 electron-optical transfer functions of. XXXIll A. 563 electrostatic, two-stage, XVI, 105 electrostatic zoom, XXVllI A, 537 with fiber-optic coupling, XXVllI A, 105 for r-ray scintillation camera, XL A, 41 high-contrast channel, design and performance, XXXIll A. 133 high gain astronomical research utilizing, XXII B, 705 image quality, XXXIII B, 639 information transfer with, XXVIII B. 615 input field, for scintillation cameras, XXXlIl B. 1031 magnetically focused with evaporated field electrodes, XXXIII A, 93 magnetically focused cascade, experiences with, XVI, 119 measuring image sharpness and noise of, X L B, 577 microchannel, characteristics and performance, XXXIII, 145 microchannel wafer, X L A, 141 multi-stage, XVI, 567 noise measurements and, X L B, 553 observation of Cerenkov Ring with, XXII B, 801 for observation of rapid luminescence phenomena, XXII B, 949 proximity focused, X L A. 1, 11, 21 proximity focused diode, XXXIII A, 83 in scanning radioactive objects, XXll B, 94 1

secondary emission, XII, 59 sensitivity of, XXII. 38 with simple electron optics, X L A, 83 with streamer chambers, XXII B, 813 thermal-neutron, XXII €3, 781 with thii. end-window, XII, 31 for track recording, XVI, 113

379

transfer function measurements on, XXXIII B. 593 transmission, XVI. 141 with transmitted secondary electron multiplication. XVI, 127 use of. in nuclear physics, XVI, 501 visual performance at low light, XXVIII B, 635 wire-electrode type, experiments on, XVI, 62 1

x-ray, XVIII. 205 image quality of, XXXlll B, 1049 some experiences with, XVI, 601 Image intensifier cine angiography, XXXIII B, 1089 Image intensifier multichannel analyser, for astronomical spectroscopy, X L B, 887 Image intensifier noise, effects on visual pattern detection. X L B, 565 Image intensifier output screens, modulation transfer function, XXXIII A, 483 Image intensifier system, XVI, 75,475 for night vision, X L A, 33 Image-intensifying systems, visual thresholds using, XXXIII B, 631 Image intensifying tubes, astronomical, uses of, XVI, 403 Image isocon, use with computer, XXXIII B, 819 Image isocon tube, XXVlIl B, 827 Image multiplier, electron. XII, 87 Image orthicon(s), XVI, 447, 581 applications to astronomy, XXII B, 713 bombardment-induced conductivity targets for. XVI, 247 comparison of SEC camera tube and. XXII A, 291 improvements in application, to astronomy, XXXIIIB, 789 with magnesium oxide targets, XVI, 213 stellar photometry with, XVI, 431 using slow readout, XVI, 419 Image photon counting system, University College London, X L B, 877 Image quality. of photoelectronic imaging systems, X L B, 519 Image recording, comparison of efficiency, XXVIII B, 725 Image scanning in astronomy, potentialities and limitations of, XVI, 409 Image sensors, for solid state cameras,

3 80

CUMULATIVE SUBJECT INDEX, VOLUMES 1 4 5

XXXVII, 181 Image storage techniques, XVI, 593 Image transducers low energy quanta, XXII A, 189 photoemitter-membrane light modulator, XXXIII A, 493 Image tube@) analysis of, XXVIII B, 603 characteristics in x-ray diffraction, XL B, 923 computation of imaging properties, XXVIII A, 535 computation of symmetric electrostatic fields in, XL A, 485 demagnifying, for nuclear physics, applications, XVI, 99 and detective quantum efficiency, XL B, 539 development of electronographic, XXXIII A, 37 effect of pulse height distribution, XXXIII B, 617 electronographic type, in astronomy, XXXIII B, 697 for experimental electron optics, XL A, 493 field emission in, XII, 73 for high-speed photography, XXVIII B, 989 high-resolution, for integrated circuit fabrication, XXVIII A, 47 intensifier, evaluation, XXII A, 369 laboratory evaluation, for astronomical purposes, XVI, 391 Lallemand, modified, XVI, 47 Lenard window, XVI, 47 for astronomical spectrophotometry, XXII B, 741 low background, for electronography, XVI, 37 magnetic focusing of, XXII A, 617 modulation transfer function of, XXVIII B, 567 orthicon, XII, 379 astronomical photometry and, XL B, 901 photochron 11, XL A, 51 picosecond chronography, XXXIII B, 1145 proximity-focused, XXVIII A, 129 research, XII, 17 resolving power of, XXVIII B, 553 serial read-out from, XXXIII B, 873

