1-i2 I ffflllllffff -EEEEEEEEEEE -EEEEmmEEEEE -EI ...AD A089 343 TEXAS LNV AT AUSTIN ELECTRONICS RESEARCH CENTER F/6 9/5 UNC MAY ANNUAL SO REPORT E J POWERS, ON ELECTRONICS S I MARCUS

AD A089 343 TEXAS LNV AT AUSTIN ELECTRONICS

RESEARCH CENTER F/6 9/5

ANNUAL REPORT ON ELECTRONICS RESEARCH AT THE UNIVERSITY OF TEXA-ETC(U)UNC MAY SO E J POWERS, S I MARCUS F49620-77-C-0101NCLASSIFIEO N

1-i2 llllllfl i

I ffflllllffff

-EEEEEEEEEEE-EEEEmmEEEEE-EI-EmmmEEEE--- EmmEEmmE

LEEILAnnual Report on Electronics Researchat The University of Texas at Austin

No. 27

For the period April 1, 1979 through March 31, 1980

JOINT SERVICES ELECTONICS PROGRAM

Research Contract AFOSR F49620-77-C-0101v'/

40,May 15, 1980

i '

ELECTRONICS RESEARCH CENTER

Bureau of Engineering ResearchU.- The University of Texas at Austin

Austin, Texas 78712

80 9 1• ... . . . .. .. ... ... ,,j _,l rr

' -' '

II

The Electronics Research Center at The Universityof Texas at Austin consists of interdisciplinary laboratoriesin which graduate faculty members, Master and PhD candidatesfrom numerous academic disciplines conduct research. Thedisciplines represented in this report include informationelectronics, solid state electronics, quantum electronics,and electromagnetics.

The research summarized in this report was supportedby the Department of Defense's JOINT SERVICES ELECTRONICSPROGRAM (U.S. Army, U.S. Navy, and the U.S. Air Force) throughthe Research Contract AFOSR F49620-77-C-0101. This program ismonitored by the Department of Defense's JSEP Technical Coordi-nating Committee consisting of representatives from the U.S.Army Research Office, Office of Naval Research and the U.S.Air Force Office of Scientific Research.

Reproduction in whole or in part is permitted forany purpose of the U.S. Government.

f

..-. . .. . . . . .- -

' Annual Report on Electronics Research

at The University of Texas at Austin,

No. 27

For the period April 1, 1979 through March 31, 1980

P " / / , ,

JOINT SERVICES ELECTRONICS PROGRAM

Research Contract AFOSR/F49629-77-C-0101

Submitted by Edward J. ,Powers

on behalf of the faculty and staff

of the Electronics Research Center

Technical Editor: Steven ./Marcus

May MI 80 2.J~-


Bureau of Engineering ResearchThe University of Lexas at Austin

Austin, Texas 78712

Approved for public release; distribution unlimited.

.Z/.. . . .

ABSTRACT

This report summarizes progress on projects carried

out at the Electronics Research Center at The University of

Texas at Austin and which were supported by the Joint Services

Electronics Program. In the area of Information Electronics

progress is reported for projects involving (1) nonlinear

filtering and estimation, (2) electronic multi-dimensional

signal processing, (3) electronic computer system design and

analysis and (4) electronic computer software systems.

In the Solid State Electronics area recent findings

in (1) basic solid state materials research and (2) research

on instabilities and transport near surfaces and interfaces

of solids are described.

In the area of Quantum Electronics progress is pre-

sented for the following projects: (1) nonlinear wave phe-

nomena, (2) atomic and molecular electronic processes and

(3) high power laser systems.

In the Electromagnetics area progress in (1)

electromagnetic signal analysis and identification, and (2)

Guided-Wave Devices for the far infrared-mm wave spectrum is

summarized.

y

iOced

on

iiitl

TABLE OF CONTENTS

Page

Abstract ........... ........................ iii

Personnel and Research Areas ...... ............... .vii

Publications, Technical Presentations, Lecturesand Reports ......... ....................... . xi

I. INFORMATION ELECTRONICS

Res. Unit IE9-1. Nonlinear Filtering and Estimation . . . 3

Res. Unit IE9-2. Electronic Multi-Dimensional SignalProcessing. ..... ............... . 13

Res. Unit IE9-3. Electronic Computer System Designand Analysis ..... ............. . 17

Res. Unit IE9-4. Electronic Computer Software Systems 23

II. SOLID STATE ELECTRONICS

Res. Unit SS9-1. Basic Solid State Materials Research . . 27

Res. Unit SS9-2. Research on Instabilities and TransportNear Surfaces and Interfaces of Solids . 31

III. QUANTUM ELECTRONICS

Res. Unit QE9-1. Nonlinear Wave Phenomena .. ........ .41

Res. Unit QE9-2. Atomic and Molecular ElectronicProcesses ....... ................ .. 47

Res. Unit QE9-3. High Power Laser Systems .. ....... .. 57

IV. ELECTROMAGNETICS

Res. Unit EM9-1. Electromagnetic Signal Analysis andIdentification ...... .......... .67

Res. Unit EM9-2. Guided-wave Devices for the FarInfrared-mm Wave Spectrum ........... .. 77

Research Grants and ContractsFederal Funds ........ .................... .89

Other Than Federal Funds ..... .............. 90

v

PERSONNEL AND RESEARCH AREAS


Phone: (512) 471-3954

Administration for the Joint Services Electronics Program

Professor Edward J. Powers, DirectorProfessor Rodger M. Walser, Assoc. Director

Electronics Research Center Staff

Roberta Brown, Administrative AssistantConnie Finger, Administrative SecretaryJan White, Accounting Clerk III

Coordinators for Research Areas

Professor R.M. Walser, Solid State ElectronicsProfessor J.K. Aggarwal, Information Electronics

(Information Sciences)Professor S.A. Szygenda, Information Electronics

(Computers)Professor E.J. Powers, Quantum Electronics and

Electromagnetics

Faculty

Solid State Electronics:

M.F. Becker, Assistant Professor, EE, 471-3628R.W. Bend, Professor, EE, 471-1225A.B. Buckman, Associate Professor, EE, 471-1095J.P. Stark, Professor, ME, 471-1504J.S. Turner, Assistant Professor, Physics, 471-7253R.M. Walser, Professor, EE, 471-5733

Information Electronics

J.K. Aggarwal, Professor, EE, 471-1369S.I. Marcus, Assistant Professor, EE, 471-1952S.A. Szygenda, Professor, EE, 471-7365E.W. Thompson, Associate Professor, EET.J. Wagner, Professor, EE, 471-3183

vii


Information Electronics, cont.

G.L. Wise, Assistant Professor, EE, 471-3356R.T. Yeh, Professor, Computer Sciences

Quantum Electronics:

M.F. Becker, Assistant Professor, EE, 471-3628M. Fink, Professor, Physics, 471-5747L. Frommhold, Professor, Physics, 471-5100J. Keto, Assistant Professor, Physics, 471-4151E.J. Powers, Professor, EE, 471-1430

Electromagnetics:

A.B. Buckman, Associate Professor, EE, 471-1095T. Itoh, Associate Professor, EE, 471-1072E.J. Powers, Professor, BE, 471-1430

Postdoctoral Research Fellows, Research Associates,

Research Scientists and Technical Staff Assistants

Kiyomichi Araki, Postdoctoral Res. Assoc., EEDavid Browning, EE, Technical Staff AssistantLarry S. Davis, Research Scientist, Computer SciencesLuc Devroye, Research Scientist, EE*Clinton Holder, Postdoctoral Res. Assoc., Physics

Younq C. Kim, Res. Engr. Assoc. IV, EEHarold V. Poor, Research Engineer, EEMichael Proffitt, Res. Scientist Assoc. V, PhysicsLorenz-Peter Schmidt, Postdoctoral Res. Assoc., EE

Research Assistants

Efren Abaya, EE William S. Burns, PhysicsEl-Saied Aly, EE Hyokang Chang, EEJoseph Ambrose, EE Shiuh Chao, EENader Bagherzadeh, EE Kong-Chen Chen, EERodney Barto, EE Mark Chonko, EEJohn M. Beall, EE

*Denotes persons who have contributed to JSEP projects, butwho have not been paid out of JSEP funds (e.g., students onfellowships).

viii

VT


Research Assistants (Cont.)

Kang Min Chung, EE Chang-Huan Liu, BESulaksh Gautam, Physics *Joseph Mauger, EEJoe Haas, EE Richard Mawhorter, PhysicsDon Halverson, EE *Jabez McClelland, PhysicsJae Y. Hong, EE Douglas Michaisky, EEKai Hsu, EE *Bruce Miller, PhysicsNian-Chyi Huang, EE Sung-Han Park, EMichael Kelley, Physics Bharat D. Rathi, EESuhas N. Ketkar, Physics *T.D. Raymond, Physics*Labib Khadra, EE Don Ross, BENahid Khazenie, EE Wayne Schwiesow, EStephen Koch, Comp. Sci. David Sheng, EEFederico Kuhlmann, EE Bang-Sup Song, EEChien-Yu Kuo, Physics Yihjye Twu, ELoren Lancaster, EE Richard Von Blucher, EEGan-Shu Lee, EE Hong-Yee Wang, EEWha-Joon Lee, EE *Randall 0. Withrow, ERafael Lemus, EE Wai-Fan Wong, EEChuang Li, EE *Hwa..Yueh Yang, EE

Advanced Degrees Awarded

Nader Bagherzadeh, EE, M.S., August 1979, "Quantized andLinear Detection"

John M. Beall, Jr., EE, M.S., May 1979, "Digital BispectralAnalysis of Nonlinear Effects in a RF-Excited Glow DischargePlasma"

Hyokang Chang, EE, Ph.D., May 1979, "Design and Implementationof Two-Dimensional Recursive Digital Filters"

Mark A. Chonko, EE, M.S., May 1979, "A Feedback ControlledSystem for the Precise Measurement of Weak Current-VoltageNon-Linearities"

Rang Min Chung, EE, Ph.D., December 1979, "Infrared NonlinearOptical Processes in Molecules"

Don Halverson, BE, Ph.D., August 1979, "Discrete Time Detec-tion of Signals in Dependent Non-Gaussian Noise"~

Jae Y. Hong, BE, M.S., May 1979,."Development of a PlasmaFluctuation Diagnostic Based on Digital Time Series Analysis"

ix

Vi


Advanced Degrees Awarded, Cont.

