liANAW,\\simulation,

and analysistlxxx%

of.., .."

I ' ,,/. N / .i ,, . \ / ...0

N / el,.

Cover photo: Andy Whiting, MSR, Moorestown

Haymanp. 43

Enstromp. 71

Guide. of aLp. 84

Adelson, al al.p. 56

Ashkinazyp. 10

Ashkinazyp. 10

Golubp. 32

Our cover shows a crisp still life made from outputs gen-erated by authors contributing to this Modeling, Simula-tion, and Analysis issue (the keyline above showsauthors and page numbers). These engineers and scien-tists are a special breed. They predict how engineeredproducts and systems will act.

Given a jumble of information and outputs, our authorsmust constantly bestow order and describe gray areas inblack and white terms. They can give the green light to astep in an an engineering project, or suggest that it goback to the drawing board. More rarely laying hands oncostly, time-consuming prototypes and breadboards totry out ideas, our authors today take to their terminalsinstead, and use increasingly well -developed simulationand modeling programs to tell them whether somethingwill work.

These practitioners, as gurus of the engineering com-munity, establish "photographically accurate" engineer-ing perceptions of the future-based on mathematics,computer analyses, and algorithmically disciplined logicand intuition. Though these professionals may seem towork at their green CRTs and study their green duotonedcomputer printouts in a silent twilight zone, neverthelesswe can see that the stark realism of their work bogglesthe mind.

-MRS

[ECM EngineerA technical journal published byRCA Research and EngineeringBldg. 204-2Cherry Hill, NJ C8358TACNET: 222-4254 (609-338-4254)

Tom King

Mike Sweeny

Louise Can

Frank Strobl

Betty Gutchigian

Dorothy Berry

Jay Brandinger

John Christopher

Hans Jenny

Arch Luther

Howie Rosenthal

Ed Troy

Bill Underwood

Bill Webster

Ed Burke

Walt Dennen

Charlie Foster

John Phillips

RCA Engineer Staff

Editor

Associate Editor

Art Editor

Contributing Editor

Composition

Editorial Secretary

Editorial Advisory Board

Division Vice -President and GeneralManager, "SelectaVision"VideoDisc Operations

Vice -President, Technical Operations,RCA Americom

Manager, Engineering Information

Division Vice -President, Engineeringand Product Assurance,Commercial CommunicationsSystems Division

Staff Vice -President, Engineering

Director, Operations Planning andSupport, Solid State Division

Director, EngineeringProfessional Programs

Vice -President, Laboratories

Consulting Editors

Administrator, MarketingInformation and Communications,Government Systems Division

Manager, Naval Systems DepartmentCommunications and Information,Missile and Surface Radar

Manager, Systems and Procedures,RCA Laboratories

Manager, Business Developmentand Planning, RCA Service Company

To disseminate to RCA engineers technical information of professional value To publishin an appropriate manner important technical developments at RCA. and the role of theengineer To serve as a medium of interchange of technical information between variousgroups at RCA create a community of engineering interest within the company bystressing the interrelated nature of all technical contributions To help publicize engineer-ing achievements in a manner that will promote the interests and reputation of RCA in theengineering field To provide a convenient means by which the RCA engineer may reviewhis professional work before associates and engineering management To announce out-standing and unusual achievements of RCA engineers in a manner most likely to enhancetheir prestige and professional status

A 14. Teger

Modeling and simulation:Tools for the eighties

Over the last few decades, modeling has had many meanings fordifferent people. When I was a teenager, modeling was somethingdone only by untouchable beauties in sophisticated outfits (or lackof them). As a young graduate engineer in the early sixties. model-ing had an overtone of being a last resort when a designer was notable to theoretically analyze, or directly build, his circuit or system.Today the designer often won't begin his task without a selection ofthese design tools at his disposal.

Modeling and simulation have become integral parts of thedesign process for a surprisingly wide -ange of application areas. Itwill be obvious from scanning this issue of the RCA Engineer thatthe techniques are invaluable for management analyses of newbusiness areas, as well as for technical designs ranging from largemulti -satellite communication systems to dc motors to complexintegrated circuits. Simulations have become incredible time saversin the design cycle-reducing costs of false starts in fabrication orbreadboard, and creating much more flexibility for consideringalternate approaches.

Perhaps more important, many of today's technologies cannot beused effectively at all without simulation tools. For example, bread-boards alone cannot fully characterize a final integrated circuit; ormanual textbook analyses cannot handle the number of variablescritical to a large system. In situations such as these, the computerbecomes not only an aid, but an essential to the design process.

The usefulness of any of these tools remains critically dependentupon their human interface. How easy is it to input the data, tocharacterize the parameters, to request different simulations? Howare the results reported or displayed? Is it easy to use the results tomodify the design or the approach? More and more emphasis isbeing placed on this interface, and the resulting user -friendliness islargely responsible for the enormous growth in use of modeling andsimulation tools.

Despite the recent growth, we have only barely begun to use thisaspect of the computer's power. Within ten years, it may well be thatevery engineer uses computer modeling and simulation to design,test and validate his system long before fabrication.

Alfred H. TegerStaff Vice -President, Systems ResearchRCA Laboratories

'limn EngineerVol. 271No. 6 Nov. Dec. 1982

overview 4 Simulation in an engineering environmentR.R. Barton 1K.A. Pitts

simulation 10 The MIMIC logic simulatorA. Ashkinazy

18 An HF modem simulationP.A. De Maria K.J.1 Bodzioch

24 Dynamic analysis and simulation of mechanically scannedradar systemsG.M. Sparks 1J. Liston

32 Discrete -event simulations of complex weapon systemsJ. Golub

39 Digital interface simulation for large system developmentG.W. Suhy

43 The Engineer's Notebook: Design and simulation of an intelligentmissile seekerJ. Hayman

46 ANA radio -frequency circuit simulation and analysisS.M. Perlow

modeling 56 Modeling of the human visual systemE.H. Adelson 1C.R. Carlson A.P. Pica

65 Business modeling and the engineer:An Astro-Electronics applicationR. Nigam IS. Hong I S.R. Spence

71 Modeling and simulation in mechanical and thermal designR.E. Enstrom

80 Electromechanical motor simulationR. Browne

84 Computer modeling in the RCA Satcom systemA.A. Guida IK. Jonnalagadda ID. Raychaudhuri L. Schiff

departments Patents, 92 I Pen and Podium, 931News and Highlights, 97Index to Volume 27, 102

Copyright 1982 RCA CorporationAll rights reserved

in this issue ...modeling, simulation, and analysis

Barton/Pitts: "We must describe the important questions we want toanswer and use these criteria to develop the model."

Ashkinazy: "Due to the complexity of LSI and VLSI chips, computersimulation has essentially replaced the breadboard . . . . "

De Maria/Bodzioch: "The simulation program was written to repre-sent the operation of a specific modem used in ,CIF radio transmission."

Sparks/Liston: "We have selected a mechanically scanned radarsystem as an example of a system of moderate complexity, with threedistinct areas of concern facing the designer."

Golub: "Although only weapon systems have been discussed, dis-crete -event simulations have been widely used for transportation, 'ogis-tics, and traffic -control problems."

Suhy: "Thousands of programming errors have been detected, pre-cisely documented, and corrected using such simulations."

Hayman: "Imaging seekers which utilize considerably more informa-tion concerning the target area are now possible."

Perlow: "The circuit and desired outputs can be modified easily bymeans of a simple conversational mode that does not require editing ofdata files."

4

18

s/VVVVA.A.

Adelson/Carlson/Pica: "Indeed, for all of us, seeing seems so directand effortless that we remain unaware of the complex visual processingthat underlies a statement like, 'That's a good picture.' "

Nigam/Hong/Spence: "This oaper describes a model developed tohelp Astro-Electronics analyze tne business imoact of design alterna-tives of a multisatellite system."

Enstrom: "Finite -element modeling analysis has been used at RCA inthe solution of numerous complex problems."

Browne: "The model helps design engineers to set manufacturingtolerances by providing estimates of motor -performance variation fromthese various design parameters."

Guida, et al.: "The emphasis is on what the models do, not how theydo it."

'2=3.*

in future issues...

engineering productivity: tools and techniques,technology transfer/energy,RCA technology guide

F-INPAD(0)R#UNPAD MATRIFITRAI

A J. SOLOMON. 8A THIS FUNCTION AL ItiNF

COLUMNSC A-MATRI..

A L

DA

.**NATRI._ .*MATRI.4.,LEAD^TRAIL)/[2] MATRI.

R.R. Barton K.A.1 Pitts

Simulation in an engineering environment

Simulation is a powerful tool for solving complex engineeringproblems. When is it useful?

Abstract: Simulation-developing amodel of a real system-is a powerful toolif used correctly. Concerns in simulationinclude how to formulate, develop, andverify models; how to use computers effec-tively, and how to present the results.

Why simulate? Given that modern sci-ence is based on mathematical models thatsimulate the observed behavior of physi-cal, chemical, and electrical systems, thisquestion, in the broad sense, appears to berhetorical. The reasons for using modelsare clear. It may be impractical or costlyto manipulate the real system, or one maywant to perform experiments on a com-pressed time scale. We ask the question tofocus on the more parochial definition ofsimulation we're dealing with here. Simu-lations, though perhaps sophisticated, areusually thought of as modeling by bruteforce. Rather than solving a system of dif-ferential equations to predict circuit behav-ior or a projectile trajectory, we stepthrough a series of calculations using dif-ference equations and simple arithmetic,recording the approximate voltage or posi-tion at each time increment. In this nar-rower sense, simulation models can be con-sidered as alternatives to analytical models,and the question, "Why simulate?", be-comes meaningful.

A frequent reason to simulate is that thesystem or problem being modeled is toocomplex to understand any other way.The only way to understand the interac-tion between parts of a system, to improvethe current methods, or to design a newsystem may be to model the system andsimulate reality. For example, we may bec1982 RCA CorporationFinal manuscript received September 8. 1982Reprint RE -27-6-1

enlarging our production facility. Whichmachines should we add? How many?Where will our potential bottlenecks be?

Simulation can also be used to identifytrouble spots. Suppose a piece of auto-matic test equipment is a bottleneck. Whathappens if we increase the speed of theequipment? The quality of incoming mate-rial? Another frequent reason to simulate,closely related to the complexity issue, isthat the mathematics needed to analyzethe system are intractable. In a complexqueuing system-such as a production floor,highway -traffic network, or telephone -switch-ing system-the underlying mathematicalbehavior may not lead to equations thatare easy to solve.

Although we usually think of simula-tion as a problem -solving tool, it doeshave other uses. Simulation models arefrequently used as a training tool-flightsimulation is a well-known example. Weap-ons systems have been simulated, as wellas other man -machine interfaces, Anotherreason to simulate is to check the behaviorof a new algorithm, mathematical tech-nique, or model. To answer the question,"Does the algorithm work?", you can con-struct information you know and under-stand (historical or randomly generated da-ta), and then examine the algorithm's perfor-mance when it is given the known data.This method is frequently used to test math-ematical techniques. Recently, statisticiansused computer simulations to develop morerobust measures than the simple average.The average (arithmetic mean) can givemisleading estimates if some of the obser-vations are faulty (misplaced decimal points,bad readings, disasters during the measure-ment, changes in equipment). Several alter-natives to the average have been suggested.These methods have been tested through

simulation: One generates known distribu-tions of numbers, and checks how badlyeach technique errs when supplied withcontaminated data or data from unusualdistributions.

Building simulation modelsSimulation is a modeling and problem -solving tool. In any modeling situation,each of the following four steps must beused to build good simulation tools: (1)Formulate the problem-What questionsdo we wish to answer? (2) Develop themodel-How do we estimate the param-eters? What data and formulas do weneed? (3) Verify the model-Does themodel work correctly? and (4) Validatethe model-Does the model resemble real-ity closely enough?

Problem formulation

We've said that simulation can tell us theanswers to questions we have about oursystem. Why do we rely on the abstrac-tion? How can we make sure that theinformation we get is useful and not mis-leading? To develop a simulation to helpus understand our system, we have todecide what portion of the problem wewant to study. Good problem formulationis essential. We simply cannot model ourcurrent system in all its detail, and thenhope to answer questions efficiently or well.Some detail must be lost in the transferfrom a real system to a mathematical modelof it. We must decide the important ques-tions we want to answer and use thosecriteria to develop the model.

Keep the following example in mindwhen we talk about simulation. Considera production line. Say we have a simple

4 RCA Engineer 27-6 Nov./Dec. 1982

situation-three process steps to complete,and machines for the three processes, A,B, and C. These processes vary in com-plexity and time to complete the job. Wehave incoming raw material, and outgoingfinished product. We might want answersto questions like: Where are our bottle-necks? Do we need more type -B machines?Do we need a preventive maintenanceprogram? Do we need a new priorityscheme for scheduling work?

Model development

Once we have identified the problems wewish to solve, we must construct the model.Which aspects of our production systemmust we model in detail to answer thesequestions? Which parts of our system canbe modeled at reduced levels of detail orleft out? We will need to collect data fromthe actual system we are modeling. Howlong does each process take? What is thevariability? How often do orders come in?What probability distribution do the ma-chine failures follow? As we collect thedata, we begin to develop the model bymaking assumptions about the data distri-bution, the simplifications we want to make,the parameters that are the most impor-tant, the variables we will track or control,the ways we will vary them, the ways thevariables interact, and so on.

Model development includes implemen-tation. What do we use to build the model?Its parts may be physical, analog, or dig-ital. If it is a digital computer model, wemust choose the programming languageand the computer on which it will run.

In what form will it provide answers?Tables? Graphs? Will these answers beadequate for our needs? This last pointseems common sense, but, incredibly, it isoften overlooked until the model is com-plete. In some cases this oversight meansretrofitting the model to provide the rightinformation; in other cases, the conse-quences are more severe!

Verification and validation

After we have developed the model, wemust verify it. The model should work aswe planned (debugged). Is the behaviorroughly what we expect? Is there some-thing grossly wrong? The final step, one ofthe most important and also most frequentlyneglected, is validation. How does our ab-straction compare with reality? How welldoes it do on a different data set? Is it toodetailed? Too simplistic? Does it addressaccurately or precisely the questions oforiginal interest? Can it be extended to

Queues are ontities waiting to perform activities.

other questions? At this stage we frequentlyfind problems, and repeat the process offormulation, model development, testing,and verification.

The structure of simulation modelsWhat are the general characteristics of mod-els that simulate systems? We think of asystem as being a set of objects with char-acteristics and some interdependence in-volved in actions. In simulation jargon, the"objects" are called entities, their "charac-teristics" are attributes, and zhe "actions"are called activities. Examples of systemsare shown in Table I.

The items listed in Table I for any onesystem are by no means complete. Theitems (entities, attributes, and activities) cho-sen to be included in the model, and theinterrelations chosen to be modeled, dependon the purpose the user has in mind. Whenwe model a system, we pick such a subsetof entities and interrelations. One cleartrade-off in model formulation is the levelof aggregation. A simple model with manyentities aggregated into a few major enti-ties is much easier to program, test, anddebug since there are fewer entities andinteractions. Of course, the resulting modelmay generate a poor approxmation to thereal system. A good strategy is to start

Table I. Examples of systems. Systems are sets of objects, with characteristics andsome interdependent actions.System Entities Attributes Activities

Airline counter Passengers With/without ticketWith/without baggageFlight Number

Agents Capability (tickets only,baggage only, or both)

Get informationPurchase ticketCheck inCheck baggage

Give informationSell ticketCheck inCheck baggage

Fighter -missile Fighter PositionVelocityPerformance ability

Attack missile PositionVelocityPerformance ability

Assembly line

Evasion

Pursuit

Components TypeDefective / nondefective

Assemblers Type (fast, slow)

AssembleTest

AssembleTest

Computer Programs Core requirements lan-guage (e.g., FORTRAN)

Special librariesExecution timePeripheral devices

CompileExecutePrintPlotStore on tapeStore data on disk

Barton/Pitts: Simulation in an engineering environment 5

with simple representations and augmentthe model later in areas that are found tobe too crude.

For the level of detail chosen, the state(at time t) of the system is given by acomplete description of all entities, attri-butes, and activities at one point in time, 1.

We define an event to be something whichcauses the state of the system to change.

The kind of simulation model we choosedepends on what kinds of characteristicswe try to model. We need to considerwhether the structural relationships are tobe static or dynamic; and whether theattributes and/or activities are to be deter-ministic or probabilistic. We might use astatic probabilistic simulation to study thestrength of a building made with compo-nents of randomly varying strength. Onecould simulate many such buildings beingconstructed, and simulate stresses, to deter-mine what fraction of such buildings wouldcollapse. To model waiting times at an air-line counter, on the other hand, wouldinvolve dynamic relations of arrival timesand service times. Difference equations oranalog computers are often used to per-form deterministic simulations of the per-formance of physical systems.

A broad definition of simulation modelsincludes mathematical models, physical mod-els, and electronic -analog, and digital -elec-tronic models. Each of these implementa-tions can be used to imitate the performanceof the real system being studied. Somemodels such as flight simulators combinephysical, electronic -analog, and digital simu-lation techniques. When we speak of sys-tem simulation below, we will refer to anarrow meaning-dynamic, probabilistic,digital -electronic (computer) simulation. Togive a better understanding of these mod-els, we'll review the important considera-tions in building and using them.

Issues in computer -simulationmodelingSimulation languages

One critical issue in developing a compu-ter model for simulation is whether toprogram the model with a general-purposelanguage or to use one of the simulationlanguages currently available. This decisionis based on your acres to the language,your ability and desire to program, andthe complexity of the system you are sim-ulating. The more complex the system, themore you should seriously consider usinga simulation language. There are clear ad-vantages in using a special-purpose simula-tion language like Simscript, as opposed to

A static, probabilistic computer simulationFor this three -link riveted assembly, what will be the variability inthe span (X) given the variability of the inter -hole distances Y,, YY,?

x

X = 4a+bp* (c+d)2 la+b)(+d) (a2+0_e2)ac

reality trigonometric model

If the random distances a, b. c, d, and e have normal probabil-ity distributions, we cannot find the distribution of X analytically.So we simulate the assembly of many parts.

How the computer simulation works:Step 1. Compute random distances for a, c, from distribution 1.

Step 2. Compute random distances for b. d, from distribution 2.

Step 3. Compute random distances for e from distribution 3.

Step 4. Compute X via the model and save the value.

Repeating steps 1 through 4 many times, the computer simula-tion generates a large set of X values. A histogram of valuesgives a good representation of the variability that can beexpected. The sample variance can be computed from the dataas well.

t !

X va ue

Simulation output

a general-purpose language like FOR-TRAN. The high-level representation meansa shorter, simpler program that takes lesstime to code and is more likely to passverification tests for proper operation. Onthe other hand, the main disadvantages-greater memory and execution -time require-ments-can be devastating for a frequentlyrun model. When we use a simulationapproach to identify optimal control, orplanning strategies, we will need to beable to make many repeat runs quickly,calling the simulation program from withinan optimization routine, and modifying itsoperation via parameters passed from the

optimization program. Fortunately, pack-ages written as a set of general-purposelanguage subroutines allow the engineeran intermediate approach (see box, page7).

For complex systems the output of asimulation can be a horrendous pile ofnumbers. Tables and tables of numbersare difficult to scan for interesting correla-tions, time trends, or exceptional events.It's of utmost importance to find ways tographically display the results, so this con-sideration should weigh heavily in yourchoice of a simulation language. If youdon't find something acceptable, take heart.

6 RCA Engineer 27-6 Nov./Dec. 1982

Computer -graphics capabilities are chang-ing rapidly, and packages are being con-tinually updated.

Graphical methods increase the useful-ness of simulation output.

Modeling a dynamic system

A dynamic system such as the fighter -mis-sile pursuit system is clearly continuous innature. Position, time of contact, and soon, can vary continuously. For other sys-tems the important events can be consid-ered to occur at discrete points in time.For the assembly -line system, these dis-crete events are the arrival of componentsfor assembly, completion of assembly, startof test, and so on.

Since dynamic systems evolve over time,some sort of timing scheme must be chosenfor the model. For continuous systems, afixed -increment timing approximation isused for digital simulations. For discrete -event systems, there is no need to examinethe system between events. Virtually allmodem simulation languages use a next event

scheduling -timing method.There are two methods for dynamic

discrete -event simulation. One is calledevent scheduling, the other, process inter-action. In most cases, a simulation lan-guage is structured to do one or the other.Simscript 11.5 and GASP IV use eventscheduling; Simscript 11.5, GPSS, and Sim-ula use process interaction. For event sched-uling, one models the system by focusingon what events or activities can happen.Typical events are arrival, repair, comple-tion, service interruption, and machine break-down. Since time lapses between events,one can regard an event scheduling as astring of all events that will happen to thesystem. Process -interaction provides a sepa-rate procedure for each entity in the sys-tem. Machine, operator, CPU, I/O -device,

Some common simulation languages*Simscript 11.5 is a complete high-level programming languagewith specific structures to make simulation easy. It uses a com-piler; has good, well -documented random -number generators;and has good capabilities for handling complex systems. It is,however, a programming language, not a software package. Yousacrifice ease in getting a model up and running but you gainincreased control while it is running. It is available only on largecomputers.

GPSS is a software package, working with an interpreter ratherthan a compiler. It is easy to work with, easy for beginners to use.Its defects are clumsy generation of most probability distribu-tions, and difficulties in collecting statistics. In addition, we havenot found documentation on its random -number generator. Also,GPSS is no longer supported by IBM. GPSS is available only onlarge computers.

GASP IV is a collection of FORTRAN IV subroutines to help insi-nJlation. GASP IV provides many of the necessary simulationtools; however, it is necessary to write the programs to handlethe data, call the routines, and print the results. Its advantagesare that it runs on any system with a FORTRAN compiler. It alsodoes both discrete and continuous simulations.

SIMULA is an algol-based simu.ation language. It is similar toSimscript in that it is a programming language with all that thatin -plies. (It includes a compiler to decrease computer executiontime, but one must still learn a new programming language).

RCAP is a circuit -simulation package that can provide static,dynamic, and frequency -domain simulations of integrated -circuitperformance. Arbitrary circuit elements may be supplied assubroutines.

DYNAMO is a language used for modeling large social or indus-trial systems via a set of interconnected difference equations.

* This list is by no means complete. See Reference 4, page 115-143, for more detail on some of these.

and airplane processes are typical. In theterms we defined earlier, processes are enti-ties/activity sets. In a production -schedul-ing problem one would model each pro-duction process; in event scheduling, eachevent. Both approaches base their timingroutines on a next -event basis. Althoughthe process -interaction approach is oftenconceptually simpler, there is some loss inprogramming control. Collecting informa-tion about congestion, lengths of queues,and so on, is more difficult.

Modeling random variability

Most simulation models have random var-iables (service time, interarrival time, com-ponent strength. and so on) that one mustsomehow generate either prior to or dur-ing the simulation. Practically speaking,continuous random variables with any kind

of probability distribution can be gener-ated on the computer by transforming aset of random numbers uniformly distrib-uted on the interval [0, I ]. The impor-tance of the uniform distribution stemsfrom the fact that if X is a continuous ran-dom variable with cumulative distributionfunction F. then the quantity F(X) has auniform [0, I ] distribution. Many simula-tion packages automatically provide ran-dom numbers from common distributionfamilies such as the normal, Poisson, orexponential. Some packages also allow oneto generate nonstandard distributions.

We often need to model random eventsthat are correlated; for example, assemblytime may be correlated with component-interarrival time if the assemblers are fresherafter a break. Special multivariate tech-niques are needed to generate pairs (or trip-lets, etc.) of correlated random variables.

Barton/Pitts: Simulation in an engineering environment 7

Correlated variables are also useful for im-proving the precision of the simulationresults; well talk more about this later.

Verification / validation

What techniques are used to verify andvalidate simulation models? Some func-tional checks will be unique to particularmodels, but there are others that we mustmake in most situations. The quality ofrandom variates generated for a simulationmay require examination. Random -numbergenerators can have several problems-cycling or other serial dependence, wrongdistributional form, and so on. Statisticianshave developed a battery of tests to checkfor various kinds of misbehavior. Oneshould plot histograms and perform Kol-mogorov-Smirnov tests to check for properdistributional form. An effective check forfirst -order dependence is to plot generatedrandom variates against time and againstone another. Contingency tables (chi-squaretests) are also used to check independence.

For functional tests, one can often testthe model using the following technique:Feed the model by controlled, rather thanrandom, number streams in such a waythat the outcome can be clearly predicted.For example, to test the three -link assem-

could usevalues for a. b, c, d and e that make Xeasy to check, for example.

a-b-c-d-e-1-0. X=2.It is sometimes possible to design an

experiment to test a model's faithfulness toreality. This validation technique suppliesthe same input conditions, or events, toboth the real and the simulated systems,comparing the resulting behavior. Typicallythis is not possible. The most commonway to ensure that the simulation is gener-ating a good approximation to reality is totest it on historical data. One of the bestmethods is to use a random sample of his-torical data to develop the model, andanother sample to verify the model.

Using the model

We have to be careful in deriving infor-mation from simulation data, particularlyfor probabilistic models. In estimating quan-tities like average backlog, average waitingtime, or average throughput, the commonstatistical techniques assume the observa-tions are random and independent. Clearlyfor some models, the time it takes the ithitem to be completed depends heavily onthe completion times for previous items.There are two solutions to this dependence

Authors Pitts (left) and Barton (right) have jointly worked on several projects includ-ing a seminar on statistical design of experiments and a recently completed Corpo-rate Engineering Education course on the design of experiments.

Karen Pitts joined RCA Laboratories in1976 as a Member of the Technical Staffwhere she has been working as an appliedmathematician and statistician. Her simula-tion background includes electrical cir-cuits, production lines, and scheduling, aswell as studying simulation with GeorgeFishman at the University of NorthCarolina.Contact her at:RCA LaboratoriesPrinceton. N.J.TACNET: 226-3045

Russell Barton began work at RCA in 1973.He has worked on simulation models for ICfabrication, communication -satellite relia-bility, and transportation -telecommunica-tion systems. He holds a B.S. in ElectricalEngineering, and M.S. and Ph.D. degreesin Operations Research. Dr. Barton is cur-rently active in statistical optimizationstudies of VideoDisc processes.Contact him at.RCA LaboratoriesPrinceton, N.J.TACNET: 226-2079

problem. One is based on renewal theory,where you group data into batches,called epochs, that are statistically inde-pendent. If you study the information foreach epoch separately, you then satisfy theindependence assumption and can use stan-dard techniques. For example, in the air-line -reservation system, with Poisson arri-vals and exponential service times, the timebetween the empty states (no one waitingin the system, no one in service) is inde-pendent. In this case, you collect all the

information for times the reservationist isbusy and calculate statistics using this data.If it is not possible to identify independentepochs, by renewal theory, another solu-tion is to model the time dependence expli-citly. A field of statistics known as timeseries does this explicit modeling to getvalid estimates. Without independentepochs, the estimates may also be biased ifthey are based on data taken before thesystem reaches equilibrium. It is usual tostart a simulation from an empty and idle

8 RCA Engineer 27-6 Nov./Dec. 1982

situation; but, most real systems do notstart from empty and idle. The simulationsystem must be given sufficient time toreach conditions matching the real system.

Given a simulation in working order,we can use techniques from experimentaldesign to make efficient use of our com-puting resources. Suppose we are model-ing an eight -step production process, andwant to identify the value of adding capac-ity at each of the eight locations. One -at -a -time modification would not identify anyinteractions-the effects of improvementsat one location on performance at otherlocations. On the other hand, a full factor-ial design would require 2', or 256, separ-ate simulation runs. A fractional factorialdesign, well known in statistics, could pro-vide valuable interaction estimates as wellas main effects in perhaps 32 runs.

Statistical techniques can also be usedto reduce the variability of estimates derivedfrom the model. One method, called theantithetic variate method, uses uniform [0-1] random numbers U1, U2, (4, forone stream and (I -U1), (1-U2). . . (1-U0for the second stream. Estimates based onsums of these effects will have reducedvariance. since the terms are negatively

correlated. This follows from the fact thatthe variance of a sum is equal to the sumof the variances plus twice the covariance.and the latter term here is a negativenumber.

When is simulationthe right approach?We have skimmed the surface of severalissues in simulation. The question of wheth-er or not to simulate remains unanswered.Simulation is a powerful took it also re-quires effort, time, and extensive runs on acomputer. When should we use simulation?

Let's review the costs and practicality ofsimulation. Are the data for the modelreadily available? How much will it cost(both in time and effort) to get the data?How complex a model is needed? A sim-ple model will take much less time todevelop, verify, and manipulate than a com-plicated one. How concerned are we aboutcomputer time and costs? Simulations arenot cheap. How important are the resultsof the model? If we are designing a pro-duction facility and find, through simula-tion, that the building should be 10 per-cent longer, the results are very important.

If the model's results are unlikely to beused, think twice before starting to simu-late. Finally, can you wait for the answer?The time to develop, run, and verify themodel and to analyze the results are non-trivial. Conversely, if you need to simulate,you should plan for the time it will take.

Lest all this seems too negative, let'sreiterate that simulation is a powerful toolthat solves problems that cannot be solvedby other means. For understanding com-plex systems, for evaluating new techniques,for teaching complex tasks, simulation pro-vides an excellent alternative to inefficientproduction, invalid algorithms, and expen-sive mistakes. Simulation, although power-ful, is not easy. Simulate, but don't under-estimate the complexity of the task.

BibliographyR.L.I. Lauver and G .1 Wilkinsn. Editors. Robust-ness in Statistics. Academic Press (1979).

2. Fishman. G.S.. Concepts and Methods in McraeDigital Simulation. Wiley 119733

3. Reitman. J.. Computer Simulation Applications. Wiley119713

4. Shannon. R.E.. System Simulation.. The .4rt andScience. Prentice -Hall (1975).

How may I get a copyof an RCA Engineer article?

Call or write to the author

We give authors several reprints of theirown articles. Often they order additional copies.

Contact the Technical Publications Adminis-trator (TPA) in the author's business unit

If the author i5 not available, the TPA oftencan get the paper. The TPA is ready to assistyou, either with a reprint or related material. Acurrent list of TPAs is printed on the inside backcover of the RCA Engineer.

Use your RCA Technical Library

The librarian has all back issues of the RCAEngineer and can give you photocopied articles.

Buy a back issue of RCA Engineer

They can be purchased for $2.00 a copyfrom the RCA Engineer office in Cherry Hill.Write to Dorothy Berry, Bldg. 204-2, CherryHill, NJ 08358, or call TACNET: 222-4254.

Barton/Pitts: Simulation in an engineering environment 9

A. Ashkinazy

The MIMIClogic simulator

MIMIC simulates the behaviorof complex digital circuits. andhelps engineers do it rightthe first time.

Abstract: The author describes the logicsimulation program, MIMIC, used at theSolid State Technology Center atSomerville, New Jersey, and gives thebenefits (reduction of expensive reworkcycles, for example). The needs for logicsimulation and an overview of computer -aided design capabilities at Somerville aregiven. MIMIC is examined in detail byreference to examples, and the networkdescription language, modeling capabili-ties, and run commands to control thesimulation process are covered

MIMIC is a logic simulation programdeveloped by the Design AutomationGroup of the Solid State Technology Cen-ter at Somerville, New Jersey. UsingMIMIC, the engineer can perform logicdesign verification of large -scale -integrated(LSI) and very -large -scale -integrated (VLSI)circuits in a very cost-effective manner,before mask generation, and reduce or elimi-nate expensive rework cycles due to designerrors. Designers can simulate the actualcircuit being implemented, check for races,hazards, or spike conditions, and examinecritical timing of various paths in the logic.

t 1982 RCA Corporation

Final manuscript received October 7 1982Reprint RE -27-6-2

10( 7-00?13A

MODEL

TYPE',lit

7URRENT

MIMIC

)0(BATCH

MODE- 90X

DISCOUNT

)0(

FORFEB23,82

MIMICNEWS,

ENTERPMIMNEk

..EXECUTE

MIMICPROGRAM

...VERSION

: define

ile=fa: get

f

ib

N GETNETWORK

FULL-ADDER

Cotipt6rE

O. D

1 0/

1

12:3

7(6/1,

iG

NETWORK

I

5;NO.

SIGNALS

8;NO.

INPU

f

000001

01001

FIDE FA

hazard:

lyePetermsz7pte

100101

110111

0

)f listca

change:

let

A

C AB

AGR

AIR

If errors are found, MIMIC contains nu-merous debugging aids to help the engi-neer localize problem areas.

Simulation and automated designNot too long ago, integrated circuits con-tained less than 100 gates and flip flopsper chip. It was then possible for the engi-neer to perform a paper design of the cir-cuit with reasonable confidence that anydesign error could be detected and cor-rected before committing to silicon. Thecomplexity of today's LSI and VLSI cir-cuits, containing thousands of componentsper chip, precludes a thorough manualanalysis. Computer simulation provides theengineer with detailed information aboutthe state of each signal in the circuit thatwould be difficult or impossible to obtainby hand. Thus, simulation helps the designerget it right the first time and avoid costlyand lengthy mask -generation cycles. TheSolid State Technology Center in Somer-ville has an impressive track record in thisregard; 21 of the past 22 universal arraysdesigned by the Tech Center (SOS andBulk CMOS) worked the first time. Allwere simulated with MIMIC.

Due to the complexity of LSI and VLSIchips, computer simulation has essentially

replaced the breadboard (which can beviewed as a hardware simulation). Bread-boards of complex digital systems are expen-sive, due to the cost of the hardware, thecost of specialized and sophisticated testequipment for exercising and monitoringthe hardware, and the time required topurchase the components and test equip-ment, assemble the breadboard, and debugthe system. Even then, correct operation ofthe breadboard does not necessarily implycorrect operation of the integrated circuit,since the latter's implementation may differand on -chip delays will certainly differ fromthose on the breadboard.

On -chip delays can only be determinedafter the layout has been completed. TheDesign Automation Group of the SolidState Technology Center is developing thesoftware to close the simulation loop forsemi -custom and gate -array designs. Thesedesign methodologies use standard cellswith fixed geometries, thereby simplifyingthe tasks of automatic placement and rout-ing (APAR), and extraction of the logicalconnectivity and wiring capacitance fromthe computer -generated layout. Standardcell libraries exist, or are being developed,for SOS, CCL, CMOS I, and CMOS II.Figure 1 illustrates the design cycle usingthis software. The designer (an outside cus-

10 RCA Engineer 27-6 Nov./Dec. 1982

tomer or an RCA engineer) generates aMIMIC description of the circuit as aninterconnection of parts that can ultimatelybe resolved into standard cells. After gen-erating functional tests, the designer simu-lates the response of the circuit to the testpatterns. If errors are detected, the designermodifies the circuit and/or test patterns,and iterates until the simulation results arecorrect. From there on, the designer "pushesthe button" and everything else is doneautomatically by RCA software runningon RCA -owned computers. These programsgenerate a finished layout and a simulator -compatible description of the circuit asextracted from the artwork. The designercan then compare the actual circuit de-scription to the one originally simulated,and then simulate the actual circuit usingtrue wiring delays. If there are no errors,masks are made and the chip is manufac-tured. The sidebar on page 13 details theprograms involved.

Structured design

The hierarchical (nested) structure ofCADL, MIMIC's circuit description lan-guage, allows top -down design and bot-tom -up verification. In the design phase,the overall circuit with its connections tothe outside world is at the highest level.Next, the circuit is partitioned into sub -functions, and blocks representing the sub -functions are interconnected. Then, eachsubfunctional block is represented as aninterconnection of smaller subfunctions, andso on, until the subfunctions at the lowestlevels contain only MIMIC primitives (built-in elements; see page 13). Examples of

(a)

A' B7

A B

^0 C I

FADS

A6 86

C7A B

co C.FAD

I COO57 S6

IFA08

(b) C=6 -BIT FULL ADDER

C6

A5

0

0LL

0IAJa.IA

SEMI -CUSTOM CAD SYSTEM

CUSTOMER

LOGICDESCRIPTION

MIMIC

1AUTOMATIC

LAYOUT

TEST PROGRAMGENERATION

RCA

NET DELAYS

PG. MASKS.WAFER FAB

CHARACTERIZE

TEST

FUNCTIONALIC'S

Fig.1. Semi -custom CAD system. This system consists of a number of computerprograms that automate the design cycle from simulation through layout and test -program generation. They help the engineer to minimize errors in all aspects of thedesign process.

lowest level user -defined subfunctions arethe standard cell libraries for semi -custom(APAR) and gate arrays (GUA or AUA).This approach is a top -down design.

In order to debug the circuit, the lowestlevel subfunctions are simulated, and all

65

A B

CO CI

FAD S

S5

B4

VA B

C5CO CI

FAO S

C4

AT B3

t'A B

CO CIFAD S

S3

C3

A2

logical errors and timing problems are cor-rected. Then, the next higher level sub -functions (now containing verified com-ponents) are simulated and debugged. Thisprocedure, iterated until the entire circuitis included, is bottom -up verification.

82

A B

CO CIFAD S

S2

C2

AI 81

VA

CO CIFAD

S 1

CI

AO

1

BO CIN

4

CO CIFAD S

5)_ _ J

TYPE=FAD8 1=A7,A6,A5,A4,A3,A2,A1,A0,137,136,135,B4,133,132,131,60,CIN S PART=F3 TYPE=FAD I=A3,133,C3 O=S3,C40=COUT,S7,S6,S5,S433,S2,S1,S0 PART=F4 TYPE= FAD I=A4,B4,C4 O=S4,C5

PAMT=F0 TYPE=FAD 1=AO,B0,CIN 0=SO,C1 PART=F5 TYPE=FAD I=A5,B5,C5 0=S5,C6PART =F1 TYPE= FAD I =A1 ,B1,C1 0 = S1 ,C2 PART=F6 TYPE=FAD I=A6,136,C6 0=S6,C7PAIIT=F2 TYPE=FAD I=A2,132,C2 O=S2,C3 PART=F7 TYPE=FAD 1=A7,137,C7 0=S7,COUT

Fig. 2. FAD8 circuit of example #1. (a) Decomposition of FAD8 (eight -bit full adder)as iterative array of eight FAD (one -bit full adder) cel's. (b) MIMIC description ofFAD8 corresponding to (a). Figures 2 through 5 illustrate the concept of 'lierarchi-cal design, and the description of this hierarchy to the MIMIC logic simulator.

Ashkinazy: The MIMIC logic simulator 11

(a)

B

MAJ3P.1

FAD

XOR3

(b) C=FULL-ADDER CELL

TYPE=FAD I=A,B,C1 0=S,C0PART=S TYPE=X0R3 I=A,B,C1PART=C0 TYPE=MAJ3 I=A,B,C1

0=SO=CO

CI

Fig. 3. FAD circuit of Example #1. (a) Decomposition of FAD into sum logic (XOR3)and carry logic (MAJ3). (b) MIMIC description of FAD corresponding to (a).

V w

(b)

C=3 -INPUT EXCLUSIVE -OR

TYPE=X0R3 1=U,V,W O=YPART=Z TYPE=EXOR I=U,V 0=ZPART=Y TYPE=EXOR 1=2,W 0=Y

Fig. 4. XOR3 circuit of Example #1. (a)Decomposition of XOR3 into two 2 -inputexclusive -OR gates. (b) MIMIC descrip-tion of XOR3 corresponding to (a).

Example #1: Top -down design of an8 -bit ripple -carry adder

The first step in designing an 8 -bit full -adder is to decompose it into an array ofeight individual 1 -bit full -adders. This isshown in Fig. 2(a). At this point, the inter-nal realization of each 1 -bit adder is notyet known. The MIMIC description of thecircuit at this level is shown in Fig. 2(b).The first statement is a TYPE statement,which declares the beginning of a subnet-work definition. The statement actuallyspans two lines; the dollar sign at the endof the first line is the MIMIC continuationcharacter. The TYPE statement defines atype of circuit called FAD8 having seven-teen inputs (A7-AO,B7-BO,CIN) and nineoutputs (COUT,S7-S0). The next eight state-ments are PART statements that describethe internal components of the FAD8 andtheir interconnections. Each component hap-pens to be a FAD circuit, as yet unde-fined. The first FAD circuit, named FO,has three inputs (AO,BO,CIN) and two out-puts (SO,C1). The second through eighthPART statements describe the rest of theFAD8 circuit in the identical manner.

The serond step in the design procedureis to fill in each 1 -bit adder. Figure 3(a)illustrates this next level of design. The Soutput, representing the modulo -2 sum ofA, B, and CI, is the exclusive -or of thesethree inputs. The CO output, representingthe carry -out, is the majority function of

(b)

C=3 -INPUT MAJORITY GATE

TYPE=MAJ3 I=A,B,C 0=MPART=1 TYPE=AND VALE,PART=2 TYPE=AND I=A,CPART=3 TYPE=AND 1=6,C

ODOM=10=M0=MOrM

Fig. 5. MAJ3 circuit of Example #1. (a)Implementation of MAJ3. (b) MIMIC de-scription of MAJ3 corresponding to (a).

the three inputs (that is, CO is a logical 1

if two or three inputs are logical 1, andCO is a logical 0 otherwise). Figure 3(b)illustrates the MIMIC description of the1 -bit adder. The TYPE statement definesthe FAD cell as having three inputs andtwo outputs. The next two lines are PARTstatements that define the two componentsof the FAD cell as a three -input XOR3cell (that generates the sum, S) and athree -input MAJ3 cell (that generates thecarry -out).

The XOR3 cell and the MAJ3 cellmust be filled in next. These cells areshown in Fig. 4 and Fig. 5, respectively.Since all components in these cells arebuilt-in MIMIC primitives, the design isnow complete.

The wire -tie in the MAJ3 cell, outputsignal M, acts as a wired -OR. Delay infor-mation has been omitted here, but will beincluded in the simulation of this circuit inExample #2 below.

MIMIC's simulation capabilityMIMIC is a four -state simulator. That is,each signal in the simulated circuit will bein one of four possible states: 0 (logical 0),1 (logical 1), X (unknown or uninitial-ized), or HIZ (high impedance or float-ing). MIMIC models the logical operationof each component, in addition to timingand propagation delays along signal pathsof the circuit.

The smallest interval of time in MIMIC

12 RCA Engineer 27-6 Nov./Dec. 1982

is the time -unit. That is, time is modeledas a sequence of discrete time -units, andall events that occur within the same time -unit are considered to be simultaneous.The user is free to scale time -units accord-ing to the technology (for example, equate1 time -unit to 1 nanosecond).

MIMIC rise and fall delays

All signal delays are expressed in time -units, and the user may specify indepen-dent rise and fall delays for each output ofeach element. Either delay (or both) maybe linear functions of loading. Each inputof each element may be individually as-signed a load value, and the loading on aparticular signal is automatically computedas the sum of the load values of the inputpins that the signal drives.

MIMIC spikes and pulses

A spike condition in MIMIC is synony-mous with an attempt to drive a gate'soutput faster than it can respond. Here, thegate begins to respond to a new inputstate, but before the output signal can swingtoo far, a second input change drives theoutput back to its original value. This con-dition is one cause of the ubiquitous"glitch." MIMIC will normally suppress aspike, and the simulated signal will appearclean. Optionally, the user can request prop-agation of spikes as unknown (X) pulses.In either case, MIMIC will report the occur-rence of a spike if the user requests thisinformation.

A pulse condition in MIMIC is theoccurrence of a pulse whose width is com-parable to the average propagation delay,(rise + fall)/ 2, of the signal. Narrow pulseson signals are usually unplanned, and anunexpected pulse could cause problems.Pulses could be regarded as spikes that"made it." In many cases, minor variationof element delays could transform pulsesinto spikes and vice versa. MIMIC reportsthe occurrence of pulses at the user's option.Example #2 (page 15) illustrates spikesand pulses.

MIMIC built-in models

MIMIC supports a variety of built-in, orprimitive, logic elements. In addition tothe basic combinational types (for exam-ple, inserter, AND, NOR, and so on),MIMIC models a two -input multiplexer, afour -input AND -AND -NOR, and a four -input OR -OR -NAND. It models three basictypes of latches and six different edge -trig-

gered flip flops. MIMIC also models sev-eral types of tristate input/output pads thatare particularly useful for test applications.It models complex functional blocks suchas decade counters, ROMs, RAMs, andPLAs. In addition, the bilateral transmissiongate (BTG) is a built-in primitive that mod-els two-way signal flow.