signal generating, XII, 307 silicon diode array, with serial readout, XL B, 777 single, performance in spectroscopy, XXXIII B, 639 storage application of, XII, 3 11 experimental, XII, 247 for character recognition, XXVIII B, 1043 for optoelectronic computing, XXVIII B, 1051 two stage, XXXIII B, 653 Image-tube Fourier spectrograph, XXVIII B, 899 Image tube high-speed cameras, XXII B, 957 Image tube scanner, Lick observatory, status of,XL 8,693 Images planetary, stabilization of, XXXIII B, 781 transient, in high-gain photocathodephosphor intensifier systems, XII, 135 Imaging, flat channel system for, XXXIII A, 117 Inelastic collisions, XI11 1 Infra-red converters, XXVIII B, 1087 Infrared radiation detectors, V, 1 quantum efficiency of, XI, 87 Infra-red stellar spectroscopy, with micawindow tube, XXII B, 723 Infra-red television camera tube, XII, 263 Infrared TV focal planes, XL B, 981 Instrumentation, electronic for oceanography, XIX, 1 Insulators, low density deposits of, XVI, 145 Integrating television system, XXXIII B, 795 Intensifier orthicon, performance of, XII, 183 Intensifier tube, single-crystal, XXVIII B, 93 Intensifier vidicon, development of, XXXIII B, 937 Intensifiers. thin window image resolution in, XVI, 67 with phosphor output, XVI, 61 Intensity variations, in cosmic rays, IX, 129 Interactions, weak, parity nonconservation in, XI, 31 Internal optic image converters, XXXIII B, 88 1 INTIC, image storage tube, XXVIII A, 137 Intrinsic dielectric breakdown in solids, 11, 185

CUMULATIVE SUBJECT INDEX, VOLUMES 1 4 5 Ion bombardment, sputtering by, VII, 239 Ion implantation, in semiconductors, XXXVII, 263 Ion microscopy, field, XIII, 83 Ion scattering, against metal surfaces, XXI, 67 Ion thrusters, electron bombardment, XXXVI, 265 Ionization, field, XIII, 83 Ionosphere, radio wave scattering, in XIX, 55 Ionospheric research, I, 37 Ions lifetimes of, XXIX, 115 negative, IX, 43 “small” multimolecular atmospheric, XIX, 177 Isaac Newton telescope, direct spectracon exposures, XXXIII B, 747 Isocon photon counting TV system, XL B, 711 Isophate converter, XII, 307 Isotope separators, XLII, 113

J Junction field-effect transistors. XXXI 247

K Kodak IIa-0 emulsion, comparison of image recording with, XXVIII B, 725

38 1

Lick-Stromlo electronic camera, development of, XXII A, 59 Light amplifier, with high light output, XXVIII A, 151 Light-emitting devices design and applications, XLV, 39 methods, XLII, 179 Linear accelerators, XXV, 1 Linear silicon array, for astronomical spectroscopy, XL B, 879 Low background image tube, XVI, 37 Low density deposits, transmission secondary elission from, XVI, 145 Low energy electron diffraction, in surface research, XLIII, 139 Low energy electron physics, swarm techniques in, XXVII, 1 Low light levels, direct observation, image intensifier for, XVI, 75 Low light-level camera tubes, XXXIII B, 961 Low light-level television channel image intensifiers in, XXXIII B, 945 sensor Characteristics, XXXIII B, 979 Low temperatures, metallic conduction at, VI, 1 Luminance gain, of image intensifier systems, XXVIII B, 629 Luminescence, solid-state, V, 137 Luminescent chamber, XII, 153 high energy physics experiments, XVI, 513

L Lallemand electronic camera in astronomical observation, XII, 113; XVI, 341 development of new, XXXIII A, 7 focused by superconducting magnetic coil, XXXIII A, 1 specially adapted spectrograph for, XL B, 641