Kai Hsu, EE, Ph.D., December 1979, "On Discrete Time Central-ized and Decentralized Estimation and Stochastic Control"

Michael Kelley, Physics, Ph.D., December 1979, "Determinationof the Molecular Structure of Dimolybdenum-Tetracetate andInvestigation of the Temperature Dependence of the MolecularStructure of Sulfur-Hexafluoride by Gas Phase ElectronDiffraction"

Wha-Joon Lee, EE, M.S., December 1979, "Synthesis of FanFilters by Piecewise Separable Decomposition"

Rafael A. Lemus (Mazariegos), EE, M.S., May 1979, "Synchroni-zation of Concurrent Transactions in a Distributed DatabaseManagement System"

Douglas L. Michalsky, EE, M.S., December 1979, "A RelativeEfficiency Study of Some Popular Detectors"

Ernesto Pacas-Skewes, EE, Ph.D., May 1979, "A Design Method-ology for Digital Systems Using Petri Nets"

Bharat Deep Rathi, EE, M.S., December 1979, "The Design ofIntelligent Controllers--Using the Move Processor"

Man Ho Tong, EE, M.S., May 1979, "Crash Recovery in Distribu-ted Systems"

*Randall 0. Withrow, EE, M.S., May 1979, "A Digital DataAcquisition System Used in the Determination of the SpectralIndex of Turbulent Fluctuation Data"

Production Staff for This Report

Prof. S.I. Marcus Technical EditorRoberta Brown Admin. AssistantConnie Finger Admin. SecretaryJames Saltus Offset Press SupervisorMaralin Smith Offset Press OperatorC. Thomas Smith Offset Press OperatorSuzan Offord Accounting Clerk

.

PUBLICATIONS, TECHNICAL PRESENTATIONS,

LECTURES, AND REPORTS

tI

PUBLICATIONS, TECHNICAL PRESENTATIONS, LECTURES, AND REPORTS

JOURNAL ARTICLES

*Y.C. Kim, W.F. Wong, E.J. Powers, and J.R. Roth, "Extensionof the Coherence Function to Quadratic Models," Proceedingsof the IEEE (Letters), 67, 428-429, March 1979.

T. Itoh and C. Chang, "Resonant Characteristics of DielectricResonators for Millimeter-Wave Integrated Circuits," Archi furElektronik und Unbertragungstechnik, 33, No. 4, 141-144,April 1979.

*S.I. Marcus, "Optimal Nonlinear Estimation for a Class of Dis-crete Time Stochastic Systems," IEEE Transactions on AutomaticControl, AC-24, 297-302, April 1979.

*G.L. Wise and S.I. Marcus, "Stochastic Stability for a Classof Systems with Multiplicative State Noise," IEEE Transactionson Automatic Control, AC-24, 333-335, April 1979.

J.K. Aggarwal and J.W. Roach, "Computer Tracking of ObjectsMoving in Space," IEEE Transactions on Pattern Analysis andMachine Intelligence, PAMI-I, No. 2, 127-135, April 1979.

T. Itoh and F.J. Hsu, "Distributed Bragg Reflector Gunn Oscil-lators for Dielectric Millimeter-Wave Integrated Circuits,"IEEE Transactions on Microwave Theory and Techniques," MTT-27,No. 5, 514-518, May 1979.

J.K. Aggarwal, R. Jain and W.N. Martin, "Computer Analysis ofScenes with Curved Objects," Proceedings of the IEEE, 67, No.5, 805-812, May 1979.

J.K. Aggarwal and M. Knox, "Advanced Graphics Laboratory,"Discovery, Vol. III, No. 4, 12-15, June 1979.

*Y.C. Kim and E.J. Powers, "Digital Bispectral Analysis and ItsApplications to Nonlinear Wave Interactions," IEEE Transac-tions on Plasma Science, PS-7, 120-131, June 1979.

J.K. Aggarwal, L. Davis and S. Johns, "Texture Analysis UsingGeneralized Cooccurrence Matrices," IEEE Transactions on Pat-tern Analysis and Machine Intelligence, PAMI-I, No. 3, 251-259,July 1979.

*Funded entirely or in part by the Joint Services ElectronicsProgram.

xiii


*K. Hsu and S.I. Marcus, "A General Martingale Approach to Dis-

crete Time Stochastic Control and Estimation," IEEE Transac-tions on Automatic Control, AC-24, 580-583, August 1979.

T. Itoh and F.J. Hsu, "Application of Inverted Strip Dielec-tric Waveguide for Measurement of the Dielectric Constant ofLow-Loss Materials at Millimeter-Wave Frequencies," IEEE Trans-actions on Microwave Theory and Techniques, MTT-27, No. 10,841-844, October 1979.

B.S. Song and T. Itoh, "Distributed Bragg Reflection Dielec-tric Waveguide Oscillators," IEEE Transactions on MicrowaveTheory and Techniques, MTT-27, No. 12, 1019-1022, December1979.

*J.W. Keto, F.K. Soley, R.E. Gleason and G.K. Walters, "Exci-

ton Lifetime in Electron Beam Excited Condensed Phases of Ar-gon and Xenon," J. Chem. Phys., 71, 2676, 1979.

J.S. Turner, "Computer Simulation of Chemical Nucleation,"Astrophysics and Space Science, 65, 383-395, 1979.

Luc P. Devroye and T.J. Wagner, "Distribution-Free Inequali-ties for the Deleted and Holdout Error Estimates," IEEETransactions on Information Theory, IT-25, 202-207, 1979.

Luc P. Devroye and T.J. Wagner, "Distribution-Free PerformanceBounds with the Resubstitution Error Estimate," IEEE Transac-tions on Information Theory, IT-25, 208-210, 1979.

*G.L. Wise and S.I. Marcus, "Stochastic Stability for a Class

of Systems with Multiplicative State Noise," IEEE Transac-tions on Automatic Control, AC-24, 333-335, 1979.

L.P. Devroye and T.J. Wagner, "The L, Convergence of KernelDensity Estimates," Annals of Stati sics, 7, 1136-1139, 1979.

*C.S. Penrod and T.J. Wagner, "Risk Estimation for Nonparametric

Discrimination and Estimation Rules," IEEE Transactions onInformation Theory, IT-25, 753-758, 1979.

*L.P. Devroye and T.J. Wagner, "Distribution-Free Performance

Bounds for Potential Function Rules," IEEE Transactions onInformation Theory, IT-25, 601-604, 1979.

*J.P. Stark, "Field Induced Volume Diffusion in a Two Phase

System," J. Appl. Physics, 50, 2771, 1979.

xiv


*J.P. Stark, "Evolution of Two Phase Dispersion of Volume Dif-

fusion," J. Appl. Phys., 50, 5541, 1979.

*J.P. Stark, "Solute Induced Jump Correlations During DiffusiveProcesses," Journal of Applied Physics, 50, 285, 1979.

*M.H. Proffitt, L. Frommhold, "Concerning the Anisotropy of the

Polarizability Tensor of Pairs of Methane Molecules," Chem.Phys., 36, 197-200, 1979.

*L. Frommhold, M.H. Proffitt, "Raman Spectra and Polarizability

of the Neon Diatom," Chem. Phys. Letters, 66, 210, 1979.

*M.H. Proffitt, L. Frommhold, "New Measurement of the Trace of

the Helium Diatom Polarizability from the Collision-Induced,Polarized, Raman Spectrum," Phys. Rev. Letters,42, 1473-1475,1979.

*L. Frommhold, M.H. Proffitt, "Concerning the Anisotropy of

the Helium Diatom Polarizability," J. Chem. Phys., 70, 4803-4804, 1979.

*M.H. Proffitt, L. Frommhold, "Concerning the Instrumental

Profile of a Double Monochromator," Rev. Sci. Instr., 50,666-668, 1979.

*Steven M. Zwernemann and Michael F. Becker, "Enhancement of

Third Harmonic Generation in Metal-Dielectric Waveguides,"Applied Optics, 18, 728-732, 1979.

*M. Fink, P.G. Moore and D. Gregory, "Precise Determination of

Differential Electron Scattering Cross Sections I. The Appar-atus and the N2 Result," J. Chem. Phys., 71, 5221,1979.

*M. Fink, C.W. Schmiedekamp, and D. Gregory, "Precise Determi-

nation of Differential Electron Scattering Cross Sections II.CH4 , CO2 , and CF4 ," J. Chem. Phys., 71, 5238, 1979.

*M. Fink and C. Schmiedekamp, "Precise Determination of Differ-

ential Electron Scattering Cross Sections III. The ExchangeCorrections," J. Chem. Phys., 71, 5243, 1979.

*K.M. Chung, G.J. Stevens, and M.F. Becker, "Investigation of

Multiphoton Absorption in SF_ by Third Harmonic Generation,"IEEE Journal of Quantum Electronics, QE-15, 874-878, 1979.

xv

4,


*Michael F. Becker, "Correcting Inherent Aberrations in theGrating Rhomb Beam Sampler," Applied Optics, 18, 4178-4181,1979.

Y.C. Kim, E.J. Powers, J.Y. Hong, J.R. Roth and W.M. Krawczonek,"The Effects of a Weak Vertical Magnetic Field on Fluctuation-Induced Transport in a Bumpy-Torus Plasma," Nuclear Fusion,Vol. 20, 171-176, February 1980.

*Y.C. Kim, J.M. Beall, E.J. Powers, and R.W. Miksad, "The Bi-spectrum and Nonlinear Wave Coupling," Phys. Fluids, 23, 258-263, February 1980.

*D.R. Halverson and G.L. Wise, "Discrete Time Detection inf-Mixing Noise," to appear in IEEE Transactions on InformationTheory, March 1980.

*C. Liu and S.I. Marcus, "Estimator Performance for a Classof Nonlinear Estimation Problems," to appear in IEEE Transac-tions on Automatic Control, April 1980.

G.L. Wise, "The Effect of a Zero Memory Nonlinearity on theBandlimitedness of Contaminated Gaussian Inputs," to appearin IEEE Transactions on Information Theory, May 1980.

*L.P. Devroye and G.L. Wise, "Consistency of a Recursive Near-est Neighbor Regression Function Estimate," to appear in Jour-nal of Multivariate Analysis, June 1980.

L.P. Devroye and G.L. Wise, "Detection of Abnormal Behaviorvia Nonparametric Estimation of the Support," to appear inSIAM Journal of Applied Mathematics, June 1980.

J.P. Stark, "Concentration of Vacancies in a Temperature Gra-dient," Phys. Rev., B21, 556, 1980.

K. Araki and T. Itoh, "Hankel Transform Domain Analysis atOpen Circular Microstrip Radiating Structures," IEEE Trans.Antennas and Propagation, AP-28, 1980.

T. Itoh and W. Menzel, "A High Frequency Analysis Method forOpen Microstrip Structures," IEEE Trans. Antennas and Propa-gation, AP-28, 1980.

L.P. Schmidt and T. Itoh, "Spectral Domain Analysis of Domi-nant and Higher Order Modes in Fin-Lines," IEEE Trans. Micro-wave Theory and Techniques, MTT-28, 1980.

xvi


T. Itoh, "Spectral Domain Immittance Approach for DispersionCharacteristics of Generalized Printed Transmission Lines,"IEEE Trans. Microwave Theory and Techniques," MTT-28, 1980.