Wire -ties and transmission gates

MIMIC automatically handles wire -ties with-out the user having to insert fictitious ele-ments (as was required in most older sim-ulators). Element ports (for example, theoutputs of several AND gates) are tiedtogether by assigning the same name tothe signals connected to them. Also, dueto the bilateral nature of transmission gates,distinct (differently named) signals couldbe electrically tied together at times. MIMICactually supports three types of wire -ties:wired -AND (any 0 dominates), wired -OR(any 1 dominates), and wire -tie -without -dominance (where oppositely -pulling tiedsignals are recognized and reported as aconflict). Signal M in Fig. 5(b) acts as awired -OR due to the ODOM = 1 entryin the TYPE statement.

MIMIC's simulation algorithm

MIMICs simulation algorithm is efficient,since it only performs computation whensignals change values (event driven), andonly simulates those elements whose inputshave actually changed (selective trace).MIMIC initiates simulation by setting allsignals in the network to the unknown(X) state. The user may initialize selectedsignals to known values, if desired. Thefirst primary input pattern is applied, andthe effects of this input state npple throughthe circuit being simulated at rates deter-mined by the delays of the changing inter-nal signals. When the circuit reaches astable state, the second input pattern is

applied, and so on, until the simulation isterminated either by exhausting the inputpatterns or by the occurrence of a user -specified condition. If the input sequenceis designed properly, and if the circuit isdesigned for testability, the number of sig-nals in the unknown (X) state should de-crease as the effects of the (known) inputpatterns propagate through the logic.

MIMIC run commands

MIMIC's run command language allowsthe user to control the entire simulationprocess. Run commands may be issued

CAD programs and authors

MIMIC is part of an integratedCAD system used at RCA andillustrated in Fig. 1. The programsand their sequence of use aregiven below. Program names arecapitalized and their authors'names are parenthesized.

After verifying the circuit designusing MIMIC, the MIMIC networkdescription is inputted to theCADLM program (David Tsao).The output of this program is theninputted to the MP2D (semi -cus-tom) or the AUA (gate array) pro-gram, resulting in a complete cir-cuit layout. This layout can thenbe inputted to the CONCERTprogram (Joe Mastroianni) whichextracts the logical connectivityfrom the layout. The FASTRACKsystem (Fred Heath, Dick Lydick)currently automates the path fromMIMIC through CONCERT. Thisextracted network descriptioncan then be compared to the orig-inal MIMIC description to verifyequivalence (this step is not yetautomated). The wiring capaci-tance can then be added to theextracted network description(this step is not yet automated)and the reconstructed circuit,now containing actual implemen-tation delays, can be simulated byMIMIC.

If no timing problems exist thecircuit can optionally be faultsimulated by TESTGEN (HenryHellman) to verify the effective-ness of the test patterns. Finally,the TGEN file created by MIMICcan be inputted to the AFTERprogram (Mark Turner) to gener-ate the tester (for example, Sentry,Teradyne) program.

from the terminal (interactive session), ormay be contained in files that are accessedusing the EXECUTE command (interac-tive or batch). Saving predefined run com-mand sequences in files is extremely usefulif these operations are performed repeat-edly (in one or more simulation sessions).The following is a brief overview of thesecommands.

Controlling signal values. The DEFINEcommand allows the user to define pri-mary input patterns hierarchically. It is

Ashkinazy: The MIMIC logic simulator 13

!LOGICAL

C=8 -BIT FULL ADDERTYPE=FAD8 1=A7,A6,A5,A4,A3,A2,Al.A0,B7,B6.135,B4,133,B2,131.130.CIN

0=COUT,S7,S6,S5,S4,S3,52,S1,S0PART=F0PART=F1PART=F2PART=F3PART=F4PART=F5PART=F6PART=F7

TYPE FAD 1 AO,BO,CINTYPE=FAD I=A1.B1,C1TYPE=FAD I=A2,B2,C2TYPE=FAD I=A3,133,C3TYPE=FAD I=A4,B4,C4TYPE FAD I=A5,B5,C5TYPE=FAD I=A6,136,C6TYPE=FAD I=A7,137,C7

O SO,C10-S1,C2O S2,C3O-S3,C40-S4,CSO S5.C6O S6,C70- S7,COUT

C FULL -ADDER CELLTYPE=FAD 1-A,B,C1 O=S,CO

PART=S TYPE=XOR3 I=A,B.C1 0 SPART=CO TYPE=MAJ3 I=A,B,C1 0 CO

C=3 -INPUT EXCLUSIVE -ORTYPE=XOR3 I U,V.W O=Y

PART=Z TYPE EXOR I=U,VPART=Y TYPE EXOR I-Z,W

C 3 -INPUT MAJORITY GATETYPE=MAJ3 I A,B,C 0=M

PART=1 TYPE AND 1=-A,B

PART=2 TYPE AND I-A,CPART=3 TYPE AND 1=13,C

!DELAYDELAY=DEL1D E LAY= DEL2DELAY=DEL4

RISE 1 FALL=1CHANGE 2CHANGE 4

O Z ODEL DELIO Y ODEL DEL4

ODEL=DEL20=M0=M0=M

ODOM 1

Fig. 6. MIMIC description of ripple -carry adder (Example #2)illustrate the use of MIMIC.

Figures 6 through

DEFINE FILE FADGET TYPE FADS FILE:DEFINE PRIPPLE.17= 00000000 00000000 0 $

00000000 11111111 1

APPLY PATTERNS=PRIPPLEWARN FILE: HAZARD COUT,S7,S6,S5,S4.S3,S2,S1,S0WRITE CHANGE: LIST A7,A6,A5,A4,A3,A2,Al.A0,,B7,B6.85,B4,B3,B2,B1.B0,

,CIN,-,COUT,S6,S5.S4,S3,S2,S1,S0 FILE:

Fig. 7. MIMIC run commands for ripple -carry adder (Example #2).

C= AAAAAAAA BBBBBBBB C C SSSSSSSSC= 76543210 76543210 I 0 76543210C= N U

C.0T 1: 00000000 00000000 0 X XXXXXXXX2T 1: 00000000 00000000 0 0 XXXXXXXX5T 00000000 00000000 0 0 XXXXXXX 06T 1: 00000000 00000000 0 0 000000006T 1: 00000000 00000000 0 0 00000000OT 2:00000000 11111111 1 0 0000000051 2:00000000 11111111 1 0 11111000

10T 2:00000000 11111111 1 0 11110000121 2:00000000 11111111 1 0 11100000141 2:00000000 11111111 1 0 1100000016T 2:00000000 11111111 1 1 1000000018T 2:00000000 11111111 1 1 00000000181 2:00000000 11111111 1 1 00000000

S

also used to establish certain defaults. Exam-ple 3 illustrates hierarchical pattern de-scription.

The SET command allows the user toset (initialize) selected signals to specifiedvalues (0,I,X,HIZ). These signal values maysubsequently change in the course of simu-lation. The CLAMP command is similarto SET, except that CLAMPed signals re-main at their specified values until releasedby the user.

The RESTORE command allows theuser to restore the total state of the circuitto a state that was attained at some pre-vious time, either in the same session or ina previous session. The restored state wassaved at that time as a result of MIMIC'sSAVE command. The user has total con-trol over which states are saved (if any),and therefore from which states simulationcan be resumed.

Observing signal values. The PRINT(to the terminal) and WRITE (to a speci-fied file) commands allow the user to speci-fy the signals whose values are to be listed

/0 in the course of simulation. The signalvalues may optionally be listed wheneverany of the selected signals changes state.

The TRACE command allows the userto observe transitions of specified (or all)signals. This command is useful in tracingsignal activity leading to spikes and races.

The LOOK command allows the userto observe selected signal values while simu-lation is suspended. Essentially, the usercan probe signal values while the circuitstate is frozen in time.

The WARN command allows the userto control reports about questionable cir-cuit operation. Events that are reportedinclude spikes, pulses, wire -tie conflicts, andoscillations. The user may specify signalsto be monitored for each type of question-able event.

The TGEN command controls the gen-eration of a file containing the input stateand resulting output state for each inputpattern. This stimulus/response file is com-patible with the output of TESTGEN,RCA's good-logic/fault simulator, and canbe used for test program generation inconjunction with the AFTER program.Optionally, the TGEN command may beused to generate a circuit description filethat is compatible with TESTGEN, if faultsimulation is desired.

Fig. 8. Simulation results of ripple -carry adder generated by MIMIC in response tothe WRITE command (in Fig. 7).

Simulation control. The BREAK com-mand allows the user to conditionally inter-rupt simulation. The user -specified condi-tions that could interrupt simulation include

14 RCA Engineer 27-6 Nov./Dec. 1982

C0A AAA A AA A B B B B B B B U SS XX XXX X

7 6 5 4 3 2 1 0 7 6 5 4 3 2 N 7 55 4 3 2 1 0PAT 4: TIME

1: 0.0E00 <X> <X> X> <X> <X> <X> <X> <X> <X>1: 100.0E-02 I I I

I I <X> <X> <X> <X> <X> <X> <X> <X> <X>1: 200.0E-02 I I I I <X> <X> <X> <X> <X> <7J <X> <X>1:

1:

1:

2:

2:

2:

300.0E-02400.0E-02500.0E-02600.0E-02700.0E-02800.0E-02 7

I I

I I

I

I_r 7

II I

11

I

7 I -r -r 7

I

II

i

i

i

<X><X><3>

<X><X><3>

1

I

<X><X><3>

1

I

<X> <X> <X><X> <X> <3><3> <x> <X>

<X><X><X>

<X><X>

2: 900.0E-02 I2: 100.0E-01

i2,

2:

2:

110.0E-01120.0E-01130.0E-01

I

I -rI__

I I

2: 140.0E-01I I

2:

2:

150.0E-01160.0E-01 I I

2: 170.0E-01 I

2: 180.0E-012: 190.0E-01 712: 200.0E-012: 210.0E-01 -r-2: 220.0E-012:

2:

2:

230.0E-01240.0E-01250.0E-01

-r IrI

I

Fig. 9. Simulation results of ripple -carry adder as plotted by MIME program.

(a) a change in value or attainment of aspecified value at selected signals, or (b)the occurrence of a spike, pulse, wire -tieconflict, or oscillation at selected signals.Whatever the cause of the interrupt, simu-lation can always be resumed, if desired.

Example #2. Simulation of theripple -carry adder

This example illustrates the MIMIC simu-lation of the 8 -bit full adder discussed inExample #1. The MIMIC network descrip-tion file, shown in Fig. 6, contains all thecell descriptions of Figs. 2 through 5. Inaddition, delay information has been addedto the cells at the lowest levels. For exam-ple, the exclusive -or generating U® V inthe XOR3 cell has been assigned an out-put delay (ODEL) DELI, and the secondexclusive -or (generating UO VO W) hasbeen assigned an output delay of DEL4.These symbolic delay names reference thedelay tables contained in the bottom threelines of Fig. 6. Thus, DELI specifies a risedelay of 1 and a fall delay of 1. If the riseand fall delays are identical, the singleCHANGE keyword can be used in placeof the two keywords RISE and FALL.Thus, DEL4 specifies rise and fall delaysof 4, and DEL2 specifies rise and falldelays of 2. Note that the output delay ofthe MAJ3 has been specified as DEL2, sothe carry signal will propagate from stageto stage with a delay of 2 time -units.

The worst -case response time of the rip-ple carry adder occurs when carry signalspropagate through every stage. Thus, if theA -inputs are set to all zeros, the B -inputsare set to all ones, and the carry into theleast significant bit (CIN) is set to one, the

WARN -ON -SPIKE . SIGNAL: SOTIME: 1. TEST: 2. LEVELS: 0 -> 1 -> 0

WARN -ON -SPIKE . .. SIGNAL: S1TIME: 2. TEST: 2. LEVELS: 0 -> 1 -> 0

WARN -ON -SPIKE . . . SIGNAL: S2TIME: 4. TEST: 2. LEVELS: 0 -> 1 -> 0

WARN-ON-PULSE(WIDTH= 1.25) . SIGNAL: S3TIME: 10. TEST: 2. LEVELS: 0 -> 1 -> 0

WARN-ON-PIASE(WIDTH, 1.75) . SIGNAL: S4TIME: 12. TEST: 2. LEVELS: 0 -> 1 -> 0

WARN-ON-PULSE(WIDTH= 2.25) . SIGNAL: S5TIME: 14. TEST: 2. LEVELS: 0 -> 1 -> 0

WARN -ON -P JLSE(WIDTH= 2.75) . SIGNAL: SOTIME: 16. TEST: 2. LEVELS: 0 -> 1 -> 0

Fig. 10. Spike and pulse WARN messages for ripple -carry adder generated byMIMIC in response to the WARN command (in Fig. 7).

correct output state (COUT = 1, all S -outputs are 0) can only occur after carrieshave rippled to the most significant bit andthe sum,. S7, and carryout, COUT, havereacted. Since the carry signals ripplethrough seven stages, at 2 time -units perstage, and since the XOR3 output signalshave a delay of 4 time -units, this willrequire 7 X 2 + 4 = 18 time -units. Mean-while, transient 1 -pulses will appear at theoutputs, with longer pulses occurring atthe most significant bits.

Figure 7 illustrates the run commandsissued to MIMIC, and Fig. 8 shows thesimulation results. Referring to Fig. 7, thethird run command defines a 17 -bit inputpattern (called PRIPPLE) consisting of twotests. The first test pattern sets all seven-teen inputs to 0 and the second test pat-tern changes the B -inputs and CIN to alll's. The WRITE command instructsMIMIC to list all the inputs and outputswhenever any of these signals changes state.

Referring to Fig. 8, time increases verti-cally downward. Signal A7 is listed first,then A6, and so on, then the B -inputs,then CIN, COUT, and the S -outputs. Theoutput signals (COUT,S7-S0) are initiallyin the unknown (X) state at the beginningof the first test pattern. As time goes on,known values ultimately reach all thesesignals. The response to the second testpattern exhibits progressively wider pulsesat the S -outputs of higher significance.

Figure 9 illustrates the same waveformsas Fig. 8, formatted as timing diagrams.These plots were generated by the MIMEprogram, which postprocesses MIMIC out-put files. The pulses on the S -outputs clearlycatch the eye in this representation.

Outputs SO through S2 appear to beconstant 0's, but actually they containspikes. These are shown in Fig. 10, whichlists all MIMIC warnings on spikes andpulses for the outputs in response to theWARN run command (see Fig. 7). The

Ashkinazy The MIMIC logic simulator 15

LOAD

CLOCK

RESET

Fig. 11(a). Three -stage shift register of Example #3.

RESET CLOCK LOAD D4 D2(1) 1 1 1 d d d

(2) 1 0 0 0 0

(3) 1 0 0 0 0 0

(4) 1 0 0 0 0

(5) 1 0 0 0 0 0

(6) 1 0 0 0 0

(7) 1 0 0 0 0 0

(8) 1 0 0 0 0

load the value ddd

disable load control

first clock

second clock edge

third clock edge

Fig. 11(b). Subsequence to test the three -stage shift register.

!DELAYDELAY -GATE -DELAY RISE -2 FALL=1DELAY-FF-DELAY

!LOGICAL

!FORMAT PART

TYPE,SHIFTER

RISE (0,1),(3,4) FALL (2,5),(5,8)

TYPE I 0 ODEL

I RESET,CLOCK,LOAD,D4,D2 ,D1 0=01G4 NAND LOAD,D4 G4 GATE -DELAYG2 NAND LOAD,D2 G2 GATE -DELAYG1 NAND LOAD,D1 G1 GATE -DELAY04 DCF RESET,G4,CLOCK,ZERO 04 FF-DELAY02 DCF RESET,G2,CLOCK,04 02 FF-DELAY01 DCF RESET,G1,CLOCK,02 01 FF-DELAY

Fig. 11(c). MIMIC description of three -stage shift register

first three WARN messages report spikesat these three signals.

Example #3. Parallel -to -serialconverter

The purpose of this example is to illustratehierarchical input -pattern description. Con-

sider the three -stage shift register shown inFig. 11(a). This register contains parallelinputs to the three flip -flop -SET terminals(active LOW), and can be operated as aparallel -to -serial converter by using the se-quence of eight input patterns (assumingthe three flip-flops are initially reset) shown

C DEFINE 65 -PATTERN TEST SEQUENCE'PTEST'DEFINE PSHIFT.6 110000 DO3(100000 110000)DEFINE P000.6 111000 PSHIFTDEFINE P001.6 111001 PSHIFTDEFINE P010.6 111010 PSHIFTDEFINE P011.6 111011 PSHIFTDEFINE P100.6 111100 PSHIFTDEFINE P101.6 111101 PSHIFTDEFINE P110.6 111110 PSHIFTDEFINE P111.6 111111 PSHIFTDEFINE PTEST.6 010000 P000 P001 P010P011 P100 P101 P110 P111C END OF INPUT PATTERN DEFINITIONS

Fig. 12. Hierarchical description ofPTEST, the 65 primary input patterns ofExample #3.

in Fig. I 1(b). Note that the flip-flops arepositive -edge triggered and that logical 0 isshifted into the register as the serial data isshifted out.

If the circuit is to be tested exhaustively,all eight combinations of the parallel datainputs must be applied. This requires 8 X8, or 64. input patterns plus an initialresetting pattern. One way to specify these65 input patterns is to explicitly enumeratethem in a DEFINE statement. Due to therepetitive nature of these patterns, they canbe hierarchically described as follows:

Let PSHIFT be the last seven patternsof the clocking sequence (patterns twothrough eight above).Let Pi be the entire eight -pattern testwhen the three parallel inputs are instate i (1 between zero and seven in-clusive).Let PTEST be the entire 65 -pattern testsequence.Then the ten DEFINE statements shownin Figure 12 completely define the entiretest sequence. PTEST.This example incidentally illustrates sev-

eral aspects of MIMIC's network descrip-tion language unrelated to the above. Re-ferring to Fig. 11(c), the delay called FF-DELAY contains independent rise and falldelays based on loading. Each of thesedelays specifies a pair of points on a straightline representing delay (in time -units) ver-sus loading. Thus, the rise -delay specifica-tion contains a delay of I for 0 loading.and a delay of 4 for a loading of 3. Forany other loading, MIMIC interpolates avalue along the line joining these two points.

A second item illustrated in Fig. 11(d)is the !FORMAT statement that specifies

16 RCA Engineer 27-6 Nov /Dec 1982

the order of keywords in all subsequentPART statements. MIMIC fills -in each suchstatement with these ordered keywords.For example, this !FORMAT statementcauses the first line after the TYPE state-ment to be equivalent to:

PART = G4 TYPE = NANDI = LOAD, D4 0 = G4ODEL = GATE -DELAY

The !FORMAT statement allows consid-erable reduction in the amount of typingrequired to specify the circuit.

Accessing MIMICThis paper has presented an overview ofMIMIC. The MIMIC User Guide containsa complete description of the simulator.and the MIMIC Primer contains good intro-ductory material. Anyone interested in ob-taining a copy of the Guide or the Primer.or using MIMIC, should contact GaryGendel at Somerville (TACNET 325-7399).

AcknowledgmentsMIMIC is the result of a team effort.Henry Hellman. the co-author of MIMIC.developed the simulation routines and the

Aaron Ashkinazy joined RCA in 1970. afterreceiving his Doctor of EngineeringScience degree from Columbia University.At that time. he was responsible for inves-tigating and developing algorithms forautomatic test generation and fault simula-tion. In 1974, he joined the Design Automa-tion Department at Somerville. where hedeveloped the first version of AFTER.which generates tester programs fromstimulus/response patterns. Havingalready written a simulation program priorto joining RCA, MIMIC is his second logicsimulator Dr. Ashkinazy is currentlyresponsible for enhancing MIMIC, particu-larly in the areas of fault simulation andmixed -mode (functional) modeling.Contact him at.RCA Solid State DivisionSomerville, N.J.TACNET: 325-6163

r -

bulk of the run -command processor. ChrisDavis supervised the project and providedthe valuable insight necessary to defineand implement MIMIC's capabilities. Shir-ley Chen has admirably performed theHerculean task of maintaining and extend-ing the MIMIC software. Gary Gendel hasdone an excellent job of introducingMIMIC to the engineering community and

helping users to model their circuits, inter-pret MIMIC runs, and utilize MIMIC effec-tively. He is also the author of MIME.Finally. the author would like to acknowl-edge the contribution by many MIMICusers, particularly Dick Lydick and FredHeath, of valuable suggestions and feed -bad( in debugging MIMIC and making ita more useful program.

How many of theseRCA Engineer issuesshould you have read?

Order then now.

Back issues and reprints listed here maybe ordered from the RCA Engineer Edi-torial Office. 204-2, Cherry Hill.

All back issues are $2.00 except theIndex, which is free.

Photocopy this form, include your name.RCA charge number and your internalmailing address.

THEME -RELATED ISSUES Consumer Electronics Manufacturing

VOL. 24 NO. 4 Dec 78/Jan 1979

MicroelectronicsVOL. 24 NO. 6 Apr/May 1979

Microp-ocessor ApplicationsVOL. 25. NO. 3 Oct/Nov 1979

Quality (Do it right the first time)VOL. 25. NO. 4 Dec 79/Jan 1980

RCA InternationalVOL. 25. NO. 5 Feb/Mar 1980

Colou TV ReceiversVOL. 25, NO. 6 Apr/May/Jun 1980

25th Anniversary IssueVOL. 26. NO. 1 Jul/Aug 1980

Communications TrendsVOL. 26. NO. 3 Nov/Dec 1980

Mechanical EngineeringVOL. 26. NO. 4 Jan/Feb 198'

Increasing Your EffectivenessVOL. 26. NO. 5 Mar/Apr 1981

"SelectaVision" VideoDisc, Part IIVOL. 27. NO. 1 Jan/Feb 1982

Manufacturing EngineeringVOL. 27. NO. 2 Mar/Apr 1982

Electrc-OpticsVOL. 27. NO. 3 May/Jun 1982

CCLLECTIONS OF ARTICLES Consumer Electronics

E-515 1971

Solid State TechnologyPE -552 1972

COS/MOS TechnologyPE -580 1973

Automotive ElectronicsPE -681 1976

Microprocessor TechnologyPE -702 1976

Electro-OpticsPE -703 1976

Picture TubePE -748 1980

CUMULATIVE INDEXi25 -Year Index to RCA Engineer

Aug 1980

Ashkinazy: The MIMIC logic simulator 17

P.A. De Maria K.J.1 Bodzioch

An HF modem simulation

A powerful tool for high -frequency equipment designand evaluation . . . computer simulation and modeling.

Abstract: A modular simulation archi-tecture developed on the VAX 11/780computer models the transmitter andreceiver portions of the AN/USQ-74 multi -tone modem used by the Navy's Anti -Sub-marine Warfare Operations Center. Theprogram structure is examined and func-tional descriptions of the components aregiven.

Wth the resurgence of high -frequency(HF) military communications as a backupfor satellite communication networks, com-puter simulation and analysis is becominga valuable method for evaluating the per-formance of HF communications equip-ment. Simulation offers an alternative tothe expensive method of evaluating equip-ment by field testing, and it also providesfor a common reference for performancecomparison. This is especially true at HFwhere, due to the dynamic behavior of themedia, equipment being compared wouldneed to be tested at the same time, overthe same link, so that the widely varyingeffects of the ionosphere are identical foreach test. The computer allows storage ofsets of standard ionospheric conditions forregular use when comparing differentpieces of equipment.

A result of this interest was a programfor modeling and simulating an existingelement of an HF radio facility, theAN/USQ-74 multi -tone modem used bythe Navy as part of the LINK -11 opera-tions for the Anti -Submarine Warfare Op-erations Center (ASWOC). Figures 1 and2 represent the top-level flow for both thetransmitter and receiver programs.

01982 RCA CorporationFinal manuscript received September 13. 1982.Reprint RE -27-6-3

( "TR ) 0 0G

E GENERATE READ DATAOPEN N REFERENCE FROM INPUTINPUT FILE

RFRAME FILE SET COUNTER

FOR 2 FRAMES

OPEN

OUTPUT FILE

SET COUNTERFOR 5 FRAMES

P

E

AM

GENERATEPREAMBLEFRAMES

SET COUNTERFOR 1 FRAME

SET PHASEREFERENCEPHASES

SET COUNTERFOR 2 FRAMES

SET DATATO START CODE

E

N

F

GENERATESTART CODEFRAMES

YES

NO

SET COUNTERFOR 1 FRAME

SET DATA73 INPUT

E

NF

GENERATEDATAFRAME

EOF = END OF FILE

SET DATATO STOP CODE

G

N

F

GENERATESTOP CODEFRAMES

CLOSEINPUT FILE

CLOSE

OUTPUT FILE

( STOP )

Fig. 1. Top-level flowchart illustrating transmitter functional flow and control codeprocessing.

18 RCA Engineer 27-6 Nov./Dec. 1982

O

RCVR

OPEN

INPUT FILE

OPEN

OUTPUT FILE

STATE 0

SET FREQUENCYCORRECTIONTO -55 Hz

READ 64SAMPLEVALUES FROMINPUT FILE

FREQUENCYTRANSLATE64 SAMPLES

PERFORM64 POINT FFT

COMPUTEAND STOREAMPLITU )ES

E

M0

DEMODU _ATE30 BITDATA WORC

G

P

F

S

Y

N

C

P

H

E

CHECKFOR SIGNALPRESENCE

ATTEMPTTO ESTABLISHFRAME SYNC

OBTAINPHASE

REFERENCE

Fig.2. Top-level flowchart illustrating the receiver's functional flow and controlcode processing.

The simulation architecture, written inBASIC, was developed on a VAX 11/780computer as a series of modular subpro-grams, with each subprogram simulating amajor function of the modem. The simula-tion is divided into two main programs,one performing as the transmitter portionof the modem and the other as the receiverportion.

General descriptionThe USQ-74 can transmit data at two dif-ferent rates: 75 baud per channel or 45.45baud per channel. The transmitter has six-teen separate channels: fifteen channels areused for data transmission and one is re-served for the transmission of an unmodu-

lated Doppler reference tone. All sixteenchannels are summed to produce a com-posite waveform, which is sampled by a12 -bit analog -to -digital (A/D) converter.The computer simulation generates a sam-pled waveform for each channel and thesesampled waveforms are summed to pro-duce sampled values of the compositewaveform.

Binary data is input to the modem ingroups of 24 bits, representing three 8 -bitASCII characters. The modem encodeseach 24 -bit data word with 6 error-detec-tion/correction (EDAC) bits to produce a30 -bit frame for modulation. The dataframes that comprise a message are pre-ceded by header frames, which are pro -

PROCESS ANDSTRIP PARITYBITS

ADD 2 RRORINDICATORS TO24 DATA BITS

dthxd by the modem. The header framesinclude five preamble frames, a phase ref-erence frame, and two start -code frames.The preamble frames contain a Dopplertone used by the receiver for signal detec-tion and a sync tone used to establishframe synchronization.

The start codes are 30 -bit sequencesthat contain no error -detection bits. Afterthe start codes are transmitted, data infor-mation is sent in the following frames. Themodem encodes each 30 -bit data frame byseparating it into bit pairs that modulatethe fifteen transmitted data tones. Depend-ing on the bit pair, the data tones aremodulated in phase by either ±45° or±135°. The transmitter uses a Quadrature

De Maria/Bodzioch: An HF modem simulation 19

111.11

725.11

156.16

575.1111

500.11

425.11

350.00

275.11

210.10

125.00

50.011

-25.00

-100.00

-175.00

-250.00

-325.00

-400.10

-475.00

-550.00

-625.00

-700.01

-775.00

050.00

It,

11\

1..11 tl

88k AAP OF8118888SAMPLE NUMBER

Fig. 3. Plotted sample values generated by the transmitter program for the firstframe of the preamble.

Phase Shift Keying (QPSK) modulationscheme to encode the transmitted data oneach of sixteen separate channels. Dataframes are followed by two stop -codeframes used to indicate the end of amessage.

The receiver uses the header generatedby the transmitter for several functions.The preamble frames are used to deter-mine signal presence and frame synchroni-zation. The phases transmitted in the phasereference are stored by the receiver andused to demodulate the following start -code frames. Both start -code frames mustbe sucrecsfully demodulated for the mes-sage processing to continue. The receiveranalyzes the transmitted information byperforming a Fast Fourier Transform (FFT)on the received frames. The resulting fre-quency components are examined and theiramplitude and phase information are usedto accomplish the above functions.

Transmitter subfunctionsThe USQ-74 modem simulation uses foursubprograms to perform transmitter func-tions. The subprograms are responsible forcomputing sample values for each tone,performing EDAC encoding, generating pre-amble frames, and performing QPSK modu-lation. The four subprograms used for thesefunctions are called GENTN, PREAM,GENFR, and QPSK.

GENTN

GENTN is the subprogram that generatesthe sample values for each tone transmit-ted. The values are produced by executinga table look -up of stored values of a sam-pled cosine wave. The table contains 128sample points of one complete cycle of a55 -Hz cosine. The amplitude values of thecosine range over ±127. Tones that aremultiples of 55 Hz can be reproduced byreading the points from the table in a spec-ified order. By reading every jth point,where j is the harmonic of 55, samplevalues of the j X 55 Hz tone are pro-duced. For example, to produce a 2200 -Hz tone every 40th value would be readfrom the table. When the index of thetable exceeds 128, modulo -128 arithmeticis performed on the index. For this exam-ple, sample values corresponding to tableentries 40. 80, 120, 12, 52, . . . would beread out.

The phase of the tone is produced in asimilar manner. The phase is determinedby selecting the appropriate starting pointin the table from which the points will beread. Since the wave stored is a cosine, areading of the points-starting with thefirst entry in the table-produces a tone at90° phase. To produce the same tone start-ing 180° later in phase, the 64th pointwould be the first point read from thetable. Once the starting point is selectedto determine the phase, the points are

read out using the algorithm describedpreviously.

In the example of the 2200 -Hz tone, toproduce that tone at a phase of 135°, thefirst point read from the table would be16, then 56, 96, 8, . . . . The resultingformula for tone generation can be writtenas

N = (J * M) + P(I)

where N is the entry to read from thetable; J is the harmonic being generated;P(I) is the starting phase of the /th tone (Iequals 1 to 16); and M equals 0 throughthe number of samples/frame.

As each tone is produced, a runningsum of the amplitude values for each sam-ple point is produced. Depending on therate, either 94 or 155 composite amplitudevalues will be formed by adding the cor-responding sample points for each of the16 tones. A preamble -frame -waveform gen-erated by this method is shown in Fig. 3.

PREAM

The PREAM subprogram generates thepreamble frames that precede each mes-sage. The preamble frame consists of twotones, the Doppler tone at 605 Hz and async tone at 2915 Hz. The USQ-74 gen-erates five preamble frames for each datatransmission. The sync tone is generatedusing tone 16 so that it is advanced inphase by 180° at the preamble frame bound-aries. This phase discontinuity of 180° ateach frame boundary allows the receiverto establish frame synchronization.

GENFR

The GENFR subprogram performs twofunctions. The first function involves encod-ing the 24 -bit data words with six Ham-ming -code parity bits. The second functioninvolves writing the amplitude values ofthe composite waveform to the output file.

QPSK

The QPSK subprogram simulates differen-tial phase -shift -keying modulation of theinput data. The 30 -bit frame from GENFRis broken into 15 -bit pairs, each pair cor-responding to one of the data tones. Thebit pairs are then examined to determinethe modulation phase. The modulating phaseis then added to each tone's previous phasevalue, which has indicated the phase ofthe tone at the end of the last frame. Thisproduces a phase shift at the frame bound-ary equal to the modulating phase whenthe tone is generated during the next frame.

20 RCA Engineer 27-6 Nov./Dec. 1982

Receiver functional descriptionThe receiver portion of the USQ-74 modemcan operate in several different modes. Itcan demodulate either an upper or a lowersideband signal alone or use the two side -bands to form a diversity signal. The re-ceiver demodulates the input signal(s) byperforming an FFT on a frame of sampledata. The frame is either 94 or 155 sam-ples long, depending on the modem rate(75 or 45.45 baud).

The receiver does not use all of thesample points contained in a frame fordemodulation. The abrupt phase changesthat occur at frame boundaries make it anundesirable place to look at the signal.Therefore, the receiver examines only thecenter portion of each frame.

The two frame rates are processed slight-ly differently. At the fast rate, the central64 points are used to perform the FFT.This leaves a guard band of 15 points oneach side. When the slow rate is beingprocessed, the central 128 points of the155 sample point frame are used. Thisleaves a guard band of 13 points at thebeginning of the frame, and 14 points atthe end. In the slow case, two 64 -pointFFTs are performed, instead of a 128 -

point FFT. The first FFT uses the first halfof the 128 points; the second FFT, thesecond half. The results of these two trans-formations are added together and dividedby two to produce a single set of 64 fre-quency components. These components arethen processed in the same manner as inthe fast rate.

The receiver performs the following fourfunctions for signal demodulation: (I) itestablishes that a signal is present beforedoing any further processing; (2) it pro-vides frame synchronization; (3) it estab-lishes the proper reference phases; and (4)it decodes the received signal into a binarystream. In addition, the receiver is respon-sible for various decision -making processes,as well as support processing. Support pro-cessing includes performing a Fast FourierTransform and error -detection -and -correc-tion decoding. The four basic functionshave been segmented into subprograms.Each subprogram performs a specific func-tion and passes information back to themain program to be used for furtherprocessing.

Signal presence is determined by theSIGPR subprogram. SIGPR uses the Dop-pler tone present in the preamble framesto determine if a signal is located in thereceived data. The amplitudes of the fre-quency components of the signal locatedabout the Doppler tone are compared to

Fig. 4. Window position when a minimum occurs in the DFT value, indicating aframe boundary.

the amplitudes of the frequency compo-nents located in the rest of the band. If thepower al the Doppler frequency is greaterthan that in the rest of the band, a signal isassumed to be present.

Once a signal presence is detected, theSIGPR subprogram performs a coarse Dop-pler -shift calculation. A Doppler shift ofthe received spectrum may occur if eitherthe radio transmitter or receiver is locatedon a moving platform (for example, on anaircraft). Instability in any of the frequencydevices used in the modulation/demodula-tion or in frequency translations in receiverand transmitter may also contribute to afrequency shift. This coarse correction isbased upon how much energy in the Dop-pler tone has "spilled over" into adjacentfrequency components. The sidelobe withthe larger amplitude indicates the directionin which the Doppler shift occurred.

Once signal presence has been detected,the receiver must sync to the incomingsignal. The FSYNC subprogram performsthis function in addition to calculating afine Doppler correction. Having stored thefrequency components of the Doppler tone,FSYNC computes the phase difference ofthe Doppler tone in two successive frames.This difference is the amount the phaseadvanced over 64 sample points or 9.09milliseconds (64 samples, at 7040 samples/second). The Doppler frequency correctiontherefore equals the change in phase overthe 9.09 milliseconds or

fi = = (0, - 02)1(9.09 X 10-')

where 0, is the phase of the Doppler inframe 1.

After calculating a fine Doppler correc-tion, FSYNC begins processing to locatethe frame boundary. FSYNC uses the 180°

phase change of the sync tone betweenframes in locating the frame boundary. Itlocates the frame boundary by using aDiscrete Fourier Transform (DFT) of the

sample points. Because the phase of thesync tone changes by 180°, sample win-dows of the DFT, which have an equalnumber of points of the sync tone at 0°phase and 180° phase, will produce a min-imum value for the amplitudes of the synctone. This condition occurs when the DFTwindow straddles a frame boundary, asshown in Fig. 4. FSYNC, in performingthe DFT, slides the sample window onepoint at a time, using a recursive formulato compute the DFT. By keeping track ofthe location of the DFT window, andknowing its position when a minimum isproduced, the program locates the frameboundary.

The PHREF subprogram is responsiblefor detecting the phase -reference frame.PHREF also stores the frequency compo-nen-s for each reference vector. These vec-tors will be compared with the vectors ofthe next frame. The phase differences be-tween the vectors in successive frames con-tain the information necessary to allow thedecoding of the data tones.

Message decodingThe data information contained in a mes-sage can be decoded by the DEMODsubprogram once the phase -reference framehas been found, and the reference vectorsstored. Each of the 15 data tones containstwo bits of information. The exact bit pairtransmitted for each tone is determined bythe phase difference between the presentdata vector and the corresponding vectorfrom the previous frame.

Rather than examining the real and imag-inary parts of the frequency componentsfor each tone to determine the phase ofeach, and then the phase difference betweenthe two, a complex -conjugate vector mul-tiplication is performed between the refer-ence vector and the present data vector.The resultant vector of the multiplication

De Mana/Bodzioch: An HF modem simulation 21

Paul De Maria, Communications SystemsEngineering, Government CommunicationsSystems, Camden, New Jersey, receivedhis B.E. in Electrical Engineering from CityCollege of the City University of New Yorkin 1979. Mr. De Maria joined the SystemsEngineering group upon graduation. Whileat RCA, he has been involved in develop-ing packet switching and circuit switchingsystems. He was also involved in thedevelopment of a radio relay network for aforward area tactical operation. He is cur-rently working in the area of the analysisand simulation of communications systemsusing digital signal -processing techniques.Contact him atGovernment Communications SystemsCamden, N.J.TACNET: 222-4634

Ken Bodzioch, Manager, Digital Commun-ications Systems Engineering, GovernmentCommunications Systems, Camden, NewJersey, received a B.S. in Physics from theUniversity of Massachusetts in 1970, andan M.S.E. in Systems Engineering from theUniversity of Pennsylvania in 1977. He hasalso studied mathematics at the Universityof Mississippi and the University ofNebraska. Mr. Bodzioch joined RCA in1974 after four years in the U.S. Air Forceas a communications officer. He has beendeeply involved in the analysis, design,and implementation of message-, circuit-,and packet -switched communications sys-tems, high-speed signal -processing sys-tems, and classified information-process-ing systems. Most recently, he has beenresponsible for the mathematical modelingand simulation of an end -to -end HF com-munications link and a high -resolutionlaser -beam recording system. Mr.Bodzioch taught the original microproces-sor for logic design course in the RCA After -Hours Study Program.Contact him atGovernment Communications SystemsCamden, N.J.TACNET: 222-3850

is stored in the P array. This operation isshown in the equation below.

P(k) = FO*(k) X(k)

where F0*(k) is the complex conjugate ofthe reference vector, X(k) is the presentdata vector; and represents complex multi-plication.

Because the phase for each frame ismeasured several sample points "in" fromthe beginning of a frame, the phase of theresultant vector of the complex conjugatemultiplication must now be corrected bythe amount the phase has advanced fromthe end of one sample window to thebeginning of the next. The resultant vec-

tors P(k) are phase -corrected, using theequation shown below, and stored in theZ array. The equation is

Z(k)= P(k)X e ISk

where 0 is the advance in phase and k isthe number of the harmonic of the datatone.

After phase correction, the vectors canbe used for decoding the received data.The complex -conjugate multiplication isequivalent to a translation of the axis. It is,therefore, possible to use the resultant vec-tor to determine the phase difference bynoting which quadrant the resultant vectorlies in. This quadrant of the resultant vec-tor is determined by the sign of its realand imaginary components.

The bit pairs are decoded using the fol-lowing algorithm. If the real part of theresultant vector is positive, the first bit ofthe pair is a 1, otherwise it's a 0. If theimaginary part is positive, then the secondbit of the pair is a 0, otherwise it's a 1.

The receiver then performs error detec-tion and correction by using the six paritybits at the end of each frame. The Ham-ming code is capable of double -bit detec-tion and single -bit correction. A two-bitstatus word, indicating which of the aboveconditions occurred, is appended to thedemodulated word.

Diversity processingThe modem program also simulates the di-versity processing of the receiver. The USQ-74 receiver can operate in two diversitymodes. One diversity mode forms a coher-ent combination of the upper and lowersideband signals after they have been indi-vidually corrected for Doppler shift. Theother diversity mode carries all the sig-nals-upper-sideband, lower-sideband anddiversity-through to demodulation. Aftera frame has been demodulated on all threechannels, the receiver determines the framethat will be used as output by examiningthe error -status bits for each word. Thediversity frame is examined first, then theupper-sideband frame and, finally, the lower-sideband frame. If all of the channels arefound to have an uncorrectable number oferrors, the diversity word is chosen foroutput.

SummaryThe simulation program was written torepresent the operation of a specific modemused in HF radio transmission. The simu-

22 RCA Engineer 27-6 Nov./Dec. 1982

0.1

0.01

0.001

10

SIMULATION RESULTS

1

0-5 10 15 20 25 30

Fig.5. Simulation results compared with theoretical performance in the presenceof Gaussian noise.

lation results correlate well with standardtheoretical results. Figure 5 shows the simu-lation performance when Gaussian noise isadded to the simulated signal. The simula-tion serves three purposes: (1) It offersreduced cost and effort in evaluating mo-dem performance; (2) it can be used as aconfidence -building test bed prior to mak-ing expensive changes on the existing equip-ment: and (3) its modular design allowsaddition or deletion of selected modules torepresent performance of other similar Link11 modems without a major modeling andredesign effort.

AcknowledgmentThe authors would like to express theirthanks to Dr. T.T.N. Bucher whose helpin this project as well as in past efforts hasbeen invaluable.

Have you "given a stroke" to your TEC lately?Do you know that engineers at RCA are devoting timeand effort to help you in establishing and maintaining asatisfying career? They are your Technical ExcellenceCommittee and they attempt to contribute to a stimulat-ing working environment by determining your needs inthe areas of technical education, technical information,professional activities and recognition. Once the needis established, they move on to develop and executesupportive activities.

Most of you are somewhat aware of and use thetechnical excellence activities. But have you paid yourdues by expressing your appreciation to your TECrepresentative? Or even better, do you tell him yourviews on today's engineering climate and the careerneeds you perceive?

Engireers are highly skilled professionals, but all toooften they keep their thoughts and needs to themselves,perhaps because they are busily engaged solving theirproblems. Technologies, techniques, and engineeringtools-the entire supporting environment is constantlychanging and better ways to do engineering work areevolving. Do you take the time to discover and masterthem? This is where the TEC is attempting to supportyou-you can help the TEC help you, by sharing yourthoughts with the TEC members. Then give them creditfor their efforts, every once in a while.

De Maria/Bodzioch: An HF modem simulation 23

G.M. Sparks 1J. Liston

Dynamic analysis and simulationof mechanically scanned radar systems

For complex systems problems with no direct and easilymanaged solution, digital simulation often affords theonly practical analytical approach.

Abstract: The methods by whichmechanically scanned radars are dynami-cally analyzed through digital computersimulation are summarized Three distinctareas of concern are addressed as exam-ples (1) Dynamic modeling of the radarand its antenna drive as the mechanicalinteraction of a viscous damped spring -mass system; (2) The implementation ofdigital filters for servo compensation; and(3) Detection analysis as the radarantenna beam scans past a target. Thesimulation methods that are presented arenot restricted to the analysis of radar sys-tems, but can readily be applied to otherfields.

System analysts and designers make exten-sive use of digital simulation as a means ofpredicting the dynamic performance of com-plex systems. These systems are subjectedto mechanical and environmental distur-bances that, along with design constraints,impose fundamental performance lim-itations.

Simulation usually is not warranted forrelatively simple systems that readily yielda direct mathematical solution for the vari-ables of interest. For complex systems, how-ever, a tractable mathematical solutionmay not exist, or may be so cumbersomethat its utility is compromised. In this situa-tion, digital simulation of the system be-comes a virtual necessity; simulation is espe-cially valuable in conserving engineeringtime and resources.

1982 RCA CorporationFinal manuscript received September 29. 1982Reprint RE -27-6-4

Fig.1. Mobile radar mounted on a tank (above) and its dynamic model (opposite).

Three areas for simulationThis paper illustrates a number of specifictechniques that are useful in simulating theperformance of any complex system. Wehave selected a mechanically scanned radarsystem as an example of a system of mod-erate complexity, with three distinct areasof concern facing the designer. Problemsin each of the areas cited below have beensolved by means of digital computer simu-lation. The approaches are sufficiently gen-eral to be directly applicable to the solu-tion of related problems in other areas.

The first technique is the digital repres-entation of a mechanical system. A com-

plex mechanical body (the radar, antenna,and pedestal) is modeled by springs, masses,and viscous dampers with the objective ofsimulating the dynamic response as theradar tracks maneuvering targets in noise,clutter, and multipath. The simulation tech-niques account for the effect of platformmotion of the mount to which the radar isattached, and allow for external torquedisturbances (for example, wind, gun fir-ing, and vehicle motion) at designatedpoints within the system. These analysesand simulation techniques have been usedto identify potential design problems andtheir solutions early in the design phase

24 RCA Engineer 27-6 Nov./Dec. 1982

81

D1A

62

MOTORDAMPING D2

83

AZIMUTHANTENNA

TAA

YOKE ETC.