Lallemand image tube, XVI, 25 Large-scale integrated circuits, failure in, x x x , 79 Lasers, chemical, XXXI, 1 Lead monoxide, electron bombardment induced conductivity in, XXII A, 305 Lenses electron, 11, 48 strong-focusing, XIV, 85

M Magnet, superconducting, technology of, XXIII, 385 Magnetic airborne detector, IV, 258 Magnetic beta-ray spectrometers, XXX, 43 Magnetic coherence resonances, at zero frequency, XXVII, 19 Magnetic-core memory technology, high-speed, XXI, 249 Magnetic fields, measurement of by magnetic resonance, XXIII, 36 Magnetic focus systems aberrations in, XVI, 333 photocathode resistance on resolution of, XXII A, 591

382

CUMULATIVE SUBJECT INDEX, VOLUMES

Magnetically focused electron beams, X, 1 Magnetron continuous-wave, modulation of, IV, 188 cylindrical, electronic theory of, 111, 15 microwave, 11, 220 mode transitions, VIII, 503 plane, electronic theory of, 111, 85 steady state theory of, V, 247 Manufacture of cathode ray tubes, 11, 2 Mariner IV spacecraft television system, XXII B, 849 Masers, XV, 73 Mass spectroscopy, I, 219; VIII, 188; XLII, 1

using RF quadrupole fields, XXVII, 59 Medical scintillation cameras, image intensifying for, XXII B, 927 Memories electron beam addressed, XLIII, 43 nonvolatile semiconductor, XLI, 249 Memory technology, high-speed magneticcore, XXI, 249 Meshes, use of, to reduce electron tube errors, XXXllI A, 571 Metal-insulator-metal structure, photoemission from, XXVIII A, 409 Metal surfaces, ion scattering against, XXI, 67 Metallic conduction, at high frequencies and low temperatures, VI, 1 Meteors, radio observation of, IX, 95 Mica, electron transmission through, XXII A, 31; XXIl A, 41 Microanalysis, electron probe, XIII, 3 17 Microchannel plate(s) applications to electronic imaging, XL A, 91 with curved channels, XL A, 103 Microchannel plate output, reduction of variance, XXXIIl A, 167 Microphotometer. for photographic and electronographic image tubes, XXII A, 435 Microscope, electron, VI. 296; XII, 317 Microscopy electron emission, XVIII, 251 field emission, 111, 1 field ion, XIII, 83 scanning electron, XXI, 181 Microwaves, plasma generation and amplification of, XXI, 287

1-45

Microwave frequencies Cerenkov radiation at, XIV, 265 ferromagnetic phenomena at, 11, 258 Microwave magnetron, 11, 220 Microwave optics, X, 107 Microwave power semiconductor devices, XXXIX, 291; XLIV, 141 Microwave spectroscopy, 11, 300 Millimeter wave techniques, XV, 197 Miniaturization, tube, 11, 183 Minicomputers, basic concepts of, XLIV, 283 MIS varactors, physics and applications of, XXXIV, 281 Modulation of continuous-wave magnetrons, IV, 188 pulse code, 111, 121 Modulation transfer functions electron-optical, XXXIII B, 579 of fluorescent screens, XXII A, 395 of image tube, XXVIII B, 567 measurements on channel image intensifiers, XXXIII B, 593 of x-ray image intensifiers, XXII A, 355 Molecular beams, new applications and techniques, VIII, 2 Molecular beam masers, XXXIX, 183 Molecular hydrogen, in interstellar medium, XXVIII B, 801 Molecules, lifetimes of, XXIX, 115 Multialkali photocathodes, crystal structure of, XXVIII A, 337 Multichannel radio telemetering, IV, 3 10 Multidynode electron multipliers, single electron pulse sizes from, XXII A, 71 Multi-MeV electrons, response of KCI foils to, XXII A, 635 Multiphoton ionization of atoms, XXXVI, 57 Multiple scattering and transport of microwaves, XXXII, 31 1 Multiplication, transmitted secondary electron, XXII A, 629 Multiplication rule, in O.T.F. concept, XXXIII B, 585 Multiplier electron image, XII, 87 with forty channels, XXVIII B, 955 Multisignal effects, in nonlinear beam plasma systems, XXIX, 1 Multistable semiconductor devices and integrated circuits, XXXV, 269