*J.W. Keto, T.D. Raymond, and S.T. Walsh, "A Low InductanceSpark Gap Switch for Blumlein Driven Lasers," Rev. Sci. Inst.,51, 42, 1980.

*M.H. Proffitt and L. Frommhold, "Collision-Induced Polarizedand Depolarized Spectra of Helium and the Diatom Polarizabil-ity," J. Chem. Phys., 72, No. 2, 1377, 1980.

L.P. Devroye and T.J. Wagner, "Distribution-Free ConsistencyResults in Nonparametric Discrimination and Regression Func-tions Estimation," Annals of Statistics 8, 231-239, 1980.

*A.B. Buckman and S. Chao, "Ellipsometric Characterization ofthe Glassy Layer at Metal/Semiconductor Interfaces," SurfaceScience 96, 1980.

J.L. Speyer, S.I. Marcus, and J. Krainak, "A DecentralizedTeam Decision Problem with an Exponential Cost Criterion,"to appear in IEEE Transactions on Automatic Control.

*L.W. Frommhold, "Recent Developments Concerning Diatom Polar-izabilities," Adv. Chem. Phys., Vol. 46, in press.

*L.W. Frommhold and M.H. Proffitt, "The Collision-InducedRaman Spectra of Neon and the Diatom Polarizability," Phys.Rev., in press.

L.P. Devroye and T.J. Wagner, "On the L1 Convergence of Ker-nel Regression Function Estimators with Applications in Dis-crimination," to appear in Zeitschrift fur Wahrscheinlich-keitstheorie und Verwandte Geibiete.

*D.R. Halverson and G.L. Wise, "A Detection Scheme for Depen-dent Noise Processes," to appear in Journal of the FranklinInstitute.

B.F. Dickens, C.E. Martin, G.P. King, J.S. Turner, and G.A.Thompson, Jr., "Fluorescence Polarization Measurements Corre-lating Specific Lipid Compositional Changes with Alterationsin the Discontinuous Thermotropic Response of TetrahymenaMembrane Lipids," Biochem. Biophys. Acta, in press.

xvii j- - .. . .


*C.H. Holder, Jr., D. Gregory and M. Fink, "Data CorrelationAnalysis Applied to Electron Diffraction," J. Chem. Phys.,accepted for publication.

*C.H. Holder and M. Fink, "Structure Determination of SO byElectron Diffraction," J. Chem. Phys., accepted for publication.

*B. Miller and M. Fink, "Mean Amplitudes of Vibration of SF6and Intramolecular Multiple Scattering," J. Chem. Phys.,accepted for publication.

*J.Y. Hong, Y.C. Kim, and E.J. Powers, "On Modelling the Non-linear Relationship Between Fluctuations with Nonlinear Trans-fer Functions," accepted for publication.

*J.W. Keto, C.F. Hart, and Chien-Yu Kuo, "Electron Beam ExcitedMixtures of 0 in Argon, I. Spectroscopy," accepted for publi-cation in J. hem. Phys.

*J.W. Keto, "Electron Beam Excited Mixtures of 0 in Argon, II.Electron Distributions and Excitation Rates," acepted for pub-lication in J. Chem. Phys.

*J.W. Keto, C.F. Hart, and Chien-Yu Kuo, "Electron Beam ExcitedMixtures of 02 in Argon, III. Energy Transfer Rates, acceotedfor publication in J. Chem. Phys.

J.S. Turner, "Explanation of Bursts of Oscillation, MultipleFrequencies, and Chemical 'Chaos' in the Belousov-ZhabotinskiiReaction," submitted to J. Chem. Phys.

E.V. Mielczarek, J.S. Turner, D. Leiter, and L. Davis, "Chem-ical Clocks--A Set of Demonstrations Illustrating NonlinearPhenomena," submitted to Amer. J. Phys.

*R.W. Bene, R.M. Walser, G.S. Lee and K.C. Chen," Is FirstPhase Nucleation at Metal-Semiconductor Interfaces a Electron-ically Induced Instability?", submitted to J. Vac. Soc. Tech-nol., 1980.

*Y.C. Kim, L. Khadra, and E.J. Powers, "On Wave Modulations ina Nonlinear Dispersive Medium," submitted for publication.

J.W. Keto, F.K. Soley, and R.E. Gleason, "Response to 'Commenton Exciton Lifetime in Electron Beam Excited Condensed Phasesof Argon and Xenon'," submitted to J. Chem. Phys.

xviii

. .- . .. . . . . .- - - -


*M.F. Becker, R.M. Walser, J.G. Ambrose, and D.Y. Sheng, "Pi-cosecond, 1.06 Micron Laser-Induced Amorphous Phases in Thin,Single Crystal Silicon Membranes," submitted for publication.

*K.M. Chung, G.J. Mauger, and M.F. Becker, "Third HarmonicGeneration and Multiphoton Absorption in SF6 at 193K," sub-mitted for publication.

*S.N. Ketkar, J.W. Keto, and C.H. Holder, "Correlation formeasuring Picosecond Pulses-A New Design," submitted for pub-lication.

J.P. Stark, "Thermodynamic Evolution During Diffusion," Met.Trans., in review.J.P. Stark, "An Approximate Analytical Demonstration of the

Famous Darken Experiment," Met. Trans., in review.

*J.W. Keto and 4Chien-Yu Kuo, "Collisional Quenching of A (3p)4

4 and A (3p) 3d in Electron Beam Excited Argon at High Dens-files, in preparation.

*Chien-Yu Kuo and J.W. Keto, "Dissociative Recombination ofIons in Electron Beam Excited Argon at High Densities," inpreparation.

*L.W. Frommhold, J.W. Keto, and M.H. Proffitt, "Collision-In-duced Light Scattering by the 3He Diatom," in preparation forPhys. Rev. Letters.

J.S. Turner, "Homogeneous Bifurcation Phenomena in OscillatingChemical Reactions. A Model Study of the Belousov-ZhabotinskiiReaction," J. Chem. Phys., in preparation.

J.S. Turner, "Thermodynamics, Dissipative Structures, and Self-Organization in Nonequilibrium Chemistry," J. Nonequil. Thermo-dynamics, invited review, in preparation.

J.S. Turner, "Order and Chaos in Nonequilibrium Chemical Dy-namics," Topics in Current Chemistry, invited review, in prep-aration.

*B.S. Song and T. Itoh, "An Analysis of Planar DistributedGunn Devices," IEEE Trans. Microwave Theory and Techniques,in preparation.

*L.W. Frommhold, J.W. Keto, and M.H. Proffitt, "The Diatom Po-

larizability from Colli ion Ind ced Scattering Measurementsof the Helium Isotopes, He- 3He, He- 4He," in preparation.

xix

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


TECHNICAL PRESENTATIONS AND LECTURES

1979 IEEE International Conference onAcoustics, Speech and Signal Processing

Washington, D.C.April 2-4, 1979

*J.K. Aggarwal and H. Chang, "Stabilizationof Two-Dimensional Recursive Filters."

Workshop on Computer Analysis ofTime-Varying Imagery

Philadelphia, PA.April 5-6, 1979

J.K. Aggarwal and J. Roach, "On the Ambi-guity of Three-Dimensional Analysis of aMoving Object from Its Images."

J.K. Aggarwal, R. Jain and W. Martin,"Segmentation Through the Detectionof Changes Due to Motion."

J.K. Aggarwal and W. Martin, "Occlusionin Dynamic Scene Analysis."

Texas Systems WorkshopDallas, TexasApril 21, 1979

G.L. Wise, "Nonlinear Transformations ofRandom Processes."

IEEE MTT SymposiumOrlando, FloridaApril 29-May 2, 1979

B.S. Song and T. Itoh, "A DistributedFeedback Dielectric Waveguide Oscillatorwith a Built-in Leaky-Wave Antenna."

C. Chang and T. Itoh, "Spectral DomainAnalysis of Dominant and Higher OrderModes in Fin-Lines."

*Funded entirely or in part by the Joint Services ElectronicsProgram.

xx

I


1979 IEEE International Conferenceon Plasma Science

Montreal, Quebec, CanadaJune 4-6, 1979

Y.C. Kim, E.J. Powers, J.Y. Hong, J.R.Roth, and W.M. Krawczonek, "Fluctuation-Induced Transport in a Bumpy Torus Plasma."

Y.C. Kim, E.J. Powers, R.O. Withrow, andJ.R. Roth, "Spectral Index Determinationof Dispersive Fluctuations."

J.R. Roth, W.M. Krawczonek, E.J. Powers,J.Y. Hong, and Y.C. Kim, "The Role ofFluctuation-Induced Transport in a To-roidal Plasma with Strong Radial Elec-tric Fields."

Twenty-Second Midwest Symposium onCircuits and Systems

Philadelphia, PennsylvaniaJune 17-19, 1979

*D.R. Halverson and G.L. Wise, "On Poly-nomial Nonlinearities for Detection in-Mixing Noise."

1979 International IEEE/APS Symposiumand National Radio Science Meeting

Seattle, WashingtonJune 18-22, 1979

*Y.C. Kim and E.J. Powers, "Amplitudeand Phase Modulation of a Wave due toNonlinear Wave-Wave Interactions."

*E.J. Powers and Y.C. Kim, "BispectralStudy of Nonlinear Wave-Wave Interactions."

IEEE International Symposium onInformation Theory

Grignano, ItalyJune 25-29, 1979

T.J. Wagner and L.P. Devroye, "The Asymp-totic Convergence of Kernel Rules."

xxi

- , .. . . . . . . .


IEEE International Symposium onInformation Theory

(continued)

T.J. Wagner and L.P. Devroye, "Distri-bution-Free Performance Bounds for Po-tential Function Rules."

Technical University of DarmstadtDarmstadt, West GermanyJune 26, 1979

T. Itoh, "New Millimeter-Wave CircuitApproach."

Technical University of AachenAachen, West GermanyJune 28, 1979

T. Itoh, "Dielectric Waveguide Tech-niques for Millimeter-Wave IntegratedCircuits."

University of DuisburgDuisburg, West GermanyJune 29, 1979

T. Itoh, "Quasi-Optical Techniques forMillimeter-Wave Integrated Circuits."

Euro-Physics Study Conferenceon Multiphoton Processes

Benodet, FranceJune 1979

*M.F. Becker, K.M. Chung, and G.J. Stevens,"Investigation of Multiphoton Absorptionin SF6 by Third Harmonic Generation."

Joint Automatic Conference ofInstitute of Electrical andElectronic Engineers

Denver, ColoradoJune 1979

J.K. Aggarwal and N. Huang, "On LinearShift-Variant Digital Filters."

xxii


Symposium on Machine Processingof Remotely Sensed Data

Purdue University

June 1979

J.K. Aggarwal and D. Williams, "ComputerRecognition of Citrus Infestations."