RADAR MOTORROTOR + YOKE

RADAR

TRA MOTOR+4,---"". TORQUE

J3A RADAR MOTORSTATOR

D3A 1 K3A BASE"

J4A

D4 A

05

GUN

TGA

) K4A UPRIGHTS

J5A

D I5A

'96

MOTORDAMPING D6

D7A

J6A

TURRET

_i TJA

K5A DRIVES

GUN MOTORROTOR

TMA TORQUEGUN MOTOR

VEHICLE

TVA

SUSPENSION

when corrective action can be accomplishedwith relatively minor schedule and costpenalties.

The second technique is the implemen-tation of digital filters. In modern radardesign, many functions previously per-formed with hardware are now executedwith software. One of these functions isthe compensation filtering for the servosused to drive the antenna in azimuth andelevation. Two examples are presented thatshow the importance of using simulationto solve problems that arise during systemtests. The simulation model used in thesetests precipitated a solution to specific prob-

lems associated with servo -noise transientsand also with inherent antenna drive rateand acceleration limits. The simulation al-lowed these problems to be solved "off-line" during critical periods when the actualsystem was not available.

The third technique is target detectionusing a mechanically scanned radar. Animportant element of radar design is adefinition of system parameters that willprovide a satisfactory probability of detec-tion. Simulation is a useful tool to estimatethe ability of a mechanically scanned radarto detect a target with specified character-istics in a specified environment. Methodsare presented that allow this capability.

Dynamic analysis methodology torepresent a mechanical systemThe first example of simulation as a tool-digital representation of a mechanical sys-tem-was used to model a fire controlsystem consisting of a radar mounted onthe breech of a gun. The gun and its drivesystem are mounted on a full -track vehi-cle. The vehicle, its gun, and the radar areshown in Fig. 1. Each axis of this fire con-trol system is modeled dynamically as aseven -mass system. The elevation -axis mod-el is similar to the azimuth -axis modelshown on the right of the figure. Each axiscontains both radar and gun -drive subsys-tems, with external torque disturbances in-troduced at each mass.

This complex system, with each axisdescribed by seven interacting, second -order

differential equations, can be solved as aset of N simultaneous, linear, constant -coefficient differential equations with Mtime -varying inputs. The constant coeffi-cients of the differential equations are orga-nized into an N X N matrix A, whichdefines the compliance, inertia, and vis-cous damping of the system elements.*The M time -varying inputs are representedas an N X M matrix B, which defines themanner in which the M -forcing functionsare coupled to the system.' The resultingdifferential equations are written in matrixform (see box, equations (1) and (2)).

To simulate this procPss on a digitalcomputer, we assume the input to be con-stant over a sample period T In the result-ing equation, the state variables that con-stitute the vector X are known at time kT,and the input vector M is constant within thetime interval from time kT to (k + 1)T.The resulting equation is given in the boxas equation (3).

The computational steps in calculatingthe a and $ matrices are rather complex;however, with widely available library rou-tines for matrix operation (solutions foreigenvalues, eigenvectors, and matrix inver-sion), their implementations in a simula-tion is straightforward. Also, because theirelements are invariant with time, they needonly be computed once and stored for useas required. Subsequent iterations of equa-tion (3), which involve only multiplica-

Matrices and vectors are given in bold face.

Mathematical description of a complex mechanical system

X(t) = AX(I) + Bol

2

LIm

A

aNI .

The solution to equation (1) is:

X(t) = exp [A(t - X(to)+1,+

B

a "VI ...

exp [A(t - r)] Bu (r) dr (2)

X [ (k + 1) T] = aX(k7) + pm (k7)Where a = exp (Al), and N X M is a matrix with constant elements

p= [a - I] 4t-' B, and N X M is a matrix with constant elements

I is the identity matrix

(31

Sparks/Liston: Dynamic analysis and simulation of mechanically scanned radar systems 25

n

°n EoDIGITALDRIVE

Ein

A/D COMPENSATIONFILTER

D/ASYSTEM

LIMITER LIMITERG1 (S) G2 (S)

Fig. 2. Block diagram of a digital servo -drive system.

tions and summation, are solved efficientlyin a digital computer.

In solving matrix a, the first step in-volves determining the eigenvalues A1, A2,. . . , AN of the matrix A. Then the N -element eigenvector Z, needs to be deter-mined for each eigenvalue A,. The diago-nal matrix A can then be formed:

A

e

Guillemin

Bush

0

Ein

E,,,

0

eA vi

(4)

Bin = INPUT

Bo = OUTPUT

n - NOISE

The N X N matrix, F, whose columnsare the eigenvectors of A, is formed as:

F = [Z1, Z2, . . . ZN] (5)

Finally, exp (AT) is calculated, using asimilarity transformation that involves invert-ing the F matrix. With successive multi-plications,

a = exp (AT) = FAF-1

with a calculated, the matrix $ can bedetermined, as indicated in equation (3).This step requires inverting the A matrix,which must be non-singular.*

These two matrices provide the basisfor solving the set of given differentialequations for a sequence of time -varyinginputs. Once the equations are solved, thesystem outputs are determined by simple

ST1 + 1I I

I 1KST +1-2 L _J

sT2,1L______Fig. 3. Alternative methods of implementing digital filters.

K(ST1 + 1) I

multiplications from a set of stored co-efficients.

Digital implementation of servocompensation filters

A typical digital servo -drive system is repre-sented by the block diagram in Fig. 2.Nonlinear effects occurring in the servosystem include signal, rate, and accelerationlimiting. Signal limiting occurs in the A/Dand D/A converters; rate and accelerationlimiting occur in the drive system. Com-puter simulation is used to determine theeffects of such limiting as well as the effectsof noise on system performance.

For a second -order (Type II) servo, acompensation filter is required to providethe specified performance and ensure sta-ble operation. The transfer function of thisfilter is

GI(S) = E(S)/ E,(S)= K(STI + 1)/ S(ST2 + 1) (7)

To implement G, (S) in a digital com-puter it is necessary to use a state -variablerepresentation. There is no unique set ofstate -variable equations for a given transferfunction. Discussions of some forms ofparticular interest are found in references1, 2, and 3. This paper is concerned withtwo methods,' one known as Guillemin'sform and another known as Bush's form.tThese are shown in Fig. 3 and are, hence-forth, referred to as Method 1 and Method2, respectively.

The integrators in either of the twomethods can be digitally implemented byvarious approaches, the bilinear transfor-mation being a frequently used method.'Each of these two methods were used fortwo system designs, the HR -76 radar servoand the NIDIR servo; these designs requireddifferent state variable representations be-cause of their different requirements andconditions. Specifically, the HR -76 radarwas required to provide a means of select-ing different servo bandwidths in order tooptimize performance in the presence ofnoise and varying target dynamics. In thecase of the NIDIR radar, a single servobandwidth was specified; however, unlikethe HR -76, there was the requirement toaccommodate severe antenna -scan -rate andacceleration condition& Simulation was use-ful in selecting the best filter design foreach system.

Special techniques exist that can he applied when the Amatrix is singular.

Also referred to as the Standard Controllable form in ref-erence I.

26 RCA Engineer 27-6 Nov./Dec. 1982

HR -76 radarservo -noise transients

Method 1 was initially used to implementthe compensation filters for the HR -76 ra-dar, which is a system designed and devel-oped at RCA Missile and Surface Radar.Moorestown, New Jersey. During the sys-tem tests, a transient error that occurredwhen switching from a wide to a narrowservo bandwidth was noticed. To deter-mine the cause of this problem, the servoclosed -loop system was simulated on thedigital computer. The simulated drive -sys-tem transfer function was:

G2 (S) = 1 /S [(S/ (00) + 1] (8)

The parameters for the compensation filterand the drive system transfer functions,Gi(S) and GAS) respectively, are given inTable I.

Table I. Filter and drive -system param-eters.

Parameter BW 4 Hz BW 1 Hz

K 28.8 2.56

0.374 1.25

T 0.0064 0.08

30 30

With E, = 0 (noise input only), the tran-sient effect was reproduced when switch-ing from a 4 -Hz to a 1 -Hz servo band-width. The result is shown in Fig. 4.

The filter implementation in the simula-tion was changed to Method 2. Using thesame input -noise sequence and same sys-tem parameters, the bandwidth was againswitched from 4 Hz to 1 Hz. This tran-sient error, also shown in Fig. 4, is muchsmaller than that obtained using Method1, leading to a design change in the HR -76 system.

Even though these two approaches haveidentical transfer functions, they do nothave identical state equations. At the timethe bandwidth is reduced, a non -zero valueof E2 exists. For Method 1, this valueimmediately after switching is the same asthat immediately prior to switching becauseit is maintained as the initial condition onthe integrator. For Method 2, however,the value of E2 decreases at the switchingtime because of the reduction of the KT,product.

NIDIR designation servo -limiting effects

Another system design that was aided bysimulation was the NIDIR-designation ser-

METHOD 2

0 1 2 3 4 5 6 7 8

TIME (sec)

Fig. 4. Servo output with noise only as the input

vo. This servo did not have a bandwidth -switching requirement but needed to ac-quire targets under conditions that wouldreach both system -rate and accelerationlimits during the transitional slew to themoving target. In order that the value ofE0 and El in Fig. 3 not reach extremelylarge values during saturation times, it wasnecessary to limit the output of the inte-grator described by K/S to a value equalto the rate limit. Otherwise, the integratoroutput during saturation periods could reachvalues, that could cause loss of track. Thesystem of interest had rate, acceleration,and error limits of respectively 0.5 rad/sec, 0.5 rad/sec,' and 0.4 rad. The prob-lem was to acquire a constant velocitycrossing target with the conditions given inthe inset of Fig. 5.

Systems using both Methods 1 and 2were simulated and tested, the trajectorybeing defined as in the inset of Fig. 5. Thevalues of K, T1, and T2, are, respectively,4, 1, and 0.0625. Plots of the servo error,as a function of time for each system, areshown in Fig. 5. Note that Method 1 inthis case is far superior to Method 2.

For both methods, the output of theintegrator, K/S, is limited. In Method 1,the input to the integrator senses the rateof change of the error, E, and causes theintegrator to come out of saturation assoon as the error starts to decrease. InMethod 2, however, the integrator input

9 10

does not change polarity until the error,E,, changes polarity. The integrator doesnot come out of saturation until the errorchanges polarity. This results in a largeovershoot.

As discussed, simulation is a valuableanalytic tool for selecting design approaches,not only from the two examples presentedbut for the design requirements that mayarise in the future.

Target detection simulationThe process of detecting a radar return asan antenna beam scans past a target canbe modeled as illustrated in Fig. 6. In thisfigure. the radar returns are represented asa time sequence of signal -amplitude sam-ples whose statistics reflect a specific rmssignal-to-noise ratio (S/N),), assuming thatthe target is located on the peak of theantema pattern mainlobe. The amplitudeof the signal samples vary from pulse topulse or from scan to scan depending uponthe type of target that is modeled. Figure 7shows a variety of radar -fluctuating targetmodels that are commonly used as a basisfor specifying radar detection performance.

Simple algorithms can be used in con-junction with random -number generatorsto simulate the desired target signal -returnfluctuations. Of the models shown in Fig.7, the amplitude fluctuations from thosemodes that have a suffix I orIII are com-

Sparks/Liston: Dynamic analysis and simulation of mecha lically scanned radar systems 27

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

-0.1

-0.2

-0.3

-0.4

-0.5

-0.6

METHOD 1

-07 I I I 1

-2 -1 0 1

METHOD 2

ORO = INITIAL RADAR Az = 30°

°TO = INITIAL TARGET Az = -45°

V = TARGET VELOCITY

RM = MINIMUM CROSSING RANGE

V/RM = 0.5

I 1

2 3 4 5 6 7 8

TIME FROM CROSSOVER (sec)

Fig. 5. Servo error as a function of time. Inset represents initial target and radargeometry.

pletely correlated within any given scanand are independent from scan to scan.Those models that have a suffix II or IVresult in amplitude fluctuations that areindependent from pulse to pulse as well asfrom scan to scan.

SINGLE SCAN TRIALSEQUENCE OF NINPUT SIGNAL SAMPLES, Si

A

ROUND TRIP BEAMSHAPE MODULATION

Csm (0.866 TOi/OB1 2

0.866 wOi/OB

The detection performance, which isachieved with a sequence of N constantrms-amplitude target returns processed witha square -law detector and post -detectionintegrator (as shown in the shaded blocksof Fig. 6), can be analyzed directly with-

VI;1 PHASE AND

OUADRATURE PHASE /fi NOISE ADDITIONlllllA, S, 6102 +zo?

NO

YES TARGET STILLWITHIN MAIN LOBE

OF ANTENNA

Fig. 6. Radar receiver/signal processor simulation.

out resorting to Monte -Carlo simulation.In the event a mechanism is present thatmodulates the level of the target returns,such as the gain pattern of a scanningantenna beam, a rigorous theoretical analy-sis becomes very cumbersome and Monte -Carlo methods can be presented as a vi-able alternative.

Figure 8 illustrates the degree to whicha simple 2000 -trial Monte -Carlo simula-tion of a detection process, involving thepost -detection integration of 4 pulses (exclu-sive of beam shape modulation effects),approaches the theoretical' detection per-formance of a 4 -pulse integrator. For thesecases, the threshold level is analyticallycalculated using the method of Pachares,'which assumes a square -law detector modelof the form:

Y = (1/2)x2 (9)

where x = the input voltage level

y = the output voltage level.

The threshold curves of Fig. 9 show thethreshold voltage variation with the numberof pulses integrated for different values offalse alarm probability. These values arenormalized with respect to rms noise andare directly applied to the threshold deci-sion block of Fig. 6.

Detection performance achieved by theentire process outlined in Fig. 6 has beenanalyzed by means of a 5000 -trial Monte -

SQUARE LAWDETECTION

A.2

YES

DECLARETARGET NOT

DETECTED

NEXT SCAN TRIAL

SLIDING -WINDOWPOST DETECTION INTEGRATOR

SHIFTREGISTER

N VARIATESINTEGRATED

28 RCA Engineer 27-6 Nov./Dec. 1982

STEADY TARGET

SPHERE

SWERLING 1, II

JET AIRCRAFT IIIPROPELLER AIRCRAFT WI

SWERLING III, IV

STABILIZED MISSILETANKS, RANDOMLYORIENTED PROLATE SPHEROID 11111.UNSTABILIZED MISSILE TANKS (IV)

LOG NORMAL CASE I, 11

SHIPS, RANDOMLY OPIENTEDFLAT PLATE

CHI SQUAREN DEGREES OF FREEDOM

AIRCRAFT, CERTAIN FREQUENCYDIVERSITY RADAR RETURNS

=i 2 log p

S= C S C log x = log (C21- a 2/2=

N

Zi2j=1

log u, S = C J0.5 109 lu,ud

S expUr, 4- log K

2

C RMS SIGNAL-TO-NOISE AMPLITUDE RATIOS

ui,ui- SAMPLE OF SIGNAL AMPLITUDE SAMPLES FROM UNIFORM DISTRIBUTION OVER

SINGLE SIGNAL SAMPLE THE INTERVAL 0 TO 1TO RADAR RECEIVER SIGNAL zi, SAMPLES FROM NORMAL DISTRIBUTION, ZEROPROCESSOR SIMULATION MEAN, UNITY VARIANCE

Fig. 7. Simulation of return signal for target fluctuation models. p MEAN TO MEDIAN RATIO, A PARAMETEROF THE _OG NORMAL DISTRIBUTION

Carlo simulation, and the results are plot-ted in Fig. 10. For each trial, the beam(modeled as a sin x/x one-way amplitudepattern) scans past the target, and eighttarget returns are integrated in a slidingwindow integrator. This is a process inwhich the eight most recent target returns,modulated by the round-trip antenna pat-tern are: (1) contaminated by in -phase andquadrature-phase components of whiteGaussian noise; (2) square -law detected;and (3) accumulated as a summation.

During this process, the beam positionand subsequent pattern loss are varied frompulse to pulse to reflect the antenna scanrate. After each target return is received,the summation of the last eight processedreturns are tested to see if the detectionthreshold is exceeded, in which case theMonte -Carlo trial is said to have resultedin a detection and the trial is completed.Otherwise, the beam advances to the nextangular position in the sequence, a newreturn is accepted, and the process isrepeated.

In the event no threshold crossing occursby the time the main lobe of the antennascans past the target, a missed target decla-ration is made and the Monte -Carlo trialis completed. For the cases shown in Fig.10, 5000 trials were carried out for eachdata point on the curves, with the detec-tion probability calculated as the ratio ofdetections to Monte -Carlo trials.

Figure 10 also shows the results for anideal rectangular beam -shape pattern, inwhich eight pulses are integrated withoutthe influence of an antenna pattern loss.The difference between the curves is thebeam -shape loss for a scanning radar, whichis seen to be in the neighborhood of 0.7 to0.9 dB, depending on the SNR and target -

99

98

95

90

STEADY TARGET4 PULSES INTEGRATED10 6 FALSE ALARMPROBABILITY

80

70 -

60 -50 -40

306 7 9 10

IS/N), dB

12

Fig. 8. Ccmparison of theoretical detection probability values with that obtainedfrom a Monte -Carlo simulation.

fluctuation model. The results presentedare based on an antenna -scanning rate suchthat the sliding -window integrator receiveseight pulses as the antenna scans through asector corresponding to 85 percent of the3 -dB one-way beamwidth.

Blake and Trunks show this to be anoptimum integration window length in thesense of minimizing the SNR required fora specified detection probability. The resultscompare with a pattern loss of 1.6 dB,which is frequently cited as a typical beam -shape loss in a scanning radar. The 1.6 -dB

value applies strictly to a gate positionedso that the pulses modulated by the antennapattern are accepted by the gate only withinthe gated angle centered on the beam maxi-mum 7.

For the cases simulated in this paper,the sliding window (or gate) effectivelyscans completely through the region ofmaximum antenna -gain -modulated targetreturns, thereby affording additional detec-tion opportunities. While the noise com-ponent of the post -detection integrated sig-nal -plus -noise variates is highly correlated

Sparks/Liston: Dynamic analysis and simulation of mechanically scanned radar systems 29

200

150

100908070

60

50

40

30

20

10

DETECTOR MODEL y -1-x22

FALSE ALARM PROBABILITY

1 1 1 1 1 1 1

2 3 4 5 6 7 8 9 10J I I I 111_120 30 40

NUMBER OF PULSES INTEGRATED

50 60 70 80 100

Fig. 9. Theoretical detection threshold values for square -law detector.

99

98

95

90

80

70

60

50

40

30/

.--/ /20 / /10

STEADY TARGET

/Z> /

0*

SWERLING II TARGET/16b0-/

SWERLING ITARGET

0.8dBe

'BEAM SHAPE LOSSSINGLE SCAN PASTTARGET

- RECTANGULARBEAMSHAPE

- - SIN X/XBEAMSHAPE

3 4 7 8

IS/N10 dB

10 11 12

Fig. 10. Monte -Carlo detection analysis of scanning -radar results averaged over5000 trials. Eight pulses are integrated with a false alarm probability of 10-6. Forsin x/x beam shape, the 8 pulses are integrated within a sliding window that spans85% of the one-way, 3 -dB beamwidth. Uniform weighting is applied within thewindow.

within the scan, there is still a modestimprovement in detectability to accountfor the reduction in beamshape loss.

ConclusionIn addition to solving these specific prob-lems, computer simulation was used in1970 to simulate the entire AN/MPS-36

instrumentation radar.' I0 The realism andthe modular nature of the simulation en-abled a wide variety of experiments andstudies as follows:

1. Investigation of the capabilities of theexisting radar system to support plannedfuture missions.

2. Isolation of those radar parameters that

may limit the radar's performance capa-bilities and that are the most likelycandidates for future modification.

3. Assessment of the improvements in radarperformance that might be achieved fromvarious proposed radar modifications.

4. Comparison of the existing/modifiedAN/MPS-36 radar's performance andcapabilities with the performance thatcan be achieved with other existing/newtracking systems.

5. Study of the effects of radar locationupon the mission -supporting capabilitiesof the radars.

6. Post -mission data analysis whereby ap-parent radar breakdowns/problems canbe simulated to check on the validity ofthe analysis results.

7. Integration of the present simulationprogram as a building block in a largermulti -station, multi -instrument, trackingnetwork. A larger. system such as thiscould be used to determine an opti-mum tracking configuration for a spe-cific mission.

The simulation of the AN/MPS-36 radarwas successful and has been used exten-sively by the government. These uses haveincluded general studies and investigationsof the radar's operation so as to improveunderstanding of the complex interactionsthat occur within the complete trackingsystem (the radar, the target, the environ-ment, and so on).

To summarize, we have presented thesolutions to three important problems re-lated to mechanically scanned radar sys-tems. Specific examples have been pre-sented that demonstrate how to use thesetechniques to solve radar design problems.The approaches to these problems, how-ever, are general and not restricted to radar -related problems.

ReferencesI. M. Athans and D.L. Falb, Optimal Contro4 pp. 173-

190, McGraw-Hill Book Co. (1936).

2. C. Chen and J.J. Hass. Elements of Control SystemAnalysis, pp. 298-307. Prentice -Hall Inc. (1968).

3. J.W. Brewer, Control Systems Analysis Design andSimulation. Chapter 8, Prentice -Hall, Inc. (1974).

4. W.D. Flyer and W.C. Schultz. "A Survey of Methodsof Digital Simulation of Control System," CornellAeronautical Laboratory, Inc.. Report XA-1681-E-1 (July 1964).

5. R.L. Mitchell and J.F. Walker, "Recursive Methodsfor Computing Detection Probabilities," /EEE Trans-actions on Aerospace and Electronic Systems. Vol. 7,

No. 4 (July 1971).

6. J. Pachares. "A Table of Bias Levels Useful in

30 RCA Engineer 27-6 Nov./Dec. 1 982

Radar Detection Problems," IRE Transactions onInformation Theory, Vol. 4 (March 1958).

7. L.V. Blake. "The Effective Number of Pulses perBeamwidth for a Scanning Radar." IRE Proceed-ings, Vol. 41 (June 1953) and "Addendum." IREProceedings. Vol. 41 (December 1953).

8. G.V. Trunk. "Comparison of Two Scanning RadarDetectors: The Moving Window and The Feedback

Integrator." IEEE Transactions on Aerospace andElectronic Systems (March 1971).

9. AN/MPS-36 Radar Modeling and Computer Simu-lation Final Report Contract F04701 -69-C-0131Subtask 7.3.3 (September 1971).

10. Refinement and Modification of the AN/MPS-36Radar Computer Simulation Program ContractF08606 -72-C-0031 (June 1974).

Glenn Sparks has been associated withMSR since 1954. During this period he hasbeen engaged in key radar system engi-neering assignments emphasizing control -system analysis, computer simulation,command and control, and estimationtheory.Contact him at:RCA Missile and Surface RadarMoorestown, N.J.TACNET: 224-3037

Jack Liston has held major systemsresponsibility for concept design, perfor-mance analysis, and design optimizationon advanced phased -array radar systemdevelopments. Some of the areas of hisprincipal contributions include systemcomputer modeling for cost optimizationand performance trade-offs, advancedself-duplexing receiver concepts, and sys-tem performance verification throughsimulation.Contact h,m at:RCA Missile and Surface RadarMoorestown, N.J.TACNET: 224-3037

Sparks/Liston: Dynamic analysis and simulation of mechanically scanned radar systems 31

J. Golub

Discrete -event simulationsof complex weapon systems

Using proven simulation principles, engineers can easily vary aplethora of parameters and quickly gain information aboutweapon -system behavior.

Abstract: Discrete -event simulationsmodel system functions as event subpro-grams occurring at discrete times. Thismodeling process yields a structure suitedto the representation of complex weaponsystems. An elementary example demon-strates the principles. The author describesthe initial program developed by the Simu-lation Group at Missile and SurfaceRadar, called MEDUSA, and an advancedprogram derivative He continues with de-scriptions of techniques being used todevelop a new simulation, and concludeswith a note on potential uses in modelingother complex systems such as air-trafficcontrol or transportation planning.

Modern weapon systems contain manycomplex components, not the least of whichare digital computers that control opera-tions via real-time software. The variablesand options available in designing and de-ploying these systems present a bewilder-ing array of possibilities to systems ana-lysts. Classical analyses cannot provide allthe necessary answers in the required time.However, with a well -designed, large-scaledigital computer simulation of the weaponsystem, engineers can readily explore dif-ferent system designs and deployments.

For the last eight years, the System Simu-lation group in the Naval Systems Depart-ment of Missile and Surface Radar has beendesigning and developing discrete -event1982 RCA Corporation

Final manuscript received September 28 1982Reprint RE27-6-6

simulations of complex weapon systems.These simulations, written in FORTRAN,have been applied to many problems inareas such as system performance, algo-rithm design, evaluation, testing, surviva-bility, and requirements specifications.

MEDUSA (Multi -target Effectiveness De-termined Under Simulation for AEGIS),the AEGIS Operations Analysis Simula-tion, was the initial program developed bythe Simulation Group. MEDUSA has beenin use for over seven years to aid AEGIStactical algorithm designers, predict AEGISWeapon System performance against com-plex raids, and support tests at the CSED(Combat System Engineering Development)site. MEDUSA is a discrete -event simula-tion that models significant AEGIS func-tions-from search, detection, and trackthrough threat evaluation, weapon sched-uling, and intercept in defending surfacefleets against air attacks. The success ofMEDUSA led to the development of dis-crete -event simulations of other weaponsystems. To obtain useful results quicklyfrom these new simulations, their originalversions were developed by adding newmodels to MEDUSA and removing mod-els unnecessary for the new simulations.

SIMATR (Simulation of Integrated Mul-tiple Advanced Tactical Radars), the mostsuccessful MEDUSA derivative, models theaspects of Tactical Air Control Systems(TACS) that detect and track hostile air-craft and missiles and that control fightersto intercept these targets. With emphasison a network of ground radars assisting

airborne interceptors defending WesternEurope, SIMATR contributes to the deter-mination of the relative military worth ofphased -array and rotator radars.

A current Independent Research andDevelopment (IR&D) program to developa discrete -event simulation of a naval bat-tle group provides an excellent opportun-ity to implement the techniques learnedand developed from several years expe-rience. Since the scope of this simulationmakes it impractical to generate the initialversion by modifying MEDUSA, a PROTO-type of a Naval Simulation (PROTONS)has been selected as the vehicle to deter-mine development methods, user inter-faces, and level of detail for the battle -group simulation.

Elements ofdiscrete -event simulationsA discrete -event digital computer simula-tion represents a system by event subpro-grams that model system functions, inter-actions between the system and its environ-ment, and actions in the environment. Theevents that model the system are chosenand designed to achieve a sufficient degreeof realism without using excessive compu-ter time. Since these events occur at dis-crete times, the state variables that de-scribe the components and interactions ofthe system change only at these times andnot continuously as they would in the realsystem.

Computations are performed and logi-

32 RCA Engineer 27-6 Nov./Dec. 1982

INPUT/INITIALIZATIONSECT ON

REMOVE LISTSPECIFYING NEXTEVENT TO BEEXECUTED

ADVANCESIMULATIONTIME TO TIMEOF THIS EVENT

CALL EVENT SUBPROGRAMCORRESPONDINGTO THIS EVENT

SIMULATIONCOMPLE TED

Fig. 1. Operation of the event sequenc-er, the executive subprogram that con-trols simulation cycling after initialization.

cal decisions are made when significantevents are expected to occur rather than atevenly spaced time intervals. This yieldsefficient use of computer time as eventsare executed only when their results areneeded. The operation of these event sub-programs is controlled by an executivesubprogram called the event sequencer thatinteracts with the event calendar, a time -ordered list of scheduled events.

After the input is read and the initialevents are placed on the calendar, controlis transferred to the event sequencer, whichdrives the simulation by the following steps(Fig. 1):

1. Remove the top event from the eventcalendar.

2. Advance the simulation time to thetime of this event.

3. Call the event subprogram correspond-ing to the removed event.

4. Repeat the above steps until the eventcalendar is empty or simulation time isgreater than a specified simulation stoptime.

A hypothetical shipboard antiair missilesystem (sidebar, pages 34-35) illustratesthe above concepts (refer to Fig. 2 for agraphic description of the example).

Fig. 2. An aerial view of a surface-to-air missile ship defending an area againsttwo aircraft, one that is attacking the ship and another that is passing through thedefended area. The missile coverage zone is represented by the circle, and thepoints indrcate where the two targets enter and leave this coverage. This examplewas chosen to illustrate discrete -event simulation principles.

MEDUSA, the AEGIS OperationsAnalysis SimulationMEDUSA (Multi -target Effectiveness Deter-mined Under Simulation for AEGIS), adiscrete -event simulation, was designed andimplemented by RCA's AEGIS SystemSimulation Group to support the devel-opment of the AEGIS Weapon SystemMark 7, which performs the antiair-war-fare functions of the AEGIS CombatSystem.

The AEGIS Weapon System is com-posed of nine computer -directed elementsthat can stand as a fully automatic antiairand antisurface missile system. As part ofthe Combat System, the AEGIS WeaponSystem performs the principal air and sur-face defense functions. The control elementsof this group provide track maintenanceand threat evaluation, as well as weaponassignment and control for all warfare oper-ations. It is the AEGIS Weapon Systemthat furnishes direction, commands, andautomation to the AEGIS Ship CombatSystem.

Work began on MEDUSA in early 1975when a simulation was needed to predictthe performance of the AEGIS WeaponSystem against complex multi -target threats.A discrete -event architecture in FORTRANwas chosen because: (1) AEGIS and itsinteraction with the environment can eas-ily be represented by discrete events; (2)the subroutines described above were avail-able; and (3) discrete -event simulations hadbeen successfully used to model other weap-on systems.

The example in the sidebar may beviewed as an extreme simplification of theAEGIS Weapon System and analogously,the sample simulation described is an ex-tremely simplified version of MEDUSA.Some of the important differences betweenthe two simulations indicate the vast dis-parity in complexity. In MEDUSA, forexample, a three-dimensional curved -earthgeometry is used; targets are modeled bymulti -leg trajectories; radar detection is mod-eled; the availability of shipboard resources(Continued on page 36)

Golub. Discrete -event simulations of complex weapon systems 33

Hypothetical antiair missile system example

A ship with a surface-to-air missile sys-tem attempts to intercept enemy airtargets entering its missile coveragezone but can engage only a singletarget at a time. The targets fly straighttrajectory legs at constant speeds andthe ship's missiles fly at an averagespeed, V over their effective range,Rm. It is assumed that the ship is awareof all targets in its coverage zone. Themissile system is controlled by theship's Combat Direction System(CDS) computer to fire at the closesttarget. The CDS computer executes aprogram every 10 seconds to determinethe closest target. A missile is launchedonly if an engagement is not in prog-ress. An engagement lasts from missilefiring time to intercept. Missiles aregiven a 65 -percent chance of suerPcsfulintercept, and a target hitting the ship isgiven an 80 -percent chance of destroy-ing it.

To simplify this example, two-dimen-sional (x, y) planar geometry will beused. The ship will be stationary andpositioned at (0,0). A target trajectorycan be specified by:

initial position x y,final position x1, yt

initial time t,

speed S

Therefore, its position at any time (t),is computed as:

x(I) = x, + - ts]y(t)= y, + ift - t,] (1)

where k is the target velocity in the xdirection and j, is the target velocity inthey direction and S is (5c2 ,2)i 2.

The position of the target is a continu-ous function of time, but the simulationwill compute the position only at thediscrete times when requested by anevent subprogram.

The times, ,,,,,, a target enters andleaves the missile coverage zone (a cir-cle of radius Rm) is computed by solv-ing the quadratic equation for the inter-section of the straight-line trajectoryand the coverage circle (see eqn. 2) .

The time, t, a missile intercepts a targetis computed from the intersection ofthe missile and target trajectories (seeeqn. 3) .

Although this system is a simpleone, it is a good example to demon-strate the principles of discrete -eventsimulation, and the seven events de-scribed in Table I could be used tomodel it.

The preceding principles of simula-tion operation can easily be put intopractice if subroutines and techniquesexist to:

1. Represent events.

2. Represent a calendar of events.

3. Schedule events on the calendar.

4. Cancel events from the calendar.

5. Access events on the calendar.

Events are represented by five -wordblocks of memory called events lists,with words numbered -1, 0, 1, 2, and3, as shown in Fig. 3.

Word #

-1 # Words in block (5)

0 Index of next event list

1 Event #

2 Event time

3 Data for event subroutine

Fig. 3. A list representing a simulationevent. These lists are linked to the eventcalendar when the corresponding eventsare scheduled. They are removed whenthe event is executed or cancelled.

Words -1, and 0 are for bookkeepingand calendar linkage. Word 1 containsthe event number, word 2 contains theevent time, and word 3 is available to

pass information to the event subrou-tine. Word 3 could contain the targetnumber for the target, enter, leave, at-tack, launch and intercept events in theexample.

The calendar is a set of time -orderedlists that represent events that havebeen scheduled but not executed. Apackage of subroutines that create listsfrom a dynamic storage array and thatschedule, cancel and access events isused to handle simulation bookkeepingfunctions. Table II shows the eventsequencer for the example.

Other available subroutines are use-ful for maintaining and acrP%ing theinternal database that dynamically de-scribes the status of the simulated enti-ties and their relationships. Althoughthese subroutines are not necessary forsimulation, they provide an extremelyflexible method of maintaining aninternal database and they allow therepresentation of virtually unlimitednumbers of entities of different types.

This package of subroutines is writ-ten in FORTRAN and is used daily onthe MSR DEC 2060 computer.' Ver-sions are also available for the IBM370/168, CDC 6700, Honeywell, andUNIVAC computers. These simula-tion -management and list -handling sub-routines and the techniques for usingthem are similar to GASP IV,2 a simu-lation language that consists of FOR-TRAN -callable subroutines. These sub-routines were selected at MSR overGASP because of their immediateavailability in 1975 and the subsequentsuccess with them.

[ -(x,1 + yti,) ±N. /(x,i + - S2( R,2 - R.2),,i = (2)S2

where, R, = (x,2 .01/2

t = -(xik + y15') (x + jLi')2 - (S2 - V.2) R12(S2 - V.2)

= R112( + y

v,2

1s2 = v72

where xi, y, is the target position at launch time, R,2 = xr2 + yi2, and V, ismissile average speed.

(3)

34 RCA Engineer 27-6 Nov./Dec. 1982

Table I. Events for hypothetical antiair missile system example.Events Name/# Scheduling conditions Function

TARGET/1 For each target at Computes the time the target enters and leaves missile coverage (eqn. 2) andthe time the target schedules ENTER and LEAVE events at these times. f a target is within missile cover -trajectory begins age at the beginning of its trajectory, ENTER is scheduled at this time. If a target tra-

jectory ends within the missile coverage, a LEAVE event is scheduled at the end oftrajectory time. For a target attacking the ship, an ATTACK event is scheduled atthe time the target arrives at the ship.

ENTER/2 By TARGET for each Makes the target eligible for engagement by the missile system. The first ENTERtarget entering missile event executed schedules CDS, the command and decision event.coverage

LEAVE/3 By TARGET for each Cancels all events for this target. No further computations will be madetarget leaving missile for this target.coverage

ATTACK/4 By TARGET when the Assesses the attack by comparing random number (rA) with 0.80. If rA .0.8, the shiptarget arrives at the is destroyed and the simulation ends. If rA >0.8, the attack failed.ship

CDS/5 The first CDS is Prioritizes the targets in missile coverage in order of distance from the ship andscheduled by the first schedules LAUNCH for the closest target if an engagement is not in progress.ENTER. Thereafter,CDS schedules itselfon a 10 -second cycle.

LAUNCH/6 By CDS for each tar- Computes intercept time (eqn. 3) and schedules INTERCEPT at this time.get to be engaged atlaunch time

INTERCEPT/7 By LAUNCH at inter- Assesses the intercept by comparing a random number (ri) with 0.65. If ri 0.65, the

cept time target is killed and all events for the target are cancelled. If ri >0.65, the missile missedthe target and the target is eligible for re -engagement by CDS.

Table II. Event sequencer to control cycling of simulation of hypothetical antiair missile system.

FORTRAN statement Comment

5 ICAL = 0CALL GETCAL (ICAL, $100)CALL CANCEL (ICAL)TIME = 0 (ICAL,2)*IF (TIME .GT. TSTOP) GO

TO 100J = Q °CALM*GO TO (10,20,30,40,50,60,70),J

10 CALL TARGET (ICAL)GO TO 5

20 CALL ENTER (ICAL)GO TO 5

30 CALL LEAVE (ICAL)GO TO 5

40 CALL ATACK (ICAL)GO TO 5

50 CALL CDS (ICAL)GO TO 5

60 CALL LAUNCH (ICAL)GO TO 5

70 CALL INTCPT (ICAL)GO TO 5

100 RETURNEND

INITIALIZEGET ICAL, INDEX OF FIRST EVENT, ON CALENDARREMOVE EVENT LIST FROM CALENDARUPDATE SIMULATION TIMEIS SIMULATION TIME GREATER THAN SPECIFIED

STOP TIME?GET EVENT NUMBERGO TO CALL STATEMENT FOR EVENT J

CALL THE SPECIFIED EVENT AND GET THE NEXTEVENT LIST ON THE CALENDAR.

ALL EVENTS HAVE BEEN EXECUTED OR TIMEGREATER THAN TSTOP

to store event lists and other lists that describe system entities.*0 is the dynamic storage array used

Golub: Discrete -event simulations of complex weapon systems 35

(Cone. from page 33)

is considered when scheduling targets forintercept; and realistic engageability andmissile time -of -flight data are used. Al-though the two simulations differ in com-plexity, the discrete -event architecture is

suitable for both. The selection from themany AEGIS functions to be used as MED-USA events was dependent on the intendedMEDUSA uses. Some of the events usedto model AEGIS Weapon System func-tions and interactions with the environ-ment are briefly described in Table III.Other MEDUSA events model SPY -IAradar tracking; the Standard Missile launch.mid -course guidance. illumination, and inter-cept; the Combat Air Patrol target inter-cept; and the PHALANX close -in weaponsystem.

Although MEDUSA was developed pri-marily to demonstrate AEGIS perfor-mance against complex raids, the discrete -event architecture permitted a wider rangeof applications. Some of these are listedbelow:

I. Evaluation of tactical algorithms for:

Standard missile engageability Threat evaluation Missile scheduling Variations in missile salvo doctrine Electronic countermeasures algorithms

2. Evaluation of combat system configu-rations such as:

Casualty modes Alternative illuminator positioning

3. Design of tests at the CSED site and atthe production test center

4. Verification of system qualification tests

5. Determination of new missile -system re-quirements

6. Demonstration of performance ofAEGIS derivative systems.

The versatility and portability of thesediscrete -event simulation techniques is at-tested to by MEDUSA operation on theUNIVAC SPECTRA 70. IBM 370/168.DEC system 2020, 2040, 2060, and CDC6700 computers. Naval Surface WeaponCenter personnel, who were able to mod-ify and use MEDUSA in less than fourmonths.' demonstrated that these tech-niques were easy to learn.

The original version of MEDUSA mod-eled a limited number of AEGIS functionswith nine events. The open-ended natureof the architecture allowed new events tobe added easily and quickly as they wererequired for new analyses and studies. Thecurrent version of MEDUSA includes 20events and the evolutionary nature of theprogram indicates that more will be addedin the near future.

SIMATR (Simulation of IntegratedMultiple Advanced Tactical Radars)

In addition to its value to the AEGIS pro-gram, MEDUSA has been used as thenucleus for simulations of other weaponsystems. These simulations were started byremoving events unnecessary to the model-ing of the new system and adding eventsrequired to model the new system. Thistechnique made it possible to make a work-ing simulation of a subset of the new sys-tem operational in a short time.

The most successful MEDUSA deriva-tive is SIMATR,' an air -battle simulationof USAF Tactical Air Control Systems(TACS). SIMATR demonstrates how anetwork of phased -array ground radars canincrease the effectiveness of airborne inter-ceptors in repulsing an enemy air attack.This increased effectiveness is realized inthe direct control and designation of eachairborne interceptor by extending its effec-tive range to provide the interceptor astandoff shot with an air-to-air missile inthe initial engagement phases. RCA neededa TACS air -battle simulation to evaluatethe effectiveness of phased -array radars, toassess C' system loading, and to determinethe other resources needed to wage an airbattle.

Many of the models required forSIMATR were derived from MEDUSA.Indeed, all of the MEDUSA architecturaland simulation management features were

Table III. Selected MEDUSA events that model AEGIS functions and interactions with the environment.Event Function modeled Event operationDETECT Response of SPY -1A

to the return of a sur-veillance pulse froma target.

Computes probability of detection (Pd) as a function of radar parameters, target radarcross section, target position, multipath, anc the position and power of all simu-lated jammers. The psuedo-random number rD is generated from a uniform distri-bution. If r0 < Pp, the detection is successful and TRACK is scheduled. If rD>PD, thetarget is not detected and DETECT is scheduled to occur one radar scan time in thefuture.

TEWS Execution of the threat -evaluation (TV) andweapon -selection al-gorithms by the AEGISCDS computer.

Computes TV for all tracks as a function of target indentification, position and velocity.The targets are prioritized by TV. Targets beyond a specified range from the AEGISship are selected for engagement by Combat Air Patrol and CAPINT events arescheduled at the expected intercept times. An engagement -order transmission issimulated for targets engageable by SM-2 by entering them in the Engagement OrderQueue (E00). TEWS schedules itself cyclically.

ENGAGE Execution of the SM-2scheduling algorithmby the AEGIS WCScomputer.

Targets in the EGO with the highest threat values are placed in the smaller pre -engagement queue (P/0). A target will be scheduled for SM-2 engagement if alauncher is available within its launch window; an illuminator is available for thehoming period; and the maximum allowable missiles in flight will not be exceeded fromlaunch through intercept.

To determine the above conditions, there are models to compute earliest andlatest launch times (launch window): solve the fire control problem for SM-2; computethe homing interval; maintain periods of launcher and illuminator availability; and deter-mine launcher and illuminator tie-up times. A successful schedule for a target con-sists of a launcher available for the required tie-up time and an illuminator avail-able for the homing interval. A list of successful scheduled engagements is main-tained and ENGAGE schedules LAUNCH, the SM-2 launch event, at the scheduledlaunch times.

36 RCA Engineer 27-6 Nov./Dec. 1982

directly applicable to a TACS simulation.In addition, the MEDUSA target -genera-tor, jamming, radar, threat -evaluation, CAP -assignment, and CAP -intercept models ful-filled the corresponding TACS modelingrequirements with minor modifications.Thus, after the MEDUSA Standard Missile,PHALANX, and SPY -1 ECM logic modelswere removed, the MEDUSA architectureand a subset of the MEDUSA modelswere sufficient for elementary TACS analy-sis. In order to perform more extensiveanalyses, models were implemented for sur-face-to-air missile systems, short-range de-fense systems, attrition of interceptors, andnetted multiple radars.

Although SIMATR models distributedweapon systems and sensors described byanalytic algorithms, whereas MEDUSA rep-resents the shipboard AEGIS Weapon Sys-tem controlled by tactical algorithms, thegenerality of the discrete -event techniquesmade it possible to derive the initial ver-sion of SIMATR from MEDUSA in lessthan two months.

Simulation of a naval battle groupIn 1982 an lR&D project was undertakenat MSR to develop a digital computersimulation of two opposing naval battlegroups. The Operations Analysis and Sys-tem Simulation groups combined to pro-vide a tool to: determine the effectivenessof new weapon systems in a battle -groupenvironment perform comparative analysesbetween existing and new ships; evaluatefleet -level tactics and doctrine; and deter-mine the effects of different types of com-munications and coordination on battle -group performance. These capabilities willdirectly support follow-up AEGISprograms.

A survey of existing battle -group simu-lations developed by government agenciesand private industry showed that most ofthem used a discrete -event architecture.Some used special simulation languages,but most were written in FORTRAN. Thisconfirmed our decision to write the battle -group simulation in FORTRAN using dis-crete -event architecture.

The battle -group simulation provided anopportunity to apply the techniques devel-oped over the past seven years. The crea-tion of a simulation to accomplish thegoals of a complete battle -group simula-tion is an enormous undertaking, and it isdifficult to decide where to begin and howto proceed. A prototype simulation wasthe vehicle chosen to evaluate developmentmethods, input formats, levels of model

PROTONS

Fig. 4. A sample PROTONS output showing five ships being attacked by threegroups of aircraft. The smaller circles represent coverage zones and the largercircles represent radar coverage.

detail, combinations of warfare areas, andoutput formats.