CUMULATIVESUBJECT INDEX, VOLUMES

N Navigation, electronic aids to, I, 425 Negative electron affinity secondary emitters, XXXIIl A, 447 Negative ions, IX, 43 Network synthesis, methods of, nl, 261 Neural control, possibilities for, XXXVIII, 55 Nightglow. XVIII, 1 Noise, cosmic radio, 1, 347 Noise measurements in camera tubes, X L B, 585 in image intensifiers, X L 9, 577 Nonconservation, parity, in weak interactions, XI, 31 Nuclear emulsions, linearity and optimum working density, XXXIll A, 53 Nuclear magnetic resonance spectroscopy, XXXIV, 1 Nuclear physics amplitude and time measurements in, VIII, 256 demagnifying image tube for, XVI, 99 image intensifiers in, XVI, 501 Numerical field plotting, and ray tracing, in electron optics, 11, 102

0 Observation(s) astronomical, photo-electronic imaging devices for, XII, 1 of meteors. radio, IX, 95 Observational radio astronomy, VII, 299 Oceanography electronic instrumentation for, XIX, 1 electronics in, IX, 239 Operational amplifiers, XI. 225 Optical diffraction patterns, XII, 31 i Optical emulsions, linearity and optimum working density, XXXIIl A, 53 Optical images, low contrast, detection of, XII, 247 Optical imaging. with acoustic waves and photo-excited charge carriers, X L B, 993 Optical systems, for image tubes, XXVIII 9, 759 Optical television methods, XII, 363 Optical transforms, in electron micrograph analysis, XLIII, 1

1-45

383

Optics electron, 11, 102 microwave, X, 107 Opto-electronic image recording, limitations of night-sky, XXXlIl B, 999 Oscillations, plasma, XX, 59 Oscilloscope, cathode ray, X, 239 Oxide-cathode receiving tubes, electrical life of. VII, 404 Oxide coated cathodes, I, 1

P Paramagnetic resonance, XV, 327 Paramagnetic salts, diluted, relaxation in, XV, 163 Paramagnetism, VI, 463 Parity nonconservation, in weak interactions, XI, 31 Particle accelerators, 1, 269; XXIX, 223 Penning discharges, XXVII, 295 Phosphor output, thin window image intensifier, XVI, 61 Phosphor screens, high resolution, XXIl A, 55 1 for cascade image intensifiers, XXll A, 571 preparation of, XXII A, 565 Photocathode composition, study of, by microbalance methods, XVI, 329 Photocathode response, linearity and dynamic range, X L A, 457 Photocathode sensitivity, decay of, XXVIII A, 357 Photocathode surfaces, research on, XXIl A, 477 Photocathode transfer system, X L A, 419 Photocat hode(s) antimonide, synthesis of, XXXIII A, 357 Sb-Rb-Cs, XXIl A, 449 Cs-Sb and Na-K-Sb, microbalance study, XXII A, 459 Cs,Sb, quantum yield of, X L A, 397 cesium telluride, proximity focused imaging from, X L A, 449 effects of high electric fields on, XL A, 441 GaAs, measurement of diffusion length in, XXXlIl A, 389

384

CUMULATIVESUBJECT INDEX, VOLUMES

gallium indium arsenide, X L A, 377 image dissector in, XXII A, 507 improvements for pulse operation, XXVIII A, 375 interference, XXVlIl A, 433 for image tubes, XXVIII A, 433 multialkali, XXXIII A, 339 multialkali, optical and photoelectric properties, XXXIII A, 369 near infra-red spectral response, XXII A, 493

new technology for transferring, XXVIIl A, 367

pre-formed introduction into vacuum systems, XVI, 325 properties of, liberated in high vacuum, XVI, 5 reflective, image intensifier system using, XXVIII A, 443 research in Czechoslovakia, XXVIII A, 323

response quantum efficiency of, XXII A, 535 S-20, and fibre optic plates, XXVIII A, 461 stability of, and residual gases, X L A, 427 time response of, XXII A, 499 Photochron 11, for subpicosecond chronography, X L A, 51