Swiss Federal Institute of TechnologyZurich, Switzerland

July 3, 1979

T. Itoh, "Dielectric Waveguide Approachfor Millimeter-Wave Integrated Circuits."

Selenia S.P.A.Rome, ItalyJuly 11, 1979

T. Itoh, "Dielectric Waveguide TypeMillimeter-Wave Components."

IEEE South Central Italy Section

University of RomeRome, ItalyJuly 22, 1979

T. Itoh, "Millimeter Wave Circuits."

Herzberg International Conferenceon van der Walls Molecules

Universite Laval

Quebec, CanadaAugust 1-3, 1979

*L.W. Frommhold and M.H. Proffitt, "Collision-

Induced Scattering of Light."

Ninth Conference on StochasticProcesses and their Applications

Evanston, IllinoisAugust 6-10, 1979

G.L. Wise, "The Non-Bandlimitedness of aClass of Random Processes."

xxiii


Sagamore Conference VIMont TremplantQuebec, CanadaAugust 19-25, 1979

M. Fink, "Charge Densities of Free Molecules."

Fourth International Conferenceon Ellipsometry

Berkeley, CaliforniaAugust 20-22, 1979

*A.B. Buckman, "Ellipsometric Character-ization of the Glassy Lager at Metal/Semiconductor Interfaces."

International Conference on thePhysics of Atomic Collisions

Kyoto, JapanAugust 29-September 6, 1979

*L.W. Frommhold and M.H. Proffitt, "Colli-sion-Induced Scattering of Light."

International Conference onArtificial Intelligence

Tokyo, JapanAugust 1979

J.K. Aggarwal, W. Martin, and R. Jain,"Extraction of Moving Object ImagesThrough Change Detection."

9th European Microwave ConferenceBrighton, EnglandSeptember 17-21, 1979

T. Itoh, "Spectral Domain ImmittanceApproach for Dispersion Characteristicsof Shielded Microstrips with Tuning Sys-tems."

xxiv

C kC


International Discussion Meeting onKinetics of Physico-Chemical Oscil-lations

Aachen, GermanySeptember 1979

Jack Turner, "Periodic and NonperiodicOscillations in the Belousov-ZhabotinskiiReaction."

Seventeenth Annual Allerton Conferenceon Communication, Control, and Computing

Monticello, IllinoisOctober 10-12, 1979

*D.R. Halverson and G.L. Wise, "On thePerformance of a Modified Sign Detectorfor M-Dependent Data."

G.L. Wise and N.C. Gallagher, "On theDetermination of Regression Functions."

Workshop on Printed CircuitAntenna Technology

Las Cruces, New MexicoOctober 17-19, 1979

T. Itoh and W. Menzel, "A High FrequencyAnalysis for Open Microstrip Structures."

National Radio Science MeetingBoulder, ColoradoNovember 5-8, 1979

T. Itoh and D. Zimmerman, "Characteristicsof Microstrip Slot Lines."

Twenty-First Annual Meeting ofthe Division of Plasma Physicsof The American Physical Society

Boston, MassachusettsNovember 12-16, 1979

*E.J. Powers, Y.C. Kim, and J.M. Beall,"Experimental Study of Nonlinear WaveCoupling of Drift Waves."

xxv


Twenty-First Annual Meeting ofthe Division of Plasma Physicsof The American Physical Society

(continued)

Y.C. Kim, L. Khadra, and E.J. Powers,"Nonlinear Wave Modulation in a WeaklyIonized Plasma."

T.P. Kochanski, R.D. Bengtson, Y.C. Kim,L. Khadra, and E.J. Powers, "Mirnov Oscil-lations and Other MHD Fluctuating Phenom-ena on PRETEXT."

Purdue UniversitySchool of Electrical EngineeringWest Lafayette, IndianaNovember 28, 1979

G.L. Wise, "Noise, Nonlinearities, andBandlimitedness."

ColloquiumTexas A&M UniversityCollege Station, TexasDecember 7, 1979

J.W. Keto, "Two Photon Spectroscopy ofXenon."

1979 IEEE Conference onDecision and Control

Ft. Lauderdale, FloridaDecember 10-12, 1979

K. Hsu and S.I. Marcus, "DecentralizedControl of Finite State Markov DecisionProcesses."

*C.H. Liu and S.I. Marcus, "EstimatorPerformance for a Class of NonlinearEstimation Problems."

1979 International Colloquium onCircuits and Systems

Taipei, Taiwan1979

J.K. Aggarwal, "Dynamic Scene Analysis."

xxvi

L............


International Conference onNonlinear Dynamics

New York Academy of SciencesNew YorkDecember 1979

Jack Turner, "Two Paths to Chaos in theBelousov-Zhabotinskii Reaction."

U.S. Air Force Aero Propulsion LaboratoryWright Patterson Air Force BaseDayton, OhioJanuary 3, 1980

J.W. Keto, "Dissociative Recombinationof Electrons and Ions in Electron BeamExcited Argon at High Density."

University of MuensterMuenster, GermanyJanuary 7, 1980

M.F. Fink, "Images of Atoms and Molecules."

University of MarburgGermanyJanuary 8, 1980

M.F. Fink, "Images of Atoms and Molecules"

Free University of BerlinGermanyJanuary 10, 1980

M.F. Fink,"Images of Atoms and Molecules"

University of KaiserslauternGermanyJanuary 14, 1980


University of TuebingenGermanyJanuary 17, 1980


xxvii


Seventh Annual Conference onPhysics of Compound Semi-conductor Interfaces

Estes Park, ColoradoJanuary 29-31, 1980

R.W. Ben4, R.M. Walser, G.S. Lee and K.C.Chen, "Is First Phase Nucleation at MetalSemiconductor Interfaces an ElectronicallyInduced Instability?"

Department of Mathematics SeminarUniversity of GroningenGroningen, The NetherlandsFebruary 5, 1980

S.I. Marcus, "Modeling and Analysis ofStochastic Differential Equations."

Workshop on Compound Semiconductorsfor Microwave Materials and Devices

San Francisco, CaliforniaFebruary 11-12, 1980

*B.S. Song and T. Itoh, "Comparative Study

of Microstrip and Coplanar Type PlanarGunn Devices."

Laboratoire des Signaux et SystemsEcole Superieure d' ElectriciteGif Sur-Yvette, FranceFebruary 20, 1980

*S.I. Marcus, "Modeling and Approximation

of Stochastic Differential Equations Drivenby Semimartingales."

Lehigh University ColloquiumBethlehem, PA.March 3, 1980

T. Itoh, "Millimeter Wave IntegratedCircuits."

xxviii


8th Austin Symposium onMolecular Structure

Austin, TexasMarch 3-5, 1980

M. Fink and M. Kelley, "Current ElectronDiffraction Studies on SF 6."

The American Physical SocietyThird Topical Conference on

High Temperature Plasma DiagnosticsLos Angeles, CaliforniaMarch 17-19, 1980

Y.C. Kim, L. Khadra, E.J. Powers, T.Kochanski, and R.D. Bengtson, "Utili-zation of Complex Energy Density Functionsin the Analysis of Nonstationary Fluctua-tions in Tokamak Plasmas."

1980 March Meeting of the APSNew YorkMarch 24-28, 1980

M.H. Kelley and M. Fink, "The Importanceof Vibrational Anharmonicity in SF6."

Fourteenth Annual Conferenceon Information Sciences and Systems

Princeton, New JerseyMarch 26-28, 1980

*D.R. Halverson and G.L. Wise, "Zero

Memory Detection of Random Signalsin -Mixing Noise."

N.C. Gallagher and G.L. Wise, "Passbandand Stopband Properties of Median Filters."

*G.L. Wise and H.V. Poor, "Stochastic

Convergence Under Nonlinear Transformationson Metric Spaces."

xxix


Workshop on Instabilities, Bifurcations,and Fluctuations in Chemical Systems

Austin, TexasMarch 1980

Jack Turner, "Bifurcations Leading toChemical Chaos: Theory and Experiment."

Twelfth Southeastern Symposiumon System Theory

Virginia Beach, VirginiaMay 19-20, 1980

G.L. Wise, "Recent Results Concerningthe Effects of Nonlinearities on RandomInputs."

xxx


CONFERENCE PROCEEDINGS

*J.K. Aggarwal and H. Chang, "Stabilization of Two-Dimensional

Recursive Filters," Proceedings of the 1979 IEEE InternationalConference on Acoustics, Speech and Signal Processing, Wash-ington, D.C., pp. 9-12, April 2-4, 1979.

J.K. Aggarwal and W. Martin, "Occlusion in Dynamic SceneAnalysis," Proceedings of the Workshop on Computer Analysis ofTime-Varying Imagery, Philadelphia, PA., April 5-6, 1979.

J.K. Aggarwal, R. Jain and W. Martin, "Segmentation Throughthe Detection of Changes Due to Motion," Proceedings of theWorkshop on Computer Analysis of Time-Varying Imagery, Phil-adelphia, PA., April 5-6, 1979.

J.K. Aggarwal and J. Roach, "On the Ambiguity of Three-Dimen-sional Analysis of a Moving Object from its Images," Proceedingsof the Workshop on Computer Analysis of Time-Varying Imagery,Philadelphia, PA., April 5-6, 1979.

B.S. Song and T. Itoh, "A Distributed Feedback DielectricWaveguide Oscillator with a Built-in Leaky-Wave Antenna,"International Microwave Symposium Digest, Orlando, Florida,pp. 217-219, April-May 1979.

C. Chang and T. Itoh, "Spectral Domain Analysis of Dominantand Higher-Order Modes in Fin-Lines," International MicrowaveSymposium Digest, Orlando, Florida, pp. 344-346, April-May 1979.

*D.R. Halverson and G.L. Wise, "On Polynomial Nonlinearitiesfor Detection in -Mixing Noise," Proceedings of the Twenty-Second Midwest Symposium on Circuits and Systems, Philadelphia,Pennsylvania, pp. 504-508, June 17-19, 1979.

*J.K. Aggarwal and N. Huang, "On Linear Shift-Variant Digital

Filters," Proceedings of the Joint Automatic Control Confer-ence of Institute of Electrical and Electronics Engineers,Denver, Colorado, June 1979.

J.K. Aggarwal and D. Williams, "Computer Recognition of CitrusInfestations," Proceedings of the Symposium on Machine Pro-cessing of Remotely Sensed Data, Purdue University, West La-fayette, Indiana, June 1979.

*Funded entirely or in part by the Joint Services Electronics

Program.

xxxi I


*M.F. Becker, K.M. Chung, and J.G. Stevens, "Investigation ofMultiphoton Absorption in SF by Third Harmonic Generation,"Proceedings of the Euro-physics Study Conference on Multipho-ton Processes, Benodet, France, p. 93, June 1979.