PROTONS (PROTOtype of a NavalSimulation) was initiated in early 1982and will evolve into a comprehensive bat-tle -group simulation of two opposing bat-tle groups waging air, surface, and subsur-face warfare. Extensive models of commu-nication and coordination will be included.

A key simulation technique successfullyevaluated with PROTONS was the use ofan event -sequencer generator program. Theevent sequencers used in MEDUSA andSIMATR were created manually andchanged manually when events were addedand deleted. This approach involves re-moving or deleting subroutine calls andadding or deleting arrays defining and cor-relating event names and numbers for inter-nal simulation use. Changes are requiredin a few subprograms and although it is adirect procedure, it is prone to error as aremost manual operations. However, this pro-cedure is ideal for automation via a com-puter program, GENSIM.

A new PROTONS event sequencer isgenerated by the GENSIM program when-ever an event name is added or deletedfrom the names of all PROTONS eventsin a disk file. Although. GENSIM waswritten for PROTONS, it is usable for asimulation of any system and quickly pro-vides an executive subprogram for simula-tion control.

A program has been developed to createinput subroutines from a disk file describ-ing input variables. It allows the additionof input variables and default values to thesimulation without programming. This pro-gram is applicable to simulations of anysystem and for any programs that requireseparate input subroutines.

Other software techniques being evalu-ated via PROTONS are increased use ofgraphics for both simulation input and out-put. The most useful SIMATR andMEDUSA outputs are multicolor plots show-ing the results of key system events, andarea maps showing trajectories and facili-ties. Figure 4 is the initial PROTONS

Golub: Discrete -event simulations of complex weapon systems 37

Jerry Golub is Unit Manager of the SystemSimulation and Data Reduction group inthe Naval Systems Department at Missileand Surface Radar in Moorestown. Hisgroup designs and develops simulations ofthe AEGIS Weapon System and writesdata -reduction programs to analyze AEGIStests. SPECTRM, the AEGIS dynamic simu-lation, and MEDUSA, the AEGISoperations analysis simulation, havebeen important system engineering toolsthroughout the AEGIS Program. Currently,a major group project is the design anddevelopment of a battle group simulation.Contact him at:Missile and Surface RadarMoorestown, N.J.TACNET: 224-2426

Jerry Golub (right) reviews simulation graphics produced by a member of the simu-lations group.

graphic output showing ship position, radarand missile coverages, and the trajectoriesof attacking air targets. The battle -groupsimulation will be the simulation group'sfirst program that will allow users to createinput by interacting with data on the screenof a graphic CRT terminal.

PROTONS provides an effective meansof evaluating and exercising the simulationtechniques learned and developed over thepast eight years. The ultimate models tobe included in the final version of the bat-tle -group simulation have not been deter-mined, but the discrete -event architecture,and the tools developed to facilitate theiruse, ensure that this program will easily bedeveloped to meet simulation needs.

ConclusionDiscrete -event simulations of complex weap-on systems have been successfully appliedto a variety of problems for the past eightyears. The discrete -event architecture is wellsuited to model complex weapon systemsthat are composed of many elements inter-acting with the environment and attackingforces. Techniques and software developedby the Naval Systems Department simula-

tion group make initiation of new simula-tions much easier than previously. Althoughonly weapon systems have been discusser[,discrete -event simulations have been widelyused for transportation, logistics, and traf-fic -control problems.

Anyone designing or analyzing complexsystems with many elements and interac-tions can beneficially use the methods dis-cussed in this paper. All of the softwarediscussed is operational on the DEC 2060computer system, and the calendar man-agement and list -handling subroutines haveoperated on the RCA IBM 370/168. More-over, since all programs are written inFORTRAN, their conversion to other sys-tems would not be difficult. MEDUSAhas been operational on UNIVAC, IBM,and CDC computers, and SIMATR hasbeen converted to operate on UNIVAC,CDC, and Honeywell computers.

Discrete -event simulations of complexweapon systems have been successfully usedfor system design, algorithm design, testdesign, test analysis, predicting system per-formance, evaluation, requirements speci-fication, operations analysis, mission analy-sis and education. Without simulation, themyriad weapon -system parameters that

were varied for these analyses would haverequired many more years of engineeringmanpower.

AcknowledgmentMany people have contributed to the suc-cess of the simulations described in thispaper. Of all these, Beryl Levy is the onemost responsible for converting the ideaspresented here into concrete, usableprograms.

ReferencesI. J. Golub, "FORTRAN Simulation with Dynamic

Lists,"Record of Proceedings of the Seventh AnnualSimulation Symposium. Tampa. Florida (March1974).

2. A. Alan and B. Pritsker, The GASP IV SimulationLanguage, John Wiley & Sons (1974).

3. J. A. Clancy and J. Golub. "Dual Development of aSimulation," 1979 Summer Computer ConferenceProceedings, Toronto, Ontario (July 1979).

4. J. Golub, D.M. Greeley, and B.B. Levy. "SIMATR.an Air Battle Simulation of the USAF Tactical AirControl System (TACS) with Advanced TacticalRadars," 1981 Summer Computer Simulation Con-ference Proceedings. Washington. D.C. (July 1981).

38 RCA Engineer 27-6 Nov /Dec. 1 982

G.W. Suhy

Digital interface simulationfor large system development

Techniques used to simulate critical hardware -softwareinterfaces are essential to cost-effective development and testof very large real-time systems.

Abstract: Digital real-time interfacesimulation has become an important toolin developing large military systems, par-ticularly those incorporating embeddedcomputer subsystems. This paper discussesthe application of interface simulation as ameans of reducing programming errors,improving the technical quality of devel-opmental systems, and significantly cuttingthe cost and complexity of system testing.

The in-process testing of incremental base-lines has always been a vital part of thedevelopment process for large electronicsystems. As these systems have made increas-

ing use of embedded computer subsystems,the testing technology to support them hasalso become increasingly computer -based.This condition has led to the applicationof digital interface simulation as a signifi-cant element of large system development.

At RCA Missile and Surface Radar, theextensive use of real-time interface simula-tion technology has become an integralpart of the engineering development pro-cess for the U.S. Navy's AEGIS CombatSystem as well as for a range of otheradvanced, highly computer -based radar sys-tems. This paper describes the applicationof interface simulation to this type of pro-ject, with particular emphasis on the AEGISSystem because of its large size and thediversity of technical issues involved.

'1982 RCA CorporationFinal manuscript received October 22. 1982Reprint RE -27-6.6

AEGIS Combat System applicationThe Ticonderoga (CG 47) class of guided -missile cruisers armed with the AEGISCombat System, is the primary protectionfor the Navy's aircraft carrier battle groups.Armed to provide anti -air, anti -surface, andanti-submarine protection, Ticonderoga andher AEGIS Combat System create an "en-velope of defense" for the battle group,

SEAHAWKSYSTEM

RADARADT

SYSTEMS

IDENTIHCATIONSYSTEMS

ELECTRONICSENSINGSYSTEM

TACTICAL DATA

LINK LINK11 14

AEGISDISPLAY SYSTEM

MARK 1

providing protection that ranges from belowthe surface to the stratosphere. The quickreaction and the high firepower of AEGISrepel missiles and aircraft attacking fromvery low and very high altitudes. The shipcan counter a saturation missile attack,making her one of the most survivablecruisers ever built. Figure I is an overviewof the AEGIS Combat System.

NAVIGATIONSYSTEM

SONARSYSTEMS

AIR CONTROL

LINK4A

iiLINK

WCS

MK I

SEAHAWKSYSTEM

ELECTRONICWARFARESYSTEM

UNFIRE CONTROL

SYSTEM

ANISPY-1A

RADAR SYSTEM

sat 53AORTSMK I

TOMAH/MIKMISSILESYSTEM

HARPOONMISSILESYSTEM

PHALANXWEAPONSYSTEM

MISSILEFIRE CONTROL

SYSTEM

UNDERWATERARE CONTROL

SYSTEM (1-4-64f4

Fig.1. Elements of the AEGIS Combat System. The AEGIS system includes a widerange of sensor and weapons subsystems under computer control. all of whichmust be extensively simulated prior to ship del; very.

RCA Engineer 27-6 Nov /Dec 1982 39

Development of the many computer pro-grams required is a major part of theAEGIS shipbuilding program. Hundredsof thousands of lines of source code areneeded to control and process sensor infor-mation, control the weapons systems, moni-tor equipment status, and provide the crewwith a powerful and flexible commandand control capability. Major computerprograms must be developed and testedbefore the equipment with which they inter-face is fully available; frequent, and some-times major, design changes occur afterthe equipment becomes available; and theentire development process involves largenumbers of subcontractors, all with differ-ent approaches to equipment and comput-er -program development.'

Tens of thousands of detailed test itemsmust be demonstrated in acceptance proce-dures; the cost of personnel, aircraft, mis-siles, and targets is prohibitive for all butthe final stages of testing. In this environ-ment, the AEGIS Interface Simulator Sys-tem filled a need for an affordable, highlyversatile test bed for the development ofthe tactical computer programs.

Requirements

The general requirement for the AEGISInterface Simulator System (ISS) is to pro-vide for each major computer -program ele-ment (or group of elements) a real-timeinterface simulation sufficiently realistic totest each of the functions of that element(or group of elements) over its entire rangeof operating conditions. In addition to pro-viding a realistic interface at each channelof the computers in which the tacticalprograms operate, other requirements arethat the ISS be reconfigurable for a widerange of test cases using the same basicprograms, that it provide monitoring anddata recording at these interfaces, and thatit have a minimal dependence on special-purpose equipment (that is, it must be ableto operate in an off -site computer labora-tory without requiring special shipboardequipment).`

Design

The computer programs that control theship equipment are organized into threemajor groups, or tactical elements: AN/SPY -1A (radar), Command and DecisionMark I (C&D), and Weapons ControlSystem Mark I (WCS). Three ISS con-figurations were developed to support test-ing of each of these elements individually(Fig. 2). Other configurations were then

Table I. Principal and derivative ISS configurations.

Configuration PurposeNumber ofcomputers

AN/SPY-1A

C&D

WCS

CDS

System

SPY /VLS

Stand-alone testing 01 AN/SPY-1A tactical element 2

Stand-alone testing of C&D tactical element 3

Stand-alone testing of WCS tactical element 3

Testing of combined C&D and WCS tactical 3elements

Testing of all three tactical elements 5

Testing of AN/SPY-1A tactical elements, modified 2for vertical launching system*

CDV Testing of combined C&D and WCS, modified 3for VLS

Testing of all three elements, modified for VLS 5

Testing of the Operational Readiness and 2Test System

Testing of the Gyro Data Converter 2

Testing of the special display system 1

ADG ISS reconfigured to operate in MAC 1

maintenance computer

Scenario generator (See text) 1

System /VLS

ORTS

GDC

ADG

ADG/MAC

* VLS is a missile launching system to be used in later AEGIS ships.

derived for combinations of these groupsand for special functions, as shown in Ta-ble I. The ISS consists of approximately

lines of source code in the CMS -2 language and operates on one to fiveUnivac AN/UYK-7 computers, dependingon configuration.

In addition to meeting its functionalrequirements, the design of the InterfaceSimulator System had to resolve a numberof other computer -program engineering is-sues such as cost, host equipment charac-teristics, precision, communications, andmaintainability. These issues were resolvedin the form of two fundamental designtrade-offs: cost and processing time versusmodularity, and degree of off-line prepro-cessing versus precision and real-timeresources.

In the first relationship, there are strongdesign and maintainability advantages inwriting a set of functionally decoupled mod-ules (or subprograms) rather than one verylarge program. Separate modules are eas-ier to design, and it is easier to isolate thesource of a system problem when eachmodule performs only a specific function.As module sizes become smaller, however,the total number of modules increases, asdo the computer time and memory over-head for inter -module communications, con-figuration control, and maintenance. Fur-thermore, as retroactive design changes aremade throughout the system's later life,

the number of modules that must be mod-ified for each change also increases. Thenature of this relationship is shown in Fig.

optimum modular partitioning forISS was one module for each equipmentsubsystem to be simulated; one module foreach group of equipment subsystems wouldhave been the next higher step, and onemodule for each function of each subsys-tem, the next lower. This resulted in 80modules with an average module size ofabout 4,000 lines of source code.

The second relationship, on-line versusoff-line processing, is somewhat more com-plex. Many of the phenomena being simu-lated (such as radar reflections from mul-tiple targets in an electronic counter-measures environment) occur very quickly,on the order of several microseconds, andthe tactical programs expect to receive thesesignals at very high data rates. However,the analytical solution of these complexelectromagnetic phenomena by a simula-tor requires computer time that is severalorders of magnitude greater than that re-quired by the tactical programs to measureand process these signals. Furthermore,many of these external electromagneticevents cause the simultaneous reporting ofsignals by sensors having widely differentcharacteristics; as a result, the simulatormust synchronize events precisely.

The use of very -high-speed computersand special-purpose analog simulators to

40 RCA Engineer 27-6 Nov./Dec. 1982

AN/SPY1 A ISS,14 -

Signal processorsimulationC&D simulation I ;

WCS simulation "

C&D ISSWCS simulationRadar simulationExternal systemsimulation

. C

;

`,

,t

I

AN/SF se -1ATactical Element

C&D Mark 1Tactical Element

WCS ISSC&D simulation ,,

Radar simulation i "Weapons equipment Li.simulationII

..1 IlLtrr

Interface Simulator System

display andcontrol -.-consoles

WCS Mark 1Tactical Element

Tactical Computer Programs

Fig. 2. Principal configurations in the AEGIS Interface Simulator System The ISS isdesigned to be reconfigurable to support the AEGIS system at various stages ofdevelopment.

solve the computational loading and mul-ti -sensor synchronization problems was noteconomically feasible. The general approachto solving these problems within availableresources required that the simulation pro-cess operate on standard Navy computersand be divided into a real-time phase andan off-line preprocessing phase. During thepreprocessing phase, all external events areanalyzed in terms of the effects they wouldhave on all possible sensors. This informa-tion is re -computed in discrete steps acrossboth the range of normal operating condi-tions and the length of the test case, and

then stored to a high-speed disk file.During the real-time phase, this data is

read back from the disk, corrected for fac-tors not predictable in advance, and trans-mitted to the tactical computer programsExamples of the types of corrections per-formed in real-time are: suppression ofsignals for sensors that are not enabled;interpolation of the data to the value ofthe real-time clock; response to humanoperator actions; and generation of flightdata for manually initiated missile launches.

A roughly inverse relationship exists be-tween the sizes and other cost factors of

Inter modulecommunica!ions

Source codemaintenar ce

Initial design,fault isolation

1/Module size

Fig. 3. Program modularity trade-offs Op-timum subprogram module size dependson communications and maintenancecosts as well as functional partitioning.

Size and running timeof real-time program

Size and runnirg timeof off-line program

Disk storage

50% 100%

Fraction of computing done off-line

Fig. 4. Real-time/off-line trade-offs fora single configuration. For an interfacesimulator, cost savings can be achievedby preprocessing functions that use anoff-line program.

the real-time and off-line programs. asshown in Fig. 4. The size of the real-timeprogram tends to decrease in a linear fash-ion (down to some absolute minimum) asfunctions are transferred to the off-line pro-gram. The size of the off-line programincreases linearly with the increase of pre-cision and the addition of relatively un-coupled functions. A point is reached, how-ever, where the further transfer of real-timefunctions to the off-line process requires ahigh degree of anticipation and the com-pounding of interrelated conditions. At thispoint, the size of the off-line program in-creases much more rapidly.

If the interpolation required in real-timewere halved, the size of a target trajectoryfile produced off-line would double. How-ever. transference of the responses to opera-tor actions from the real-time programwould require the anticipation of all pos-sible operator actions and would increaseboth size and computing time by far morethan a proportional amount. In manuallyinitiated missile launches, for example, thecomputation of a missile trajectory is af-fected by firing doctrine, launch time, type

Suhy Digital interface simulation for large system development 41

Size and running timeof single real-timeprogram

Size and running timeHof multiple real-time

programs

Size and running timeof off-line program

Disk storage

50% 80% 100%Fraction of computing done off-line

Fig. 5. Real-time/off-line trade-offs usinga common off-line program with multi-ple real-time programs. Additional costsavings can be achieved by sharing ageneral off-line program with severalreal-time programs.

of missile selected, and the cumulative effectof previous launches. In addition, the flightpath is altered in real-time by guidancedata from the Weapons Control System.The anticipation of all possible flight paths,even with real-time interpolation, wouldinvolve the generation of hundreds of dis-tinct trajectory files. This kind of nonlin-earity determines the optimum trade-offpoint between real-time and off-lineprocessing.

Within the functional requirements andequipment characteristics of the AEGISISS, the optimum fraction of total process-ing for a given configuration to be per-formed off-line was 50 to 60 percent. How-ever, an additional economy was realizeddue to the inherent similarity among thedata files used by different sensor simula-tions. For example, although the AN/SPY -IA phased -array radar, the RDP (Re-mote Digital Plotter) radars (Fig. 1), andthe AN/UPX-29 IFF (identification, friendor foe) radars have different ranges, datarates and interface formats, they have sub-stantial overlap in the targets that they candetect simultaneously. By modifying theoff-line program to produce a disk file (or"scenario") general enough to be used byall sensors and by adding simple discrimi-nation logic to each of the real-time pro-grams to suppress inapplicable data, it waspossible to have one off-line program driv-ing several real-time programs.

When seen in this light (that is, the costof one off-line program versus the totalcost of several real-time programs), thetrade-off point shifts considerably in thedirection of a more sophisticated off-lineprogram (Fig. 5). In the case of the ISS,the trade-off point was approximately 80

percent. An additional advantage in thiscommon -data -file approach is the consid-erable simplification of intersensor synchron-ization processing in the real-time programs.The use of several files would have requiredcompensation logic for the timing differ-ences between the different disk drives.

Development

Once the overall design of the InterfaceSimulator System was completed, a numberof detailed development policies were estab-lished to assure the technical integrity andtimely availability of the ISS to the overallAEGIS Program. With the large numberof programmers participating in the devel-opment of the ISS and the number of tac-tical programs (peak levels of 30 and 150,respectively), it was necessary to assurethat the two efforts did not drift apart. Aparallel development plan was establishedwherein both projects were governed bythe same set of formal interface designspecifications and supported by the samesystem engineering team under similar, con-current schedules.

Because of the large size of the InterfaceSimulator System (approximately 320,000lines) and the contract requirements, spe-cial attention was given to a formal devel-opment methodology. Use of CMS -2 (astructured, higher -order Navy computer lan-guage) was specified, as were formal doc-umentation and coding standards, and ap-proval of both the initial designs and thesubsequent changes by external reviewteams. High productivity levels were sup-ported with the extensive use of program-ming tools such as a text editor, source -code indenter, automatic flowcharter, filemanagement system, and module -level testdrivers. Additional cost and reliability ad-vantages were achieved by using the sametypes of computers, the same operatingsystems (one for real-time and one for off-line operation), and the same configura-tion control and quality assurances servi-ces as those used by the tactical program-ming team.

With Ticonderoga due to commissionearly in 1983, the role of a sophisticatedinterface simulation facility in reducing costsand assuring technical integrity is well estab-lished. Throughout the AEGIS engineer-ing development process, the Interface Sim-ulator System provided a number of valu-able functions:

A controlled and repeatable means formeeting thousands of detailed formal testrequirements;

An effective bridge between system -leveltest engineers and software developers;

Early identification of the specificationconflicts;

A basis for long-term "regressiontesting";

A basis for testing otherwise untestable"worst -case" tactical scenarios;

Reduced dependence on equipment -avail-ability schedules; and

A major system -performance and confi-guration -control mechanism.

The Interface Simulator System will con-tinue to be used to support the productionof subsequent Ticonderoga -class ships andwill be used at the Naval Surface Wea-pons Center, Dahlgren, Virginia, to repro-duce, investigate, and correct problems en-countered in ships at sea as well as tosupport any computer -program retrofittingof future ships.

Related applications

Interlace simulation techniques have beenapplied to the development of a wide rangeof systems such as conventional and phased -array radars and also for such non -radarapplications as communications, electroniccountermeasures, navigation and fire con-trol. The best candidates for interface sim-ulation are systems having an embeddedcontrol computer (or distributed system ofcomputers) and well-defined digital inter-faces to peripheral equipment. The mosteconomically advantageous applications ofinterface simulation have been for real-time target generation, particularly for co-ordinated multiple -sensor systems.

Another benefit has been the resolutionof hardware/software specification conflicts.Since an interface simulator, although itselfa computer program, is usually derivedfrom the specifications of the external equip-ment being simulated, the operation of theembedded computer against the simulatorcan often reveal specification errors or am-biguities that, although minor in a techni-cal sense, can be very costly to isolate andcorrect at a later stage.

In several projects in the early stages ofdesign, simulators are being used to replacesome equipment subsystems and are beingoperated with the embedded control com-puter to optimize computer timing as wellas overall system performance. This is beingdone at an early enough stage to allowmodifications in both the hardware andsoftware designs. In other systems, analyti-

42 RCA Engineer 27-6 Nov./Dec. 1982

woo

George Suhy is Unit Manager, SimulatorSystems. in the Software Design andDevelopment Activity at Moorestown; he iscurrently responsible for the AEGIS Inter-face Simulator System. After joining RCA in1973 he worked on a variety of systemsoftware projects before his currentassignment on the AEGIS Program. Hereceived the BSEE degree (1970) and theMS in Systems Engineering (1975) from theUniversity of Pennsylvania. He is active inthe Computer Society of the IEEE and theAssociation of Old Crows.Contact him at:RCA Missile and Surface RadarMoorestown, N.J.TACNET: 224-2542

cal simulators, which are much larger andusually do not operate at real-time speeds,are used to develop detailed critical testcases. These test cases then are repro-grammed as input data to interface simula-tors, thereby combining the analytical pow-er of a large computer with the real-timecharacteristics of an interface simulator.

Although many of the early applicationsof interface simulation were developed withIndependent Research and Developmentfunding, this approach is gaining increas-ing acceptance and many newer systemprojects now require the use of interfacesimulation for development and acceptancetesting as part of the development contract.

SummaryThe use of digital real-time interface simu-lation has repeatedly demonstrated bothtechnical and economic benefits in the devel-opment of large system projects. Thousandsof programming errors have been detected,precisely documented, and corrected usingsucb simulations. It also has enhanced flexi-bility of project scheduling, has improvedtechnical quality by providing precise andreproducible external conditions, and hasgreatly reduced many of the costs asso-ciated with full -system testing. As the sizeand complexity of modern electronic sys-tems continue to grow, interface simula-tion will be a required engineering tool forlarge system development.

References

I. F.G. Adams. "Puffing AEGIS to Sea," RCA Engi-neer, Vol. 26, No. 7. p. 40 (July/August 1981).

2. C.N. Falcon. "Testing of Systems Developed by theIncremental Build Method," Pro ceedings. of the Work-shop on Software Testing and Test Documentation.R. Lauderdale. Florida (December 1978).

3. G.W. Suhy. "Real -Time Simulation as a Tool forSystem Development," Ninth Annual Modeling andSimulation Conference Proceedings. University ofPittsburgh (April 1978).

The Engineer's Notebook

Design and simulation of an intelligent missile seeker

Julius HaymanRCA Government Systems Division

Cherry Hill, N.J.

An intelligent tracking algorithm for an infrared (IR) imagingmissile seeker and a method of evaluating its performance insimulated flight is described in this paper. The missile is a fire -and -forget, shoulder -launched, anti-tank weapon that lofts to analtitude of approximately 150 meters and then homes towardthe target using proportional guidance. A gyro -stabilized seekeris precessed to the target by pointing comands. These pointingcommands are developed from an imaging sensor by a newtracking algorithm that performs well in a highly cluttered back-ground.

Both the tracking algorithm design and its evaluation in mis-sile flights were accomplished using a digital computer simula-tion called HUGGER. This program includes a 6 -degree -of -free -

'1982 RCA CorporationFinal manuscript received October 7. 1982Reprint RE -27 -6 -Technical Note

dom missile simulation, a detailed seeker model, a three-dimen-sional model of the target, a two-dimensional background model,and the intelligent tracking algorithm.

Sensor images of the target as seen from the missile areformed on a 64 X 128 pixel IR-charge-coupled-device (CCD)array. The algorithm uses three features within the image: inten-sity, spatial frequency, and internal gradients. A multithresh-olding technique is also described which enhances target dis-crimination.

Tactical missiles of today have definite operational limitationsbecause seekers are apt to lose track of targets against highlycluttered backgrounds. Such scenarios may be attacks on tanksagainst a background of roads, foliage, rocks, and so on; attackson aircraft against a varied cloud background; or attacks fromabove on aircraft against a terrain background.

The next generation of tactical missiles may use imaging

Suhy: Digital interface simulation for large system development 43

The Engineer's Notebook

trackers because of the large amount of information in the imagethat can be utilized by an intelligent tracking algorithm. Thealgorithm is used to separate (or segment) the target from a clut-tered background, and then track this target. With the advent ofsmall, solid-state CCD arrays and high-speed microprocessors, itis now practical to utilize imaging seekers with intelligent track-ing algorithms. Two recent developments by RCA initiated thisstudy. These are the ATMAC microprocessor developed by theAdvanced Technology Laboratory,' and the Schottky -barrierIR-CCD array developed by the David Sarnoff Labora-tories'. 4 C. The latter is a 64 X 128 element staring array thatoperates in the 3- to 5 -micrometer wavelength band with excel-lent uniformity, sensitivity, and dynamic range. Since it uses sil-icon technology, its potential production economy appears veryattractive.

Computer simulation work verifies successful operation of amissile system containing the IR-CCD array and tracking algo-rithm. Both the tracking algorithm design and its evaluation inmissile flights were accomplished using program HUGGER whosescenario is illustrated in Fig. 1. Simulated sensor images of the

target as seen from the missile are displayed with the trackinggates superimposed to assist in the algorithm development. HUG-GER also provides each corresponding segmented image so thatthe algorithm's discrimination capability can be evaluated.

The intelligent tracking algorithm encloses the target with anadaptive gate and uses multifeature and multithreshold tech-niques to achieve superior image segmentation. A rectangularadaptive gate is servoed to enclose only the target vicinity, there-by eliminating most of the extraneous background clutter withinthe field of view. A background gate is located around theperimeter of the centroid gate. There are three features extractedfrom the image: intensity, spatial frequency, and internal gra-dients. Each of these features is thresholded using the multi -threshold technique where up to four thresholds are automati-

Fig.1. Program HUGGER isa simulation of a missile -target system in which aseeker forms images on a 64X 128 CCD array. The missileis guided by a processing ofthese images with an intelli-gent tracking algorithm.

tally employed, allowing target pixels to be identified if they areabove, below, or in between background value bands. Enhance-ment is accomplished by changing the intensity value of a pixelif either the spatial frequency or the gradient value is over itsrespective threshold. The enhanced intensity feature is then finallythresholded to provide the segmented image. Adaptive gate size,seeker pointing commands, as well as missile guidance signals,are all derived from this segmented image.

During terminal navigation, when the target fills the entirefield -of -view, the adaptive gate tracker is replaced by a centrallyweighted correlation tracker. Correlation is performed along ahorizontal strip for yaw and a vertical strip for pitch. These twostrips form a cross at the center of the field -of -view. Because ofthe high rate of range closure and the high -line -of sight ratesduring terminal flight, correlation must be done on a frame -to-frame basis. Drift due to the range closure effect is minimized bycentrally weighting each correlation strip. This low -drift algo-rithm is used until impact.

The simulations showed that the 64 X 128 CCD array, withthe intelligent algorithm, would track targets in high clutter andcould be made to consistently impact the upper surface of a tankwhere armor is at a minimum. Program HUGGER was anexcellent tool for evaluating algorithm modifications during devel-opment. Algorithms can be compared over the same trajectory,with the same target intensity and background statistics. It wouldbe impossible to do this otherwise, even with actual hardware.HUGGER also allows target tracking with realistic line -of -sightrates and accelerations, range closure, and a changing aspectangle.

Program HUGGER

Program HUGGER is a modularized FORTRAN program con-sisting of 18 subroutines. Output data is provided in the follow-

ACTU AT

MISSIL (6 DOF)

LIU ATRN

SEEK (SPINNING MASS SEEKER)

'Ite ATRACKR (3 FEATURE ADAPTIVE THRESHOL DING)

ATMOS

WINOYN -11

MUG

TARGET(3 DIMENSIONAL TANK MODEL)

Jo,

.M4Clir .1

Ilt rotol

- -

,..."1:1

44 RCA Engineer 27-6 Nov./Dec. 1982

The Engineer's Notebook

Fig. 2. 2. Simulated images of a tank are formed by Program HUGGER with the track-ing gates superimposed. These images are formed every 1/60 second and areprocessed to obtain missile guidance commands.

ing formats:

Printed data, where 56 variables are printed.

Striplot, where up to 10 variables are plotted against time.

Segmented image plots, indicating target/nontarget pixels asdetermined by the intelligent algorithm.

Image plots (Fig. 2), where images formed on a 64 X 128IR-CCD array are plotted using a gray scale.

Considerable detail is modeled in HUGGER so that manyparts of the missile system can be studied during flight. To elimi-nate excessive execution time, the following options are provided:

A detailed gyro -stabilized seeker with gimbal friction, magnetictorques and input errors, or a simplified seeker model.

A three-dimensional target model, the tracking algorithm andimage forming, or a simple point target.

An optical point -spread function, or a simple one -ray -per -pixelprojection.

Image plots, or no image plots.

Steady or gusty wind or both, or no wind.

A moving or a static target.

ConclusionRecent military events have vividly shown that precision guidedmissiles are so effective they have changed concepts of modernwarfare. With the advent of small, high-speed microprocessorsand solid-state staring arrays, this trend will continue. Imagingseekers which utilize considerably more information concerning

the target area are now possible. The recently -developed Schottky -barrier IR-CCD sensor array has the potential for being econom-ically produced since it is fabricated using conventional silicontechnology. Intelligent tracking algorithms can be implementedeconomically with software using presently available digitalprocessors.

Development of intelligent algorithms can be accomplishedefficiently by using computer simulations of the entire missile -target system. All the variables of the algorithm and missile areobservable, and both the target intensity and the backgroundstatistics are controllable.

ReferencesI. W.A. Helbig, W.J. Davis. and A.D. Feigenbaum, "ATMAC Microproixs.sor."

RCA Engineer. Vol. 23. No. 1, pp. 36-39 (June/July 1977).

2. W.F. Kosonocky, H. Elabd. et at. RCA Laboratories. Princeton. N.J.; and M.J.Cantella, et at. RCA Automated Systems, Burlington. Mass.. "Design and Per-formance of a 64 x 128 -Element PtSi Schottky -Barrier IRCCD Focal PlaneArray." SPIE Technical Symposium East, 1982. Arlington. Va (May 3-7. 1982).

3. R.W. Taylor. L. Skolnik, et al. "Improved Platinum Silicide IRCCD FocalPlane." SPIE Technical Symposium, Advances in Focal Plane Technology. LosAngeles. Calif. 217. 103 (February 4-5, 1980).

4. F.D. Shephard, R.W. Taylors. er at. "Schottky IRCCD Thermal Imaging."Advances in Electronics and Electron Physics. Vol. 52. p. 495. 7th Symposium onPhoto -Electronic Image Devices. (1979).

5. J.A. Tegnelia. "DARPA Unveils Staring Focal Plane Array." Defense Electronics,Vol. 12. 101 (1980).

Contact: J. Hayman TACNET: 222-5944

Suhy: Digital interface simulation for large system development 45

S.M. Perlow

ANA radio -frequency circuitsimulation and analysis

Computer -aided circuit simulation permits the development andadjustment of circuit elements without costly construction ofa laboratory model.

Abstract: Computer -aided design ena-bles the radio -frequency (RF) or micro-wave circuit designer to design a circuitand modify element placement quickly andcost effectively. The ANA program,designed primarily for minicomputers, isan extremely simple approach to com-puter -aided design of linear RF circuits.With it. coding of elements and theirplacement is in the form of easily remem-

circuit can be modified by use of an on-line conversational mode.

Computer-aided design provides the mod-ern radio -frequency (RF) circuit designerwith a tool that allows initial circuit ele-ment and topology design and adjustmentbefore a laboratory model is constructed.In addition to providing a means for pro-ducing designs quickly and at a relativelylow cost, computer simulation allows thecreation of circuits that may have beenimpossible to design otherwise.

The circuit analysis program, ANA, canbe used to design and model RF circuits. Itis easy to use because it requires only asimple description of elements and circuittopology. The circuit and desired outputscan be modified easily by means of a sim-ple conversational mode that does not re-quire editing of data files. Data, or com-plete circuits, may be written to or fromdisks or magnetic -tape -based libraries. Thisprogram was expected to be used on a con-tinuous, daily basis; it was, therefore, writ-ten to be run on a minicomputer ratherthan a large, time-shared mainframe com-

1982 RCA Corporation

Final manuscript received September 29. 1982Reprint RE -27-6-8

puter with prohibitive running costs. ANAis presently installed on HP1000 computersat RCA Laboratories, Princeton; Astro-E-lectronics (AE), Princeton; Missile and Sur-face Radar (MSR), Moorestown; Consu-mer Electronics Division (CED), Indiana-polis; RCA Laboratories Limited, Zurich;and Advanced Technology Laboratories(ATL), Camden.

Table I. Two -port element descriptors.

Describing the circuit elementsCircuit elements may be divided into threecategories, as shown in Tables I, II, and III.The coding for each element type consistsof either a simple acronym or, as in the caseof impedance and admittance, the symbolconventionally used in texts. The simpletwo -port elements shown in Table I havefixed parameter values.

Element type Coding* Parameters

R L C

Series RLC

R

X

K

L1 L2

zo

0

0

Parallel RLC

Impedance

Admittance

Coupled coils

Transm ssion line

SRLC LIn1-1) ClpF1

PRLC R(tI), L(nH), C(pF)

Z Zo(n), RiZo, X /Zo

Y Yo(15). G/Yo' B/Yo

cci K. L1(nH), L2(nH)

TL F(GHz).ATT(dB/A). Zo(!.!). L(deg)

* Only the first two characters for SRLC and PRLC are needed

+ For ideal transformer. use K z 1 and then L1:L2 is square of turns ratio.

46 RCA Engineer 27-6 Nov./Dec. 1982

Table II. Tabulated element descriptors.

Element type Cooing Parameters

S11 S12

S21 S22Scattering parameters SPara S11. S21. S12. S22

(MAG. ANGLE)

Y -parameters YPara Y11, Y21, Y12, Y22[MAG (mmhos), ANGLE]

ABCD-parameters ABcd A. B. C. D(REAL. IMAGINARY)

Tabulated impedance TZ Z0(1!). R/Zo, X/Zo

Tabulated admittance TY Yo(u), G/Yo, B/Yo

Reflection coefficient GAmma (MAG. ANGLE)

The tabulated elements of Table II consistof parameters that must be tabulated foreach frequency specified in the analysis.The impedance and admittance elementsmay also be tabulated for each frequencyby specifying TZ or TY, instead of Z or Y.This feature is useful if the impedance oradmittance of an element is known as afunction of frequency, but is difficult to syn-thesize as R. L, and C values.

The multiports of Table III cannot beconnected as two ports, but must always beconnected in a nodal configuration. This isdiscussed in the next section.

Describing the circuit topologyA complete circuit description requires spec-ification of how its elements are connectedand what its input and output are. Thismay be accomplished by using standard,two -port connections such as cascades,branches, and paths, or by specifying nodesto which the elements are connected.

Two -port connections

Two -port connections are preferred overnodal connections because they are compu-tationally more efficient. A designer ac-customed to two -port connections can usu-ally generate a complete circuit mentally,because node numbers do not have to beremembered.

Cascade connections. Cascade of ele-ments is easily accomplished by specifyingthat the element is connected as either aseries element (SE) or a parallel element(PE), as shown in Fig. 1.

Branch connections. Branch connectionsare used if more complex networks areconnected in series or parallel with themain signal path. A series branch is initiatedby the symbol SB, while a parallel branchis initiated by the symbol PB. All topologyand element entries following these sym-bols are considered part of the branch untilthe symbol EB, indicating the end of branch,

is used. Entries following EB are consi-dered to be main -path entries. A secondbranch may not be imbedded in a branch.Branch connections are shown in Fig. 2.

Path connections. Path connections indi-cate circuit elements that depart from themaim path, but eventually return (unlikebranches that just terminate). The symbolPP indicates the beginning of a parallelpath. All entries following this symbol areconsidered part of the first path until eitheranother PP or an SP for series path isentered. The second path symbol indicatesthe end of the first path and the beginningof the second. All entries following thesecond path symbol are considered part ofthis second path until a third path symbol isentered. This third symbol indicates the endof the second path and the rejoining of bothpaths into a single or main path. Note thatthe first- and second -path symbols do nothave to be the same. For example, themain path can break into two parallelpaths. initialized by PP. The two paths canthen be reconnected in a series path ratherthan a parallel connection, in which casethe second- and third -path entries are SP.The second- and third -path entries mustagree because they are the same conditions.The four possible path combinations areshown in Fig. 3.

An incomplete path will cause the ter-minal to show an error message when ananalysis is performed.

Structure connections. Series structures(SS) and parallel structures (PS) are identi-cal to the path connections, except forname and symbol. The change in nomen-clature is to allow imbedding a structurewithin a path, or vice versa, without ambi-guity. For example, if a series path were tobe imbedded (or nested) within a parallelpath, initiation of the parallel path wouldbe PP. Initiation of the imbedded seriespath would be SP. Unfortunately, the useof SP as the second -path entry indicates theconclusion of the original parallel path in a

SE SRLC PE SRLC

Fig.1. Cascade connections of two portsprovide the simplest interconnection ofelements. The interconnection consistsof a single main path.

Perlow. ANA radio -frequency circuit simulation and analysis 47

Table Ill. Multiport element descriptors.Element type

Coupled transmission lines

0 1 ° N2N1

N30

N4

N1,N2 N3,N4 CL

Inhomogeneous coupled lines* N1,N2 N3.N4 ICL

Dependent current generator

ni

N2

N3

N4

F(GHz). ELE LGTH(DEG), ATT(dB/ A)Line Between Nodes N1 and N2:

YOO(MMHO), YOE(MMHO)Line Between Nodes N3 and N4:

YOO(MMHO), YOE(MMHO)

F(GHz). C MODE: LGTH(DEG), ATT(dB/A), RCPI MODE: LGTH(DEG), ATT(dB/ A), RPI

Line Between Nodes N1 and N2:YC(MMH0), YPI(MMHO)

Line Between Nodes N3 and N4:YC(MMHO), YPI(MMHO)

N1.N2 N3,N4 GM GMF(MMHO.DEG). GMR(MMHO.DEG)

* C and PI modes are as defined in the paper by V.K. Tripathi, "Asymmetric Coupled Transmission Lines in an Inhomogeneous Medium."IEEE Trans. Microwave Theory & Techniques. Vol. MTT-23. No. 9, pp. 734-739 (September 1975). Determination of these parameters can bemade using ELOP, as described by F.J. Campbell in PTR-2649. "An Electron Optics Computer Program for the RCA Spectra, 70/45-55"(February 12. 1969).

SB

Fig. 2. Branch connections represent sim-ple departures from the main path. Aseries branch (SB) is connected in se-ries with the main path. A parallel branch(PB) is connected across the main path.

series connection rather than the initiationof a series path in the parallel path. Toovercome this shortcoming, the second pathis initiated by the use of the structure srm-bol (SS). In this case, the next two symbolsmust be structures to complete the nestedpath. followed by two path symbols tocomplete the outer paths.

Up to five paths or structures may be

Fig. 3. Path connections represent thebreaking of the main path into two sub -

paths. The junctions of these paths maybe connected in parallel (PB) or in series(SB).

nested. If this number is exceeded, an errormessage appears on the terminal whet theanalysis is performed. A summary of alltwo -port connections is presented in TableIV.

Nodal connections

Many circuits cannot be represented by thepreceding two -port connections. To allowfor these networks, nodal descriptors maybe used in place of the two -port topologyconnections.

Circuit nodes may be arbitrarily num-bered by integers used in any sequence.Entry of nodal connections into ANA isaccomplished by simply replacing the topo-logical descriptor. SE or PE. with twointegers representing the nodes to whichthe element is connected. A series RL('element, connected between nodes 5 andII, is designated as: 5. I I SRI.('. Nodalconnections may be imbedded in any ofthe two -port branches, paths, or structures.ANA automatically sets up the followingconditions:

Input nodes are always I and 0.

Reference node is always 0.

The following two rules must he adheredto:

An integer cannot be skipped.

The last nodal entry must define the out-put nodes. It must be: NI, N2 OUTwhere NI and N2 represent the two out-put nodes. Note that the reference nodedoes not have to be an output node, eventhough it is an input node.

48 RCA Engineer 27-6 Nov./Dec. 1982

Table IV. Two -port connections.

Connection type

Series element

Parallel element

Series branch

Parallel branch

End branch

Series path

Parallel path

Series structure

Parallel structure

Coding

SE

PE

SB

PB

EB

SP

PP

SS

PS

Notes on elementsconnected by nodesSome of the elements shown in Table I arerestricted by the use of two -port connec-tions. For example, coupled coils (CC) arerestricted to a common lead between inputand output when SE or PE is used as thetopological connection. This restriction isunnecessary when nodal connections areemployed. A four -terminal device can beconnected to four different nodes.

The multiport elements of Table III can-not be used with two -port connectionsbecause they have more than three termi-nals. The way in which the nodal descrip-tors are used for these elements is shown inTable V.

Modeling coupledtransmission -line arrays

In many microwave circuits, coupled trans-mission -line arrays are required. Modeling

REFERENCE NODE "0"(GROUND PLANE)

Fig. 4. Coupled lines can be arrangedto form arrays of coupled lines, whichare used to analyze microwave filtersand couplers.

Table V. Nodal descriptors.Element type Nodal descriptors Comments

SRLC

PRLC

TZ

TY

TL

SP

YP

AB

CC

GM

CL*

ICL*

N1,N2 SRLC

N1,N2 PRLC

N1,N2 Z

N1,N2 Y

N1,N2 TZ

N1,N2 TY

N1,N2 N3,N2 TL

N1,N2 N3,N2 SP

N1,N2 N3,N2 YP

N1,N2 N3,N2 AB

N1,N2 N3.N4 CC

N1,N2 N3,N4 GM

N1,N2 N3,N4 CL

N1,N2 N3,N4 ICL

First pair: input nodes

Second pair: output nodes

Note: repetition of N2

First pair: first trans line

Second pair: second trans line

* Multiple coupled lines are accommodated by repeating N3,N4 of apreceding set of coupled lines as N1,N2 of a new set.

comb -line and interdigitated structures re-quires the coupling of more than two linestogether. ANA allows complicated line struc-tures to be constructed very easily.

The first two nodes of a coupled line(CL) designate the first transmission line,and the second two nodes designate thesecond transmission line. Simply repeatingeither set of these nodes for another CLelement couples the line represented bythese nodes to a third line. To illustrate thisconstruction technique,

3.4 5,6 CL3,4 7,8 CL5,6 10,11 CL

10,11 12,13 CL

represents the structure shown in Fig. 4.The order is purposely scrambled to showthat it is of no importance to the analysis.These lines can also be represented in amore orderly fashion:

7,8 3,4 CL

3.4 5.6 CL5,6 10.11 CL

10,11 12.13 CL

In repeating the line nodes, ANA as-sumes that the repeated line is only one lineand not two lines in parallel. Some modelsrequire that lines really be connected inparallel. In those cases, it is important notto make the mistake of repeating the sameset of nodes in the CL designations.

Generator and loadtopological descriptionsANA automatically assumes a 50 -ohmsource or generator impedance and a 50 -ohm load impedance. If any other sourceor load impedance is required. the descrip-tors, GEN for generator and LOAD forload must be used. The function of GENand LOAD is to set the topological refer-ence plane of the source and load in theanalysis routine. Any network followed bythe entry GEN indicates that the programshould calculate the output impedance ofthe network preceding the GEN and usethat as the source impedance. Likewise, theentry LOAD followed by any networkindicates that the input impedance of thenetwork following LOAD must be used asthe load impedance. These networks maycontain nodally described subnetworks. butGEN and LOAD cannot be imbedded in anodal network. This is consistent with thefact that GEN and LOAD are themselvestopological descriptors and cannot haveother topological descriptors.