Photoconductive image converter, electron optics of, XXVIII A, 545 Photoconductive image detectors, electronically scanned, potentialities of, XVI, 451 Photoconductive tube, x-ray sensitive, XII, 345

Photoconductivity, problems of, XIV, 37 Photodetectors, for 0 . 1 to 1 .O pm spectral region, XXXIV, 95 Photoelectric emission statistics, in soft x-ray region, XXXIII, 433 Photoelectric image devices, survey of work on, XVI, 311 Photoelectric image detectors, signal generating memory systems for, X L B, 729

Photoelectron counter, using spectracon and diode array, X L B, 765 Photoelectron energy spectrophotometry, X L B, 973 Photoelectron spectroscopy, atomic, XLI, 73; XLIV, 1

145

Photoelectronic detector, for weakly luminous stars, XXXIII B, 1017 Photoelectronic device, development and research at BARC, XL A, 409 Photoelectronic image recording device, XXXIII B, 903 Photoelectronic imaging devices image transfer and conversion, X L B, 601 potentialities of, XII, 1 image quality of, XL B, 519 Photoelectronic storage tube, experiments with, XII, 235 counting, with semiconductor diodes, XXXIII B, 863 spectrographic studies of energy, XXXIII A, 415

Photoemission at long wavelengths, XXVIII A, 393 recent advances in, XI, 1 Photoemissive devices, solid-state application, XXII 8 , 671 Photographic images, recorded with image intensifiers, XXVIII B, 589 Photography astronomical, television methods, XII, 195 of double stars, by electronic camera, XVI, 357 electronic, XII, 5 of extensive air showers, in atmosphere, XVI, 531 high-speed, electron tubes for, XVI, 249 low brightness, by image intensification, XVI, 85 Photometric applications, and electronic photography of planets, XVI, 371 Photometry bidimensional with etectronic camera, X L B, OOO of galaxies with spectracon, X L B, 847 of Type N radiogalaxies, X L B, 867 Photomultiplier, imaging, astronomical tests of, XVI, 383 Photon counters, positive-sensitive, XXVIII B, 965

Photon-counting detector, for stellar spectroscopy, XXXIII B, 851 Photon counting system, for optical astronomy, XXXIII B, 835 Photon interference, XXVIII B, 939 Photoresponse, in silicon p n junctions, XXXIII A, 409

CUMULATIVE SUBJECT INDEX, VOLUMES

Physical properties, of ferrites, VI, 70 Pick-up tube infrared, XXXIII A, 279 infra-red vidicon-type, development of, XVI, 217 with linear light transfer, XXVllI A, 281 small high-precision, electrostatic, X L A, 183 storage, XXXIII A, 263 television, I, 131 Picture resolution, figure of merit measuring, XXII A, 341 Plane magnetron, electronic theory of, Ill, 85 Plasma flame, XX, 9 for generation and amplification of microwaves, XXI, 287 harmonics and combination frequencies in, XXVII, 187 Plasma oscillations, XX, 59 Plasma turbulence, experimental, XXX, 1 Plasmas, radio-frequency confinement and acceleration of, XXIII, 153 Plumbicons, dark current in, XXXIII A, 319 Propagation in the FM broadcast band, I, 318 tropospheric, XX, 199 Protective coating layer, against alkali vapours, XXXIII A, 381 Pulse amplitude analysis, VII, 317 Pulse code modulation, Ill, 221 Pyroelectric television tubes. thermal imaging, XXXIll A, 285 Pyroelectric vidicon, high vacuum, X L A, 301 thermal diffusion limitations of resolution, X L A, 313

Q Quadrature spatial-frequency Fourier analyser, XXVIII B, 653 Quantum efficiency of detectors, XI, 87 Quantum mechanical amplifiers, XV, 73

R Radar applications, bright displays for, XVI, 265 Radar signal processing, XLV, 203