J.K. Aggarwal, W. Martin and R. Jain, "Extraction of MovingObject Images Through Change Detection," Proceedings of theInternational Joint Conference on Artificial Intelligence,Tokyo, Japan, August 1979.

T. Itoh, "Spectral Domain Immittance Approach for DispersionCharacteristics of Shielded Microstrips with Tuning Septums,"Proceedings of the 9th European Microwave Conference, Brighton,England, pp. 435-439, September 1979.

*D.R. Halverson and G.L. Wise, "On the Performance of a Modi-fied Sign Detector for M-Dependent Data," Proceedings of theSeventeenth Annual Allerton Conference on Communication, Con-trol, and Computing, Monticello, Illinois, pp. 143-151, Octo-ber 10-12, 1979.

G.L. Wise and N.C. Gallagher, "On the Determination of Regres-sion Functions," Proceedings of the Seventeenth Annual Aller-ton Conference on Communication, Control, and Computing, Mon-ticello, Illinois, pp. 616-623, October 10-12, 1979.

T. Itoh and W. Menzel, "A High Frequency Analysis Method forOpen Microstrip Structures," Proceedings of Workshop on PrintedCircuit Antenna Technology, Las Cruces, New Mexico, October1979.

*J.W. Ke o and Chien-Yu Kuo, "Dissociative Recombination Rates

for Ar-," Proceedings of the 32nd Annual Gaseous ElectronicsConferance, October 1979.

*J.W. Keto, Chien-Yu Kuo, and C.F. Hart, "Production of O('S 0

in Electron Beam Excited 02 Doped Argon," invited paper,Proceedings of the 32nd Annual Gaseous Electronics Conference,October 1979.

K. Hsu and S.I. Marcus, "Decentralized Control of FiniteState Markov Decision Processes," Proceedings of the 1979 IEEEConference on Decision and Control, New York, IEEE Press,December 1979.

*C.H. Liu and S.I. Marcus, "Estimator Performance for a Classof Nonlinear Estimation Problems," Proceedings of the 1979

xxxii


IEEE Conference on Decision and Control, New York, IEEE Press,pp. 518-522, December 1979.

J.L. Speyer, S.I. Marcus, and J. Krainak, "A DecentralizedTeam Decision Problem with an Exponential Cost Criterion,"Proceedings of the 1979 IEEE Conference on Decision and Con-trol, New York , IEEE Press, pp. 952-957, December 1979.

J.K. Aggarwal, "Dynamic Scene Analysis," Proceedings of the1979 International Colloquium on Circuits and Systems, Taipei,Taiwan, 1979.

*L. Frommhold and M.H. Proffitt, "The Scattering of Light byCollisionally Interacting Pairs of Atoms," Proceedings of theSociety for Atomic Collision Research, Japan, pp. 922-923,1979.

*J.K. Aggarwal and H. Chang, "Implementation of Two-DimensionalSemicausal Recursive Filters," Proceedings of the InternationalSymposium on Circuits and Systems, Tokyo, Japan, 1979.

*G.L. Wise and H.V. Poor, "Stochastic Convergence Under Non-linear Transformations on Metric Spaces," Proceedings of theFourteenth Annual Conference on Information Sciences and Sys-tems, Princeton, New Jersey, March 26-28, 1980.

N.C. Gallagher and G.L. Wise, "Passband and Stopband Proper-ties of Median Filters," Proceedings of the Fourteenth AnnualConference on Information Sciences and Systems, Princeton,New Jersey, March 26-28, 1980.

*D.R. Halverson and G.L. Wise, "Zero Memory Detection of RandomSignals in -Mixing Noise," Proceedings of the Fourteenth An-nual Conference on Information Sciences and Systems, Princeton,New Jersey, March 26-28, 1980.

G.L. Wise, "Recent Results Concerning the Effects of Nonlin-earities on Random Inputs," Proceedings of the Twelfth South-eastern Symposium on System Theory, Virginia Beach, Virginia,May 19-20, 1980.

T. Itoh and B. Adelseck, "Trapped Image Guide for Millimeter-Wave Circuits," International Microwave Symposium Digest,Washington, D.C., May 1980.

L.P. Schmidt, T. Itoh and H. Hofmann, "Characteristics of Uni-lateral Fin-Line Structures with Arbitrarily Located Slots,"

xxxiii


International Microwave Symposium Digest, Washington, D.C.,May 1980.

K. Araki, B.S. Song and T. Itoh, "Nonreciprocal Effects in anOpen Dielectric Waveguide with Grating Structures," Interna-tional Microwave Symposium Digest, Washington, D.C., May 1980.

T. Itoh and B. Adelseck, "Trapped Image Guide Leaky-Wave An-tennas for Millimeter-Wave Applications," IEEE/AP-S Interna-tional Symposium Digest, Quebec, Canada, June 1980.

T. Itoh and L.P. Schmidt, "Characteristics of a GeneralizedFin-Line for Millimeter-Wave Integrated Circuits," Interna-tional URSI Symposium Digest, Munich, W. Germany, August 1980.

*T.J. Wagner and L.P. Devroye, "The Strong Uniform Consisten-cy of Kernel Density Estimates," Proceedings of the FifthInternational Symposium on Multivariate Analysis (P.R. Krish-naiah, Ed.), North Holland Publishing Co., pp. 59-77, 1980.

J.S. Turner, "Thermodynamics, Dissipative Structures, and Self-Organization in Biology: Some Implications for BiomedicalResearch," Proceedings of the "First John Lawrence Interdis-ciplinary Symposium on the Physical and Biomedical Sciences,"G.P. Scott, Ed., in press.

J.S. Turner, "Self-Organization in Nonequilibrium Chemistry,"Proceedings of a Workshop on Dissipative Structure in the So-cial and Physical Sciences," P.M. Allen and W.C. Schieve, Eds.,University of Texas Press, Austin, in press.

J.S. Turner, "Bifurcations Leading to Chemical Chaos: Theoryand Experiment," Proceedings of a Workshop on "Instabilities,Bifurcations, and Fluctuations in Chemical Systems," invitedpaper, in preparation.

BOOKS

*Digital Signal Processing, edited by J.K. Aggarwal, Western

Periodicals Co., North Hollywood, California, 91605, 1979.(Includes an article by Hyokang Chang and J.K. Aggarwal,"Fundamentals of Two-Dimensional Recursive Filtering."

J.K. Aggarwal and N.I. Badler edited the Abstracts of theWorkshop on Computer Analysis of Time-Varying Imagery, April1979, Philadelphia, Pennsylvania.

xxxiv

Al

INFORMATION ELECTRONICS

Research Unit IE9-1. NONLINEAR FILTERING AND ESTIMATION

Principal Investigators: Professor S.I. Marcus (471-3265)Professor T.J. Wagner (471-3183)Professor G.L. Wise (471-3356)

Research Engineer/Scientists: L. Devroye, H.V. Poor

Graduate Students: E. Abaya, D. Halverson, K. Hsu,N. Khazenie, F. Kuhlmann, C.H. Liu,D. Michalsky, H.Y. Wang

A. OBJECTIVES AND PROGRESS: This research unit is concerned withanalytical investigations of the statistical aspects of nonlin-ear systems. Specifically, the design of nonlinear systems forsignal detection, the analysis and design of nonlinear estima-tors, the nonparametric estimation of regression functions, thestability of stochastic systems, and the performance of adap-tive delta modulators have been investigated.

Nonlinear Systems for Signal Detection: The objective ofthis research effort was the analysis and design of nonlinearsystems for the discrete time detection of signals in corrupt-ing noise. We were concerned with the situation where thesampled data were not mutually independent, thus taking theproblem away from the usual statistical framework and puttingit into a more realistic engineering framework. In practicalsignal processing situations, data is often sampled at a ratehigh enough to rule out the assumption of mutually independentsamples.

Some recent work addressed this situation by study-ing the design of memoryless detectors for a constant signalin m-dependent noise. By memoryless detection, we refer toa zero memory nonlinearity followed by an accumulator whoseoutput is fed into a threshold comparator which announcesthe detection decision. In [1] we extended this technique tothe case where the corrupting noise was assumed to be @-mixing.Modeling the noise as a '-mixing process allows a great dealof flexibility in the dependency structure of the noise and,loosely speaking, only requires a "decrease" in the dependencyas samples are more widely separated in time. In this work,a detailed analysis of the detector was made and various con-sequences of approximating the optimum zero memory nonlinear-ity were investigated. In [2) the work done in [1] was extendedto the case where the signal was random and not constant. Inthis extension, the signal was required to be a %-mixingrandom process, an extremely weak restriction.

3.

¥I


The situation considered in [3] ahd [4] was alsoconcerned with memoryless detection of a constant signal inadditive 0-mixinq noise. However, in this work we constrainedthe zero memory nonlinearity in the memoryless detector tobe a polynomial. With this constraint, the design of thedetector was considerably simplified. For example, the methodgiven in [1] involved the solution of an integral equation ofnonstandard form. However, when the nonlinearity was con-strained to be a polynomial, the technique simplified andinvolved only a system of linear algebraic equations, whichcan be solved in a straightforward manner with the use of adigital computer. Also, it was shown that in many situations,the performance of a polynomial memoryless detector approachedthe performance of the optimal memoryless detector as thedegree of the polynomial became sufficiently large.

In [5] we considered the design of nonlinear systemsfor the nonparametric detection of constant signals in addi-tive m-dependent noise. Nonparametric detectors are appeal-ing because they offer a fixed structure which will maintaina constant false alarm probability over a wide class of noisedistributions. Both the small sample and large sample situ-ations were considered, and discrepancies between these twosituations were discussed.

The investigation of nonlinear systems for signaldetection is continuing. The research in this area wascomplemented by the Grant AFOSR-76-3062 from the Air ForceOffice of Scientific Research.

Nonlinear Estimation: This research is concerned with anumber of basic questions in the area of state estimation fornonlinear dynamic stochastic systems. The estimation probleminvolves the extraction of information about the state of thesystem from nonlinear noisy measurements. The eventual object-ive is the design and analysis of high-performance optimaland suboptimal estimators which operate recursively in realtime.

Aside from the linear (Kalman) filter, there arefew known cases in which the conditional mean (the minimumvariance estimate) of the system state given the past observa-tions can be computed recursively in real time with a filterof fixed finite dimension. However, in [6] we have provedthat for certain classes of discrete-time systems, describedeither by a finite Volterra series or by certain types ofstate-affine realizations, the minimum variance estimator isrecursive and of fixed finite dimension. Furthermore, theseoptimal estimators possess the interesting property of beingdriven by polynomials in the innovations. This phenomenon

4

-

. .


does not occur in continuous time, and it may have signifi-cance in the design of suboptimal estimators.