If the first or last element is series con-nected (SE), and GEN or LOAD is speci-fied. ANA will automatically assume thatthe off -common lead is returned to com-mon through a short circuit (in the case ofGEN, the Thevenin equivalent voltagesource representation). Similarly, if the firstor last element is a parallel -connected ele-ment, PE, the off -common lead will be leftfloating (Norton's equivalent). The GENand LOAD elements may be moved about

Perlow: ANA radio -frequency circuit simulation and analysis 49

Table VI. Output options.

GAINVOLTCURRZIN(N)

ZOU(N)YIN(N)

YOU(N)ZS(N)ZL(N)ISOL

STABFLATGM1GM2GA1GA2GASGAL

DELYVSWRRLOSTLOS

PE1

PE2

GAIN (AVAILABLE POWER GAIN)VOLTAGE GAINCURRENT GAININPUT IMPEDANCE (NORM TO 50 OHMS)OUTPUT IMPEDANCE (NORM TO 50 OHMS)INPUT ADMITTANCE-MMHOS (NORM TO .02 MHO)OUTPUT ADMITTANCE-MMHO (NORM TO .02 MHO)SOURCE IMPEDANCE (NORM TO 50 OHMS)LOAD IMPEDANCE (NORM TO 50 OHMS)ISOLATION (REVERSE GAIN)STAB FACTOR (ROLLET'S-K) AND GMAX(dB)GAIN FLATNESS (MAX, MIN, AND ACCUMULATIVE)SOURCE REFL COEF REQUIRED FOR GMAXLOAD REFL COEFFICIENT REQUIRED FOR GMAXINPUT REFLECTION COEFFICIENT - GAMMAOUTPUT REFLECTION COEFFICIENT - GAMMASOURCE REFLECTION COEFFICIENT (ZO = 50 OHMS)LOAD REFLECTION COEFFICIENT (ZO = 50 OHMS)DELAY (GROUP AND PHASE)VOLTAGE STANDING WAVE RATIORETURN LOSSTRANSMISSION LOSSPARALLEL RC EQUIVALENT OF INPUT IMPEDANCEPARALLEL RC EQUIVALENT OF OUTPUT IMPEDANCE

Two -port parameters available: -S- -Y--Z- -H- -ABCD-; ask for desired parame-ter (that is. S11, H21, and so on).

Note 1: If GEN and/or LOAD REFplane are not specified, SOURCEand/or LOAD are assumed to be50 ohms.

Note 2:

Note 3:

Note 4:

CONJ of any REFL COEF-make 4th character *.

Reciprocal of any REFL COEF-make 4th character R.

Y, Z, H, and ABCD are in polarform. RECT form is obtained bymaking last character an I.Example: Y221,AI

to find various input and output parametersthroughout the circuit.

Output optionsAt present, there are over 70 different out-put options. These outputs may be ob-tained in tabulated form with as many asfive different quantities printed during anysingle analysis. In addition to the tabulatedoutput, a graphics display can be obtainedfor any output on a Hewlett-Packard gra-phics terminal. All outputs, with the excep-tion of reflection coefficients, are plotted onrectilinear coordinates, automaticallyscaled to fit the data plotted. The scales aresegmented into decimal divisions. Reflec-tion coefficients are plotted on a computer -generated Smith chart. Table VI provides asummary of output options.

The desired outputs are calculated byassuming that all the elements precedingthe GEN element form the generator orsource impedance. All elements after theLOAD element establish the load impe-dance. If the GEN element is missing, thesource impedance is assumed to consist of asimple 50 -ohm resistance. Similarly a miss-

ing LOAD element means that the load isa pure 50 -ohm resistance.

The GEN and LOAD elements alsoestablish the reference plane for networkinput and output. The input impedance, oradmittance, is the impedance or admittancelooking into the network at the plane ofGEN. The input of the network, if GEN ismissing, is taken to be the input of the firstelement.

Voltage gain, VOLT, is calculated byplacing a voltage source at the input termi-nals of the network and calculating the vol-tage at the output terminals. This makesthe voltage gain independent of source im-pedance. This is also true for current gain,CURR.

The input and output reflection coeffi-cients, GA1 and GA2, are referenced to thereal part of the source and load impedan-ces, respectively. However, the source andload reflection coefficients, GAS and GAL,are referenced to 50 ohms.

The two -port parameters (S, Y, Z, H, orABCD) are independent of source andload impedance because they are definedfor specific terminating conditions. The char-acteristic terminating resistance for S -pa-rameters is 50 ohms.

A very useful application results fromthe different reference impedances for thereflection coefficients GA I , GA2 and S11,S22. A network can be designed by use ofGA1 and GA2 to match between a com-plex source and load. The outputs can thensimply be switched to S11 and S22 so thatthe calculated network can be compared tothe real network, as measured on a net-work analyzer, without reconstructing oreliminating the complex source and loadimpedances.

The output, STAB, prints Rollets' stabil-ity factor in column one and the maximumgain in column two, if the stability factor isgreater than unity. GM1 and GM2 providethe source and load reflection coefficientsthat produce the maximum gain. If the sta-bility factor is less than unity, GM1 andGM2 are undefined and the printout con-tains zeros. These three outputs are particu-larly useful in designing networks that arenonreciprocal and active, such as two -portamplifiers.

VSWR, RLOS, and TLOS are all calcu-lated from the reflection coefficients GA1and GA2. The VSWR is one plus themagnitude of the reflection coefficient, divid-ed by one minus the magnitude of thereflection coefficient. The return loss, RLOS,is the reflected power in dB and is calcu-lated as the magnitude of the reflectioncoefficient squared. Transmission loss,TLOS, is one minus the squared magnitudeof the reflection coefficient expressed in dB.The first column for all three of these out-puts represents the values at the input of thenetwork, while the second column repres-ents the values at the output.

In many design situations, the conjugatesor reciprocals of a reflection coefficient arerequired. These are easily obtained for anyreflection coefficient, including GM1, GM2,S11, and S22, by making the fourth char-acter an asterisk (*) for the conjugate or Rfor the reciprocal. The Y, Z, H, and ABCDparameters ae normally outputted in polarform. If the rectangular form is required,we make the fourth character I-for ex-ample Y121.

Establishing the initial networkThe initial network may be established byreading it from a disk file or magneticcaw -Re file previously generated by ANA,or by typing the data on the user's terminal.If a previously saved file is to be used,ANA will prompt for the name of the diskfile or the number of the magnetic cassettefile and retrieve the data.

If the terminal is used to initially estab-

50 RCA Engineer 27-6 Nov /Dec 1982

lish a network. ANA prompts the user withthe following statements:

TITLE

FREQS. (MHz) =

OUTPUT #1:

OUTPUT #2:

OUTPUT #3:

OUTPUT #4:

OUTPUT #5:

ELE NO 1: CONN, TYPE =

ELE NO 99: CONN, TYPE =

ANALYZE? CHANGES? WRITE?

The response to TITLE is any group ofcharacters limited to a maximum of 44. Itis used as a personal means of identifyingthe analysis.

The FREQS. (MHz) may be inputted asdiscrete values in the start, stop, step fashion,or in any combination of these methods.For example, an entry of the form,

1.35,1.76 2,10,2 10.15 20,100,10

results in an analysis at the frequencies,

1.35,1.76,2,4,6,8,10,10.15,

2,10,2

20,30,40,50,60,70,80,90,100

20,100,10

The maximum number of frequencies islimited to 21 at any one time. If thisnumber is exceeded, a message is issuedindicating that more than 21 frequencieswere asked for and that the frequency listwas terminated after the 21st. The frequencylist may be modified later.

The outputs are then selected. If less thanfive outputs are desired, a simple spacereturn will leave the output blank. Theoutput may also be changed after the initialdata has been inputted.

The final stage of the input phase con-sists of establishing the elements. After eachelement is specified by the user's responseto connection CONN, TYPE, the programprompts with the parameters required tocompletely define the element. By typingEND in response to the ELE NO XX:CONN. TYPE = question, the input phaseis completed, and ANA responds with:

ANALYZE?, CHANGES?, WRITE?

This is the prompt that ANA returns toafter each task is completed.

Table VII. Replies.

Replies to "ANALYZE?. CHANGES?, WRITE?" are:

ADDDELETE

ANALYZECONVERT

EL #FREQGEN

LOADHELP

INTERPOLATEOUTPUT

STARTTITLE

TERMINALWRITEPURGE

DATATABULATED

OFF END"r?

REACTANCETUNE

TTPLOT

VIDERAS

ADD MORE ELEMENTSDELETE ELEMENTSANALYZECONVERTS Z AND Y ELEMENTS TO SRLC AND PRLCELEMENT CHANGEFREQUENCY CHANGEGENERATOR REFERENCE PLANE CAN BE MOVEDLOAD REFERENCE PLANE CAN BE MOVEDLISTS PROGRAM PARAMETERSINTERPOLATES TABULATED ELEMENT DATAOUTPUT PARAMETER CHANGERESTARTS PROGRAMTITLE. CHANGETERMINAL CHANGEWRITES DATA FILE TO TERMINAL. DISK. OR MAG TAPEPURGES DISC DATA FILES CREATED BY ANAWRITES DATA TO TERMINAL ONLYWRITES TABULATED DATA TO TERMINALTERMINATES PROGRAMGIVES PARAMETERS FOR SINGLE ELEMENT IN DATA FILECALC L&C FOR X(F1) AND X(F2)ALLOWS TUNING OF ELEMENTSTUNE TABLE LISTINGPLOTTING ON VERSATECPLOTTING ON VIDEO MONITORSTOPS VIDEO PLOTTING

Note 1: Only the first two letters of each reply are needed.

Note 2: The reply "FREQ" automatically interpolates tab data.

Reply optionsMost of the replies to the prompt, summa-rized in Table VII, are self-explanatory.The reply ANALYZE obviously results inan analysis of the modeled network. SinceANA responds to the first two characters ofa reply, only two are required. An unrec-ognized reply is ignored.

The reply DATA produces a list of thenetwork elements and parameters on theuser's terminal. It does not produce a listingof the actual parameters of the tabulateddata, such as S -parameters. They may beobtained by replying TAB.

Any element connection, type, or param-eter may be changed by replying ELE-MENT XX, or simply EL XX. If the con-nection and type are to remain the same, aspace return is all that is required. Any ofthe parameters may be kept the same bycommas used to preserve the correct order.For example, if element I I is a series -con-

nected, series RLC, and the inductance is tobe changed, we can write

ANALYZE?, CHANGES?, WRITE? EL 11

ELE NO I I: CONN, TYPE = (Space Return)

R, L(NH), C(PF) = .120

ELE NO = 0

ANALYZE?, CHANGES?, WRITE?

Note that after the first element is changed,ANA responds by asking for an additionalelement number. This continues until an-swered with a zero. The addition of a newelement is achieved using the ADD reply.Any element may be completely eliminatedfrom the network using DELETE.

The generator and load reference planesmay be moved about by using the repliesGEN or LOAD. This reply will result inANA's requesting:

ANALYZE?. CHANGES?. WRITE?. GEN

MOVE FROM ELE#A TO AFTER ELE#B

If element A does not contain the elementGEN, the question is repeated. A generatoror load can be created by making A equalto zero.

Any element that is not tabulated, orcoupled lines, may be tuned by setting it upin the tune table. The tune table defineswhich elements and parameters are to bevaried during the analysis phase of ANA.When analysis is performed, the variableelements and their parameter values areprinted out just before the calculated out-put quantities. To create the tune table,

Perlow: ANA radio -frequency circuit simulation and analysis 51

Fig. 5. Schematic of wideband amplifierused in Example 1. Shunt and seriesfeedback are used, along with high -fre-quency peaking, to obtain flat gain andgood input and output VSWR charac-teristics.

reply TUNE. To check on the variableelements, a printout of the tune table maybe obtained by replying with TT. Up tofour element parameters may be tuned atone time. After the analysis is complete. theelement values are reset to their valuesbefore tune was initiated.

The title. frequencies. and outputs maybe changed by simply replying with TITLE.FR EQ, or OUT. Video plotting on the HPgraphics terminal is accomplished by thereply VIDEO, followed by the output andcolumn numbers associated with the desiredoutput. The command. ERASE, stops theplotting.

Two very helpful replies that can beused when designing circuits are CONVERTand REACTANCE. CONVERT automat-

Fig. 6. Circuit topologies used in analyzing the wideband amplifier of Fig. 5. (a)two -port topology: (b) nodal topology: and (c) mixed topology consisting of bothtwo -port and nodal topology.

ically converts all Z and Y elements in thenetwork to equivalent SRLC or PRLCelements. REACTANCE provides the Land C values required to obtain a specifiedimpedance or admittance at two differentfrequencies.

The reply HELP allows listing of themost recent tables of elements, connections.outputs. replies, and limitations. STARTallows restarting ANA so that a new diskfile may be read. END or OFF terminatesexecution of ANA.

Stewart Perlow, Member of the TechnicalStaff. RCA Laboratories, received his BEEfrom City College of New York and MSEEfrom the Polytechnic Institute of Brooklyn.His nineteen years of professional expe-rience include RF and microwave compo-nent development. contributions to studiesof distortion relationships in RF signal pro-cessors, and computer -aided design andmeasurements. Mr. Perlow is presentlywith the Consumer Electronics ResearchLaboratory, where he is involved in devel-opment of low-cost front ends for 13-GHzdirect broadcast satellite receivers. andcomputer -aided design. He received anindividual RCA Laboratories OutstandingAchievement Award in 1980. Mr. Perlow isa member of Eta Kappa Nu and the IEEE.Contact him at:RCA LaboratoriesPrinceton, N.J.TACNET: 226-3168

52 RCA Engineer 27-6 Nov./Dec. 1982

ANAIF LATEST INFO ON ANA IS REQUIRED TYPE 'HELP' .(REV. 0.8>

TITLE:WIDEBAND AMPLIFIER 0.1-2.0 GHZDATA INPUT-TERM.DISCrMAG TAPE7TERMFREOS.(MHZ)=100.2000.100OUTPUT 1:GAINOUTPUT 0 2:GA1OUTPUT 3:0A2OUTPUT * 4:VSWROUTPUT 0 5:STABELE NO 1:CONN.TYPE=PPELE NO 2:CONN.TYPE=SE SRLC

R.L(NH),C(PF)-183.3.1E6ELE NO 3:CONN.TYPE=PPELE NO 4:CONN.TYPE=SSELE NO 5:CONN,TYPE=SE SPARTABULATED DATA FROM TERMINAL7rDISC7rMAG TAPUDISCDISC FILE NAME=SPSNO. OF FIRST DATA LINE=48ELE NO 6:CONN,TYPE=SE SRLC

R.L(NH).C(PF)-0.2.5,1E6ELE NO 7:CONN.TYPE=SSELE NO 8:CONN.TYPE=PBELE NO 9:CONN.TYPE=SE SRLC

R.L(NH).C(PF)=0.1.75.1E6ELE NO 10:CONN.TYPE=PE PRLC

R,L(NH).C(PF)=12.1E6.2.6

ELE NO 2:CONN.TYPE=1,2 3.2 SPTABULATED DATA FROM TERMINAL7rDISC/rMAG TAPE7DIDISC FILE NAME=SPSNO. OF FIRST DATA LINE=48ELE NO=3ELE NO 3:CONN.TYPE=2.4 SRLC

R.L(NH),C(PF)=0.1.75.1E6ELE NO=4ELE NO 4:CONN.TY0E=4.0 PRLC

R.L(NH),C(PF)=12.1E6.2.6ELE NO=5ELE NO 5:CONN,TYPE=3.5 SRLC

R.L(NH).C(PF)=0.2.5.1E6ELE NO=6ELE NO 6:CONN.TYFE=5,0 SRLC

R.L(NH).C(PF)=0,0.1.23ELE NO=7ELE NO 7:CONN.TYFE=5,6 SRLC

R.L(NH),C(PF)=0.3.526.1E6ELE NO=8ELE NO 8:CONN.TYPE=6.0 OUTELE NO=9ELE NO 9:CONN.TYPE=ENDELE NO=0ANALYZE7rCHANGES7,WRITE?

ELE NO 11:CONN.TYPE=EBELE NO 12:CONN.TYPE=SSELE NO 13:CONN.TYPE=PP

ANALYZE'rCHANGES7PWRITE7DATATITLE: WIDEBAND AMPLIFIER 0.1-2.0 GHZ

ELE NO 14:CONN.TYPE=PE SRLC 100.000 2000.000 100.000

R.L(NH).C(PF)=0.0.1.233AIN

ELE NO 15:CONN.TYPE=SE SRLC 3A1

R.L(NH),C(PF)=0.3.526.1E6 GA2

ELE NO 16:CONN,TYPE=ENDVSWRSTAB

ANALYZE7rCHANGES/rWRITE7DATA 1 1. 5 SR 183.000 3.000 1000000.000

TITLE: WIDEBAND AMPLIFIER 0.1-2.0 GHZ 2 1. 2 3. 2 SP3 2. 4 SR 0.000 1.750 1000000.000

100.000 2000.000 100.000 4 41 0 PR 12.000 1000000.000 2.600

GAIN 5 3. 5 SR 0.000 2.500 1000000.000

GA1 6 5. 0 SR 0.000 0.000 1.230

GA2 7 5. 6 SR 0.000 3.526 1000000.000VSWR 8 6. 0 OU

STAB 9 EN1 PP ANALYZET.CHANGESTrWRITETAN2 SE SR 183.000 3.000 1000000.000 9:27 AM THU.. 9 SEPT. 19823 PP TITLE: WIDEBAND AMPLIFIER 0.1-2.0 GHZ4 SS5 SE SP6 SE SR 0.000 2.500 1000000.000 FRED ORIN GA1 GA2 VSWA STAB

7 SSMill DP .DEG MAG DEG NAG DEG IN OUT R GMAX

8 PB100.00 7.9012 168.7 .01632 -60.9 .05153 42.51 1.0331 1.109 1.1997 7.914200.00 8.0203 158.2 .03004 -94.3 .08821 55.37 1.0619 1.193 1.1925 8.057

9 SE SR 0.000 1.750 1000000.000 300.00 8.1068 147.9 .04392 -106. .12574 53.26 1.0918 1.288 1.1796 8.182

10 PE PR 12.000 1000000.000 2.600 400.00 8.1273 137.6 .05601 -114. .16155 47.41 1.1186 1.385 1.1690 8.254

11 EB500.00 8.2287 127.4 .06626 -118. .19374 39.95 1.1419 1.481 1.1521 8.415

12 SS600.00 8.1796 117.4 .07610 -127. .21531 32.06 1.1647 1.549 1.1465 8.416700.00 8.0844 107.8 .08748 -134. .23026 23.73 1.1917 1.598 1.1423 8.361

13 PP 800.00 8.1170 98.08 .09894 -140. .24158 14.58 1.2196 1.637 1.1324 8.430

14 PE SR 0.000 0.000 1.230 900.00 8.1983 88.45 .10829 -146. .24906 4.463 1.2420 1.663 1.1157 8.553

15 SE SR 0.000 3.526 1000000.000 1000.0 8.2038 79.17 .11750 -156. .24825 -6.08 1.2662 1.660 1.1015 8.597

16 EN1100.0 8.3249 69.14 .12478 -161. .24216 -18.8 1.2851 1.639 1.0837 8.7451200.0 8.3358 59.50 .13610 -171. .23090 -32.3 1.3150 1.600 1.0676 8.816

ANALYZE?rCHANGEST,WRITE? 1300.0 8.3885 50.06 .14229 -178. .21494 -47.9 1.3317 1.548 1.0500 8.9331400.0 8.4128 40.18 .15579 171.0 .19709 -68.0 1.3690 1.491 1.0284 9.1341500.0 8.4320 30.48 .16631 160.8 .19211 -91.2 1.3989 1.476 1.0028 9.6951600.0 8.4776 19.26 .17465 155.6 .20703 -119. 1.4232 1.522 .97357 0.0001700.0 8.4004 9.282 .18145 145.3 .23397 -143. 1.4433 1.611 .94698 0.000

ANALYZETTCHANGES7rWRITUANA 1800.0 8.2070 -1.01 .18659 138.1 .27146 -164. 1.4587 1.745 .92281 0.000

9:20 AM THU.. 9 SEPT, 1982 1900.0 7.9212 -11.1 .18901 128.2 .31268 176.8 1.4661 1.910 .90204 0.000

TITLE: WIDEBAND AMPLIFIER 0.1-2.0 GHZ 2000.0 7.6705 -22.2 .18843 121.3 .36456 160.1 1.4643 2.147 .87510 0.000

FRED GAIN GA1 GA2MHZ DP DEG NAG DEG MAO DEG

USIAIN OUT

STABP. GMAX

ANALYZET.CHANGESTrWRITET

100.00 7.9012 168.7 .01632 -60.9 .05153 42.51 1.0331 1.109 1.1997 7.914 ANALYZE7rCHANGES7rWRITE7DATA200.00 8.0203 158.2 .03004 -94.3 .08821 55.37 1.0619 1.193 1.1925 8.057 TITLE: WIDEBAND AMPLIFIER 0.1-2.0 GHZ300.00 8.1068 147.9 .04392 -106. .12573 53.26 1.0918 1.288 1.1796 8.182400.00 8.1273 137.6 .05601 -114. .16155500.00 8.2287 127.4 .06626 -118. .19374

47.4139.95

1.11861.1419

1.3851.481

1.16901.1521

8.2548.415 400.000 2000.000 100.000

600.00 8.1796 117.4 .07610 -127. .21531 32.06 1.1647 1.549 1.1465 8.416 GAIN700.00 8.0844 107.8 .08748 -134. .23026 23.73 1.1917 1.598 1.1423 8.361 GA1800.00 8.1170 98.08 .09894 -140. .24158 14.58 1.2196 1.637 1.1324 8.430 GA2900.00 8.1983 88.45 .10829 -146. .249061000.0 8.2038 79.17 .11750 -156. .24824

4.463-6.08

1.24281.2663

1.6631.660

1.11571.1015

8.5538.597 VSWR

1100.0 8.3248 69.14 .12478 -161. .24216 -18.8 1.2851 1.639 1.0837 8.745 STAB1200.0 8.3358 59.50 .13610 -171. .23090 -32.3 1.3150 1.600 1.0676 8.816 1 PP1300.0 8.3885 50.06 .14229 -178. .214941400.0 8.4128 40.18 .15580 171.0 .19709

-47.9-68.0

1.33171.3690

1.5481.491

1.05001.0284

8.9339.134

2 SE SR 183.000 3.000 1000000.000

1500.0 8.4320 30.48 .16631 160.8 .19211 -91.2 1.3989 1.476 1.0028 9.695 3 PP

1600.0 8.4776 19.26 .17465 155.6 .20703 -119. 1.4232 1.522 .97357 0.000 4 1. 2 3. 2 SP1700.0 8.4004 9.282 .18145 145.3 .23397 -143. 1.4433 1.611 .94698 0.000 5 2. 4 SR 0.000 1.750 1000000.0001800.0 8.2069 -1.01 .18659 138.1 .271461900.0 7.9212 -11.1 .18901 128.2 .312682000.0 7.6705 -22.2 .18843 121.3 .36456

-164.176.8160.1

1.45871.46611.4643

1.7451.9102.147

.92281

.90204

.87510

0.0000.0000.000

6 4. 07 3. 0

PROU

12.000 1000000.000 2.600

8 SE SR 0.000 2.500 1000000.0009 PP10 1. 0 SR 0.000 0.000 1.230

ANALYZET.CHANGES7rWRITETEL 1 11 1. 2 SR 0.000 3.526 1000000.000

ELE NO 1:CONN.TYPE=1.5 SRLC 12 2. 0 OU

R.L(NH).C(PF)=183.3.1E6 13 EN

ELE NO=2 ANALYZE7FCHANGESTTWRITE'

Perlow: ANA radio -frequency circuit simulation am: analysis 53

Program -size limitations

The present version of ANA is limited to amaximum of 100 elements. The maximumnode designator is 30, which means thatany network or subnetwork can have amaximum of 30 nodes. If, however, thereare several nodal subnetworks, each mayhave up to 30 nodes. The nodal numbersfor each subnetwork are completely unre-lated to those of the other subnetworks.

The maximum number of frequenciesfor any analysis is 21. The product of fre-quencies and tabulated elements (the numberof total lines in the TAB data file) is limitedto 84. If all 21 frequencies are used for ananalysis, this value limits the number oftransistors to four. However, reducing thenumber of frequencies to 10 allows the useof eight transistors.

The maximum number of coupled lines,or unhomogeneous coupled lines, is limited

to 20 pairs. If any of these limits areexceeded, ANA informs the user of theviolation.

Example: Wideband amplifier,Figure 5

The following analysis is done in three dif-ferent ways. The first method uses a pure,two -port analysis. It is shown here as anexample of a parallel branch within a seriesstructure that is within a parallel path. Thetopology is shown in Fig. 6a.

All the program prompts and the userresponses are shown. The initialization ofthe input data ends at element 16 by respond-ing with END. The input data is then visu-ally verified by using DATA, and an analy-sis is performed with the AN command.

The elements are then changed by usingthe EL XX command so that the secondmethod of analysis, purely nodal topo-

graphy, is used. This configuration is shownin Fig. 6b.

The third method of analysis uses bothtwo -port and two -nodal subnetworks, asshown in Fig. 6c. The data listing is shown(page 53), but the actual changing of ele-ments and analyses has been left out.

In this particular case, it is evident thatthe nodal configuration is far simpler thanthe two -port: this is not true in general. Thiscomplex case is included here to indicateANA's versatility in analyzing circuittopology.

AcknowledgmentThe writer wishes to thank the users ofANA for the many helpful suggestionsincorporated into the program. His specialacknowledgments go to R.M. Evans, A.Presser, B.S. Perlman. and G.E. Theriaultfor their continuous flow of ideas.

RCA Review available nowThe two most recent issues (June 1982 and September1982) of the RCA Review are now available. For additionalinformation, contact Ralph Ciafone, Editor, RCA Laborato-ries, TACNET: 226-3222.

Contents Page, June 1982

An Optical Communication Link in the 2.0-6.0 GHz BandDaniel W. Bechtle and Stefan A Siegel

Fluorescent Tracers-Powerful Tools for Studying CorrosionPhenomena and Defects in DielectricsWerner Kern. Robert B. Comizzoli. and George L. Schnable

Simplified Analysis of Kinescope Electron GunsD A deWolf

Miniature Beryllia Circuits-A New Technology for MicrowavePower AmplifiersF.N. Sechi, R. Brown. H. Johnson. E. Belohoubek,E. Mykietyn. and M. Oz

Interpreting the Beta Versus Collector Current and TemperatureCharacteristics of a TransistorRobert Amantea

Positive and Negative Tone Near -Contact Printing of ContactHole MaskingsLawrence K. White

Contents Page, September 1982

Chemically Vapor -Deposited Borophosphosilicate Glasses forSilicon Device ApplicationsWerner Kern and George L. Schnable

A Radiation Hardened 256x4 Bulk CMOS RAML.S. Napoli, R.K. Smeltzer, R. Donnelly, and J. Yeh

Single Sideband, Amplitude Modulated, Satellite VoiceCommunication System Having 6000 Channels per TransponderKrishnamurthy Jonnalagadda

Broadband Microwave Power Amplifiers Using Lumped -ElementMatching and Distributed Combining TechniquesR.L. Camisa and A. Mikelsons

Offset Near -Field Gregorian Antenna Scanning Beam AnalysisEugene C. Ngai

A Simplified Real Frequency Technique for Broadband Matchinga Complex Generator to a Complex LoadB.S. Yarman

The Photoresponse of Thin -Film PtSi Schottky -Barrier Detectorswith Optical CavityHammam Elabd and Walter F. Kosonocky

Theory of Large -Angle Deflection of Electrons in the MagneticField Inside a Television TubeBasab B. Dasgupta

54 RCA Engineer 27-6 Nov./Dec. 1982

OnlineLiterature

Searching

Did you know thatthe following databases

are available to youfor online searching?

Conference Papers Index,1973 to present-provides

access to records ofmore than 100,000 scien-

tific and technical paperspresented at over 1,000

major regional, nationaland international meetings

each year.

RCA TAD (TechnicalAbstracts Database),

1968 to present-providesaccess to records of more

than 18,000 RCA technicaldocuments, including tech-

nical reports, engineeringmemoranda, unpublished

manuscripts, patents, andtechnical notes.

Your RCA librarianwill be pleased

to assist youin searching

these databases.. .

and many others, too.

E.H. Adelson C.R. Carlson A.P. Pica

Modeling the human visual system

High -quality television pictures are the goal, but whatconstitutes quality? What kind of image processingproduces the best visual effect? To find out, theseRCA researchers are trying to quantify psychologicaland physiological aspects of the human visual system.

Abstract: Considerable progress hasbeen made recently in understanding cer-tain aspects of early information process-ing in the human visual system. For exam-ple, there is good evidence that the visualsystem is sensitive to information localizedin both space and spatial frequency. Usingthis and other results, we now can buildmodels applicable to several interestingimage -quality and image -processing prob-lems. Examples discussed include the pre-diction of visible changes in image sharp-ness and the development of efficientalgorithms for image encoding.

The final test of a television picture comeswhen a viewer looks at it and says, "That'sa good picture." For the viewer, this judg-ment is simple and requires no understand-ing of operations of the eye and brain.Indeed for all of us, seeing seems so directand effortless that we remain unaware ofthe complex visual processing that under-lies a statement like, "That's a good pic-ture." Coming to understand some of thatprocessing is part of the function of theImage Quality and Human PerceptionGroup at RCA Laboratories.

RCA has a long tradition of vision re-search. In the 1940s, Al Rose introducednoise analysis into models of vision, estab-lishing a law that bears his name (theRose-DeVries law) and a conceptual frame-work that is now basic to much work invisual discrimination. And in the 1950s,Otto Schade showed how we could con-sider the eye as part of an overall optical

1982 RCA CorporationFinal manuscript received October 8. 1982Reprint RE -27-6-9

system, and applied the techniques of lin-ear systems analysis to characterize its per-formance. Schade's work laid the founda-tions for much of our current thinkingabout both biological and nonbiologicaloptical systems: A symposium in Schade'shonor was held at this year's conventionof the Optical Society of America.

I

The contrast -sensitivity function

Schade's approach was to quantify humanvisual sensitivity by using gratings withsinusoidal luminance profiles-the same in-puts he had used for studying optical andelectro-optical systems.' A sinusoidal grat-ing is illustrated in Fig. 1. By varying thespatial frequency, and then measuring the

Fig. 1. Patch of a one-dimensional lurn nance sine -wave grating. At a viewing dis-tance of 18 inches, its frequency is approximately 3 cycles/degree.

56 RCA Engineer 27-6 Nov./Dec 1982

minimum contrast necessary for detection,Schade derived what is now known as the"contrast -sensitivity function" (CSF), whichis closely related to the "modulation -transferfunction" (MTF) familiar to engineers. Fig-ure 2 shows a typical contrast -sensitivityfunction: Note that sensitivity peaks atabout 3 cycles/degree, and drops at bothhigher and lower spatial frequencies.

You can visualize your own CSF bylooking at Fig. 3, which shows a sinusoi-dal grating whose frequency is swept loga-rithmically along the abscissa, and whosecontrast is swept logarithmically along theordinate. You should see an inverted, U-shaped envelope in which gratings are vis-ible. This envelope is not present in thepicture, but only in your perception of it.Move the page either closer or fartherfrom you, and the peak will shift.

The CSF is a natural quantity to mea-sure if one conceives of the visual systemas containing a single, broadly tuned, lin-ear filter, followed by a detector stage witha fixed threshold. But the visual system ismore complicated than that; more compli-cated measurements must be made if oneis to model the eye effectively. The currentview, pioneered by Campbell and Robsonand others around 1968, is that the visualsystem contains a bank of spatial -frequency -tuned channels, which function more or lessindependently of each other.'

In this model, the overall CSF of theeye is the outer envelope of the CSFs ofthe individual channels; Fig. 4a illustratesthe idea. Support for this scheme comesfrom adaptation experiments, in which one"fatigues" the channels' responses to a givenspatial frequency by having a subject stareat a sinusoidal grating for a long time.After such adaptation, it is found that anotch has been cut in the CSF, as shownin Fig. 4b. This indicates that the channelssensitive to the adapting stimulus have hadtheir sensitivities reduced.'

You can convince yourself of the adap-tation effect by staring at the grating ofFig. 1, and then examining Fig. 3. Adapta-tion takes about one minute, and youreyes should be constantly moving duringadaptation to avoid build-up of a staticafter -image. When you look at the sweptsine -wave pattern, a flattening or a dipshould be visible near 3 cycles/degree.The subset of filters tuned to around 3cycles/degree have been fatigued by theadaptation. and so they respond less strong-ly than before.

To understand how the filtering comesabout, we must begin at the retina, whichconsists of an array of about 100 million

0.2

2.0

20

3.0

SPATIAL FREQUENCY (CYCLES/DEGREE)

Fig.2. Contrast-sensit vity thresholds fcr the riurran eye for hign-luminance dis-plays (that is, above 30 fL). The ordinate is the threshold contrast necessary fordetection of a sine -wave grating.

0.2

oe

ii2.0

CCCC

C.,

20

0.3

0.3 3.0

SPATIAL FREQUENCY (CYCLES/DEGREE)

Fig.3. Plot of a sinusoidal grating whose frequency is swept logarithmically alongthe abscissa and whose contrast is swept logarithmically along the ordinate. Usingthis plot, you can see your own contrast -sensitivity function, which should be sim-ilar to that given in Fig. 2. The units of the abscissa are ralid when the figure isviewed from a distance of 18 inches.

30

30

Adelson/Carlson/Pica: Modeling the human visual system 57

photoreceptor cells. Each photoreceptor re-sponds to the light at one point in the vi-sual field. Next, the outputs of many pho-toreceptors in a small patch of retina arecombined into a new set of signals withinthe eye.

0.1

1.0

10

10001 10 10 100

SPATIAL FREQUENCY (CYCLES /DEGREE)

These signals are passed from the eye tothe brain through the optic nerve, which isa bundle of about I -million nerve fibers.Each fiber corresponds to a local regionon the retina,' and thus to a local regionin visual space. The output of a fiber repre-

Fig. 4a. A bank of spatial -frequency filters in the human visual system. Overallresponse determines the contrast -sensitivity function of Fig. 2.

01

ztj

oj I.04CC

O

0u)I 10

trF-

10.1 I 10 100

SPATIAL FREQUENCY (CYCLES /DEGREE )

Fig. 4b. Measured results from an adaptation experiment that support the repre-sentation indicated in Fig. 4a. In this experiment, an observer's contrast -sensitivityfunction was first measured, as indicated by the solid line. Next, the observerviewed a high -contrast, 6.4-cycle/degree "adapting" grating for an extended periodbefore again having his contrast -sensitivity function remeasured, as indicated bythe data points. It may be seen that only near the adapting frequency is theobserver's threshold sensitivity reduced. That is, the adaptation effect was fre-quency specific, supporting the idea that spatial -frequency information is pro-cessed by the human visual system in relatively narrow spatial -frequency bands,as indicated in Fig. 4a. These data are from reference 19, after reference 4.

seas. in effect, a weighted sum over space,which is called its "receptive field." Recep-tive fields of neighboring fibers overlap.The weighting functions have both posi-tive and negative parts; a typical cell'sreceptive field will resemble a "difference-of-Gaussians" shape. sometimes known asa "Mexican -hat function." Such a functionis shown in Fig. 5.

The difference-of-Gaussians functionserves as a simple bandpass filter, selectingout information in a certain spatial -fre-quency range. A large array of fibers, whosereceptive fields cover the visual scene onthe retina, can represent a bandpassed andsampled version of the scene. Of course,to fully represent the scene, all regions ofthe spectrum must be represented. whichis to say that there must be cells withreceptive fields of all sizes. This is true inthe eye: it is also true in the brain, wherecells are similarly found to be tuned forspecific hands of spatial frequencies. local-ized within finite regions of visual space.

ApplicationsBuilding on the concept of receptive fields.solutions to a number of interesting prob-lems can be obtained. For example. it hasbeen found that when the outputs of thereceptive fields are combined over spaceaccording to a simple summation rule, thatthe threshold visibility of arbitrary lumi-nance patterns can be accurately pnxiicted.'

O

I

4

(a)

(b)

Fig. 5. A receptive field. Figure 5a showsthe relative response of the receptivefield through its midline. Figure 5b showsthe receptive field response from above.In the center of the field, the receptivefield's response is positive. In a ringsurrounding the center, the receptivefield's response is negative.

58 RCA Engineer 27-6 Nov./Dec. 1982

(a) Original (b) 9J = -3 find'sFig 6. Examples of images that differ in their image sharpness by roughly 3 and10 Ind's. When viewed from about 18 inches, there is a 3-jnd difference betweenFig. 5a and Fig. 5b, and a 10-jnd difference between Fig. 5a and Fig. Sc.

In the two examples given below, thereceptive field concept is used differently.In the first example, where noticeable dif-ferences in image sharpness are predicted,it is assumed that the outputs of the recep-tive fields are combined over space in away that suggests the visual system per-forms a Fourier -like decomposition of spa-tially complex scenes. This simplificationallows a large number of problems to besolved analytically with reasonableprecision.

In the second example, where imageprocessing is achieved by use of a pyramidstructure, an image transform is constructedthat contains features of the receptive fieldstructure of the human visual system.' 10

Image transforms of this type have someinteresting advantages when compared tomore conventional approaches.

Predicting just -noticeable differencesin image sharpness

One of the principles of display design isthat the performance of a display shouldmatch the perceptual requirements of theobservers. One aspect of this principle isthat information that cannot be seen shouldnot be transmitted. Thus, a basic display -design issue is knowing when a change ina display variable will result in a perceiv-able change in display performance. Inthis section we describe a simple signal -de-tection model for the visual system thatpredicts answers to questions of this type.We will illustrate the application of themodel by applying it to the problem ofpredicting noticeable differences in dis-played image sharpness.

Since the work of Schade, researchers

have known that the overall spatial -fre-quency response of a display determinesits perceived image sharpness.' I One aspectof a display's frequency response is givenby its modulation -transfer function (MTF),which specifies the ratio of the displayedcontrast of sine -wave gratings divided bytheir input contrasts. Phase variations arenot included in the modulation -transferfunction.

When will a change in a display MTFresult in a perceivable change in displayedimage sharpness? The type of analysis re-quired to answer this question is calledjust -noticeable -difference (jnd) analysis. Onejnd is defined as the change in a stimulusnecessary for a viewer to see that change75 percent of the time. Perceptually, onejnd is obviously a small unit. But a 3-jndchange can be seen 99 percent of the timeand a 10-jnd change, or more, representsvery significant perceptual effects. Exam -

IMAGEI(X)

ETC

NOISE

Fig. 7. The visual signal -detection model

(c) eJ = -10 find's

pies of images that differ by roughly 3 and10 jnd's are shown in Fig. 6.

Fig. 7 shows a schematic of a signal -de-tection model for the visual system thataccurately predicts when changes in dis-play MTF can be seen. There are fiveelements in the model. The first element isa bank of spatial -frequency filters (or chan-nels) as is suggested by Fig. 4a. These actto decompose an arbitrary scene into rela-tively narrow, contiguous bands of spatial -frequency information. Each channel is cen-tered about a retinal frequency P, given incycles/degree, and has a bandwidthWe assume that luminance informationfalling within a channel is processed indepen-dently of information falling within anyother channel. We also assume that thechannel bandwidths, for frequencies above1.5 cycles/degree, are an octave wide: that

is, p/p equals 2/3. This estimate is inagreement with much psychophysical evi-

SQUARELAW

SPATIAL -TEMPORALINTEGRATOR

ETC

DETECT IFS/N K

Adelson/Carlson/Pica Modeling the human visual system 59

dence.I2 In making these assumptions wehave, in effect, taken the large number ofreceptive fields that exist over visual spaceand combined them into a relatively smallnumber of frequency -specific channels. Thissimplification allows us to develop an ana-lytic model for visual processing that iseasily applicable to a wide range of prob-lems and is still reasonably accurate. Thesecond element within each channel is theaddition of several sources of visual noisethat interfere with the perception of a sig-nal. The third element is a square -lawnonlinearity. The motivation for the nonlin-earity is provided by our interpretation ofbasic contrast -detection experiments. Al-though the explanation of these experi-ments requires a nonlinear stage, the actualchoice of a square law is motivated in partby its mathematical simplicity. The fourthelement of the model is a spatial -temporalintegrator to smooth the squared outputfrom each channel. The fifth and finalelement is a detection stage, which speci-fies the probability of detecting a signalchange occurring within the channel.

Whether a change in display MTF willresult in a perceived change in image sharp-ness obviously depends on the scenes beingdisplayed. We have found that for pictor-ial scenes a very good test image is a one-dimensional luminance edge transition.' Ob-servers are usually very sensitive to smallchanges in display MTF when looking atedges, and edges are often the most impor-tant feature in pictorial scenes. Other impor-tant display variables that determine wheth-er a noticeable change in image sharpnesswill occur are the display brightness anddisplay signal-to-noise ratio. We will assume

10

0.1

0 01

0 001

0205

0

here high -brightness displays (that is, above30 -foot lamberts) with high signal-to-noiseratios.

Once a scene has been selected, the sig-nal -detection model (Fig. 7) can be usedto predict when a change in display MTFwill result in a perceivable change in dis-played image sharpness. The operation ofthe model is straightforward, since changesin display MTF will result in changes inthe signals existing within the appropriatefrequency -specific channels. The model sim-ply converts these signal changes withineach channel to the appropriate number ofperceived jnd's.

To facilitate the application of the sig-nal -detection model, we have developed agraphical representation of it called a jnddiagram." A jnd diagram is shown in Fig.8a, where the quantity m(v)R(v) is plot-ted as a function of retinal frequency v, incycles/degree. Retinal frequency is givenby

v = rd. 180,where r is the viewing distance in inchesand where f is the display frequency incycles/inch. As before, R(v) represents thedisplay MTF; the quantity m( v) is a func-tion that represents the scene being viewed.For a one-dimensional edge transition, withedge height A/ and mean luminancem( v) can be shown to be given by

m(v) = (1 /1471-2)(..1/ 1)2(-1v v),

where -1v is the bandwidth of a frequency -specific channel and where v is the centerfrequency of the channel. If, as stated ear-lier, we let v v equal 2/3. and take a100 -percent contrast edge (that is, A / / =2.0) then m( v) = 0.14.

301

6 0 12

1024 4

RETINAL FREQUENCY, wr f /180 (CYCLES/DEGREE)

100

Fig. 8a. Jnd diagram appropriate for high -luminance displays.

1.0

0.I

0.01

0.001

02

Returning now to the jnd diagram, eachvertical line, located at the key frequenciesof 0.5, 1.5, 3.0, 6.0, 12, 24, and 48 cycles/degree, defines the center of a frequency -specific channel in the human visual sys-tem (as indicated in Fig. 4a). On each ver-tical line the distances between the smalltic marks indicate the change in m(v)R(v)

necessary over that channel for an observerto perceive a 1-jnd change in displayedimage sharpness. When jnd changes occurover more than one channel, the total per-ceived change, to a first approximation,can be obtained by simply adding thejnd's occurring over each frequency -spe-cific channel.'

Figure 8b illustrates two examples ofthe application of the jnd diagrams. Thisfigure is identical to Fig. 8a except for thefour curves, which represent different valuesof m(v)R(v). In all the cases, we have as-sumed a 100 -percent contrast -edge transi-tion as our image, so that m( v) equals

0.14. Thus, the differences between thecurves are only due to differences betweenthe MTFs, R(v).

Figure 8b, case A, represents the MTFof an extremely good display for whichthere exists only one full jnd (at 24 cycles/degree) in image sharpness between thisdisplay and the perfect display where R(v)

equals 1.0. Or, said differently, if the band-width of this display were increased infi-nitely, an improvement of only one jnd inthe sharpness of the image would be real-ized. Case B is more typical of televisionviewed at normal viewing distances. Forthis situation the increase in display band-width, defined where R(v) equals 1/2,necessary to produce one additional jnd in

10 10 100

RETINAL FREQUENCY, u rrf /180 (CYCLES/DEGREE)

Fig. 8b. Two applications of the jnd diagram shown in Fig. 8a.

60 RCA Engineer 27-6 Nov./Dec. 1982

image structure is only about 0.7 cycles/degree. This example indicates that, forpractical displays, perceivable increases inimage sharpness can result from relativelysmall changes in display MTF.