145

385

Radiation detectors for, V, 1 far-infrared, generation of, XXVI, 171 Radio astronomy observational, VII, 299 technology and observations, XXXII, 1 Radio isotope scanners, evaluation of image quality, XXXIII B, 1041 Radiology, diagnostic, image orthicon in, XVI, 581 image storage techniques applied to, XVI, 593 Radio noise, cosmic, 1, 347 Radio observation, of meteors, IX, 95 Radio telemetering, XI, 287 multichannel, IV, 301 Radio-wave propagation, IX, 187 Ray tracing, in electron optics, 11, 127 Read-out, from image tubes, XXXIII B, 873 Receiving tubes, oxide-cathode, electrical life of, VIII, 404 Reflex discharges, XXVII, 295 Relaxation, in diluted paramagnetic salts, XV, 163 RF quadrupole fields, mass spectroscopy using, XXVII, 59 River and ocean technology, electronic engineering, XXXI, 267

S Satellite communications, XXXI. 119 Scalar diffraction, in electron optics, XXX, 139 Scattering electron, in solids, IV, 2; VIII, 183 radio wave, in ionosphere, XIX. 55 Scintillation, in CsI(Na) and Csl(TI). due to low energy, XXVIII A, 451 Scintillation camera(s), input-field image intensifier for, XXXIII B, 1031 Scintillation chambers application to space research, XVI, 535 fibers versus Nal, XVI, 469 image intensifiers versus orthicons, XVI, 469 in space research, XXII B, 823 Scintillation counter, IV, 69 Scintillation detector, statistical behaviour of, XXVI, 251

386

CUMULATIVE SUBJECT INDEX, VOLUMES

SEC camera tube application to astronomy, XXVIII B, 807 and image orthicon, comparison, XXII A, 29 1 optically scanned, XXII A, 241 SEC device, optically scanned, characteristics of, XL A, 253 SEC target, XXII A, 229 photometric statistical performance, XL A, 239 point-source imaging with, XXII A, 251 SEC vidion for astronomy, XXXlll B, 801 orbital operation and calibration, XXXlIl B, 925 Secondary electron conduction, and photoelectronic image devices, XXII A, 219 Secondary electron emission, 1, 66; XI, 413 Secondary electron yield, from reduced lead glasses, XL A, 167 Secondary emission dynode, caesium activated, XXXIll A, 469 Seismology, contributions of electronics, IX, 291 Semiconducting materials, investigation of, XI, 355 Semiconductor crystals, diamond-type, defects in, X, 71 Semiconductor device(s) comparison with gaseous electronic devices, VI, 257 electron-bombarded, XLIV, 221 evaluation, XVIII. 167 microwave power, XXXIX, 291; XLIV, 141 noise in, XXIII, 303 Semiconductor materials, physics of, VII, 1 Semiconductor memories, nonvolatile, XLI, 249 Semiconductors 111-IV, acoustoelectric interactions in, XXXI, 161 interband magneto-optical studies in, Part I, XXXVII, 1; Part 11, XXXVII, 1 ion implantation in, XXXVII, 263 type 11, XXIII, 1 Sequency theory, and Walsh functions, XXXVI, 195 Shutter tubes, gas discharge application, XXVIII B, 1033

145

Signal-to-noise ratio, XII, 277, 291; XXVIII B, 577 with S . 1 photocathodes, XXVlIl B, 677 Silicon, theory of electrical properties, VII, 87 Silver-magnesium alloy dynodes, in water vapour, XXII A, 661 Single electron pulse sizes, distribution of, XXII A, 71 Single-electron response, reduction of variance in, XXXlII A, 167 Single photon detection and timing, XXXI, 39 Single-stage image converter, photography, XXVIII B, 999 Skeletal control systems, XXX, 273 “Small” multimolecular atmospheric ions bioclimatic action of, XIX, 177 measurement of, XIX, 177 properties of, XIX, 177 Solar cells, silicon, for space use, XLII, 41 Solar photometry, image orthicon applied to, XVI, 447 Solar radio astronomy, XX, 147 Solar wind, theoretical studies of, XXXVI, 1 Solid state camera(s), image sensors for, XXXVII, 181 Solid state image amplifiers, XVI, 607 Solid state image converters, XVI, 613 infra-red, XXVIII B, 1073 Solid-state luminescence, V, 137 Solid state microwave devices, XXXVIII, 147 Solid state radiographic amplifiers, XXVIII, B, 1087 Solids, characteristic energy losses of electrons in, VII, 183 electrical breakdown in, XXVI, 309 electron scattering in, IV, 2 emission of polarized electrons from, XLI, 113 intrinsic dielectric breakdown in, 11, 185 kinetic ejection of electrons from, XXI, 101 secondary electron emission from, XI, 413 Space-charge-limited currents, VI, 138 Space harmonic traveling wave electron interaction, general perturbational theory of, XVII, 1 Space research, application of scintillation chambers, XVI, 535