In [7] we have investigated nonlinear discrete timeestimation and stochastic control problems in which the obser-vations take a finite or countable number of values. Theapproach involves the application of recent developments inmartingale theory, resulting in a generalization of the workof Segall, Bremaud, and Van Schuppen. General methods forconstructing system models and deriving optimal estimatorsare presented, and the previously obtained estimation equa-tions of the researchers mentioned above are exhibited asspecial cases. The key concept is the use of modern martin-gale theory and the judicious choice of certain sigma-algebrasand martingales. Real time finite-dimensional recursive

estimators are derived and constructed for systems in whichthe state is a finite state Markov process. In addition,these methods are also applied to stochastic control problemsinvolving finite state Markov processes.

In [8] and [9], by employing the finite dimensionalestimators which we derived previously [10], we have for thefirst time been able to analyze the performance of suboptimalestimators and the tightness of estimation lower bounds fora nonlinear system by comparison with the optimal estimator.A system for which we can construct the optimal estimator wasstudied; the optimal estimator, extended Kalman filter (EKF),constant gain extended Kalman filter (CGEKF), best linearestimator (BLE), and Bobrovsky-Zakai lower bound were com-pared both analytically and via Monte Carlo simulations. Theresults indicated that the performance of the EKF is virtuallyas good as that of the optimal estimator, and the Bobrovsky-Zakai lower bound is tight for very high signal-to-noise ratiobut is less effective for large values of state and observa-tion noises. As far as suboptimal filter design is concerned,the CGEKF is probably preferable, in most of the cases studied,to the optimal estimator and the EKF, due to its simple com-putational requirements.

In a study related to the robustness of nonlinearfilters, we have also considered some questions of convergenceof the inputs and outputs of nonlinear stochastic systems.Consider a nonlinear system with a given input and the cor-responding output. If a sequence of inputs converged to thatparticular input, it would often be of interest to know whenthe corresponding sequence of outputs converged to the parti-cular output. In one sense, we might estimate the desiredoutput of the system by the output due to an estimate of thedesired input. In [11] we considered this problem in astochastic framework. We considered the output convergenceproperties of Borel measurable (but not necessarily continuous)

5


mappings defined on separable metric spaces. Generalresults were presented pertaining to the Lp convergence ofthe outputs of systems when subjected to sequences of inputrandom quantities converging in various modes.

The research in this is continuing and has beencomplemented by Grant AFOSR-79-0025 from the Air Force Officeof Scientific Research and Grant ENG76-11106 from the NationalScience Foundation.

Nonparametric Estimation of Regression Functions: It isreasonable to expect that with a large amount of empiricaldata we could achieve a good estimate of a regression func-tion. However, with a large amount of data, we may be facedwith computational burdens in processing them. Therefore,a recursive method of estimation may seem attractive. Inthis research effort we investigated distribution-free con-sistency results for the recursive nonparametric regressionfunction estimation problem.

Assume that (X,Y), (XI,YI), --- , (XNYN) are in-dependent identically distributed ]Rd xR - valued randomvectors with E{JYJ} < . Consider estimating the regressionfunction

m(x) E YIX = x

from the data, (X1 ,Y1 ), -'-, (XNYN). We proposed the fol-

lowing estimate. Break the data up into disjoint blocks oflengths bl,b2 , -. , bn, and among all X i in the j-th block,

find the one that is closest to x using the kq norm 11-11

on ]Rd (in case of a tie, pick the Xi with the lowest index i).Call the corresponding ]R x R - valued random vector(XI, Yt). The dependency on x is suppressed for the sake of

brevity.If '{wni,***, wnn}, n 1 i is a triangular array of

positive weights, then we estimated m(x) by

n .Sw n]y.j

m (x) = j=l (1)n nSWnj

j=l

when N = b + + bn observations (XiYi) were available.

6


Notice that when Wni = v i for all n, i, then the computation

in (1) can be performed recursively. That is, there is noneed to store all the observations (Xi,Yi), and if we are

not satisfied with mn we can collect more observations andupdate our estimate. Also, (1) retains the flavor of thenearest neighbor estimates, but the processing burden aris-ing from the ranking procedure is less. The conditions whichwe put upon bn and wni were weak:

b - +n

sup wni W 0l~i~nj=l

In [12] we investigated which consistency properties of mnhold without additional restrictions on the joint distributionof (X,Y). Also in [12] the discrimination problem was con-sidered and the first distribution-free strong Bayes riskconsistency result in the literature was given.

In addition, some previous research on nonparametricestimation and discrimination performed under this unit waspublished during the past year [15,161.

The research in this area was complemented by theGrant AFOSR 76-3062 from the Air Force Office of ScientificResearch.

Stochastic Stability: The objective of the work in stochas-tic stability was a tractable method to determine the sto-chastic stability of linear systems with multiplicative noise,that is, systems described by equations of the form

X(t) = A + Bifi(t) X(t)

where the fi(t) are random processes. Notice that the noiseis multiplicative instead of additive, accounting for thenonlinear structure of the system. Such systems are popularmodels for many practical problems, such as circuits withrandom parameters and the effect of switching jitter on sam-pled data system performance. In the case that the noise pro-cesses are jump processes, such models find application inmodeling faulty systems, or systems subject to abrupt randomchanges. A major concern is the stability of such systems.The objective of this work was the establishment of usefulmethods for determining when the statistical moments of thestate components tend to zero.

I 1


A new method for investigating the stochastic sta-bility of the above form of system was introduced. It isbased upon the use of the characteristic functional of a ran-dom process. That is, if X(t) is a random process, its char-acteristic functional is given by

(1) = E lexp [if X (t) p (dt)]

where W is a suitably restricted generalized measure. It wasshown that for certain types of bilinear systems, the charac-teristic functional afforded a convenient method for investi-gation of stochastic stability. In particular, this methodwas particularly appropriate for situations where the noiseprocesses were filtered Poisson processes. This form ofnoise provides a good model for a wide variety of phenomenasuch as shot noise, ELF and VLF atmospheric noises, and otherrandom sporadic events such as the noise generated by a faulty

component. The results of this research are given in [13],where several examples are presented.

The research in this area was complemented by theGrant ENG 76-11106 from the National Science Foundation.

Adaptive Delta Modulation: This investigation is concernedwith how the adaptive delta modulator (ADM) of Figure 1performs with a stationary Gaussian input X(t). The measureof performance is the limiting average squared error betweenthe input X(t) and its digital approximation Y(t), namely,

T

lim 1f (Y(t)-X(t))2 dt (2)

Assuming that X(t) is Gauss-Markov and making an assumptionabout the time averages of [X(t),Y(t)}, it was shown that(2) is the same as

1 N )2lim (X .n)-Y( (3)

N- n=l

The impact of this result is that in simulating the adaptivedelta modulator, the points X(0),X(T),X(2T),-" are the onlyones that need to be generated in order to estimate (2). Asimulation study revealed that the ADM achieves its best per-formance for PQ=l and 1.002 < P < 1.1. While this performanceis essentially equal to that of the ordinary delta modulator

8

I


(P=Q=l) with an optimally chosen A (which depends on T o andX(t)), the ADM achieves its performance regardless of theinitial A used, the sampling time To, and the distributionof the underlying Gauss-Markov-process X(t) ([14]).

X( I) 1 To Sampler = ±1

Y ~~


5. D.R. Halverson and G.L. Wise, "On the Performance of aModified Sign Detector for M-Dependent Data," Proceedingsof the Seventeenth Annual Allerton Conference on Communi-cation, Control, and Computing,,Monticello, Illinois,143-151 (October 10-12, 1979).

6. S.I. Marcus, "Optimal Nonlinear Estimation for a Class ofDiscrete Time Stochastic Systems," IEEE Transactions onAutomatic Control, AC-24, 297-302 (April 1979).

7. K. Hsu and S.I. Marcus, "A General Martingale Approach toDiscrete Time Stochastic Control and Estimation," IEEETransactions on Automatic Control, AC-24, 580-583 (August1979).

8. C.H. Liu and S.I. Marcus, "Estimator Performance for a

Class of Nonlinear Estimation Problems," to appear inIEEE Transactions on Automatic Control (April 1980).

9. C.H. Liu and S.I. Marcus, "Estimator Performance for aClass of Nonlinear Estimation Problems," Proceedings ofthe 1979 IEEE Conference on Decision and Control, NewYork, IEEE Press, 518-522 (December 1979).

10. S.I. Marcus and A.S. Willsky, "Algebraic Structure andFinite Dimensional Nonlinear Estimation," SIAM Journalon Mathematical Analysis, Vol. 9, 312-327 (April 1978).

11. G.L. Wise and H.V. Poor, "Stochastic Convergence Under

Nonlinear Transformations on Metric Spaces," Proceedingsof the 1980 Conference on Information Sciences and Sys-tems, Princeton University (March 26-28, 1980).

12. L. Devroye and G.L. Wise, "Consistency of a RecursiveNearest Neighbor Regression Function Estimate," to appearin Journal of Multivariate Analysis (June 1980).

13. G.L. Wise and S.I. Marcus, "Stochastic Stability for aClass of Systems with Multiplicative State Noise," IEEETransactions on Automatic Control, AC-24, 333-335 (April1979).

14. Hong-Yee Wang, "Delta Modulation with Stationary GaussianInputs," M.S. Thesis, The University of Texas at Austin,Austin, Texas (May 1980).

10


15. L.P. Devroye and T.J. Wagner, "Distribution-FreePerformance Bounds for Potential Function Rules," IEEETransactions on Information Theory, IT-25, 601-604 (1979).

16. C.S. Penrod and T.J. Wagner, "Risk Estimation for Non-parametric Discrimination and Estimation Rules," IEEETransactions on Information Theory, IT-25, 753-758 (1979).

11

I


Research Unit IE9-2 ELECTRONIC MULTI-DIMENSIONAL SIGNALPROCESSING

Principal Investigator: Professor J.K. Aggarwal (471-1369)

Graduate Students: N.C. Huang and S. Park

A. PROGRESS: The broad objective of this research unit isto develop new and efficient techniques for the processing oftwo-dimensional (2D) images and shift-variant signals. Sig-nificant progress has been made in the synthesis andimplementation of 2D recursive digital filters. In addition,we have made some progress in the analysis and synthesis oflinear shift-variant digital filters.

Linear shift-variant (LSV) digital filters are ageneralization of linear shift-invariant (LSI) digital filtersand are important in processing seismic trace and speech sig-nals whose frequency content changes significantly with time.By using LSV digital filters, we can change filter character-istics at each time instant as desired. In [1) we have inves-tigated some fundamental properties of recursive LSV digitalfilters in both the time and frequency domains. We presentthe notion of a generalized transfer function and discuss thefrequency characteristics of shift-variant digital filtersin terms of the generalized transfer function. In the timedomain, the necessary and sufficient conditions for a filterto be realizable as an LSV difference equation have been ex-amined in terms of its impulse response. Furthermore, sev-eral techniques to obtain difference equations from realizableimpulse responses are also proposed. From the properties ofthe impulse response, we derive the relationship between theclass of systems characterized by LSV difference equationsand the class of systems characterized by rational general-ized transfer junctions. In doing so, we have established somebasic notions regarding LSV digital filters.