Figure 9 shows the results of actualexperiments with pictorial scenes and MTFssimilar to those shown in Fig. 8b.7 Thesolid line in Fig. 9 shows the predictionsfrom Fig. 8b, assuming that the pictorialimages are approximated by a 100 -percentcontrast -edge transition. It may be seenthat the predictions of the model, as sum-marized in the jnd diagram. correspond tothe measured values obtained with pictor-ial scenes.

In conclusion, this example has shownthat it is possible to incorporate a numberof important visual properties into a rela-tively simple model that accurately pre-dicts when a change in display MTF willbe seen. A large number of other applica-tions have also been computed, includingthe perceptual effects resulting from eithersampled or raster displays.

Image processing based on thevisual system

By understanding the visual system, wecan also develop techniques for the effi-cient digital encoding of images. The goalis to represent images in a way that empha-sizes image information that is importantto the eye, at the expense of informationthat is redundant or visually insignificant.Digitized images are normally representedin terms of the intensities of individualpixels. A popular alternative is to store atransform representation in which the coef-ficients of, say, the Fourier transform ofthe image are stored and the image is laterreconstructed via the inverse transform. Inan interesting variant on the transform cod-ing theme, we can generate a representationin which the basis functions resemble thereceptive fields in the human visual sys-tem. The image is then represented in termsof the responses of many receptive fieldsof many sizes, located at many positions.

The basis functions in such a schemeare not orthogonal, which leads to somecomplexities in generating and invertingimage representations. However, Peter Burtof Rensselaer Polytechnic Institute has devel-oped a convenient computational structurefor manipulating this kind of encodingscheme.9 We have been collaborating withBurt in applying the scheme to the effi-cient encoding and decoding of images."'

The basic idea is shown in Fig. 10. A256 X 256 pixel image is blurred by aGaussian blur function in successive stages.

I0

FREQUENCY, v

1 e 0.1

xo x

0

0010.11.0 10

SPATIAL FREQUENCY, v1 (CYCLES/DEGREE)

Fig. 9. Measured data for the change in display bandwidth, from P to v., neces-sary to produce a 1-jnd difference in image sharpness as a function of the initialdisplay bandwidth vl. The bandwidths are all defined where R(v) equals 1 /2. Dif-ferent symbols represent different scenes The solid line is the predicted resultfrom Fig. 8b obtained when the change in any one channel was 1 Ind.

a

0x

CROWD SCENEo MANIKIN SCENE 82% CONTRAST EDGE

At each step, the effective size of the Gauss-ian blur function is doubled. At the sametime, the linear sampling density is halved(so the number of samples per image areais quartered). Next, difference-of-Gaussiansimages are formed by subtracting the valuesof adjacent blurred images. The result is asequence of images representing differentspatial frequency bands. Information withineach image is localized in both space andspatial -frequency. The bands are spacedlogarithmically in frequency. since eachimage represents frequencies an octave be-low that of its predecessor. And the sam-pling density becomes sparse as the fre-quency becomes lower.

The resulting representation is called a"pyramid," because each image is com-puted from the one below and becausethere is a 4 -to -1 convergence of pixelsfrom one level to the next. The number ofsamples in the highest frequency bandpassimage (that is, the image at the bottom,left, of Fig. 10d) is the same as the numberin the original image. The next -lower levelhas 1/4 as many, the next has 1/16 asmany, and so on. Thus, the image is repre-sented by 1 + 1/4 + 1/16 + . . . or4/3 times as many samples as were in theoriginal image. But note that 75 percent ofthese samples represent the high frequen-cies. and only 25 percent represent themedium and low frequencies.

The high -frequency image is densely sam-pled, but its sample values can be coarsely

quantized. That is, in digitizing the samplevalues, one can use a small number of bits.There are two reasons for this. First, thereis relatively little high -frequency energy inmost scenes, so that the intensity values inthe high -frequency image tend to be small,clustering around zero. Thus, the dynamicrange and the entropy are small to beginwith. Second, the eye is fairly insensitiveto contrast errors in the high frequencies,and so the errors due to coarse quantiza-tion tend not to be noticed.

The medium- and low -spatial -frequencyimages require finer quantization. There isrelatively more energy in these frequencybands, and the visual system is more sensi-tive to the errors. However, these imagesare sampled more sparsely. Only 25 per-cent of the samples in the image code cor-respond to the medium and low frequen-cies. Thus, the high -frequency image isdensely sampled but coarsely quantized,while the other images are sparsely sam-pled but finely quantized. The result is anefficient encoding that puts the informa-tion where it is needed.

This encoding scheme may be used inconcert with other techniques to store ortransmit an image at a rate of about 1.5bits/pixel, which is considerably lower thanthe standard rate of 8 bits/pixel. An exam-ple is shown in Fig. 11.

The pyramid encoding is also quite con-venient for a procedure called "progressivetransmission." Suppose that an image is

Adelson/Carlson/Pica: Modeling the human visual systerr 61

GAUSSIAN PYRAMID

LAPLACIAN PYRAMIDFig. 10. The sequence of steps by which the pyramid repre-sentation is built. In the top row, the image is convolved witha sequence of Gaussian blur functions, leading to a se-quence of low -passed images. In the bottom row, "difference-of-Gaussians" images are generated by subtracting adja-

being transmitted over a low -bandwidthchannel, such as a phone or teletext line. Itmay take as much as a full minute totransmit the fully detailed image, duringwhich time the viewer patiently (or impa-tiently) waits. In progressive transmission,a spatially coarse version of the image issent initially, and finer and finer details areprovided as time progresses."' Is Thus, theviewer quickly gets a sense of what theimage is. He can wait longer to see it infull detail or, if he prefers, he can skip to anew image as soon as he sees what thecurrent one is. Figure 12 shows how thepyramid representation can be used in trans-mitting a progressive sequence of images.The information in the low frequencies issent first, followed by successively higheroctaves of frequency. In the example shownhere, additional data -compression tech-niques were used and the blur functionsused were somewhat different from thedifference-of-Gaussians. The numbers belowthe images indicate the number of bits/pixel required for each stage of the pro-gressive transmission. Even at 0.1 bits/pixel,we can get a good sense of what the pic-

cent low-pass images. The resulting images represent infor-mation within a given spatial -frequency band, and the bandcenter drops by one octave from one image to the next.Low -frequency images require fewer samples, in proportionto the square of their band center.

ture portrays. The lady can be recognizedat 0.31 bits/pixel; most of the details areavailable at 0.81 bits/pixel; and the lastdetails are filled in at 1.58 bits/pixel. The

pyramid appears to be well suited for avariety of other image -processing tasks, in-cluding preprocessing for pattern recogni-tion and image enhancement.

(a) (b)

Fig. 11a. The original image before encoding. It is represented as a 256 by 256array of pixels, with 8 bits/pixel. Figure 11b is the same image reconstructed froman encoded representation, based on the pyramid structure of Fig. 10. High -fre-quency information is coarsely quantized but finely sampled; low- and medium -frequency information is finely quantized but coarsely sampled. The result is anencoding requiring only 1.5 bits/pixel, which remains faithful to the original.

62 RCA Engineer 27-6 Nov./Dec 1982

0.03 0.10 0 31

Fig. 12. Progressive transmission of an image by means ofthe pyramid scheme. The low -frequency information is sentfirst, and high frequencies are sent as time goes on. Onecan get a good impression of the final image content quite

ConclusionsA central concept in much of current visionresearch is that the visual system performsan early decomposition of images by theuse of receptive fields, which are filtersselectively responsive to limited ranges ofspatial frequency and space. The outputs

0.81 1.58quickly. without waiting for all the bits to be transmitted. Thenumber beneath each picture indicates the accumulatednumber of bits/pixel required at that stage of transmission.

of filters are further limited by inherentnoise before they are sent to a detectorstage, which combines their outputs overspace and spatial frequency. Although thesegeneral properties of the receptive fieldsare established, the specific properties ofthe receptive fields remain areas of active

research in our laboratory and in manyothers.

As the two examples outlined here demon-strate, however, we now have sufficientunderstanding of the receptive field prop-erties to realize some interesting results.These ideas have also been applied to a

011001~041

Authors Edward Adelson, Albert Pica, and Curtis Carlson.

Edward Adelson is a Phi Beta Kappagraduate of Yale University, where hemajored in Physics and Philosophy. Hereceived his Ph.D. in experimental psy-chology at the University of Michigan andperformed postdoctoral research at NewYork University. His research areas havebeen in vision, visual perception, andimage encoding. Currently. Ted is amember of the Image Quality and Human

Perception Research Group at RCALaboratories, where he is involved indeveloping models for human motion pe--ception and in developing image -process-ing algorithms based on aspects of thehuman visual system.Contact him atRCA LaboratoriesPrinceton, N.J.TACNET: 226-3036

Curtis Carlson is a Tau Beta Pi graduatefrom Worcester Polytechnic Institute,where he received his B.S. degree in Phys-ics in 1967. He received his M.S. and Ph.D.degrees from Rutgers University in 1969and 1973. respectively. Since joining thetechnical staff of RCA Laboratories in 1973,Curt has been involved in the generalareas of imaging devices, image analysis.and human perception. He is a member ofthe SMPTE Vision Committee on HighDefinition Television and was recently aSpecial Editor of an issue of the SID Pro-ceedings on "Advances in Visual Informa-tion Processing." In 1979, he received anRCA Individual Outstanding AchievementAward. Curt is currently Head of ImageQuality and Human Perception Researchat RCA Laboratories.Contact him at:RCA LaboratoriesPrinceton, N.J.TACNET: 226-2436

Albert Pica joined RCA Laboratories in1977. He received his B.S. in Physics fromthe City College of New York in 1974. andhis Masters from Rutgers University in1976. Since joining RCA he has worKed onthe CED VideoDisc System helping to builda digital scanning capacitance microscopethat measures disc -stylus capacitancevariations. He has also performed researchin visual perception. Albert is presentlyworking in the Image Quality and HumanPerception Research Group. where hespecializes in computer and image pro-cessing systems as well as computer sim-ulations of human visual performance.Contact him at:RCA LaboratoriesPrinceton, N.J.TACNET: 226-2859

4Adelson/Carlson/Pica: Modeling the human visual system 63

number of other television problems, suchas predicting the required signal-to-noiseratio of different displays°: determiningthe number of phosphor -stripe triads re-quired to produce a high -quality color -television image' and establishing the con-vergence requirements for color television.''

One future objective is to use some ofthese ideas to perform image enhancementwithin commercial receivers. As the costof digital memory comes down, more im-age processing within the receiver will be-come practical. The basic question is then:What kind of image processing will pro-duce the best visual effects? Since repre-sentations like the pyramid structure describedearlier incorporate the receptive field con-cept of the visual system, they promise tobe useful in answering this question.

AcknowledgmentsWe wish to express our warm apprecia-tion to our colleagues. C.H. Anderson. P.J.Burt, R. Klopfenstein, and R.Sverdlove,for their many contributions to this research.

We also give special thanks to E. Cuomoand G. Zak for making Figs. I and 3.

ReferencesI. O.H. Schade. Sr.. "Optical and Photoelectric Analog

of the Eve." J. Opt. Soc. Amer., Vol. 46, p. 721(1956).

2. F.I.. VanNess and M.A. Bowman. "Spatial Modula-tion Transfer in the Human Eye." J Opt. Soc.Amer.. Vol. 57. p. 401 (1967).

3. F.W. Campbell and J.G. Robson. "Application ofFourier Analysis to the Visibility of Gratings," .1.

Physiot (London). Vol. 197. p. 551 (1968).

4. C. Blakemore and F.W. Campbell. "On the Exist-ence of Neurones in the Human Visual SystemSelectively Sensitive to the Orientation and Sim ofRetinal Images." .1. Pitysiol. (London). Vol. 203. p.

237 (1969).

5 E.C. Carterette and M.P. Friedman. Handbook ofPerception. Vol. V: Seeing. Academic Press. Nev..

York. N.Y. (1975).

6. C.S. Harris. Ed.. !Isnot Coding and Adaptability.Laurence Erlbaum Ass. Inc.. Hillsdale. N.J. (1980).

7. C.R. Carlson and R.W. Cohen. "Visibility of Dis-played Information." Technical Report to the Officeof Naval Research. Cont. No. N0014 -74-C-0184(July 1978).

8. C.R. Carlson and R.W. Cohen. "A Simple Psycho -physical Model for Predicting the Visibilit!, of Dis-played Information." Pro. Soc. IplOr. Display. Vol.21. p. 229 (1980).

9. P.J. Burt. "Fast Filter Transforms for Image Process-ing." Comp. Graphics and Image Pro.. Vol. 16. p.10 (1981).

10. P.J. Bun and E.H. Adelson, "The Laplacian Pyramidas a Compact Image Code." Rensselaer PolytechnicInstitute I.P.L. Report No. 25 (March 1982).

I I. O.H. Schade, Sr., Image Quality. RCA Laboratones.Princeton. N.J. (1975).

12. H.R. Wilson and J.R. Bergen. "A Four MechanismModel for Threshold Spatial Vision." Vision Res,Vol. 19. p. 19 (1979).

13. C.R. Carlson and A.P. Pica. "Predicting Just Notice-able Differences in Image Sharpness: The Universal

JND Diagrams." Private Communication.

14. V.R. Sloan, Jr. and S.L. Tanimoto, "ProgressiveRefinement of Raster Scan Images." IEEE Trans.Comm Vol. ('-28. p. 871 (1979).

IS. K. Knowlton. "Progressive Transmission of Gray -Scale and Binary Pictures by Simple. Efficient andLossless Encoding Schemes." Pro. IEEE, Vol. 7. p.885 (1980).

16. R.W. Cohen. "Applying Psychophysics to DisplayDesign." Phoi Sci Eng. J. Vol. 22, p. 56 (1978).

17. C.R. Carlson, "Application of Psychophysics to Dis-play Evaluation," Pro. of the /982 Int. Display ResConf. Cherry Hill. N.J. (October 1982).

18. C.R. Carlson. "Convergence Requirements for Color

Television." Private Communication.

19. R.W. Cohen. C.R. Carlson. and G.D. Cody. "ImageDescriptions for Displays." Office of Naval ResearchTech. Report No. N00014-74-C-0184 (May 1976).

Schade's book available from RCA

Image Quality -A Comparison of Photographic and Tele-vision Systems, by Otto H. Schade, Sr., gives a technicaloverview of the concepts developed by the author, andnow in universal use. that permit a quantitative evaluationof image quality. Dr. Schade describes in some detail thethree basic parameters that determine image quality: theintensity -transfer function, which is a measure of the grayscale; the modulation -transfer function, which is ameasure of sharpness and definition; and the particle orquantum density that can be stored in the sensor of thecamera, which is a measure of granularity, or noise.

A unique feature of the book is a series of 54 unusuallyhigh -quality reproductions of photographic and televisionimages that dramatically illustrate the effects of variousparameters on image quality. For example, the same sub-ject is shown as reproduced by television systems having525 lines, 4.25 MHz; 625 lines, 5 MHz; 525 lines, 7 MHz;625 lines, 9 MHz; and 1760 lines, 60 MHz. Another set of

reproductions shows photographic images of a subjectmade with films of different speeds and with formats

requiring different magnifications. The story told by theseillustrations will be readily perceived by the expert and thelayman alike.

Most of the reproductions measure 81/2 by 6'/4 inches.Some have appeared in various of Dr. Schade's papersscattered throughout the technical literature (althoughusually in a much smaller format and with less detail) andothers have not been published before. Here, for the firsttime, they are brought together in a single volume.

Otto Schade has been active in the field of television formore than thirty-five years. His pioneering work in the1940s and 1950s led to the concept of ModulationTransfer Functions and Noise Equivalent Pass Bands,which can be applied equally to amplifiers, lenses, and thehuman eye. He made the first measurements on thehuman visual system in terms of these parameters. Dr.Schade's work has received worldwide acclaim.

For your copy send $20.00 (U.S.A.) or $22.50 (else-where) to:

Scientific PublicationsRCA LaboratoriesPrinceton, New Jersey 08540

64 RCA Engineer 27-6 Nov./Dec. 1982

R. Nigam IS. Hong 'S.R. Spence

Business modeling in an engineering context:An Astro-Electronics application

Operations Research developed a successful model to analyzecomplex mission requirements. The focus was on businessissues. while technical underpinnings were retained.

Abstract: This paper describes a modelthat has been developed to assist Astro-Electronics in analyzing the businessimpact of design alternatives of a multi-satellite system. The model simulates theaggregate -level technical systems and bus-iness requirements-satellite design, orbi-tal placement, launch timing, and sparespolicy. To ensure that the business aspectsare reflected well, the model also capturesthe technical microstructure-the detailsas they relate to operation, reliability, andcost. The model demonstrates how engi-neering and business aspects can beblended together to produce a total per-spective for decision making.

The primary use of the model is to testand refine various alternatives until a bal-ance is struck among cost, risk, and avail-ability requirements Results from themodel can then be used in setting initialbidding prices on contracts and in produc-tion scheduling. This model was used forthe conceptual design study of the NationalOceanic Satellite System (NOSS)-aproposal in the 700 -million -dollar rnnge-and the Defense Meteorological SatelliteProgram (DMSP) Block 5-D satellites.

RCA Astro-Electronics designs, develops,and manufactures satellites on a contrac-tual basis. It has been doing so since theearly 1960s and has successfully partici-<1982 RCA CorporationFinal manuscript received October 1, 1982Reprint RE -27-6-10

paced in TIROS, ITOS, Satcom and numer-ous other satellite systems ranging frommeteorological and communications to de-fense -related systems. Astro-Electronics iscurrently experiencing the largest backlogof work in its history.

The contracts are awarded on a verycompetitive basis. Hence, a key element inthe success of Astro's business is the abilityto design a satellite system that meets thecustomer's mission requirements-overalloperational and reliability requirements-in a very cost effective manner. A concep-tual design is a necessary first step in theprocess of bidding for satellite systems.

In a given design, alternatives can differin the number, configuration, and reliabil-ity of the satellites built, in the schedule ofbuilding and launching the satellites, andin the replacement policy of failed satel-lites. Other operational policies consideredare the use of standby satellites and satel-lite retrieval by the space shuttle. Thispaper describes a model developed to helpAstro-Electronics analyze the business im-pact of design alternatives of a multisatel-lite system. The model simulates aggregate -

level technical, systems, and business re-quirements. For a given alternative, themodel determines the system availabilityby trying to meet the operational require-ments with the specified resources and cal-culates the time -phased costs. The modelcan be used iteratively to arrive at the bestbalance among cost, risk, and availabilityrequirements.

Complexity of the situation

The primary concern in any contractualarrangement is the mission objectives, andthere are numerous ways in which theycan be satisfied. Before proceeding further,some definitions are necessary. A satelliteconsists of essential bus systems (for ex-ample, power, attitude control) and sensorsystems (for example, atmospheric temper-ature -profile measurement). Each of thesesystems has several identical subsystems(for redundancy) performing the function.Figure 1 gives a simplified schematic of asatellite's systems. To meet the missionobjectives, Astro-Electronics has to answerquestions relating to the number and typeof satellites to be used, their overall design,and general performance and lifetime char-acteristics at an aggregate level. At a finerlevel, trade-offs between redundancy in sen-sor and bus systems, and trade-offs betweenredundancies and satellite lifetime, amongothers, have to be resolved. At a still finerlevel, issues of which specific componentsshould be chosen, their lifetime and relia-bility characteristics, costs and learningcurves, all have to be addressed. It thusinvolves looking at the entire spectrum-from the business issues, to the satellitesystem issues, to the technical issues ofcomponent reliabilities and redundancies.

Increasingly, a customer's requirementsmust be or are best satisfied with multiplesatellites, operating simultaneously and/orconsecutively. The number of satellites thatmust be operating simultaneously are usu-

RCA Engineer 27-6 Nov./Dec. 198265

Fig.1. This satellite schematic shows redundancies in bus and sensor systems.The satellite model consists of systems in series (the bus) and the sensor systemsin parallel. Each of the systems has a number of identical subsystems in parallel.Subsystems may be redundant in either active or standby mode.

ally determined from the satellite system'scapacity or coverage requirements. For ex-ample, a meteorological satellite has a lim-ited payload of instruments and can moni-tor only selected areas of the earth. Consecu-tive launch and operation of satellites maybe necessary to maintain the system dur-ing its life.

With a multiple satellite system, the num-ber of system designs is greatly increased.Alternative designs can differ in the number,configuration, and reliability of the satel-lites built; in the schedule of building andlaunching the satellites; and in the replace-ment policy of failed satellites. With theadvent of the space shuttle, which allowsthe retrieval of satellites and the use of in -orbit standby satellites, planners are facedwith the increasingly difficult task of assess-ing many possible system configurations.

Compounding the complexity of theseissues is the fact that the performance ofsatellite systems is subject to a multitude ofuncertainties. For example. the orbital place-ment of an individual satellite involves some

elements of chance. These chance eventsare usually addressed by adding redundantcomponents to each individual satellite andby including entire redundant satellites inthe system. Therefore, the difficulties in-volved in making cost-effective designchoices are amplified in order to reduceuncertainty. However, competitive biddingprocedures dictate that such choices bemade.

New approachOver the last twenty-five years, satellitetechnology has made major advances sothat it is indeed a different world now.This period has seen growth from small-satellite/single-objective missions to multi-functional satellites with mission definitionsspanning several years. Over the years,Astro-Electronics had developed rather so-phisticated computer programs to assistthem in designing satellites. But the per-formance of an individual satellite doesnot address the complexity inherent in a

more complex mission objective. To addressthis situation, Astro's Engineering manage-ment asked the Corporate Operations Re-search group, whose charter is to promotethe use of quantitative methods in man-agement, to assist them.

The approach we took was to explicitlyshift the focus to the business issues, whileretaining the technical underpinnings. It

was more of a top -down approach. Thebusiness issues played a vital role in thedevelopment of the model.

Business issuesAs mentioned earlier, a satellite design isjudged only in the context of a mission orsystem requirement. A typical mission re-quirement may be gathering certain infor-mation (cloud pictures, infrared images,temperature profiles, and so on) over aperiod of several years. In responding tothis requirement, Astro has the followingalternatives:

It can build inexpensive short-lived satel-lites that will be replaced again and againover time, or it can build more costlylong-lived satellites (even up to the lifeof the mission).

It can put all the sensor systems in all thesatellites (but with a reduced reliabilitybecause of weight limitations), or equipcertain satellites with certain sensor sys-tems, thereby spreading the load to sev-eral satellites.

To ensure that a certain minimum cov-erage will be provided, Astro-Electronicscan allow for an in -orbit spare (an orbit-ing satellite that will be fully activatedwhen a satellite failure occurs), or it canreduce the risk of not meeting the mis-sion requirements over an extended periodof time by having satellite inventory onthe ground.

Astro-Electronics can provide for differ-ent levels of on -ground satellite inven-tory. The basic decision is how manysatellites of each type should be in inven-tory. The inventory costs are traded offagainst satellite system availability.

Now that the shuttle is operational (ornearly so), the question of what to dowith satellites that have catastrophic fail-ures is one that adds to the decisioncomplexity. The satellite can be broughtback, repaired, refurbished and launchedagain. There are costs and risks in thisprocedure but there are also savings tobe achieved.

66 RCA Engineer 27-6 Nov./Dec 1982

Note that in dealing with these broadermission issues, the underlying technical is-sues of reliability and redundancy cannotbe ignored. In fact, several of these mis-sion objectives have become possible onlybecause of the technological advances insubsystem weights, costs, and reliabilities.As a result, the mission can be accomp-lished in many techically feasible ways,and thus choosing the best alternative is adominating business issue. In the past, thetechnical issue was itself the business issue.The challenge to Astro-Electronics is tofind the satellite system designs that arecost effective and will stand the challengeof competition.

These business issues were explored ingreat detail with Engineering management:it took several rounds to define the scopeof the problem. The Operations Researchgroup then responded by developing a Monte -Carlo -based computer -simulation modelcalled the Life Cycle Cost model (or, LCC,for short).

The Life Cycle Cost model

For a proposed satellite ,tein design, LCCgenerates the availability of the satellitesystem, the spacecraft reliability, and thecost data from a number of trials, eachone of which represents a possible historyof the satellite system. The model usesMonte -Carlo simulation-a technique forstudying the performance of systems involv-ing random variables-to accomplish this.The model has the flexibility to simulatediverse satellite systems-systems that differin the number, configurations, and reliabil-ities of the satellites built; in the scheduleof building and launching the satellites:and in the replacement policy of failedsatellites. For a given systems design, themodel simulates in detail the satellite -launchoperation, satellite performance, and neces-sary satellite -replacement operations, andit produces reports on the system perfor-mance and on all costs incurred.

To evaluate any alternatives, the satel-lites are considered in detail from the pointof view of reliability and cost. Each satel-lite is modeled as a number of support sys-tems in series (the bus) and a number ofsensor systems in parallel. Each bus orsensor system is composed of a number ofidentical subsystems in parallel (Fig. 1).

Some of the subsystems may be redun-dant, either in active or standby mode.More specifically, the user specifies-foreach system-the total number of subsys-tems installed, the number of these subsys-tems desired to be active, and the min-

imum number of active subsystems requiredfor the system as a whole to be still calledoperational. Also specified is the numberof subsystems that should be active whenthe satellite is an in -orbit standby.

The life of an active subsystem is deter-mined by its reliability function, such asan exponential decay with normal wear -out or a probability distribution table, ob-tained through an independent analysis.When the subsystem is in standby, its lifeis determined by a scaled reliability distri-bution of the active case. Standby subsys-tems are automatically activated, with aspecified probability of success, to replacefailed subsystems that were active. Reli-ability and cost distributions are specifiedat the subsystem level.

Because of the series/parallel structure,satellite failure is caused by the failure ofeither a bus system or all the sensor sys-tems. When an active satellite fails, an in-

( START ):LEAR VARIABLES

TO ZERO

READ INPUT

INITIALIZATIONS

orbit standby satellite is activated if avail-able, and a new launch may be scheduled.Here, the policy governing the numberand usage of standby satellites and launchschedules is controlled by the user input.

The launch sequence is a function ofthe satellite inventory arrival times andsensor requirements and will vary in eachtrial in response to the different failurepatterns that occur. When a particular satel-lite has to be launched, the model sched-ules the earliest possible launch. The simu-lation of the launch process, the insertioninto the right orbit, the turn -on of subsys-tems, all determine the initial state of thesatellite in the sky. These events surres-fully occur according to user -specified prob-abilities, based on prior design experience.

The model captures the time cycle ofthe manufacture of satellites in considera-ble detail. The model progresses on a month-to -month basis, while getting all the coin-

NTRIAL

GENE RATEREPORTS

STOP

DETERMINE RECURRING ANDNONE ECURRING MFG. COSTS

I\ITIALIZATIONS

MONTH

BLAST

COL .ECT OPERATIONALRESULTS OF TRIAL

STORE COSTSINCURRED IN TRIAL

B

MONTHLYEVENTS

Fig. 2 This simplified system -level flowchart of LCC shows the two primary loops:one for the trial number and the other for the launch month. This structure iscommon to many simulation models.

Nigam/Hong/Spence: Business modeling in an engineering context An Astro-Electronics application 67

MONTHLY EVENTS

PERFORM LAUNCHES SCHEDULEDFOR THIS MONTH

UPDATE OPERATIONAL INFORMATION

TURN ON STAND BY SATELLITESTO MEET SYSTEM REQUIREMENT

SCHEDULE LAUNCHES TO MEETSYSTEM REQUIREMENT

SCHEDULE LAUNCHES TO MEETSTAND BY LAUNCH TRIGGER UNITS

RECORD SYSTEM STATUS

IF A SATELLITE SHOULD BERETRIEVED, ATTEMPT RETRIEVAL

UPDATE OPERATIONAL IMFORMATION

FAIL COMPONENTS WHOSELIFETIMES END THIS MONTH

ACTIVATE STANDBY COMPONENTSTO REPLACE FAILED COMPONENTSCHECKING SWITCH PROB.

DETERMINE FAILURE TIMES OFSUCCESSFULLY ACTIVATEDCOMPONENTS

INSERT TIMES OF COMPONENTFAILURES TO THE EVENT LIST

CHECK IF ANY SATELLITESHAVE FAILED

Fig. 3. Every month in the simulation, LCC checks the status of subsystems, sys-tems, and satellites and schedules launches to determine the required actions.

ponents (and incurring related costs), start-ing satellite assembly, and putting it throughtests before it is ready for inventory. Thenumber of satellites, the satellite configura-tion built, and the lead time to completethe task are all controlled via user inputs.

The model keeps a tab on the bus andsensor subsystem failures and also recordswhether or not they are still operational.From this, after looking over all the satel-lites, one can determine whether overallmission objectives are being met or not.

The costs in the model are of two types.The nonrecurring costs for a subsystemcapture the development cost required be-fore a subsystem may be manufactured.This cost is incurred once and may beallocated over consecutive quarters. Therecurring costs in the model are meant toinclude subsystem manufacturing, launch,inventory, refurbishment, and repair costs.These costs are incurred every time theassociated operation occurs. The cost ofmanufacturing every subsystem is consid-ered to be probabilistic, and the mean costis adjusted downward as the number ofpreviously manufactured components of the

same type increases, according to a learn-ing -curve rate input by the users. The manu-facturing costs of components are also cor-related to reflect the reality that improve-ments in the design or manufacture of onesubsystem may spill over in the cost improve-ments of another subsystem. All costs inthe model are normally distributed. Sincethe cost expenditures occur at differenttimes, the means and standard deviationsof the discounted costs are also calculatedto provide a meaningful comparison amongalternatives. The standard deviations of thecost allow for an assessment of the riskprofile and enable the largest sources ofcost variation to be identified.

The input to the model consists of a setof parameters that describe the missionobjectives for operational sensors as a func-tion of time, the subsystem and launchvehicle reliability, the mean and standarddeviation of nonrecurring and recurringcosts, the number and configuration ofeach satellite type built, the schedule ofbuilding the satellites, the launch vehicleused for each satellite type as a function oftime, the prespecified launches of each satel-

lite type, and the replacement policy offailed satellites.

Model logicA flowchart of the program's logic is shownin Figs. 2 and 3. The time increment inthe simulation is one month. First, LCCsimulates all launches scheduled for thecurrent month. This launch may have beenprescheduled by the user or dynamicallyscheduled by the model. When a launch isattempted, the launch vehicle may fail orthe final orbital placement may be unsuc-cessful. To take advantage of the possibili-ties offered by the shuttle, recovery of thesatellite can be attempted where applic-able. If recovery is successful, the satelliteis refurbished and returned to the satelliteinventory.

After all scheduled launches for the cur-rent month have been simulated, LCCchecks the status of the entire system. If,due to bus or sensor -system failures, sys-tem demands are not satisfied, the modelwill activate available in -orbit standby satel-lites that contribute to meeting the missionrequirements. A satellite is activated byswitching on enough of its standby com-ponents to bring it up to its active configu-ration. Successfully activated systems aregiven a new lifetime from their unsealedlifetime distribution. If mission demandsare still unsatisfied, one or more replace-ment satellite(s) may be scheduled forlaunch. The model selects the satellite(s)from inventory whose type will satisfy thelargest number of unmet demands (as deter-mined by the different sensor systemsaboard an available satellite).

Finally, the standby satellite requirementsare checked, and additional launches arescheduled if necessary. The model allowsfor launch preparation time when it sched-ules the launches in the simulation. Thecurrent model logic reacts to unsatisfiedmission demands only as they occur ratherthan anticipating wear -out failures. Becauseof this, it is meant to be used in an itera-tive mode. For a given scenario, if theachievement of mission objectives is un-satisfactory (as in Fig. 6), then this indi-cates that the replacement policy needs tobe changed by the addition of a standbysatellite or a prescheduled launch to avoida drop in availability. The model wouldthen be rerun to test the revised replace-ment policy.

Model outputs and usesAfter completing the specified number oftrials, usually around one hundred, the

68 RCA Engineer 27-6 Nov./Dec. 1982

CAUSE OF SATELLITE FAILURE

SAT ONE SAT TWO

ADACS 0.03 0.0

PWR 0.29 0.0

THERM 0.0 0.0

SOLAR 0.21 0.0

TLM 0.02 0.0

COMM 0.0 0.0

CC 0.30 0.0

DLS 0.07 0.0

ADACS2 0.0 0.05

PWR2 0.0 0.27

THERM2 0.0 0.0

SOLAR2 0.0 0.22

TLM2 0.0 0.02

COMM2 0.0 0.00

CC2 0.0 0.24

DLS2 0.0 0.12

SENSORS 0.0 0.0

AT LAUNCH 0.08 0.07

Fig.4. For each satellite type, the frac-tion of the satellite failures caused bythe failure of each bus system is shown.Also listed is the fraction of satellitefailures caused by the failure of all thesensors or by an unsuccessful launch.

program computes descriptive statistics(mean and standard deviation) on theevents (launches, retrievals, satellite failures,demand failures, and the like) that havebeen observed. The output reports producedby the model fall into the following majorcategories:

Reports on the technical features have todo with the bus/sensor failures, the num-ber of bus/sensor subsystems in the activeand standby mode every month by eachsatellite configuration, the causes of satel-lite failure, the lifetime of satellites failed,and so on. A sample report is shown inFig. 4.

Reports on the system features show thenumber of launch attempts each month,the number of active and in -orbit standbysatellites, the number of satellites inground inventory, and the number ofsatellites ordered by each satellite con-figuration type, the shuttle launches andrefurbishments invoked, and so on. Atypical report is shown in Fig. 5.

Reports on measures that relate to theoverall business/mission performance in -

elude the mission objectives achieved ev-ery month by each sensor type, the degreeand type of shortfall, if any, the overallperformance over the entire mission dura-tion, and the nonrecurring and recurringcosts expended over the life of the sys-tem. In addition to these reports pre-sented in a tabular form, graphical out-puts are used to enhance understanding.Figure 6 shows a sample graph of theprobability of meeting the mission objec-tives over time.The primary use of the model has been

to test and refine various alternatives untila balance is struck among the cost, risk,and availability requirements. Results fromthe model can then be used in settinginitial bidding prices on contracts and inproduction scheduling.

In spite of all this flexibility in handlingthe tremendous diversity of satellite sys-tems, this model, written in FORTRAN,requires only 1 megabyte of core, and a100 -iteration simulation costs only $20 inthe batch mode. This low cost of runningthe model and evaluating the scenariosplays an important role in its use.

The model was developed by the Opera-tions Research group and then turned overto Astro-Electronics. Because they were

1.1

0N 0.74

° al

0B 0 .A

I 0.3-1

0.2-1

3

0 .

°

MEAN AND ST. DEV. OF LAUNCHMONTH AND PROB. OF LAUNCHOF THE 1ST. 2ND. 3RD, ... LAUNCHOF EACH SATELLIE TYPE

SAT ONE SAT TWO

1 1.00 6.000.0 0.01.0000 1.0000

2 14.25 48.532.49 14.291.0000 1.0000

3 39.91 87.5414.64 22.77

1.000C. 1.0000

4 55.64 97.0212.68 22.23

1.000C 0.4100

5 78.33 0.021.01 0.0

1.000C 0.0

Fig. 5. The pattern of the launches isshown in this report. The first launch ofeach satellite type was prespecified sothat the standard deviation of the launchmonth is zero. The other launches weredynamically scheduled by the model.This shows that a fifth satellite of thesecond type will not be required.

'-rT 9Try 'T"""Tr r'"TT`"T - ..70 10 20 30 40 50 80 70 SO

MONT -I

ti

T "7"T`T90 100 110 120

Fig. 6. This graph shows the fraction of the time the system requirements are metin each month -the monthly probability of mission success. The graph indicatesthat a satellite launch should be scheduled near month 50 in a subsequent sce-nario to avoid a dip in the availability curve.

Nigam /Hong/Spence Business modeling in an engineering context An Astro-Electronics application 69

Authors (lett to right) Hong. Spence. and Nigam.

Raj Nigam is currently OperationsResearch Scientist with the CorporateOperations Research group. He joinedRCA in 1970 and has worked with practi-cally all RCA MOUs and Corporate Staff.He has been applying quantitative tools-modeling and data analysis-to aid inbusiness decision making. His currentinterests include performance and produc-tivity measurement, strategic analysis, andthe application of fuzzy logic.Contact him atRCA Corporate Operations ResearchPrinceton, N.J.TACNET: 254-9629

Saman Hong is Manager of OperationsResearch at the Corporate OperationsResearch group. He joined RCA in 1978,and has since been working on the analyt-ical approaches to business problems forthe Corporate Staff and divisionalmanagement.Contact him at:RCA Corporate Operations ResearchPrinceton, N.J.TACNET: 254-9616

Stephen Spence joined the RCA BusinessSystems and Analysis Career Program in1980. He is currently employed as an ana-lyst in the Corporate Operations Researchgroup. He has completed projects for Ser-vice Company, Records, Hertz, and Corpo-rate Staff.Contact him at:RCA Corporate Operations ResearchPrinceton, N.J.TACNET: 254-9690

constantly involved in outlining the scopeof the model and in the development pro-cess, the Engineering group at Astro hadlittle difficulty in using the model. Soonthereafter they were making inputs and,more importantly, analyzing outputs anddeveloping alternatives for sensitivity anal-yses on their own. The transfer of themodel has been so successful that the roleof Operations Research has now been re-duced to general maintenance and minorenhancements.

Business evaluationsThis model was successfully used by Astro-Electronics in studies of the National Ocean-ic Satellite System (NOSS)-a systemwhose overall cost was publicly acknowl-edged as being in the 700 -million -dollarrange-and in the DMSP Block 5-D pro-gram. It helped in analyzing several alter-natives of satellite configurations, bus/sen-sor subsystem reliabilities and redundan-des, in -orbit standby and ground -inventorypolicies, shuttle retrieval -and -refurbishment

issues, among others. It proved its worthin shifting the emphasis to the business/mission objectives while retaining the prop-er emphasis on the technical underpinnings.

AcknowledgmentThe authors thank Dr. Warren Manger ofAstro-Electronics, for his assistance, sup-port and encouragement throughout thisproject.

70 RCA Engineer 27-6 Nov./Dec. 1982

R.E. Enstrom

Modeling and simulationin mechanical and thermal design

Finite -element computer -modeling programs have been usedto accurately predict the response of engineering structuresto complex mechanical and thermal loads.

Abstract: Finite -element methods arereviewed and applications within RCACorporation are discussed The computer -

modeling programs based on thesemethods have been used to solve complexengineering problems related to mechani-cal design and thermal analysis In mostcases, solutions could not have beenachieved by other means because of thecomplexity of the structure, loads, orboundary conditions. Finite -element model-ing is extremely useful both in the designof new structures and in the analysis ofthose that have failed or that needimprovement.

Finite-element modeling is a powerful meth-od for the simulation of heat transfer andmechanical stresses in various structuralbodies. The power of the finite -elementmethod lies in the a priori prediction ofthe response to a complex mechanical,thermal, electrical, or fluid load from ablueprint of the structure and its physicalproperties. In this way, design changes canbe made before molds are produced orconstruction is begun. A number of finite-

element methods available to RCA scien-tists and engineers are being used in proj-ects throughout RCA to solve otherwisehopelessly complex mechanical -design andthermal -analysis problems.

1982 RCA CorporationFinal manuscript received September 17 1982Reprint RE- 27-6-11

Finite -element modelingthecry and proceduresIn this method, a complex solid continuumis divided up into a series of discrete blocks-finite elements-that have preprogrammedmathematical descriptions for their behav-ior. The number and locations of discreteblocks chosen to describe the structure influ-ence the accuracy of the final solution andthe cost. Therefore, a finer mesh is used ina rapidly varying stress or temperature re-gime. and a coarser mesh is used in otherregions to minimize computer costs.

The elements are geometrically describedby their node points. For the case of twodimensional structures, this is generally doneas 3 or 4 node points per element forsolid structures, this is 8 nodes per element(or up to 20 nodes for some elements,including the 8 corners and the centers ofthe 12 edges). The node -point spatial loca-tions (discretization) can be readily dig-itized from a blueprint of the structurewith the aid of commercial finite -elementprograms or with the aid of RCA -devel-oped software programs' used in con-junction with a Tektronix graphics termi-nal and digitizing tablet.

A force acting on a node causes a dis-placement, much like that described byHooke's law. For all the nodes of an ele-ment, this response is described by theelement stiffness equation':

{F1= [IC] ful

Here the components of IF, and I u I are

the nodal force and displacement vectors,and [K] is the element -stiffness matrix.Included in If, are the vectors for appliednodal loads, pressures, temperature gradi-ents, creep,swelling, large displacements, and so on.The node points locate the displacementsin structural analyses and temperatures inthermal analyses. Each node point has onedegree of freedom in a thermal analysissince it is a scalar function of position, andup to 6 degrees of freedom (3 translationaland 3 rotational) since displacement is avector quantity. The analogy between themechanics and thermodynamics of struc-tures is used to extend finite -element meth-ods to heat -transfer problems, and the re-sulting matrix equation is similar:

iQi = [R] ITIwhere I Q1 is the heat -flow vector, [K] isthe thermal conductivity matrix, and 71

is the vector of node -point temperatures.The matrices for the individual elements

are next combined to form a complete setof matrix equations (global) for all theelements in the continuum with the addi-tional constraints of equilibrium, boundaryconditions, and continuity of nodal displace-ment conditions. There are as many simul-taneous equations as degrees of freedom,and this can range up to 10,000 or morefor large or complex structures. The gen-eral solution for the displacement I Ed ortemperature I 71 in the global equation isachieved by sophisticated wave -front meth -

RCA Engineer 27-6 Nov./Dec. 198271

ods that can be thought of symbolically asinverting the matrix. Stresses in each ele-ment are obtained from the displacementscalculated for each element.

Time -dependent parameters such as thenatural frequency and shape of the vibrat-ing structure can be determined as well astransient dynamic and transient heat -transferproperties. For these calculations, the damp-ing (C), mass (M), velocity (k), accelera-tion (il), specific heat (C), internal heat -generation rate (4), film coefficients (h,),and the emissivity (e) must be known.The basic matrix equations then includeadditional terms as follows:

= [K]1(41 + [C] {id + [M]

IQI=[K]i71+[C]lflFrom a procedural point of view, one

type of element is used for stress analysis,and up to three other elements for conduc-tion, convection, and radiation are usedfor the heat -transfer solution. If both a struc-tural analysis and a thermal analysis are tobe performed on the same structure, thesame discretization (that is, assembly ofnode points described by their Cartesian,cylindrical, or spherical coordinates) canbe used for both analyses. Thus, the ther-mal analysis can be performed first todetermine the heat transfer and tempera-ture distribution, and then the structuralanalysis can be conducted to determinethe stresses generated by the induced tem-perature distribution.

The properties of the materials consti-tuting the structure are needed as inputdata for the finite -element solution becausestresses and deformations induced and theheat transferred are a function of the mate-rials properties. The mechanical propertiesrequired are the modulus of elasticity, Pois-son's ratio (which is a measure of the lat-eral contraction when a material is extendedin the axial direction), the coefficient ofthermal expansion, and the density. Forspecial analyses involving creep, swelling,plasticity, and damping, these values willalso have to be input as material proper-ties. The heat -transfer properties requiredfor solutions include the heat capacity, theheat -transfer coefficients (as a function oftemperature, if necessary) for convection,conduction, and radiation. For radiationproblems, the radiation area is requiredand this can be calculated from the view-ing angles and the geometry of the struc-ture. This tedious procedure can be fa-cilitated using a program developed atAstro-Electronics, entitled RVFAC.4

Once the structure has been discretizedand the materials properties added, then

the structures can be analyzed for stresses,thermal heat transfer, and so on, by usinga general-purpose finite -element computerprogram. These programs include ANSYS,MARC, STRUDL, NASTRAN, DYNA-3, TVTOWER, and PLAFOM. Some ofthese programs have been developed withinRCA in response to specific needs: DYNA-3 for dynamic analyses': TVTOWER forguyed broadcast -tower analyses': and PLA-FOM for modeling of metal -forming pro-cesses.' The other programs have beendeveloped by commercial software housesand are continuously maintained, upgraded,and documented.