CUMULATIVE SUBJECT INDEX, VOLUMES Spectracon applications in astronomy, XXVIII B, 783; X L B, 803 in astronomical spectroscopy, XXVIII B, 773 comparison of image recording with, XXVIII B, 725 detective quantum efficiency of, X L B, 661 electronographic image recording tube, XXIl A, 1 1 electron transmission through mica and, XXIl A. 31 exposures on Isaac Newton telescope, XXXIII B, 747 extended field, XXXIII A, 13 further developments of, XXVIII A, 61 observations of planetary nebulae, X L B, 817 photometry of galaxies with. X L B, 847 sources of spurious background, XXXIII A, 27 use of for direct electronography, X L B, 829 Spectrograph, high-gain, for simulated re-entry, XXVllI B, 1021 Spectrometers, beta-ray, V, 97 Spectroscopy atomic photoelectron, XLI, 73; XLIV, 1 mass, I, 129; VIII, 188; XLII, 1 microwave, II, 300 time resolved interference, XXIl B, 985 Sputtering, by ion bombardment, VII, 239 Stationary afterglows, low temperature, rare gas, XXXIX, 121 Steady-state theory of magnetron, V, 247 Stellar photometry, XVI, 431 Storage tube(s) photo-electronic, XU. 235 viewing, VIII. 448 Striations, moving, XXIV, 155 Strong-focusing lenses, XIV. 85 Structure analysis, electron diffraction, XI, 355 Subminiaturization techniques, 111, 195 Superconducting magnet technology, XXIII. 385

T Target glass, effects of caesium vapour on, XXVIII A, 309

1-45

387

Target(s) conductivity, bombardment-induced for image orthicons, XVI, 247 lead sulphide, VXI, 217 magnesium oxide, XVI, 213 Teaching of electronics to scientists, XLV, 253 Telemetering multichannel, IV, 301 radio, XI, 297 Television closed-circuit, and field-ion microscopy, XXVIII B, 875 at low light-levels, XXVIII B, 837 Television camera noise, measurement of, XXVIII A, 289 Television camera tube, XII, 203; XXVIII A, 2 65 with gallium arsenide target, X L A, 185 infrared, XII, 263 a review, XIII, 387 using potassium chloride target, XXII A, 2 73 x-ray sensitive, XXVlIl A, 273 Television channels, experiences with, XVI, 6 01 Television color, V, 291 Television methods, optical, XII, 363 Television photometer, XXVIII B, 891 Television pickup tube(s), I, 131 electronic zooming with, XVI, 195 image orthicon effect in, XVI, 171 minimizing black-border effect in. XVI. I71 Television sensors, for space astronomy XXVIII B, 851 Television storage tubes, signal-to-noise ratio in, XXI, 277 Television systems for satellite-borne ultra-violet photometer, XXII B, 865 for stratoscope use, XXll B, 885 for ultra-violet astronomical photometry, X X l l B, 875 Television x-ray diffractometer X L B, 913 Tenicon, high resolution information storage tube, XVI, 287 Thermal energy ion-molecule reactions, XXIV, 1; XXXIV, 223 Thermal image detection, infra-red threshold in, XXVIII B, 685

388

CUMULATIVESUBJECT INDEX, VOLUMES

Thermionic cathodes, XXV, 211 Thermionic emission, from negative electron affinity silicon, XL A, 387 Thermionic energy conversion, XVII, 125 Thermoelectricity, XVII, 207 Thin film growth, in glow discharge, XVII,

velocity modulated, 111, 43 x-ray, XII, 327 Turbulent plasmas, multiple scattering and transport of microwaves, XXXII, 3 11