The synthesis of 2D recursive digital filters isprimarily concerned with the problem of approximating thespecified frequency characteristics in magnitude and/or phaseby 2D stable rational polynomials. In the past, the diffi-culties with spectral factorization and stability in two-dimensions has led to synthesis techniques in which thesedifficulties could be alleviated by synthesizing a filter inseparable form. In 1978, Hirano and Aggarwal developed asynthesis technique for approximating nonseparable frequencycharacteristics by sums and products of separable transferfunctions. We have generalized this technique to synthesize2D separable fan filters. In general, an ideal fan filterhas triangular pass or stop regions, and its frequency

13

I


characteristic is nonseparable. We have developed an approachto decompose the triangular pass or stop regions into severalrectangular sections. With this approach, the nonseparablefrequency characteristics can be approximated by the combi-nation of several separable transfer functions. The resultsof this effort are documented in [2]. One major advantageof the present technique is that the difficulties associatedwith 2D nonseparable fan filters can be avoided by using suchan approximation.

Another approach to the synthesis of 2D recursivedigital filters, which in a certain sense mimics the synthe-sis philosophy of one-dimensional (1D) recursive filters,involves the spectral factorization of 2D magnitude-squaredfunctions. However, synthesis techniques for 1D filters donot generalize easily to 2D filters. In one-dimension thestabilization of unstable filters can be carried out by usingleast square inverse polynomials. In two dimensions, however,there are two forms of least square inverse polynomials:planar least square inverse (PLSI) polynomials of causal form,and PLSI polynomials of semicausal form. It is now knownthat PLSI polynomials of causal form do not lead to the sta-bilization of 2D unstable filters. The reason for this isthat 2D causal recursive filters are inadequate for the syn-thesis of filters with arbitrary magnitude functions. There-fore, we have developed a new procedure [3) for synthesizing2D semicausal recursive filters by taking advantage of PLSIpolynomials of semicausal form which can be spectrally fac-tored in an approximate way. The new procedure requires onlythe single operation of finding the PLSI polynomials of semi-causal form; this procedure is much simpler and more accuratethan the previous methods. In addition, we present a newstabilization procedure [41 which offers an effective meansfor stabilizing 2D unstable filters when incorporated withany stability checking algorithm in the literature. Further-more, we propose a new measure of the amplitude distortiondue to stabilization to help judge the acceptability of theamplitude response of the resulting filter.

In general, semicausal recursive digital filtersrequire much more storage than causal recursive filters. Theproblem of implementing semicausal recursive filters has beeninvestigated and reported in [5]. We first generalize thestate-space implementation scheme for causal transfer func-tions to semicausal transfer functions. We then present amethod for simplifying the exact implementation, since itrequires an output frame much larger than the input frame.Finally, we give an example which compares the output signalfrom the simplified implementation with the output signalfrom the exact implementation.

14


The paper [6] published in the book edited by J.K.

Aggarwal reviews recent developments of 2D recursive digital

filters in terms of semicausality and spectral factorization,

thus establishing a fundamental theory of 2D recursive fil-

ters. In addition to the work documented above, the paper

[71 has been accepted for publication.This has been a brief summary of the principal

results we have achieved in the past year. Several aspects

of the problems are still under investigation. Specifically,

the research concerning the analysis and synthesis of LSV

digital filters will be continued. The applications of LSV

digital filters for the processing of LSV digital signals

will be considered. Furthermore, the results on the one-

dimensional LSV digital filters will be generalized to two-

dimensional LSV digital filters which have potential applica-

tions to space-variant images.

B. REFERENCES

1. N.C. Huang and J.K. Aggarwal, "On Linear Shift-VariantDigital Filters," Proceedings of 1979 Joint Automatic

Control Conference, Denver, Colorado (June 17-20, 1979).

2. W. Lee, "Synthesis of Fan Filters by Piecewise Separable

Decomposition," M.S. Thesis, The University of Texas atAustin (December 1979).

3. H. Chang, "Design and Implementation of Two-Dimensional

Recursive Digital Filters," Ph.D. Dissertation, The

University of Texas at Austin (May 1979).

4. H. Chang and J.K. Aggarwal, "Stabilization of Two-

Dimensional Recursive Filters," Proceedings of 1979 IEEE

International Conference on Acoustics, Speech, and Signal

Processing, Washington, D.C. (April 1979).

5. H. Chang and J.K. Aggarwal, "Implementation of 2D Semi-

causal Recursive Filters," Proceedings of 1979 IEEE

International Symposium on Circuits and Systems, Tokyo,Japan (July 1979).

6. H. Chang and J.K. Aggarwal, "Fundamentals of Two-

Dimensional Recursive Digital Filtering," in Digital

Signal Processing, edited by J.K. Aggarwal, Western

Periodicals, California, 211-260 (1979).

7. N.C. Huang and J.K. Aggarwal, "On Linear Shift-Variant

Digital Filters," to appear in IEEE Transactions on

Circuits and Systems (1980).

15 I


Research Unit IE9-3 ELECTRONIC COMPUTER SYSTEM DESIGNAND ANALYSIS

Principal Investigators: Professor S.A. Szygenda (471-7365)Professor E.W. Thompson

Graduate Students: Rodney Barto, Rafael Lemus, Don Rossand Richard Von Blucher

A. OBJECTIVES AND PROGRESS: Logic simulators can performseveral useful functions for the design engineer [1]. By

simulating his circuit he can verify the correctness of hisdesign and determine whether it possesses any undesirable be-havior due to device delays, without having to build any hard-ware. He can also examine its performance under fault condi-tions and devise tests to locate devices that have failed.The increasing proliferation of digital devices in common ma-chines makes their correct operation in both normal and faultconditions much more important. The rising cost of manpowermakes any device that can hold down development costs lookvery attractive. Logic simulators can fill these needs andare a potentially useful tool for the design engineer [23.

In order for a logic simulator to be useful it mustpossess two attributes. First, it must accurately predictthe operation of the logic circuit being simulated. Second,it must do this in a cost effective manner. These require-ments are interrelated, as will be shown.

A logic simulator performs its functions by con-structing a working model of the network under consideration[3,4]. The devices, Boolean and sequential gates, that thenetwork is constructed from must be modeled as must the top-ology of the network, that is, the interconnections of thedevices. The network topology is stored in a set of tablesthat the simulator creates. The most difficult job of thesimulator is to model the devices in the network and the elec-trical activity that they produce.

Even though logic devices are considered to be bi-nary valued, that is, producing outputs that are either logic0 or logic 1, in reality they are all analog in nature. Assuch, they are capable of a considerable amount of non-binarybehav'or. Signals propagate through them in finite delaytimes, and this produces undersirable logic level behavior,such as hazards. The effects of output loading and improperlyterminated connecting lines will produce poorly defined rising

17

NPAU$ BIAMGP40 7lJ6D


and falling edges. All this activity should be modeled bythe simulator.

There are two basic types of logic simulators inuse today. The most common is the logic level simulator.This type models logic signals as being either logic 0, logic1, or unknown, and some will have representations of risingand falling signals and hazards. The most accurate of thesewill have a number of possible signal representations. Delaymodeling capabilities range from simulators that assume eachgate has one time unit of delay, to those allowing each gateto have an unique ambiguous delay.

The second type of simulator models the electricalactivity of the network. Each gate is modeled by consider-ing the transistor circuit that it is made from. The MOTISsimulator [5], which is of this type, has capability of rep-resenting 64 signal levels and is thus more accurate than thelogic level type mentioned above.

The more accurate a simulator is the slower it per-forms its simulations. The most common measure for simula-tor speed is the number of gate evaluations per second (GES).This is the number of gate output values the simulator cancalculate in one second. Logic level simulators can performfrom 400 to 4,000 GES, while the MOTIS simulator can perform400 GES. Since only 10% to 20% of a network is active at anytime, most simulators are event driven. Such a simulator willevaluate a gate only if it is active; this saves considerableamounts of simulation time.

The limitation on simulator speed stems mostly fromthe fact that all simulators in use today are computer pro-grams run on general purpose computers. The common computerarchitecture does not provide a natural framework for the com-plex data structures a logic simulator requires. The purposeof this work was to design and validate a computer architec-ture for logic simulation. This architecture incorporatedinto its hardware the building blocks of a software simulator.Thus the tables and flows of data required by a simulator arenatural to the computer, rather than defined and maintainedby a computer program. The simulator would run with the aidof a host computer, that would load the simulator and handleits output.

There were three objectives in this design. First,the computer resulting from this design should be faster atsimulation than a software simulator. The "hardware" simula-tor should be at least one and possibly two orders of magni-tude faster than a software simulator. Second, the hardwaresimulator should be more accurate than a software simulator.The computer should support, yet not be limited to, the MOTIS

18


type of simulation. Third, the initial cost of the hardwaresimulator should be comparable to the initial cost of a soft-ware simulator.

The constraints on this design fall into two cat-egories. The first category limits the components which canbe used in the design and the second limits the extent of thedesign itself.

1. Constraints on Hardware Used

The decision was made at the beginning of thedesign to limit the components used to those that are readilycommercially available. There would, therefore, be no design-ing of custom made integrated circuits or other devices. Thelogic family chosen for the control circuitry is the TTL fam-ily. There are two reasons for this. First, the initial costof TTL logic is much less than that of the faster ECL logic.Second, it is much less expensive to construct a circuit withTTL than with ECL. This is because the higher speeds providedby ECL make the design of interconnections in the circuit muchmore critical than they are with TTL.

The major memories are constrained to operate atspeeds obtainable with standard MOS memory ICs. It would bepossible to build the memories with TTL or ECL ICs, but thecost of this would be prohibitive.

The processors of the computer, the units thatdo the actual work, may be made of either TTL or ECL. Theformer would be preferable, for reasons mentioned above.None of the processors, however, will be physically large,and thus, would not be either expensive or difficult to buildwith faster logic.

These hardware limitations placed a heavy demandon the design. Any increase in speed will be obtained by vir-tue of an improved architecture, not from the use of fasterlogic. The larger of the computers a software simulator mayrun on rely more heavily on the use of the faster logic fami-lies for their speed increases.

2. Constraints on the Design

There are a number of items of the design thatwere not dealt with in this study. All of them result fromthe fact that the purpose here is to present only the basicarchitecture of the computer, not a design for a f~nishedproduct.