A new feature is to incorporate finite -element modeling with mechanical designso that one can both design the structureand do a computer analysis for stress, dis-placements, frequencies of vibration, heattransferred and temperature gradients, usingthe same descriptor for the structure. Thisis then a powerful method of both design-ing the structure and calculating the re-sponse of the structure to the externallyapplied factors of force, pressure, and ther-

mal gradients. In turn, the structure canthen be readily redesigned to reduce criti-cal values of the induced parameters ofstress, deformation, heat transferred, andso on. Computervision is one of a numberof systems that are particularly useful forthis type of application. Here, the structureis developed locally on the built-in mini-computer and then the finite -element struc-ture with the appropriate materials proper-ties and job -control cards is sent to a large-scale scientific computer for execution ofthe solution run using a general-purposeprogram such as NASTRAN.

Other trends in finite -element modelinginclude the tendency to model structuresinteractively from a video terminal ratherthan by batch processing, the use of large-scale minicomputers such as Prime or DECVAX rather than a large mainframe com-puter, and the increasing use of graphics todisplay the analytical results rather thanpages of numerical output.

Finite -element modeling analysis hasbeen used at RCA in the solution of numer-ous complex problems. In the following

Fig. 1. View of a 25V100° bulb from the electron -gun end and looking toward theviewing surface. The structure is modeled in quarter symmetry.

72 RCA Engineer 27-6 Nov./Dec 1982

sections, specific applications of the finite -element method to a few of these mechan-ical and thermal problems are describedand discussed.

Mechanical modelingMechanical modeling includes the responseof a structure to applied pressure, appliedforces, gravitational forces, wind loads, andtemperature gradients. These externally ap-plied loads can induce stresses, changeshapes, and cause buckling, plastic defor-mation, and vibration. All of these exter-nal forces can cause the structure to fail;indeed, finite -element analysis can deter-mine the effect of various external parame-ters on the response of the structure. In thecase of a failed part, a systematic investiga-tion will thereby show the principal male-factor causing the failure. Thus, finite -ele-ment analysis can be used both for design-ing new structures to avoid failure and forthe analysis of existing structures that mayhave failed. It should be emphasized thatfinite -element modeling does not provideremedies to a breakage problem; alterna-tive designs must be created by the humanmind. Rather, it does facilitate the investi-gation of these alternate designs and quan-tifies the results.

In modeling the structure, the finite -element mesh can very closely resemblethe structure (TV picture tube) or lumpedmass elements can be used with a resultingloss of visual similarity to the structure toreduce the complexity and computer com-putational costs (satellite); examples willbe shown of both approaches. Many struc-tures are symmetrical about a plane or anaxis so that 1/2 or Vs of the structure can bemodeled, by employing appropriate bound-ary conditions at the edges of the model.without any loss in accuracy but with anappreciable savings in computer cost.

Model of a TV tube

The finite -element model of a TV tube is

shown in Fig. 1. Here, the model closelyresembles an actual 25 -inch diagonal tube.With this model, it is possible to predictthe stresses and displacements induced byevacuation of the bulb. Further, this can bedone for a variety of geometry variationsto investigate the effect of funnel and paneldesign, manufacturing tolerances, applica-tion of implosion protection bands, andtemperature gradients on the stresses.' Thestress distributions calculated over the view-ing face of the kinescope and along itssides were found to agree with subsequent

strain -gage measurements within about 10percent. In this way, a much more detailedstress and distortion distribution can bedetermined than could be done experi-mentally.

Model of a spacecraft

The finite -element model of a spacecraft iscreated somewhat differently than thatshown for the TV kinescope. For theTIROS -N spacecraft, the model was con-structed from equivalent masses and stiff-nesses to describe the mechanical behaviorof the structure even though it does notnecessarily resemble the actual spacecraft:the structure and the equivalent finite -ele-ment model' are shown in Fig. 2. Aftertests of the TIROS -N spacecraft were com-pleted, the predicted and correspondingactual test results could be compared; excel-lent agreement was found between thepredicted and observed natural frequenciesof vibration for the various parts of thespacecraft. Finite -element modeling for theAdvanced TIROS -N spacecraft is now com-plete and the experimental vibration testsare in progress. In this case, the finite -ele-ment model more closely resembles theactual spacecraft as shown in Fig. 3a; Fig.3b is the spacecraft second lateral mode ofnatural vibration at 38 Hz."I To help devel-op the finite -element model, a CDC pro-gram called UNISTRUC was employed

to generate both the Cartesian descriptionfor the node -point locations and the ele-ment connectivity. In this way, much timecan be saved developing the model com-pared to manual methods. UNISTRUC isalso being used for the latest Astro com-munications satellites, GSTAR and SPACE-NET, which are being built for GTE andSouthern Pacific, respectively.'"

Model of a hardened -steel press part

Engineers encountered a problem, at onepoint, in the press used for RCA's Video -Disc manufacture. One of the steel partsthat creates the center hole during com-pression molding of the disc would breakprematurely. The part is steam heated andwater cooled during the pressing cycle andwas found to fail by cracking after a rela-tively short time. In this case, the part iscylindrical, and is modeled as an axisym-metric solid so that only half of the struc-ture is discretized. The program rotates thestructure around its axis to mathematicallyapproximate the entire structure, so that a2 -dimensional representation of a 3 -dimen-sional structure is obtained.

To try to understand the effect of thevarious process parameters on the stressinduced in the part during the pressingoperation, the parameters were applied oneat a time, and the stress distribution wascalculated.'' It was found that the internal

IMP --Instrument Mounting Platform

ESM--Equipment Support Module

ass-Reaction Cortrol Equipment SupportStructure

ARM-Apogee Kick Motor

STIFFASSOFFSET

SUBSTAUCTIMTES

MASSOFFSETS

AN,I4

STIFFNESS

OFFSET

.,,SEPARATIONPLANE

CONE

Fig. 2. TIROS -N spacecraft, dynamic, finite -element model showing the relation-ship between the actual structure and the limped -mass finite -element model Herepoint masses are used tc represent the spacecraft rather than a detailed represen-tational model.

Enstrom: Modeling and simulation n mechanical and thermal design 73

Fig. 3a. Finite -element model of the Ad-vanced TIROS -N spacecraft.

411111111111111111'111111111111,111111111111111111111111111111USIIIII1111111111.111111111

1.=1Mlanri.. 11/ 11111A1

Fig. 3b. Distortion of the spacecraft inthe second lateral mode of vibration.

steam pressure and the axial load createinsufficient stress to cause fracture of thehardened steel part. However, the calcula-tions showed that thermally induced stressescaused by the impingement of the coldwater on the heated surface could create

Fig. 4. Cross section through a press part exposing the internal crack generatedduring disc pressing. The dark area represents the cracked region. The stresscontour lines are shown in the accompanying finite -element model plot, wherelines spaced closer together represent a steeper stress gradient.

tensile stresses on the order of 60,000 psiand that the presence of a crack nucleuscould increase this stress to about 75,000psi tensile. Here, both the magnitude ofthe stress and the tensile nature are impor-tant in causing failure. Further, the stressdistribution calculated for the thermallyinduced stress corresponded very well withthe crack geometry observed in the failedpart, as may be seen by comparing Figs.4a and 4b. The back pressure created bythe plastic pushing up against the bottomof the part was also found to be an impor-tant parameter inducing tensile stresses, es-pecially when combined with the thermallyinduced stresses. Fracture was thereforepostulated to arise from the induced ten-sile stresses, the flexing of the part duringthe various portions of the process cycle,

and the corrosion of the inside surfaceunder the influence of the tensile stresses.

Model of a VideoDisc caddy

Another VideoDisc-related finite -elementmodeling problem involved the caddy thatthe disc comes in. In this case, it wasdesired to determine the force needed tocontact the video play area of the disc asmight be done by a consumer when squeez-ing the caddy between his fingers duringnormal handling. Figure 5 shows the fi-nite -element model of the disc, spine, andcaddy in half symmetry." The calculationsshow that the force is about 2 pounds, andsubsequent experimental measurements arein virtual agreement with this value. Nowthat the model has been developed, it is

74 RCA Engineer 27-6 Nov./Dec. 1982

Fig. 5. Finite -element model of a VideoDisc record, spine, and caddy shown in halfsymmetry.

possible to investigate the effects of geome-try changes, material changes, and externalstatic and dynamic loading on the responseof the caddy structure. In this particularmodel, a new feature was introduced, name-ly the presence of three separate parts. Thethree parts have to be mathematically con-nected to achieve a solution. This is doneby having gap elements joining oppositenodes in the structure. The gap elementsare nonlinear elements, though, so that anumber of iterations have to be performedto arrive at the correct solution, and con-vergence may be difficult to achieve if theproblem is not defined correctly.

Model of silicon wafer strength

Silicon wafer breakage concerns semicon-ductor manufacturers because it is an impor-tant component of device yield and cost.To understand differences in silicon waferstrength that could affect breakage, it wasnecessary to determine first the elastic prop-erties of perfect wafers, free of defects. Inthis way the deflection of wafers subject toa central point load could be establishedas a function of the value of the load, andexperimental values could be comparedwith this theoretical curve. The mechani-cal properties had to be carefully chosensince Young's elastic modulus andPoisson's ratio are a function of directionfor a (100) -oriented wafer.

The finite -element model and the stressat the bottom of the wafer'' are shown in

Fig. 6. Since the wafer is a flat circle, itcould be modeled with 14 axisymmetricconical shell elements to describe the com-plete wafer. As shown in the accompany-ing plot, the stress rises very rapidly, goingfrom the edge toward the center of thewafer. It was further found that the agree-ment was quite good between the calcu-lated and experimentally determined curvesfor deflection as a function of the load,and for the relative relationship of thecurves for the (100)- and (111) -orientedwafers.

Model of a plastic -packaged device

Another semiconductor -related finite -ele-ment study involved the plastic packagefor a power device. In this case, electricalleads had been breaking for unexplainedreasons and the reliability was not as highas desired. A finite -element model of thepackage and leads was developed." Thepackage consists of a hard silicone plasticcase and a soft rubber compound that sur-rounds the chip to protect it.

It was postulated that high stresses coulddevelop as the package cools from theencapsulation temperature to room tempera-ture. Thus, differential thermal contractionbetween the various components-silicon,silicone, rubber, and metal-could be impor-tant. This possibility was investigated andwas found to develop high stresses in theelectrical lead. A different geometry forthe lead was investigated and calculations

3 200

03 06DISTANCE FROM CENTER ( INCHES)

11111100

IITE ELI.MENT 1100FI,

1

I2

Fig. 6. Finite -element model and stressdstribution across a silicon wafer. Thewafer is supported near the rim and apoint load is applied at the center. Thedeformation at the center as a functionof the applied load is also calculatedand can be compared with experiment.

incorporating the different lead geometriesshowed a reduced stress level. Experimentsverified the desirability of the modifiedgeometry, and the package incorporatingthese changes was put into production withgood results and higher reliability.

Model of towers

The need to predict the response of over-the -horizon radar antennas in any winddirections or tall TV masts to comply withElectronic Industries Association or Amer-ican Institute of Steel Construction specifi-cations led to the development of a finite -element program entitled TVTWIC by R.I.Pschunder at Moorestown. It is used tocalculate deflections and stresses as well asthe first and some higher natural frequen-cies of vibration for both guyed and free-standing structures. In a wind loading, thetower deflects until the horizontal guy andtower reactions equal the wind load. Theanalysis is highly nonlinear and requiresiteration to achieve a final solution; herethe criteria are obtaining all the calculatedtower levels within 0.010 inches of theirtrue position. A plot of tower deflection asa function of height is given in Fig. 7.

Model of ship structures

Another program that has been developedby R.J. Pschunder is DYNA-3.2 This fi-nite -element program has been used toanalyze ship structures for the AEGIS pro-

Enstrom: Modeling and simulation in mechanical and thermal design 75

GUYLEVEL

13 . .

10

8

7

6

3

2

.1.:.* I

. I .'

0 ..5 ;

...%

Fig. 7. Broadcast -tower deflection calculated with the TVTWR finite -elementprogram.

C - LEVEL AT MUER ff /BRAY

ATOM BLAST

vl'I`II II1 T I T F IT Tl/5 L 115 L3TIME (SEC)

Fig. 8. Response of an antenna structure to atomic blast

gram. a high -power laser pedestal. the anten-nas for the Mars and moon -landing vehi-cles, the Mt. Sutro TV tower for earthquakeresistance, and the TV antennas atop theWorld Trade Center for wind -inducedstress. The capabilities include steady-state.3-D shock, and random input dynamicanalysis as well as the standard static anal-ysis for stress. DYNA-3 can handle up to2000 degrees of freedom, and has bothpreprocessing conveniences to generate thefinite -element mesh and post -processing plot-ting capabilities. The program is maintainedand improved by its users. A refinement ofDYNA-3 presently includes interactive plot-ting of the finite -element structure and thereporting of results in graphical form. Theresponse of an antenna structure to anatomic blast calculated and computer plot-ted with DYNA3 is shown in Fig. 8.

Model of metal -forming processes

A finite -element program has been devel-oped by C.C. ChenA ' for the analysis ofmetal forming of non -steady-state axisym-metric or plain strain metal -forming pro-cesses, such as deep drawing. coining, forg-ing, or stretch forming. Here the materialat its forming temperature is more like afluid (specifically a rigid -plastic) rather thanan elastic solid material considered in allthe finite -element programs and problemsdescribed above. The finite -element discreti-zation produces a set of nonlinear equa-tions in terms of nodal and elemental veloc-ities as well as hydrostatic pressure. Theprogram calculates geometry changes, veloc-ity. stress, strain, punch load, and die -pres-sure distributions.

Because of the material characteristicsand the nature of the forming processes,the calculations must be iterated to achievea solution and the finite -element mesh isupdated at each step of the forming opera-tion to accommodate the large deforma-tion changes. The input data include: nodalcoordinates, boundary conditions, die geo-metries. coefficients of friction and veloci-ties. material properties, and an initial as-sumption of the velocity field. Several exam-ples have been compared with literatureresults and the agreement has been excel-lent. A punch -stretching operation is de-picted in Fig. 9. It is expected that thisfinite -element program will have extensiveapplication at RCA Lancaster. where manyparts must be formed for electron gunsand other components.

L 25 Thermal modelingThermal problems have been modeled instructures ranging from the electron gun in

76 RCA Engineer 27-6 Nov./Dec. 1982

0.375 0 75

R (INCHES )

PUNCH TRAVEL:

(1) 0

IZ 0107"3 0 203"® 0 355"

1125 15

Fig. 9. Finite -element model ana deformation of a metal blank in a punch -stretch-ing operation.

a 25V color TV kinescope to a 25 -mega-watt power tube used for the nuclear -fu-sion test reactor. The finite -element methodof analysis is useful for the solution oftransient and steady-state thermal and ther-mal -stress problems, especially since anynonlinear material properties or heat trans-fer coefficients can be readily expressed asa function of temperature or heat flux."

Model of power -tube anode

\ 25 -megawatt power tube and an 8 -megawatt ion beam heat sink" to be incor-porated into the Tokamak Fusion TestReactor (TFTR) at the Princeton Univer-sity Plasma Physics Laboratory were de-signed with the aid of finite -element analy-sis." The 25 -megawatt power tube is usedto switch an ion beam that, after neutrali-zation, ignites the tritium plasma. The anodeof this tube must dissipate up to 2 mega-watts of power. Two alternative designswere considered: a water-cooled, low -tem-perature copper anode section having alow thermal inertia and lag, shown in Fig.10a, and a high -temperature, solid -molyb-denum, high -thermal -inertia integrator witha lower cooling -water requirement.

For each design. it was necessary topredict the transient temperature profile atthe surface and in the interior, the cyclicthermally induced stress that might inducefatigue failure, the transient heat flow across

the coolant boundar, and the effect ofmaterial and manufacturing tolerances. Forthese calculations, the ANSYS finite -ele-ment general-purpose program was usedbecause it easily handles transient thermaland thermal stress problems, and becauseit can express the material properties as afunction of temperature and can give con-vection coefficients as a function of wall

temperature. The copper design reaches amaximum temperature of 234°C, as shownin Fig. 10b, and the molybdenum designreaches 1390°C. These temperatures reflectthe differing thermal conductivities of thecopper and molybdenum, and the proxim-ity of the cooling water to the incidentbeam.

The copper anode reaches steady statein 0.4 second while the molybdenum takesmore than 15 seconds to reach equilibrium(peak heat flow occurs about 2 secondsafter the I -second pulse is input). Temper-ature gradients produce stresses up to 8860psi tensile in the copper design and up to19,000 psi compressive in the molybde-num design.

Model of circuit board

Porcelain -enameled -steel printed -circuitboards have been developed at RCA Labo-ratories that have attractive features foruse in electronic applications."' The par-tially devitrified glass porcelain can be re-fired at temperatures as high as 950°Cafter application of thick -film patterns andcan withstand high voltages because it incor-porates alkali -free materials. ANSYS finite -dement calculations were initiated to deter-mine the stresses resulting from firing, ap-plied loads, and the geometry of the holeregion: and to determine the thermal flowresulting, from thick -film resistor Jouleheating.'

Porcelain materials are weak in tensionbut strong in compression: therefore, therelative expansion coefficients have to be

Fig. 10a. Power tube copper anode sections using a thin -wall high -coolant flowriesicp,

BEAM

153.*C

LANG ZONE INCIDENT ELECTRON BEAM

193. 127.

155.4

160.COOLANT TEMPi:1.°C

195. 187

- BEAM LAND ZONE

166

I

149.

121

230

ILI166 128

"1

INCIOENT ELECTRON BE AM

Fig.10b. Finite -element model of a thin -wall copper anode section showing thetemperature distribution resulting from an incident electron beam flux.

Enstrom Modeling and simulation n mechanical and thermal design 77

005°I

T

PORCELAIN

inch5

06 07 08 085 125 165 205 245 in

-4" X ( R )

DETAIL OF COINED HOLE

25 5 .75 10 1 5 175 2 0 in

8 = 2 2 5' P/4y P= 0,12 in -lb.PORCELAIN

STEEL

STEEL

T

PORC AIN

SUPPORT

Fig. 11. Finite -element model of one quadrant of the porcelainized steel substratehaving a well-rounded (coined) hole.

tDGE 'c

120X 13.6 in ,YRESISTOR BOARD 13.Bin

RESISTOR CENTERS100 EDGE A al in FROM ADJACENT EDGES

B 0.3in FROM ADJACENT EDGESSTEEL C CENTER OF BOARD

80 RESISTOR 0.2 x0 2 in 2 W32W UNIFORMLY DISTRIBUTEDAMBIENT -25C

60

40 I I I I

-4 0 10 20 30 40DISTANCE IN Y DIRECTION FROM RESISTOR CENTER

Fig.12. Finite -element calculation of the thermal distribution for a resistor locatedat several different regions on a large substrate. The resistor contributes 50 W persquare inch of localized heating while the background general heating is 0.25 Wper square inch.

adjusted to a difference of about 1.4 XIV per °C to achieve compression in theprocelain with the porcelain coefficientbeing lower. In this way, the steel is stressedto below its yield point and the porcelaincan be bent without cracking. The finite -element model of the board and an expan-ded view of the hole region for a well-rounded hole are shown in Fig. I I.

Less -well-rounded and nonrounded holeswere also investigated; the computer resultsshowed that maximum roundness of thehole is desired. Another finite -element mod-el was constructed for a board 13.6 incheslong by 8.8 inches wide having resistorsplaced 0.1 or 0.3 inch from the edge ofthe board or at the center of the board.The results of the finite -element calcula-tion are shown in Fig. 12. Note that, asthe 2 -watt resistor is placed farther awayfrom the corner of the substrate, the temper-ature decreases substantially (going from

case A to C), suggesting that the resistorshould be at least 0.3 inch from any edge.Further, up to 16 uniformly spaced 2 -wattresistors should be tolerated without appre-ciable heat build up since the results inFig. 12 show that the temperature is quitelow at a distance of 2 to 3 inches from theresistor. In many applications, therefore,the porcelainized steel substrates should bethermally superior to alumina. The highthermal conductivity of the steel, the highemissivity of porcelain, and the thinness ofthe porcelain are favorable to thick -filmresistors operating at high thermal densities.

Model of a cathode

The thermal characteristics of cathodes ina kinescope electron gun have been inves-tigated by finite -element methods. Sincethe structure is centrosymmetric, an axi-symmetric finite element can be used to

describe it. The finite -element model isshown in Fig. 13, and has been used topredict the effect of material and geometrychanges on the cathode temperature as theheater power is changed." These calcula-tions are important in the design of cath-odes that have improved thermal and me-chanical stability.

References

I. K.W. Coffee and R.E. Enstrom. "Interactive Digitiz-ing and Graphics for Analysis of Random Lines,Surfaces, and Solids." RCA internal report.

2. R. Pschunder. "DYNA-3." presented at RCA Fi-nite -Element Symposium." DYNA-3 User's Manual.Vol. I -General Description: Vol. 11 -Program De-scription (November 1979).

3. R.N. Gallagher. "Finite -Element Analysis Funda-mentals." Prentice -Hall. Englewood Cliffs. N.J.(1975). Also. O.C. Zienkiewia. "The Finite -ElementMethod" McGraw-Hill. London. 3rd edition (1977).

4. F.W. Gorman. private communication.

5. R.J. Pschunder. "TVTOWER." presented at RCAFinite -Element Symposium.

6. C.C. Chen. "Computer -Aided Design of Metal -Form-ing Processes." RCA Engineer. Vol. 27. No. 2. pp.45-49 (March/April 1982).

7. C.C. Chen "PLAFOM2D-A Two Dimensional Axi-symmetric Finite -Element Program for Metal Form-ing Analysis," RCA internal report.

8. R.E. Enstrom, R.S. Stepleman. and J.R. Appert."Application of Finite -Element Methods to the Analy-sis of Stress in Television Picture Tubes," RCAReview, Vol. 39, pp. 665-698 (1978).

9. G.R. Niederoest. "Finite -Element Methods in Space-craft Dynamic Analysis." RCA Review. Vol. 39. pp.632-647 (1978).

10. G.R. Niederoest, private communication.

I I. R.E. Enstrom. RCA internal memos.

12. R.E. Enstrom and D.A. Doane. "A Finite -ElementSolution for Stress and Deflection in a Centrally -Loaded Silicon Wafer." RCA internal report, pub-lished in Semiconductor Characterization Techniques.edited by P.A. Barnes and G.A. Rozgonyi, Proceed-ings. Vol. 78. No. 3. The Electrochemical Society.Princeton. NJ.. pp. 413-422 (1978).

78 RCA Engineer 27-6 Nov /Dec 1982

Fig. 13. Finite -element model of the gunstructure of a 25V kinescope. Emissionis from the front of the cathode; theother parts hold the cathode in place.

13. R.C. Bauder and R.E. Enstrom. RCA internal memos.

14. R.0 Bauder. "Finite -Element Analysis of a Twenty -

Five Megawatt Power Tube and a High -EnergyWater -Cooled Heat Sink for Fusion Research." RCARenew Vol. 39. pp. 699-717119781.

15. W.F. Harbaugh. "The Development of a High-Eneigy Density Heat Sink." presented at the 8thSymposium on Engineering Problems of Fusion Re-search. San Francisco. Calif. I Nosember 3. 1979).

16. K. Hang. J. Andrus. and W.M. Anderson. "High-

Temperature Rweelain-Coated Steel Substrate.; Com-position and Properties." RCA Reriew, Vol. 42. pp.159-17711981).

17. J.H. McCusker. "Finite -Element Analysis of Stressesand Thermal Flow in Porcelain -Enamelled Steel PCBoards." RCA Reniew, Vol. 42. pp. 281-297 (19811.

18. R.C. Bauder. S. Opresko. and R.E. Enstrom. RCAinternal memos.

Ron Enstrom received the S.B., S.M.,and Sc.D. degrees from M.I.T. in 1957,1962, and 1963 respectively. Alldegrees were awarded in the field ofmetallurgy. He joined RCA Laboratoriesin 1963 where his research hasincluded superconducting materialsand devices, silicon and III -V com-pound semiconducting materials anddevices, glass technology, polymerprocessing and high -conductivitypolymers, metallurgy, and finite -element

computer modeling of semiconductor,VideoDisc, and kinescope -related prob-lems. His research has resulted in twoRCA Laboratories OutstandingAchievement Awards, in 1965 and1972, and a David Sarnoff Award forOutstanding Technical Achievement in1967.Contact him at.RCA LaboratoriesPrinceton, N.J.TACNET: 226-2309

Enstrom: Modeling and simulation in mechanical and thermal design 79

R.G. Browne

Electromechanical motor modeling

Computer modeling aids calculation of the start-up torque ofa novel motor design, intended for use in varioushigh-performance systems, including high -reliability spaceapplications and consumer electronics.

Abstract: The primary purpose of com-puter modeling of electromechanical motordesign is to estimate allowable tolerancesin mechanical design parameters. Inputparameters such as coil -rotor gap spacingand radial coil positioning are easily variedin the model. permitting estimation of thesensitivity of motor performance to thoseparameters. Start-up torques and forcesare the primary output of the model. Theyare related to start-up time and runningperformance of the motor. The geometricshape of the motor coils and their relativepositioning are variable inputs to themodel, along with the permanent magneticfield of the rotor. The model divides thecoil into individual current elements; theprogram then analyzes the interaction ofeach current element with the magneticfield of the rotor.

Computer modeling of electromechanicalmotor design aids the manufacturing designengineer. By selectively changing the designparameters entered as inputs into the modelcalculations, the effects of manufacturingtolerance variations upon motor -perfor-mance characteristics may be estimated. Inparticular, one very important perfor-mance characteristic is the maximum pos-sible torque output. The torque deliveredduring the start-up period is directly relatedto the time necessary to reach the requiredoperating rotational velocity.

The result of an effort to develop a

1982 RCA CorporationFinal manuscript received September 8 1982Reprint RE -27-6-12

80

model that calculates the start-up torqueof a novel motor design is described here.The motor in question is a brushless, iron -less two-phase dc motor. It consists of astationary set of ironless armature coils

SIDE VIEW

1/16"AIR GAP

TOP VIEW

HALL DEVICES

mounted below a rotating ferrite perma-nent magnetic field structure as shown inFig. 1. The rotor was magnetized intoeight evenly spaced salient poles. Insteadof hrw.hes. Hall -effect field sensors are used

NON-MAGNETIC MOUNT

22 5°

Fig.1. Two views of a brushless, ironless motor.

RCA Engineer 27-6 Nov /Dec 1982

FERRITEROTOR

430TURN COIL

2" INNER DIAMETER4" OUTER DIAMETER

to sense field crossover points and to pro-vide signals for commutation of the cur-rent. Since the motor construction is iron -less, there are no cogging torques and nohysteresis losses, and eddy -current lossesare very low. These features make thismotor design suitable for use in varioushigh-performance systems. Much of theearly development work on this type ofmotor was done by NASA for use inhigh -reliability space applications. Cur-rently. such motors are extensively used inthe United States and in Japan for con-sumer electronics appliances such as video-tape recorders and record players.

The model

The first step in calculating the magneticinteractions of this motor is to study theinteraction of one coil with one pole -pairof the rotor. The effects of the other coilsand poles and the phasing of the drivecurrent are discussed later. The fundamen-tal torque equation is:

T=L

JDIr X (J X 8)] d X (1)

where g is the magnetic flux density. J isthe current density in the coil, and r is themoment arm.

The shape of the coil is a smoothedwedge that approximately matches the dis-tribution of field on a pole face. This wedgeshape is approximated for computer inputby listing 10 points each around the innerand outer peripheries of the coil. This out-line is subdivided by the program into afiner grid of 48 elements (each 6 X 8).The thickness of the coil is represented byassuming there are four evenly spaced lay-ers on this grid in the z -direction coveringthe 3/16 -inch thickness. Thus, one coil ismodeled by 1920 elements (6 X 8 X 4 X10). The fineness of this grid is less than1/16 inch in each dimension.

For approximation, equation (1)becomes:

T = (a, Idl,) B,R,. (2)i elements

Here n, is the number of turns througheach element (this assumes a uniform wind-ing n, equal to 430/24), is the current,dl, is the radial component of the currentelement length. B, is the vertical compo-nent of the field at the center of the ele-ment, and R, is the distance from the cen-ter of the element to the axis. If dl, and R,are in meters, I in amperes, and B, in testa.then T is in newton -meters. The torque

constant k, of the motor is defined as thetorque -per -unit -current in the coils:

k,,, = (3)

A detailed calculation of the magnetiza-tion of the rotor would require knowledgeof the magnetizing apparatus. For the pur-pose of calculating torque in this model,the magnetic field at the face of the rotoris a fixed input. The field of the rotor wasmeasured over the whole region in ques-tion. All three components of the fieldwere mapped out over an approximate1/8 -inch fine grid over the face of therotor. The fall -off of the field as the dis-tance away from the rotor increased wasalso measured. This fall -off could be fitquite well into an equation of the form:

C(r.0)B=+ 7.2)'

where e(r.0) is the field at the surface, z isthe distance measured from the rotor sur-face facing the coils, and a is approxi-mately 0.4 inch. The fit is shown in Fig. 2;within the range of interest, the fit is betterthan 4 percent. This equation reflects thez -dependence of the field of a magneticpole of radius a.

The components of the field at the sur-face are input for each point on a gridcovenng1/8 inch radially. To find the field at anygiven point, the program linearly interpo-lates on this grid. The z -dependence of the

2.0

1.5

FLUX DENSITY

(KG) 1.0

0.5

0.0

(4)

field, as given in equation (4), is used tocalculate the fall -off of the field away fromthe surface of the magnet.

Results of torque calculations

The program was written in FORTRANand is implemented on CMS. A sampleoutput plot of the torque constant of onecoil as a function of the relative anglebetween coil and pole field is shown inFig. 3. The curve plotted was calculatedusing as input parameters an air gap of0.0625 inch and an inner radius of 0.65inch. The inner radius parameter is the dis-tance from the inner edge of the coil to therotational axis. The torque constant is givenin a practical unit of oz.in./A (1 Nnri =141 orin.). The usefulness of the model isshown in Figs. 4 and 5. By simply varyingthe input parameters of gap spacing or coilradial position, the model estimates thesensitivity to these changes of the torqueconstant.

An experimental check of these torqueconstant calculations for one set of parame-ters was made by taking back electromo-tive force (EMF) measurements on thecoils. The back EMF (V,.,,,,) is measuredby recording the voltage across the coilswith no drive current applied as the rotor

rotational speed byanother motor. The back-EMF constantk, is defined by = k, w, where w isthe rotational speed. The back-EMF con-

FIELD MEASUREMENTS

Range of Interest

0.1 0.2 0.3

- Calculated Curve

- Experimental Values

0.4

Z - DISTANCE FROM ROTOR (INCHES)

0.5

Fig. 2. Calculated and measured fall -off of rotor field away from the rotor surface.

Browne Electromechanical motor modeling 81

15

le

III 11II10. 20 30 40

ANGLE (DEGREES) RAD - 65, GAP - .0625 IN58

Fig. 3. Sample output plot of torque versus angle for one coil and one pole -pair.

scant is directly related to the torque con-stant; in fact, in MKS units, they are numer-ically equal (I NIm/A = 1 V/(rad/s)).The back EMF measured for two coils inseries was 7.8 V, at a speed of 7.5 revolu-tions per second and gap of 0.0625 inch.Thus, the hack-EMF constant at the peakis:

k7.8V

,

2w (7.5 s)= 0.165 V (rad s)

This is to be compared with the peaktorque constant from Fig. 3;

(11.03 oz in. A)= X 2 coils(141 opin. Nm)

= 0.16 1%1.m/A

The agreement between the back-EMFcurve and the torque -constant calculationis better than 5 percent over the wholecurve.

Average start-up torqueand start-up timeAddition of the interactions of the othercoils is straightforward. The drive currentis switched in response to the Hall devicessensing the field crossover points. Thereare two Hall devices, one controlling eachphase, spaced 90 electrical degrees apart(which means the Hall spacing is 22.5°,physically). For maximum start-up torque.the Hall devices should be spaced electri-

cally 45° from the centers of the coils theydrive. This corresponds to a physical sepa-ration of 56.25° from the Hall device tothe center of the nearby coil (see Fig. 1).

For other design reasons, this spacing isusually closer to 60°. A curve showing thecalculated torque for both coils, assuminga current of 0.5 A, is shown in Fig. 6. Theaverage torque is 10.5 min. The effectupon average torque of changing the Hallposition relative to the coils is easilycalculated.

Given an average torque constant, the

12.

10.

10

time required to reach a given angularvelocity can be estimated in the followingmanner. First, the relevant physical parame-ters of the motor are:

J, = 0.0047 kgm2 (moment ofinertia of motor and load)

R,,, = 40 12 (effective coil resistance)

ice = 0.16 V/(rad/s)

km = 0.16 Islm/A (motorconstant calculated)

The dynamic equation of the motor is(neglecting bearing friction):

= 1I ilUL (5)

The relational equations are:

V = k,.0) + I R,,,

T = km/with V, being the maximum voltage ap-plied across the coils. For maximum drivecurrent of 0.5 ampere, V. is 20 volts. Inte-gration of equation (5) gives:

w = ( I -el T)with (6)

r R J'"= 7.3 s

Based on equation (6), the time necessaryto reach a given running speed of 7.5revolutions per second is about 4 seconds.

Once the motor approaches this run-ning speed, its mode of drive is changedsuch that, while running, only short cur-rent pulses are applied to maintain speed.Extensions of the model to simulate be-

n

s

e

-

...

0 TM. III( i I I MI IIII0.03 004 0.05 006 007

GAP (INCHES) RED= 75 IN, ANGLE =22.5*

0.08

Fig. 4. Peak -torque constant as a function of rotor -to -coil -gap spacing variation.

82 RCA Engineer 27-6 Nov /Dec 1982

I2.

10.

I0.

y

5--,

----..-_,......,...

--....,,

0

5

0 ilwi 1 1 1 1111 111i 11110 5 0 6 0.7 0 8 0 9 1 0

RADIUSCINCHES) GAP -.E625, ANGLE =22.5.Robert Browne joined the Technical Staffin the Manufacturing Technology

Fig. 5. Peak -torque constant as a function of relative coil -to -rotor lateral positioning Research Laboratory at the David SarnoffResearch Center in late 1978. Heparticipated in a variety of manufacturingsystems research projects, some withConsumer Electronics and others relatedto Picture Tube Division work. Dr. Brownereceived his Ph.D. from Princeton Univer-sity in 1979 and is a member of both the

15

American.Physical Society and the IEEE.Contact him atRCA LaboratoriesPrinceton. N.J.10

"ACNET: 226-23630w0CC

0 Summary/-

This paper has presented typical examplesillustrating the usefulnes.s of computer mod-eling in calculating start-up torque andstart-up time as a function of input parame-ters. The model helps design engineers toset manufacturing tolerances by providing

5

0

111118

111120

11138

111148

111T58 Oz -In

estimates of motor -performance variationAN5LEMEGREES) RAD...75,GAP 8625,MH22.5:H0.60.0: A18.47 from these various design parameters.

Torque for 0.5 amp thrc Lgh :wo coils.

AcknowledgmentsFig. 6. Sample output torque curve for a motor with Hall sensors sweching current Useful discussions with W.G. McGuffin,through each pair of coils. T. Christopher. E.A. Goldberg. and K.

Kelleher were held on motor design. R.W.havior during the running operation of the time. were done in a fashion similar to Burgen provided the experimental hack-motor, as well as estimates of stopping modeling described here. EMF data.

Browne: Electromechanical motor modeling 83

A.A. Guida K. JonnalagaddaD. Raychaudhuri L. Schiff

Computer modeling in the RCA Satcom system

Different computer simulations are used for different types ofservices carried on the RCA Satcom system.

Abstract: Each of the four RCA Satcomsatellites now in orbit has 24 transpondersthat carry a wide variety of communica-tions services such as television/frequencymodulation (TV/FM), frequency -divisionmultiplexing/frequency modulation(FDM/FM) voice, time -division multi-access (TDMA), and frequency -divisionmulti-access (FDMA), with a variety ofcarriers. Each type of service is subject todifferent impairments from the trans-ponder channel. and this has necessitatedcreation of a number of computer modelsthat simulate performance. This paperpresents some of these models.

A transponder in the Satcom system maybe assigned to carry one of a large numberof possible services. Each service has itsown modulation plan and system parame-ters. Each type of service is impacted in adifferent way by the transponder channelit passes through. This has resulted in dif-ferent computer models or simulations fordifferent types of services. These modelsare used by Americom as traffic assign-ment aids for scheduling which servicesare placed on which transponders and forplanning accompanying ground systems.New services are being developed contin-uously, and computer models are typicallyused before physical simulation and over -the -air satellite tests.

The basic satellite power amplifier isusually modeled as a iero-memory or instan-taneous nonlinearity. That is, the ampli-tude and phase of the radio -frequency (RF)signal out of the amplifier is a nonlinearfunction of the amplitude of the RF input,

t 1982 RCA CorporationFinal manuscript received August 25. 1982Reprint RE -27 -6 - 13

but independent of past values of RF input.The signal arrives at the power amplifierafter passing through a filter that separatesthe signal destined for this transponder fromthe signals destined for other transponders.Likewise, before transmission to earth, theoutput of the power amplifier is refiltered.This model of input filter, instantaneousnonlinearity, and output filter is the kernelof most Satcom computer models. Some-times some elements of the ground systemor other elements of the satellite must beincluded in a model, in other cases someparts of the kernel model may be elimi-nated, but the kernel remains the startingpoint. Because it is easiest to calculate theoutput of a filter from a frequency domaindescription of its input, and, likewise, tocalculate the output of an instantaneousnonlinearity with a time -domain descrip-tion of its input, it is necessary to flip backand forth between time and frequency do-main descriptions of a signal as it transitsthe transponder. Hence, large fast Fouriertransforms (FFTs) are used extensively inthese models.

Even when it is clear that the simplekernel model of a transponder suffices, dif-ferent computer models are needed for dif-ferent types of signals sent through thetransponder. To begin with, the informa-tion expected from the model is differentfor different types of signals.

If the signal is high-speed digital, themodel should generate "eye patterns," pat-tern -dependent timing jitter, bit error rate,and so forth. If the signal is television/fre-quency modulation (TV/FM), the modelshould generate differential phase, differen-tial gain, T -pulse response, and so forth.The computer models must be signal orservice specific. Sometimes creating a com-

puter program to specifically analyze onetype of service allows for simplification inthe basic kernel model, either because someresults can be calculated analytically andbuilt into the model or because the signalstructure allows for simplification. An exam-ple of the latter is the case of multicarrierfrequency -division multiaccess (FDMA)when each of the individual carrier band-widths is small (or equivalently, the carrierspectra do not get near the filter bandedge). In this case, filter amplitude anddelay are fairly flat across any one carrierand, hence, have little effect. The primaryimpairment is classical intermodulation dis-tortion, and the kernel model reduces tothe effect of the nonlinearity above. Thissimplification allows optimization of car-rier levels and frequencies that would beimpractical with the more general trans-ponder model.

The following is a description of a fewsample computer models created for usein the Satcom system. The emphasis is onwhat the models do, not how they do it.

Computer simulation andoptimization of multicarriertransponder operationTo evaluate and compare various trans-mission arrangements for the many signalsthat go through the RCA satellite, a com-puter program was developed that simu-lates any given multicarrier arrangementwith a transponder. To reduce the compu-ter time needed for any one calculation,certain simplifications were employed. Asa result, some special interference effectsthat occur in the actual satellite system donot show up in this particular computermodel. The result is a computer program

84 RCA Engineer 27-6 Nov /Dec. 1982

that tells the user the carrier -to -noise (C/N)ratios achievable with a particular arrange-ment of carriers in one transponder. Whenthe optimization feature is used, the modelcycles through many possible carrier ar-rangements in a systematic manner, whilemoving toward the best arrangement interms of C/N. It is important. however,for the user to know the limitations of themodel before making a final decision.

A transmission system modeledby the computer programThe transmission system consists of a trans-mitting earth station, a 22.300 -mile uplinkpath, a satellite traveling -wave tube (TWT)(linear and nonlinear) with one transponder,a 22,300 -mile downlink path, and a receiv-ing earth station (see Fig. 1). The input tothis system is a set of two or more carriersthat enter the uplink path from the trans-

_ mitting earth station. In the uplink path,Gaussian thermal noise is added to thecarriers. In the TWT, the carriers undergogain changes. and intermodulation distor-tion carriers (nonlinear product terms) areadded. In the downlink path, more Gauss-ian thermal noise and some cross -polariza-tion noise, signals from adjacent cross -po-larized transponders, are added to the car-riers. In some cases, the input carriers origi-nate from more than one earth station andmay also terminate at more than one earthstation. As a result, the uplink and down-link path losses for all the carriers may beunequal.

The five components of the transmis-sion system noted above are modeled inthe computer program by entering certainspecific parameters of each component.Table I lists the parameters associated with

each component. The following is a par-tial list of parameters not included in themodel:

1. Satellite location (longitude)

2. Transmitting earth -station location (lat-itude. longitude)

3. Receiving earth -station location (lati-tude, longitude)

4. Rain data for each earth station

These parameters can be important for afew special cases.

Carrier types

At the present. satellite transmission of fivetypes of carriers can he analyzed with theOPTX program:

FDM/FM (voice/data channelcarrier)

INPUTCARRIERS- -

22,300 MILEUPLINK PATH

INTERMODULATIONNOISE

UPLINK THERMAL 22,300 MILEDOWNLINK PATHNOISE

DOWNLINK THERMAL CROSSPOLARIZATIONNOISE NOISE

TRANSMITTINGEARTH STATION

OUTPUTCARRIERSPLUS NOISE

RECEIVINGEARTH STATION

Satellite transmission system. The diagram shows a typical path in a satellitetransmission system, including all sources of interference (noise plus intermodulationdistortion) to be considered in the optimization process.

TV/FM (TV carrier)

SCPC (single channel percarrier-many smallcarriers)

BPSK (digital data carrier -2 -phase)

QPSK (digital data carrier -4 -phase)

For each carrier type listed above, the usermodels the carriers by entering the follow-ing parameters:

Power level

Center frequency

Signal bandwidth (for noise calcu-lations)

Allocated bandwidth (for frequencyspacing)

Desired carrier -to -noise ratio

Deviation (FDM/FM and TV/FMcarriers)

Computer model of thetransmission system

Based on the actual transmission systemshown in Fig. 1, a computer model of thetransmission system, shown in Fig. 2, wascreated. The two most important blocks inthe system are the computer models for thesatellite power amplifier and the noise.Further details on modeling the satellitepower amplifier, are provided in references1 and 2. The noise model is based on asimple addition of the four noise sources inthe system. With this simplification, theC/N value for each carrier is obtained byadding together the uplink thermal -noisepower times carrier gain, the downlink ther-mal -noise power, the amplifier intermodpower, and the cross -polarization noise pow-er, then dividing the result into the outputLanier power. In the optimization mode,the computer program tries to maximizethe minimum margin for each carrier where

Table I. Parameters associated with each component.Component Parameters1 Transmitting earth station None

2. Uplink path None

3a Satellite linear devices Noise (kTB)Uplink antenna G/TSaturation flux densityEffective isotropic radiated power

3b. Satellite nonlinear device(power amplifier) Amplifier data

4. Downlink path Path lossCross -polarization noise power

5. Receiving earth station Antenna G/T

Guida/Jonnalagaddai Raychaudhuri/Schiff: Computer modeling in the RCA Satcom system 85

!INPUTCARRIERPOWER

DATA

SATELL ITEPOWER

AMPLIFIERMODEL

AMPLIFIERDATA

OUTPUTCARRIER

POWERLEVELS

OUTPUTINTERMOO

POWERLEVELS

filCARRIERGAIN

MARGI

c N

N

1-9,_NOISE

MODEL DESIREDC/N

CROSS -POLARIZATION

NOISEDATA

L

CARRIERSPECTRAL

DATA

OPTIMIZATIONROUTINES

SATELLITEDATA

EARTHSTATION

DATA

DOWNLINKPATHLOSS

Fig. 2. Computer model of the satellite transmission system. This block diagramshows the sequence of operations that occur in OPTX to calculate the margins.given the input carrier data (power levels and spectrums). OPTX tries to optimize themargins by manipulating the power levels and center frequencies.