245

Thin-film semiconductor structures, electrodynamic concepts of wave interactions in, XLIV, 99; XLV, 1 Thorium oxide, and electronics, V, 169 Thyratrons, hydrogen, XIV, 207 Time measurements, in nuclear physics, VIII, 256

Timekeeping, and applications of,XLIV, 33 Track recording, image intensifier for, XVI, 113

Tracking television system, for medical applications, XL B, 951 Transmission secondary emission, XVI, 145 amplification, recent, XVI, 557 image intensifier, XII, 59; XVI, 141 statistics, XXVIll A, 513 Transmission secondary image intensifiers comparison with cascade type, XXII A, 129

performance of, XXII A, 63 Transistor applications, junction, V, 367 Transitions, at zero frequency, XXVII, 19 Traveling wave tubes, VI, 372 Trialkali Sb-K-Rb-Cs photocathode, properties of, XXVIII A, 347 Tropospheric propagation, XX, 199 Tube miniaturization, 111, 183 Tube reliability, X, 185 Tube@) axial-beam, high-power, XIV, 299 camera, integrating ultraviolet sensitive, XL A, 201 image, XII, 73 image converting, XVI, 91 image intensifier, mukipactor principle, XVI, 163 image orthicon, XII, 379 infra-red television camera, XII, 263 photoconductive pickup, XI1, 345 television camera, XIII, 387 C.P.S. Emitron, XII, 203 television pickup, I, 131; XVI, 171 television storage, XII, 277 traveling wave, VI, 372

1-45

U

Ultra-fast shutter tube, XXXIII B, 1131 Ultra high-speed photography, XXXIII B, 1137 Ultrahigh vacuum, XVII, 323 Ultra-violet camera tubes, incorporating SEC,XXII A, 261 Ultra-violet imaging, electron bombardment conductivity application, XVI, 235 Ultraviolet sensitive camera tube, XL A, 201 Ultra-violet sensitive vidicon, XVI, 227 Ultraviolet television detectors, in astronomical satellite, XXXIII B, 91 1 U.S. Navy electronic camera, XXVIII A, 1

V

Vacuum breakdown, high speed photographic study of, XXVIII B, 1041

Vacuum pumps, in clectronics manufacturing, V, 213 Vacuum systems, preformed photocathode introduction into, XVI, 325 Velocity distribution, XIII, 181 Velocity modulated tubes, 111, 43 Vidicon digital read-out of image intensifier using, XXVIII B, 981 eliminating spurious signals in, XXXIII A, 307

high-resolution reggedized half-inch, XXII A, 211 interplanetary imaging device, XXII B, 835 silicon diode array target, XL A, 287 ultra-violet sensitive, improved, XVI, 227 x-ray sensitive, with beryllium window, XL A, 209 Viewing storage tubes, VIII, 448 Visible radiation detectors, V, 1 quantum efficiency of, XI, 87 Vision, problem of, I, 121 VLF electromagnetic waves, terrestrial propagation of, XXV, 145

CUMULATIVE SUBJECT INDEX, VOLUMES 1 4 5 W

Walsh functions reasearch and development in, XXXVI, 195 and sequency theory, XXXVI, 195 Wave interactions in thin-film semiconductor structures, electrodynamic concepts of, XLIV, 99; XLV, 1 Wave techniques, millimeter, XV, 197

X-ray image intensification system, for x-ray microscope, XXIl B, 919 X-ray image intensifier, XII, 379; XXXIII B, 1049; XLIII, 205 asymmetrical astigmatism of, XL A, 507 moving structures with, XXVlII B, 647 X-ray sensitive photoconductive pickup tube, XXII, 345 X-ray sensitive vidicon, large diameter, with beryllium window, XL A, 209 X-ray tube, flying-spot. XII, 327

X X-ray camera tube with SEC target, XXXIII 9, 1061 X-ray diffraction, image tube characteristics in, XL 9, 923 X-ray diffraction patterns, XXXIII 9, 1069 X-ray image intensification, XII, 363; XVI, 567

developments on, XVI, 91

A

B C B

0 € F C H 1

9 0 1 2 3 4 1 5

389

2 Zero frequency magnetic coherence resonances at, XXVlI, 19 transitions at, XXVII, 19

This Page Intentionally Left Blank