B. SIMULATOR DESIGN OVERVIEW AND RESULTS: The major func-tional blocks of the simulator are shown in Figure 1. Eachtable that a simulator requires is confined in a separatememory. These are the Fan-in Memory (FIM), Fan-out Memory

19

E-.E-4

< 0

0 400H

0to- -


(FOM), Status and Data Memory (SDM), Activity Flag Memory(AFM), and Event Queue Memory (EQM). The FIM contains alist of the inputs to each gate, and the SDM contains itsstatic and dynamic data such as value, type, and number ofinputs. The FOM and AFM are used to allow the simulator tobe event driven. The EQM is used to schedule events.

The simulator operates in two phases, Evaluationand Update. Each of these involves a separate processor, theEVAL and UPDATE processors, respectively. The Event Queueprocessor has only minimal dities to perform in its handlingof the EQM.

During the Evaluation phase, gates that require e-valuation have their input values sent from the SDM to theEVAL processor. This phase makes use of the interleaved na-ture of the FIM and SDM. The FIM can provide the addressesof the inputs to a gate to the SDM at a high rate since theyare accessed 16 at a time and sent in a word serial, bit par-allel manner over a high speed data path. The SDM is inter-leaved 8 ways and can thus provide the input values to theEVAL processor in a relatively short amount of time. TheEVAL processor itself can be optimized for the type of gatemodeling desired.

During the Update phase, events scheduled duringthe Evaluation phase occur. The information for these eventsis stored partly in the SDM and partly in the EQM. The UP-DATE processor is a 7 stage pipeline capable of handling gatesat the rate of four gates in each cycle time of a SDM block.During this phase, the fanouts of active gates are followedby sending the contents of the FOM to the AFM. During theEvaluation phase, the AFM is searched for active gates, andtheir addresses sent to the FIM. Further details of the de-sign are provided in Reference 6.

A possible implementation of the simulator, inwhich all of the memories were constructed from NMOS inte-grated memory circuits was presented. The speeds at whichthe various components operated were compatible with TTL con-trol circuitry. The simulator was shown to be capable of per-forming from 10,000 to 200,000 gate evaluations per seconddepending on complexity of the devices modeled. This isfrom one to two orders of magnitude faster than software sim-ulators.

Because the simulator has not been designed to gatelevel, only a cost estimate of the major memories can bemade. This estimate is still meaningful, since these unitsform the bulk of the machine.

Prices for the ICs for the major5 memories were ob-tained in March of 1980. NMOS dynamic RAM- made by the Nip-pon Electric Company were selected for the FIM, FOM, AFM, and

21


SDM. The total cost of these components was $17,750. TheEQM could be constructed from memory made by the ElectronicMemory and Magnetics Corporation at a cost of $1,536. Thisbrings the total memory components cost to $19,286. A reas-onable estimate for the total system cost would be two tothree times the memory cost, roughly $60,000 to $90,000.This compares favorably with software simulators which costfrom $40,000 to $170,000.

The estimates given here must be viewed in lightof the following considerations. First, the performance ofthe machine is largely dependent on the efficiency of theEVAL processor. Its performance was given here assuming itwould perform evaluations much as a software simulator would,using stored programs. It is quite likely that special hard-ware could be designed to speed up the EVAL phase considerably.

Second, the cost estimates given here are for onemachine. In a production effort, where several simulatorswere to be built, the cost of components would decrease. Cur-rent demand for software simulators and the growth of the dig-ital electronics industry could justify such a product effort.

C. REFERENCES

1. S.A. Szygenda, "Simulation of Digital Systems: Where weare and where we may be headed," Proceedings of the Sym-posium on Computer Aided Design of Digital ElectronicCircuits and Systems (1979).

2. S.A. Szygenda, "VLSI-A Challenge to CAD," Proceedings ofthe International Conference on 'computer aided designand manufacture of electronics, components, circuits andsystems,'" (1979).

3. S.A. Szygenda and Y.H. Jea, "Mappings and Algorithms forGate Modeling in a Digital Simulation Environment," IEEETrans. on Circuits and Systems, CAS-26, No. 5, 1979.

4. S.A. Szygenda and Y.H. Jea, "Boolean Vector AlgebraicStructures; Properties and Applications to Digital LogicSimulation," Computer Aided Design, Nov. 2, No. 2, Novem-ber 1979.

5. B.R. Chawla, H.K. Gummel, P. Kozak, "MOTIS - An MOS Tim-ing Simulator," IEEE Trans. on Computers, Vol. CAS-22,No. 12, December 1975.

6. Rodney L. Barto, "A Computer Architecture for Logic Sim-ulation," Ph.D. Dissertation, Univ. of Texas at Austin,1980.

22


Research Unit IE9-4 ELECTRONIC COMPUTER SOFTWARE SYSTEMS

Principal Investigator: Professor Raymond T. Yeh

Graduate Student: T. William Mao

A. PROGRESS: This work is concerned with an approach to dis-tributed programming and involves the design and validationof a programming language for a distributed environment wheremany processors cooperate to complete a single task. The pro-cessors share no common memory and communicate with each otherthrough hardware busses. This environment can be seen as anabstract model of many networks and microprocessor networks.

We considered the design of a distributed language,DLCP, which draws most of its features from Algol-descendentlanguages such as Pascal. The basic components in the lang-uage are processes, which are assumed to be executed in par-allel. Strong emphasis is placed on communication facilitiesfor processes. Since no shared variables exist, communicationis by message passing. To simplify the implementation of thelanguage, we further assume that the distributed environmentdoes not support automatic buffering of messages. Informa-tion exchange therefore must be totally synchronized. Com-munication Ports (CP), [1,2] influenced by Hoare's communica-tion sequential processes [3], were proposed as a generalizedmechanism for both information exchange and synchronizationamong processes. During communication, the processes behaveas if they merge into a single process. To provide greaterconcurrency, another feature was introduced to allow earlydisconnection of communication. CP also provides compiler-time message type checking and a run-time mechanism to sched-ule call acceptances. Many examples have been given showingthe equivalence of CP and the communication and synchronizationfeatures of other proposed languages.

In order to formally define the semantics of DLCP,Hoare's deductive system of axioms and assertions was extendedto include distributed programs. Semantic rules were devel-oped to define the semantics of CP. We also developed proofrules which, when combined with semantic rules, can be usedto prove functional correctness of distributed programs writ-ten in DLCP. In proving functional correctness, the invari-ant property, which is essential to the verification of mul-ti-process programs, was identified. The invariant propertyis generally considered to be a property global to the whole

23


program, and requires all assertions in the program to meeta condition, called speed independence. We show that invari-ance may be proven only if a subset of the program assertionsare speed independent. The effort to discover proper asser-tions and invent auxiliary variables for proving a programwas thus reduced. Further, the deductive system was provedto be consistent with a state machine model of program execu-tion, which assumes machine states instead of logical asser-tions as elements.

Finally, the feasibility of implementing DLCP on aprocessor network was assessed. An algorithm, using messagesend and receive as primitives, was developed to enable aprocess in DLCP to communicate with other processes efficient-ly. The runtime support system needed to interface a processwith the processor it resides and other processes was consid-ered. A prototype of such a system, embedding the algorithm,is designed, illustrated, and coded in Concurrent Pascal.

B. REFERENCES

1. T.W. Mao and R.T. Yeh, "Communication Port: A LanguageConcept for Concurrent Programming," Proc. of DistributedComputing Systems Conf., Oct. 1979, 252-259, also IEEETSE,March 1980.

2. T.W. Mao and R.T. Yeh, "An Illustration of SystematicDesign of Parallel Programs for Real-Time Applications,"Proc. of COMPSAC, Nov. 1979, 392-397.

3. C.A.R. Hoare, "Communicating Sequential Processes," CACM21, 8, August. 1978, 666-677.

24

11. SOLID STATE ELECTRONICS

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

SOLID STATE ELECTRONICS

Research Unit SS9-I. BASIC SOLID STATE MATERIALS RESEARCH

Principal Investigators: Professor R.W. Ben4 (471-1225)Professor R.M. Walser (471-5733)Professor A.B. Buckman (471-1095)

Graduate Students: S. Chao, K.C. Chen, L. Lancaster,G.S. Lee, H.Y. Yang

A. OBJECTIVES: Our main objectives are to develop an under-standing of the kinetically selected reaction paths andphases at low temperatures and to increase our understandingof the relationships between atomic rearrangements and elec-tronic, magnetic and optical properties of the interfaceregion in thin film reaction couples. In previous studieswe have hypothesized the existence of a glassy membrane layerwhich forms initially upon metal deposition on a Si or Gesurface [1,2,31, and which acts as the controlling elementin compound nucleation. We have observed the formation ofa disordered layer for ultrathin deposits of metals on Si[4,5] and have shown that the possible existence of a modi-fied disordered region (behaving as a negative U glass),even after compound nucleation, is consistent with Schottkybarrier results[6].

Some of our present objectives are to check andextend these hypotheses by correlating surface resistance,magnetic moments (EPR), noise and dielectric response mea-surements with transmission electron diffraction indicationsof thin film structure. Also, ellipsometric studies arebeing used to obtain the depth resolved optical constants(and optical spectra over limited wavelength ranges) to cor-relate with the preceding measurements.

B. PROGRESS: We have found that nucleation of the firstcompound phase out of the initially formed disordered phaseat low temperatures (room T) is quite possibly due to anelectronic instability in the interfacial glassy region. Wehave made measurements on the Co-Si system, where we findthat the onset of Co2Si nucleation occurs after a depositionof a critical effective thickness of Co which depends onsubstrate preparation. (In general, the cleaner the sub-strate, the smaller the critical thickness.) This same crit-ical thickness is just the thickness at which the surfacedisordered region attains metallicity (defined by dRs a 0aTover two dimensions or less, indicated by the value of R swhere this occurs (104 Q)

0Since the first stages of Co 2 Si nucleation appear

27

- f= pECD1N PM .B"W_1i4OT nUkD

SOLID STATE ELECTRONICS

to occur in separate islands, the compound formation doesnot produce the metallicity-- rather, metallicity apparentlystimulates compound nucleation. As a further check on thisinterpretation we have deposited thin Co films on Si sub-strates which had been doped into degeneracy, i.e., whichwere metallic to start with. We find compound (Co 2 Si) nu-cleation down to the thinnest Co which we can experimentallysputter. Thus the critical thickness, which typicallyoccurs at about 70 sec. deposition for 4-10 Qcm Si, has beensuppressed below 5-10 sec deposition for degenerately dopedSi. It should be noted that we have used bo

Documents

1-i2 I ffflllllffff -EEEEEEEEEEE -EEEEmmEEEEE -EI ...AD A089 343 TEXAS LNV AT AUSTIN ELECTRONICS RESEARCH CENTER F/6 9/5 UNC MAY ANNUAL SO REPORT E J POWERS, ON ELECTRONICS S I MARCUS