RUN OPTXDMSLI0740I EXECUTION BEGINS...ENTER THE DATA FN FT (FM) :CAR3 DATACAR3 DATA FOR GLOBAL BEAMr3 CARRIERS ( FULL)

INTERMOD LINE DATANO. OF LINES= 48 LOWEST LINE= -63 DB CUTOFF= -80FIRST= 3 THIRD= 9 FIFTH= 15 SEVENTH= 21CARRIER DATAMIN MAR= 2.70 MAX MAR= 4.49 TIDO= -5.00 PEAK V= 1.38MARGIN= 2.70CAPI= -9,77 -9.57 -9.97MARG= 2.70 3.15 4.49OPTINIZE=IrOPTIMIZE(N)=2rSLIDE=3rPERMUTE=4rSTOP=5,CHANGE(TIB0)=6

MIN MAR= -1.84 MAX MAR= 5.25 TIDO= -4.53 PEAK V= 1.43

MIN MAR= 1.91 MAX MAR= 6.05 TIBO= -4.61 PEAK V= 1.43

MIN MAR= 2.65 MAX MAR= 4.72 TIDO= -5.24 PEAK V= 1.34

MIN MAR= 2.89 MAX MAR= 3.80 TIBO= -5.29 PEAK V= 1.32

MIN MAR= 0.79 MAX MAR= 7.78 TIDO= -2.39 PEAK V= 1.76MIN MAR= 0.88 MAX MAR= 7.64 TIBO= -2.28 PEAK V= 1.80

MIN MAR= 0.54 MAX MAR= 7.53 TIBO= -2.25 PEAK V= 1.81

MIN MAR= 2.97 MAX MAR= 3.91 TIDO= -4.59 PEAK V= 1.44

MIN MAR= 1.20 MAX MAR= 6.44 TIBO= -2.59 PEAK V= 1.79

MIN MAR= 1.61 MAX MAR= 6.31 TIBO= -3.66 PEAK V= 1.59

MIN MAR= 1.87 MAX MAR= 5.13 TIDO= -3.32 PEAK V= 1.67MIN MAR= 3.21 MAX MAR= 3.49 TIDO= -5.08 PEAK V= 1.36

MIN MAR= 3.21 MAX MAR= 3.31 TIDO= -4.48 PEAK V= 1.46

NO. OF OPTIMIZE RUNS= 1 MARG= 3.210 TIBO= -4.48 NCAL= 14

7

C0

CONTINUE OPTIMIZATION=1rEND OPTIMIZATION=0

CARRIER DATATIBO= -4.48 MARG= 3.21038CAPI= -8.67 -8.90 -10.36MARG. 3.21 3.23 3.31

C4

OPTIMIZE=1rOPTIMIZE(N)=2,SLIDE=3rPERMUTE=4rSTOP=5rCHANGE(TIBO)=6

PERMUTE THE ORDER OF THE CARRIERS...MAX MAR...AVE MAR..PERM NUM...CARRIER SEQUENCE

2.81073 3.53822 1 1 2 3

0.76188 4.43621 2 1 3 2

-0.98594 4.62566 3 2 1 3

(3

C5

OPTIMIZE=IrOPTIMIZE(N)=2,SLIDE=3rPERMUTE=41,STOP=5,CHANGE(TIB0)=6

SLIDE(SUCCESS)...AVMARG(MIN)rAVNARG(MAX). 3.483 4.058

FREO 5.000 19.893 30.000MARG 2.399 4.578 4.815OPTIMIZE=IrOPTIMIZE(N)=2rSLIDE=3,PERNUTE=4.STOP=5.CHANGE(TI90)=6

REAL CPU TIME(USEC)= 1354580 COST=f 1.79RI

CLOGCONNECT= 00:02:38 VIRTCPU= 000:01.22 TOTCPU= 000:02.30LOGOFF AT 10:00:57 EST TUESDAY 09/07/82

Fig. 3. Optimization with the OPTX program. In this sample run of the OPTX pro-gram, calls are made to the routines OPTIMIZE (adjust the carrier power levels tomaximize the minimum margin), PERMUTE (rearrange the frequency order of thecarriers), and SLIDE (shift the center frequencies of the carriers within the availablefrequency space) to try to maximize the minimum margin.

margin is defined as

Margin = C/N (dB) - desired C/N (dB)

Depending on the user's choice, this can bedone by varying input carrier power levels,rearranging the spacing between carriers, orchanging the order of the carriers in thefrequency domain. A typical calculationusing the OPTX program and a data filecalled CAR3 DATA is shown in Fig. 3.

Communication of televisionsignals over satellite

Frequency modulation is usually employedto transmit TV signals over satellites. Thesignals are distorted by transmit and receiveearth -station filters, satellite input and out-put filters, and the nonlinear behavior ofthe satellite power amplifier. The signal isalso corrupted by noise received along withthe desired signals.

Noise performance of FM systems canbe accurately determined by using well-known mathematical models. However, nosimple analytical model exists that can pre-dict the distortion performance satisfactor-ily. Consequently, a computer simulationmodel was developed that consists of a testwaveform as the baseband signal. Manytypes of test waveforms, such as a compos-ite test signal, combination test signal, stair-case signal, and color bars, are developedand standardized by the television industry.Figure 4 shows the composite test signal, apart of which is shown in Fig. 5 to an ex-panded scale. In the computer model, thiswaveform is generated by use of the mathe-matical functions that accurately define allthe components. In Fig. 5, the waveformshows the trailing edge of the T -step or linebar, at 31.75 tis, the 2T -pulse, beginning at35.0 kis, the chrominance pulse, beginningat about 36.5 and the first part of afive -step staircase. These components of thecomposite signal, as well as other test wave-forms, are designed so that the distortionobtained is representative of the distor-tion suffered by a real picture as it goesthrough the various parts of a communica-tion link.

In computer simulation, two lines of TVpicture carrying the test waveform are gen-erated, including the chrominance signalover its entire period, which is two lines.The spectrum of this signal, after it is nor-malized to I V peak -to -peak, is shown inFig. 6. The baseband signal is pre -empha-sized at the modulator and is de-empha-sized at the demodulator. These filteringactions result in nearly flat noise power atall frequency regions of the baseband. How-ever, filtering creates distortion problems if

86RCA Engineer 27-6 Nov./Dec. 1982

56

A

ITIT 1111 1111 1111 1111 1111le 28 30 40 58 ee 70

Time (micro 5econds)

Fig. 4. Composite test signal. The components of the signalare designed to produce a kind of distortion that real picturessuffer when they go through a communication link.

3 4

Frequency (MHz,

Fig. 6. Spectrum of the composite test signal. The discretespectral lines arise because TV signals have periodic linestructure.

the frequency deviation is excessive. Figure7 shows the output of the pre -emphasis fil-. ter (CCIR type) when its input is the com-posite signal. Notice the sharp transientsgenerated. Suppose such a signal is given apeak -to -peak deviation of 16 MHz, fre-quency modulated, and passed through a17.5 -MHz filter to limit the spectrum to anallocated bandwidth of 18.0 MHz. The fil-ter output does not have a constant envel-ope. Figure 8 shows the envelope of theoutput. The envelope takes a sharp dipwherever the transients produced by thepre -emphasis filter dominate. In the regionsof the TV picture where the FM signalenvelope has a sharp reduction, the noise

I ee

75

2S

27 5 38.0 %.5 35.0 37.5 40 eTime (micro seconds)

Fig. 5. A segment of the composite test signal.

f

100

se

se

IM27 5 30.9

42.5 45.0

1111 III 1111 1111 1111 1111 111192.5 35.0 37.5 40.0

lime (micro seconds)42.5 45.6

Fig. 7. Pre -emphasized composite test signal. The peak val-ues of the transients are related to the pre -emphasis filterparameters.

dominates; the channel noise creates im-pulses or clicks that result in a visibledegradation, often referred to as "tearing."Using the computer model, it is possible todetermine the impulse rate and study itsdependence on the type of pre -emphasisfilter, the deviation given to the carrier, andthe channel filter.

The filtered waveform passes throughthe satellite power amplifier, whose charac-teristics are nonlinear. To obtain maximumoutput power, the amplifier is operatednear the saturation region. The result is thatthe amplifier works as a nonlinearity thatgives varying amounts of gain as a functionof its input power. Further, the commonly

used power amplifiers have the undesirablefeature of introducing a phase componentat the output, whose value fluctuates as afunction of the input signal envelope. Thiseffect, termed amplitude modulation -to -phase modulation (AM -to -PM) conversionis especially bothersome when a satellitetransponder carries two TV -FM signals.As shown in Figs. 7 and 8, the envelopeof the output of the channel filter is afunction of the baseband signal. Whentwo carriers are given as input to the filter,the fluctuation of the envelope of the com-bined signal at the filter output is a func-tion of the baseband signals of both thecarriers. Through the AM -to -PM conver-

Guida/Jonnalagadda/Raychaudhuri/Schiff: Computer modeling in the RCA Satcom system 87

1.4

8.8 1111 1111 1111 1111 1111 111127.5 38.0 32.5 35.8 37.5 48.8 42.5 45.8

Time (micro seconds)

Fig. 8. Envelope of the filter output when its input is the FMsignal modulated by the composite test signal.

tee

III ITT11111 1111313 37 38 90 48 41

Time other() seconds)

Fig. 10. Demodulated chrominance pulse. The maximum val-ue and the envelope of the minimum values are used incomputing chrominance distortions.

sion, the envelope fluctuations are trans-lated into the phase components of eachof the carriers. Thus, the phase of each ofthe carriers has a component that dependson the baseband of the other carrier. Theresult is that, after demodulation, the cross-talk component of one picture appears asa background signal in the other. The vis-ible portion of the crosstalk appears as aslow color fluctuation (breathing ofcolor). These nonlinear cross -talk problemswere analyzed using computer models.

The filtering action by the channel filterscan not only introduce click noise prob-lems, but also lead to excessive distortionsin the baseband signal. Figure 9 shows theleading edge of the T -step after the receivedFM carrier is demodulated; ringing of the

Fig. 9. Demodulated T -step superimposed on a graticule.The outer graticule corresponds to 3 -percent distortion, andthe inner graticule corresponds to a lower value of distortion.

IIT t 1111 t 1111 I 11111111120 18 48 se ee 76

Time (micro seconds)

Fig. 11. Demodulated staricase signal (luminance pedestalsare removed after demodulation). The five bursts are delin-eated by spikes.

waveform is seen. The "short -time distor-tion" can be computed analytically, or, asshown in Fig. 9, a computer -generated grat-icule can be used to obtain a quick esti-mate. In Fig. 10 the chrominance pulse(also referred to as 12.5 T -modulated pulse)is shown. The shown response is that oftwo TV lines, superimposed on each other.Notice the peak, smaller than 100 IRE, andthe envelope of the bottom part, which isnot a constant. Using this waveform, the"relative chrominance level" and the "rela-tive chrominance time" may be determined.

Satellite communication links generate othertypes of distortions, most of which can bepredicted by using suitable test waveformsand processing them on the computer. Astaircase signal can be used to estimate, forinstance, "differential gain and phase" gen-erated by the link. The response to thestaircase signal is shown in Fig. I I, with theluminance component removed. The inputto the modulator consists of five chromi-nance steps, each 18 -IRE peak -to -peak andabout 9µs wide, with different values ofluminance. The luminance variations induce

88 RCA Engineer 27-6 Nov./Dec. 1982

PSK?

,SPECIFY PSK TYPE RASITO,APS40.2.111-ASK.3

2

OFFSET MODULATION, YES.1.140.11

SPECIFY SYSTEM STRUCTURE

I MODULATOR - FILTERI - RECEIVER2 NODNUELAALT ORST-RFULTUREI -T WTO-F DLT ER2C-RECEDU)ER

2

,SUPPLY THE TNT DATA FILE NAME IN INU COMMAS

'RCATMT'

,SUPPLY THE TNT INPUT WANT IN DO TAOS NO )

I II

SPECIFY FILTER TYPE

I INTERNAL IDEAL FILTER2 USER SUPPLIED FILTER DATA

SUPPLY THE SYNDOL RATE IN IT-SYTI,SEC

39

,SUPPLY THE FILTERI DATA FILE NAME IN INU COMM

'ELLIPLAD.

,SVPALY THE FILTER? DATA FILE NAME IN INV commas

'INTELSAT'

Fig.12. Typical execution sequence forPSK. PSK has been designed to exe-cute with a simple set of commandsand data. As shown above for a typicalrun, the user needs to specify the mod-ulation type, system structure, data filenames of system elements, data rate.and TWT backoff.

2

2

0

O

-1

2

1 -CHANNEL EYE DIAGRAM

11111 111711111111111-'-.7-4111 ih '11-

-0 5 0 0 0.5

TIME (SYMBOL D_ RA -IONS)

10 1.5

Fig.13. Eye pattern output from PSK for the exemplified link. The eye pattern,which is a superposition of successive two-bit intervals of the demodulated datawaveform, is a commonly used qualitative indicator for overall link performance.

**SS SCATTER DIAGRAM AT THE OPTIMUM/SPECIFIEDI AND 0 - CNANNEL SAMPLING INSTANTS OM

0

C OH

ANN

E

-2

. ).k k.........,,,,..

.4ro..-:....

d .:e: ,....i,::-.

-2 -1I -CNANNEL

2

Fig.14. Scatter diagram for the exemplified link. The scatterdiagram is a plot that shows the ensemble I and Q channelvalues at the detector's sampling instant. Increased disper-sion of the points from their nominal values (± 0.7, ± 0.7for QPSK) is an indicator of greater link degradation.

varying amounts of gain and phase to thecolor steps. Figure I I shows that the peak -to -peak values of the five steps are different.By noting these values, the differential gainfrom the communication link can be com-puted. Similarly, the "zero crossing" in -

us ERROR PROOINFLITY US CARRIER TO NOISE RATIO NIS

Fig.15. Error probability versus CNR curve for the exempli-fied link. The error probability curve is the final and mainresult of the PSK prcgram. Such curves form the basis fordesigning digital satellite systems.

formation of the sinusoids can be used toestimate the differential phase. The com-puter models are useful in analyzing theeffects of different deviation parameters,channel filters, and satellite amplifiers, onthe performance of TV -FM.

PSK-a flexible simulation/analysis program for digitalsatellite linksPSK is a hybrid, simulation -analysis pro-gram that can provide various qualitativeand quantitative performance measures fordigital communication systems. The pro-gram was written for ease of use and can beoperated with a minimum of prior knowl-edge. While the present version of the pro-gram is intended for evaluating phase -shift

Guida/Jonnalagadda/Raychaudhuri/Schitt Computer modeling in the RCA Satcom system 89

Authors (left to right) Raychaudhuri, Guida, Jonnalagadda, and Schiff.

Krish Jonnalagadda, Member of TechnicalStaff, RCA Laboratories, joined RCA in1978. His activities include analysis anddesign of satellite communication systemsand signal processing methods in Video -Disc and tape recorders. In 1980, hereceived the RCA Laboratories Outstand-ing Achievement Award for analysis ofnonlinear crosstalk effects in satellites. In1981, he received the David Sarnoff Awardfor Outstanding Technical Achievement forwork that led to a significant increase inthe capacity of RCA's domestic satellitesystem.Contact him atRCA LaboratoriesPrinceton, N.J.TACNET: 226-2025

Dipankar Raychaudhuri received theBachelor's degree in Electronics andCommunication Engineering from theIndian Institute of Technology, Kharagpur,in 1976, and the MS and PhD degrees fromthe State University of New York at StonyBrook in 1978. Since 1979, he has beenwith RCA Laboratories, where he isinvolved in the analysis of satellite com-munication systems. In 1981, he was arecipient of the RCA Laboratories Out-standing Achievement Award. His primaryinterests are in the areas of satellite multi-ple access, computer communications,and queueing theory. Dr. Raychaudhuri isa member of the IEEE and Sigma Xi andhas served as an Associate Editor of theIEEE Communications Magazine since1980.Contact him at:RCA LaboratoriesPrinceton, N.J.TACNET: 226-2311

Leonard Schiff is Head of CommunicationAnalysis Research. He joined RCA Labora-tories in 1967 and worked in a number ofcommunications areas such as cellularradio, magnetic recording, video process-ing, data transmission, and satellite sys-tems. Since 1978, his work has centeredon modulation and transmission tech-niques for use in satellite systems for thetransmission of voice, video, and data. In1981, he received the David Sarnoff Awardfor Outstanding Technical Achievement forachieving dramatic increases in the capac-ity of the RCA domestic satellite system.Contact him atRCA LaboratoriesPrinceton, N.J.TACNET: 226-2170

Allan Guida received a BS degree from theCity College of New York in 1957, and MSand PhD degrees from the PolytechnicInstitute of Brooklyn, New York, in 1961and 1968, respectively, all in electricalengineering. He joined RCA Astro-Elec-tronics Division in 1967 and transferred tothe RCA David Sarnoff Research Center in1968. In 1971, Dr. Guida received an RCALaboratories Outstanding AchievementAward for work done in the area of digitalmagnetic recording systems. He receiveda second RCA Laboratories OutstandingAchievement Award in 1982 for creatingcomputer programs to optimize traffic flowin RCA Americom communication satel-lites. He is presently working on new tech-niques for calculating intermodulation dis-tortion in TWTs and SSPAs.Contact him atRCA LaboratoriesPrinceton, N.J.TACNET: 226-2689

keyed systems such as BPSK, QPSK, and8-PSK, it can also accommodate othermodulation techniques with minor modifi-cations. The basic performance measures ofinterest in a digital satellite link are the eyepattern and scatter diagrams (qualitative)and error probability as a function of receive-earth-station carrier -to -noise ratio (quan-titative). The eye pattern is a superpositionof all possible detected binary basebandwaveforms, and is widely used by engi-neers to size up a digital link broadly. Thescatter diagram, of particular interest forQPSK and other higher -level modulation,is a polar plot of the / and Q voltages at thespecified sampling instant. The most impor-tant result is the quantitative evaluation oflink performance, for example, error prob-ability as a function of carrier -to -noise ratio.

Principles

PSK uses a combination of simulation andanalysis to obtain these performance char-acterizations. Simulation must be used be-cause there is no simple analytical approachfor a realistic channel with filters and non-linearities. The program first models thesatellite channel as a finite memory device,and obtains what can be thought of as anoise -free, input-output function. This isdone by applying a data sequence that con-tains all possible K -bit -long subsequencesto the link. The passage of the signal throughthe channel is simulated by appro-priate transformations, using FFTs whereappropriate, corresponding to the filteringand nonlinearities encountered. After thenoise -free channel behavior is obtained,analysis is used to obtain error probabilityresults. The assumption required is thatdownlink noise dominates so that noise canbe superimposed linearly on the receivedsignal. If this assumption does not hold,other techniques must be used to convertuplink noise to an equivalent down-link quantity. The program makes certainassumptions about the nature of the detec-tor. The current version of PSK can incor-porate carrier recovery effects in the errorprobability calculation. A simpler versionof the program assumes ideal synchroniza-tion at the receive end.

Using PSK

PSK has been used extensively by RCAAmericom and RCA Astro-Electronics forevaluating high-speed QPSK satellite links.The program is easy to use for such appli-cations. What is needed is a numericalspecification of the actual filter and ampli-

90 RCA Engineer 27-6 Nov./Dec. 1982

tier elements in the link. These numericalspecifications are stored in data files withunique names and are accessed by the pro-gram. Generally, the important elements ofthe satellite link are the satellite input filter,the satellite amplifier, TWT, and the receiveearth -station filter. Figure 12 shows theexecution sequence for evaluation of a typ-ical 60-Mbps RCA Satcom link. The namesELLIPLAB, RCATWT, and INTELSATcorrespond to the data file names of thesatellite input filter, the TWTA, and theearth -station receive filter respectively. Afterall the required transmission parameters aresupplied, the program provides the desiredoutputs one by one, at the option of the

user. The first output is the /-channel eyediagram, shown in Fig. 13. A similar dia-gram can be obtained for the Q -channel.The next output is the scatter diagram,shown in Fig. 14. Finally, the main resultshowing bit -error probability as a functionof receive -station carrier -to -noise ratio is

shown in Fig. 15. Such a simulation tellsthe system designer what the required car-rier -to -noise ratio (CNR) at the receiveearth station must be to maintain a desirederror probability level. The program is avaluable tool for studying various trade-offsin the link design. making it possible toanswer frequently asked questions, such as:What is the effect of equalizing the satellite

filter or replacing the TWTA with a solid-state amplifier? In addition to its primaryuse for satellite PSK links, simple modifica-tions of PSK have been used for studyingteletext systems, digital transmission overcables, and so forth.

References

I. J. Fuenalida, 0. Shimbo, and W. Cook, "TimeDomain Analysis of Intermodulation Effects Caused

by Nonlinear Amplifiers." Comsat Tech. Rev. Vol. 3.No. I, pp. 89-143 (Spring 1973).

2. R. Lyons. "A Stochastic Analysis of Signal Sharing ina Bandpass Nonlinearity." IEEE Trans.. COM-22.pp. 1778-1788 (November 1974).

Cumulative Index

available at no charge

According to the RCA EngineeringInformation Survey conducted in 1977,the RCA Engineer is the second mostimportant source of technical informa-tion about RCA (the most importantinformation source is the engineer'sassociates). The back issues of theEngineer -150 of them-provide arecord of RCA's progress in invention,development, and manufacturing. Tomake this wealth of technical informa-tion accessible to the engineers, a 25 -Year Index to the RCA Engineer hasbeen published.

The Index can help you find specificinformation that is needed in your work.

Or the Index might provide a vital con-tact in another RCA business, someonewho has related experience and whowould be willing to consult with you.The Index gives you the opportunity toprofit from RCA's technical heritage-toreuse knowledge, to build on pastaccomplishments.

The 25 -Year Index to the RCAEngineer is available in all RCALibraries. Recipients of the RCAEngineer can obtain their personalcopies by writing to:

RCA EngineerBuilding 204-2Cherry Hill, N.J.

Guida/Jonnalagadda/RaychaudhLri/Schiff: Computer modeling in the RCA Satcom system 91

Patents

Consumer Electronics

Carlson. D J TS011. S H

Multi -band antenna couplingnetwork -4352111

Gries. R J Conn. C E.IMrko. S.Television receiver high frequency regu-lated power supply including a low vol-tage ferroresonant transformer coupled toa step-up high voltagetransformer -4345188

Harlan, W ESelf-limiting video signal peakingcircuit -4350995

Harlan. W EAutomatic video signal peakingcontrol -4351003

Hicks. J ETelevision receiver high voltage generatorprotection circuit -4343028

Lagoni. W AApparatus for reducing the effect of co -channel interference on synchronizingpulses -4343019

Speer. W FRotary tuning mechanism -4347628

Waybright. G CHigh voltage protection circuit for a televi-sion receiver -4345275

West. C.E Rarnspacher. R JTransistor heat sink assembly -4344106

Wilmarth. P CMethod for making printed circuit boardswith connector terminals -4343084

GovernmentCommunications Systems

Caracappa. M GSystem and method for frequencydiscrimination -4352194

Corsover. S.L Dobbins. L.W. Pierson. P BVariable -velocity film exposing and devel-oping apparatus -4344088

Crowley. A TPrecise digitally programmed frequencysource -4349887

Laboratories

Aschwanden. FComparison arrangement for a digital tun-ing system -4352206

Bleazey. J.C.1Guarracini. JDriver arrangement for stylus lifting/lower-ing apparatus -4344166

Botez. DSingle filament semiconductor laser withlarge emitting area -4347486

Catanese. C A Bloom. SMulticolor cathode-ray tube with quad-rupolar focusing color -selectionstructure -4350922

Duffy. M T Zanzucchi. P JMethod and apparatus for determining thequality of a semiconductorsurface -4352016

Duffy. M T Corboy. J F Zanzucchi. P JApparatus for determining the quality of asemiconductor surface -4352017

Faith. Jr TJMonitor for oxygen concentration in alum-inum -based films -4348886

Flatley. D W Hsu. S TProcess for tapering openings in ternaryglass coatings -4349584

Goodman. A M Tarng. M LMethod of passivating a semiconductordevice with a multi -layer passivant systemby thermally growing a layer of oxide onan oxygen doped polycrystalline siliconlayer -4344985

Henderson. J.G.Apparatus for automatically steering anelectrically steerable televisionantenna -4349840

Jastrzebski. L.L. Levine. P ASemiconductor imagers -4348690

Johnston. L BSystem for compensating for transfercharacteristic variations of electronguns -4344021

Okamoto. F Kato. KMethod for preparing inorganicsulfides -4348299

Phillips, W Neil. C C Hammer J MMethod for making planar optical wave -

guide comprising thin metal oxide filmincorporating a relief phasegrating -4343890

Reitmeier. G A Dischert. R AAdaptive composite -component transcod-ing hierarchy for digital video -4352122

Salt. Jr W WRetainer ring for securing substrates in avacuum deposition system -4344383

Southgate. P.D. Beltz. J PInspection system for detecting defects inregular patterns -4349880

Subbarao. S N Huang. HMethod for fabricating via holes in a semi-conductor wafer -4348253

Tarng M LEtching a semiconductor material andautomatically stopping same -4343676

Tarng. M L Hicinbothem. Jr W AMethod of depositing a refractory metal ona semiconductor substrate -4349408

Toda. M 'Osaka. SFluid flow velocity sensor using a piezo-electric element -4351 192

Valachovic. J Alphonse. G AReisner. J H Etzold. K FPiezoelectric transducer for recordingvideo information -4349902

Wang. C.C. Ekstrom. LLausman. T.C.1Wielicki. H.Video disc lubricants -4342659

Woodward. 0 M Henderson. J GLoop antenna arrangements for inclusionin a televsion receiver -4342999

Missile & Surface Radar

J F

Actuation rate limiter -4349754

Landry. N RCoax to rectangular waveguidecoupler -4349790

Picture Tube Division

Hertzler. M.E.System for applying a liquid to the studs of

92RCA Engineer 27-6 No Dec 1982

a color kinescope faceplatepanel -4351265

Hughes. R.H.Electron gun with balanced lens lips toreduce astigmatism -4350923

Rang. L.L.'Alleman. R A 'Miller, D LApparatus for sensing bare metal on amoving strip of insulatively coated conduc-tive material -4351263

Wardell. Jr . M HElectron tube base with flow channelstherein -4345812

"SelectaVision"VideoDisc Operations

Crooks. H NVideo disc player having record side iden-tifying apparatus -4352175

Elliott. C AVideo disc player having carriage detentmechanism -4351046

Taylor. B KRemovable protective cover for a videodisc stylus cartridge -4342394

Solid State Division

Atherton. J H Jindra. C PSignal comparison circuit -4348596

Benner. T EMesh assembly having reduced micro -phonics for a pick-up tube -4347459

Caprari FRadiation shadow projection exposuresystem -4348105

Gillberg. J ECircuit with dual-purposeterminal -4350906

Gubitose. N.F. Schuler. M.R.Patterson. D LCrystal seed holder assembly -4348365

Harford. J RVariable load impedance gain -controlledamplifier -4344043

Harforc. J.R.Gain -controlled amplifier utilizing variableemitter degeneration and col' ector loadimpedance -4344044

Harford. J.R.Variable emitter degeneration gain -con-trolled amplifier -4345214

Harmon. J.W. !Atherton, J.HLow power voltage multipliercircuit -4344003

Holbrook. M.D.I Knapp. W.K.Precharge circuit -4352031

Marschka. F DScreen contact means for a cathode raytube -4344015

Marscnka. F.DMain lens assembly for an electrongun -4350925

Rudy. .J ETiming circuit for the digital generation of

composite luminance and chrominancevideo signal for non -interlaced televisionraster scan -line pattern -4344075

Schade. Jr_ 0 HDifferential -input amplifier circuitry withincreased common -mode voltagerange -4345213

Schade. Jr 0 HCompensation of base -current -relatederror in current mirror amplifiercircuitry -4345216

Stewart. R GPower gated decoding -4344005

Tomasetti. C M Ulaky. J APhotomultiplier tube having a gain modify-ing nichrome dynode -4347458

Special Contract Inventors

Berry. D.A. 1Preston. J R ridlenbrand. L.J.High density information disclubricants -4342660

Berry. D AHigh density information disclubricants -4351048

Hillenbrand. L.J. Preston. J.R. Berry. D A.High density information disc -4346469

Preston. J R Hillenbrand _ J Berry. D.AHigh density information disc -4346468

Pen and PodIUm Recent RCA technical papers and presentations

To obtain copies of papers. check your library or contact the author or his divisionalTechnical Publications Administrator (listed on back cover) for a reprint.

AdvancedTechnology Laboratories

A FellerAutomated Layout Techniques for CustomVLSI-Presented at the 19th Design Auto-mation Conference. Las Vegas. Nev . andpublished in Proceedings (6/19/82)

W. Heagerty !W. GehweilerG Caracciolo G Brucker (Labs)Design and Performance of Two 1KCMOS/SOS Hardened RAMs-Presentedat the Nuclear & Space Radiation EffectsConference. Las Vegas. Nev. (6/19/82)

P.T. Parrish T.G. Sollner (U. of Mass.)R.H. Mathews (U. of Mass.)' H. Fetterman(Linco'n Labs)Printed Dipole -Schottky -Diode MillimeterWave Antenna Array-Presented at theSPIE 'echnical Symposium East '82,Arlington. Va.. and published in the Pro-ceedings (5/2/82)

P. Parrish I K. Yngvesson (U. of Mass.)K. Korzenioswki (U. of Mass.)] R. Mathews(U. of Mass.)Planar Millimeter Wave Antennas WithApplication to Monolithic Receivers-Presented at the SPIE Technical Sympo-

slum East '82. Arlington. Va., and publishedthe Proceedings (5/2/821

R PutatundaA Delay Calculation Program for CustomGenerated LSI/VLSI Chips-Presented atthe Custom Integrated Circuit Conference.Rochester. N.Y., and pubtshed in the Pro-ceedings (6 /19 /82)

D Rea (ATL) H Urkowitz MISR)Sensor Design Considerations for an Anti -Ship Cruise Missile-Presented at the1982 Tri-Service Workshop on Missile

RCA Engineer 27-6 Nov./Dec 1982 93

Targeting, and published in Proceedings(May/June)

R. PutatundaAuto -Delay: A Program for AutomaticDelay Calculation for LSI/VLSI Chips-Presented at the 19th Design AutomationConference, Las Vegas, Nev., and pub-lished in the Proceedings (6/19/82)

W. SchamingAn Adaptive Gate Multifeature BayesianStatistical Tracker-Presented at the SPIE26th Annual International Technical Sym-posium & Instrument Display, San Diego,Calif.. and published in Proceedings(8/21-27/82)

D. ShazeerPerformance Measures for StatisticalSegmentation-Presented at the SPIE 26thAnnual International Techical Symposium& Instrument Display, San Diego, Calif., andpublished in Proceedings (8/21-27/82)

D.C. Smith I B.S. WagnerA Low Cost, Transportable, Data Man-agement System for R&D OrientedLSI/VLSI Design-Presented at the 19thDesign Automation Conference, LasVegas, Nev.. and published in the Proceed-ings (6/19/82)

J.R. Tower I H. Elabd (Labs) IT. Villani (Labs)High Density Schottky -Barrier IRCCDSensors for SWIR Applications at Inter-mediate Temperature-Presented at theSPIE Technical Symposium East '82,Arlington, Va., and published in Proceed-ings (5/2/82)

Astro-Electronics

W AltmanA Simplified Derivation of GeometricalDynamic Range Compression-OpticalEngineering Journal (7-8/82)

C.H.AnDesign of Lumped Element 20-GHz FETAmplifier-Proposal for MS Thesis. Mass.Institute of Technology (8/82)

K. BudlongTesting Large Spacecraft Deployables-7th Annual Testing Seminar, Los Angeles,Calif. (10/13/82)

R. BuntschuhNOVA Disturbance Control (DISCOS) Test-ing-Aerospace Testing Seminar. SanDiego, Calif. (10/13/82)

J ChangGyroless Attitude Determination & ControlSystem for Advanced Environmental Satel-lites-AIAA Guidance & Control Confer-ence, San Diego, Calif. (8/9/82)

F. Chu IB. WangOn Changing Boundary Conditions inStructural Dynamics-AIAA 23rd Structural

Materials Conference. New Orleans, La.(5/10/82)

M FeyerhermReliability of Radiation Hardened MOSRAMs in S/C Memory Application-GOMAC/82, Orlando, Fla. (11/2/82)

L. FreedmanSpace Shuttle Closed Circuit TV System-National Telesystems Conference. Galves-ton, Tex. (11/10/82)

W. FuldnerIntegration & Test of Explorer Class Scien-tific S/C Dynamics Explorer A and B-NIA - ATG Meeting, San Diego, Calif.(6/23/82)

F. GargioneUse of 8236 Plotter with 430 Offline Plot-ter-Fall Technical Mtg. Applicon UsersGroup-(11/2/82)

M. GoldbergerSwitched Mode of DC/DC Power Conver-ter-BS Thesis. MIT (5/82)

L. GombergPresentation/Discussion on the DMSPSpacecraft-National Security IndustrialAssociation, San Diego. Calif. (6/23/82)

J. KaraDual Beam Networks for CommunicationSatellite Antennas-Ben Franklin Sympo-sium. Philadelphia. Pa. (5/15/82)

J. KeiglerK -Band Spacecraft Technology -4thAnnual Satellite Users Conference. Denver,Col. (8/12/82)

J.N. LaPradeMeasurement of AM -to -PM Effects in aGaAs FET Power Amplifier-Thesis.Drexel University Library. Philadelphia. Pa.(6/82)

K MullerSpace Shuttle's Impact on Future RCASpacecraft-RCA TREND (11/82)

W. NunanSatellite Power System Optimized forLarger Load-Proposal for BS/MS Thesis,MIT (8/82)

C. Profera. et al.Appl. Method of Steepest Descent Opt.Des. of Shaped Beam Antennas -1982Joint Intl. IEEE/APS Symposium. Univ. ofNew Mexico, Albuquerque, New Mexico(5/24/82)

C. Profera E. NgaiDistortion Analysis for Satellite ReflectorAntennas -1982 Joint Intl. IEEE/APSSymposium, Univ. of New Mexico,Albuquerque, New Mexiso (5/24/82)

R. RitterThermal Radiation Characteristics of 30Amp -Hr Ni-H2Battery Cell -17th Interso-

ciety Energy Conversion Engineering Con-ference, Los Angeles, Calif. (8/8/82)

A RosenbergRemote Laser Measurement TemperatureHumidity Using Differential Absorption inAtmospheric Water Vapor-II InternationalLaser Radar Conference, Madison. Wis.(6/21/82)

T. Truong (co-op student)Des. Shaped Beam Pattern Using PhaseExcitation of Antenna Array-BS Thesis.MIT (5/82)

S Seehra W SlusarkThe Effect of Operating Conditions onRadiation Resistance of VDMOS PowerFETs-IEEE Conference, Nuclear & SpaceRad. Effects, Las Vegas, Nev (7/20/82)

Automated Systems

M.J. Cantella (Automated Systems)H. Elabd IW.F. Kosonocky (Labs)Solid State Infrared Imagers-Microstruc-ture Science and Engineering, Vol. 5

M.J. Cantella IJ.J. Klein (AutomatedSystems)H. Elabd W.F.! Kosonocky (Labs)A New Solid State Infrared Sensor for Mil-itary Applications-Military Electronics/Countermeasures (9/82)

W.F. Kosonocky (Labs)F.F. Martin (Automated Systems)Status of IR Schottky -Barrier Image Tech-nology-SPIE 26th Annual InternationalSymposium, San Diego, Calif. (8/82)

J.D. RickmanDesigning with the Intel One -MegabitBubble Memory-EDN Magazine (9/82)

Commercial CommunicationsSystems Division

J.F. McCoyA Photoconductive Telecine Camera withAutomatic Setup-Presented at the IEEEBroadcast Symposium, in Washington,D.C. (9/16/82)

J F McCoyAutomatic Burst Phase Matching Ar-rangement for Color Television CameraSystems-RCA Technical Note. PrincetonDocket No. 77060 (1982)

GovernmentCommunications Systems

H.R. Barton, Jr.User Guide -RCA Unit Production CostTracking Model UPCT5-EIA G-47 Com-mittee Mtg., Washington, D.C., and pub-lished in EIA Engineering Design Integra-tion Bulletin. No. 1 (10/15/82)

94 RCA Engineer 27-6 Nov./Dec. 1982

R.R. Bigler IJ.P. HeenenMethod for Attaching Edge Terminals-RCA Technical Note, Disclosure 76,931(10/82)

R.R. Bigler1T.G. ButtSystem for Optically Aligning a ScreenPrinter-RCA Technical Note, Disclosure76.503 (10/82)

T.T.N. BucherSpectrum Occupancy of Pulsed FSK-MILCON '82, Boston, Mass., published inthe Conference Record (10/17-20/82)

A. GarciaCleaning Audio Records-RCA TechnicalNote, Disclosure 74,911 (10/82)

J.L. LynerdTactical Military Equipment Reliability &Maintainability Demo. Programs, Planning,Demo. Test & Analysis of Results -17thIntl. Logistics Symposium. Boston, Mass.(8 / 82)

D. McClureKeying Improvements to the ICOM IC-730-QST Vol. LXVI, No. 7 (7/82)

E.J. NossenIntegrated Radar Communications -1982IEEE Military Communications Conference,Boston, Mass., published in the Confer-ence Record (10/82)

M. PackerTouch -Up Ink for Photoemulsion Art-work-RCA Technical Note (Domestic &Foreign Distribution)

P.T. PattersonE-COM-Christ Episcopal Church Men'sClub, Riverton, N.J. (9/26/82)

K. Weir IJ.J. Cohen W. RizzoOperation & Maintenance Trainers forTRIDENT Integrated Radio Room-Intl.Conf. on Simulators. Univ. of Sussex,Brighton, U.K., published in the Proceed-ings (9/26-30/82)

D.J. WebsterSoftware Management Experience in Dis-tributed C31 System Development-AF-CEA, Ft. Monmouth Chapter Symposium(9/21-22/82)

GovernmentSystems Division

J. HaymanDesign & Simulation of An Intelligent Mis-sile Seeker-AGARD - A North AtlanticTreaty Organization Guidance & ControlSymposium, Lisbon, Portugal, published inProceedings (10/12-14/82)

J. HilibrandSome Recent Perspectives in Microelec-tronics-IEEE Philadelphia Section Meet-ing (9/21/82)

Laboratories

D.P. BartonVDTIS: A VideoDisc Testing InformationSystem-ORSA.TIMS, San Diego JointNational Meeting (10/26/82)

D. Botez1D.J. Channin1M. EttenbergHigh Power Single -Mode AIGaAs LaserDiodes-Proceedings. SPIE InternationalSociety for Optical Engineering, Vol. 321,Los Angeles, Calif. (1/28-29/82)

D. Botez IJ.C. ConnollyTerraced-Heterostructure Large -OpticalCavity AIGaAs Diode Laser: A New Typeof High -Power cw Single -Mode Device-Appl. Phys Lett., Vol. 41, No. 4 (8/15/82)

A. Catalano IR.V. D'Aiello IJ. DresnerB. Faughnan IA Firester IJ. KaneH. Schade IZ.E. Smith 1G. Swartz IA. TrianoAttainment of 10% Conversion Efficiency inAmorphous Silicon Solar Cells-Presentedat IEEE PV Specialists Conference, SanDiego, Calif. (9/27-30/82)

A.H. FiresterAmorphous Silicon Module-Photovoltaic-The Solar Energy Magazine -48-9/82)

L.P. FoxRheology of Carbon -Polymer Composi-tes-Carbon Black -Polymer Composites,The Physics of Electrically ConductingComposites, ed. by Enid Keil Sichel,Marcel Dekker, Inc.

B. Goldstein IJ. Dresner D.J. SzostakThe Diffusion of Holes in Undoped Amor-phous Si:H-Philosophical Magazine B.Vol. 46, No. 1, pp. 63-70 (1982)

J.M. HammerIn -Line Anamorphic Beam Expanders-Applied Optics. Vol. 21. p. 2861 (8/1/82)

L. Jastrzebski, P. ZanzucchiD. Thebault IJ. LagowskiMethod to Measure the Precipitated andTotal Oxygen Concentration in Silicon-Journal of the Electrochemical Society. Vol.129, No. 7 (7/82)

M. Kumar IS.N. SubbaraoR.J. Menna IH. HuangMonolothic GaAs Interdigitated 90°Hybrids with 50- and 25 -ohm Impedances-IEEE Microwave and Millimeter -WaveMonolithic Circuits Symposium (5/82)

A. Okada I M. TodaInfluence of Silicone Contamination onBrush -Commutator Contacts in Small -SizeDC Motors-IEEE Transacticns on Com-ponents, Hybrids, and ManufacturingTechnology, Vol. CHMT-5, No. 2 (6/82)

D. RaychaudhuriUtilizing TDMA Idle Periods for RandomAccess Transmission-IEEE InternationalConference on Communications, ICC '82,Philadelphia. Pa. (6/13-17/82)

P.H. Robinson IR.V. D'Aiello (Labs)C.P. Khattak IF. Schmid (Crystal Systems,Inc.)

Thin -Film Epitaxial Solar Cells on Sub-strates Made From MG Silicon by the HEMProcess-Presented at IEEE PV Special-ists Conference, San Diego, Calif.(9/27-30/82)

A Rosen I M. CaultonP. Stabile IA. GombarW. Janton IC.P. Wu1C.W. MageeSilicon Technology Applicable to Monoli-thic Millimeter Wave Sources-SPIE, Inte-grated Optics and Millimeter and Micro-wave Integrated Circuits. Vol. 317 (1981)

J.R. SandercockTrends in Brillouin Scattering: Studies ofOpaque Materials, Supported Films, andCentral Modes- Topics in Applied Physics.Vol. 51: Light Scattering in Solids Ill

StaeblerStability of Amorphous Silicon Solar Cel-ls-IEEE Transactions on Reliability, Vol.R-31. No. 3 (8/82)

J.H. Thompson III IS. Hofmann (visitingScientist at RCA Labs from Germany)The Use of Plasmon-Loss Peaks in Study-ing he Epitaxial Silicon on Alumina Interla-ce-Surface and Interface Analysis. Vol. 4,No. 4 (1982)

R WilliamsThe Relation Between Contact ChargeTransfer and Chemical Donor Properties-Journal of Colloid and Interface Science,Vol. 88, No. 2 (8/82)

R. Williams I PJ WojtowiczA Simple Model for Droplet Size Distribu-tion in Atmospheric Clouds-Journal ofApplied Meteorology. Vol 12. No 7 (7/82)

Missile & Surface Radar

K. Abend I H. UrkowitzSome Sensor Design Consderations for anAnti -Ship Cruise Missile -1982 Tri-ServiceWorkshop on Missile Ship Targeting NavalResearch Laboratory, Washington, D.C.,Workshop Proceedings (8/10-12/82)

J.A. BauerLeadless Chip Carrier Packaging-SEMICON Southwest (Part of presentationon VLSI assembly, reliability, and testing),Dallas, Tex. (10/14/82)

R Blasewitz (co-author)Microcomputer Systems Hardware/Soft-ware Design-Published by Hayden BookCo.. Rochelle Park, N.J.. 550 pages (1982)

M E. BreeseA Planar Phased Array Antenna HavingIncreased Scan Coverage-TechnicalNote (8/31/82)

J.J. CampbellIR.S. JohnsonA New Integrated Radiating Element for

RCA Engineer 27-6 Nov./Dec. 1982 95

Phased Array-IEEE Transactions onAntennas and Propagation (7/82)

R.I. CreedonSupport Systems-Electrical, and SupportSystems-Mechanical Room Arrange-ments-Summer Course (Guest Lecturer):Combat Systems Engineering and ShipDesign, MIT, Cambridge. Mass. (8/82)

M.L. Dorman. JrMissiles From the Deep-Sea Classics.Vol. 15. No. 6. pp. 18-27. Part 1 of 3 parts(11/82)

R.C. DurhamRadar Design for Electromagnetic Compa-tibility-IEEE EMC International Sympo-sium, Santa Clara. Calif. (9/8-10/82)

R.F. KolcVLSI Packaging Techiques at RCA Moo-restown-IEEE VLSI Packaging Workshop,Gaithersburg. Md. (9/14/82)

S GaskellA Fixed -Beam Multilateration Radar Sys-tem for Weapon Impact Scoring-IntlConference - Radar '82, London. England,Conference Proceedings (10/18-20/82)

L.J. GrantnerAnti -Submarine Warfare: Detect, Control,Engage: and Anti -Submarine Warfare:Ship Impacts-Summer Course (GuestLecturer). Combat Systems Engineeringand Ship Design. MIT. Cambridge. Mass.(8/82)

W.C. Grubb. Jr.Minicomputers and Microcomputers forNon -Electrical Engineers-GeorgeWashington University. Presented at NavalAir Test Center, Patuxent. Md. (8/20/82)

J. HanessHow to Critique A Document -1982 Pro-fessional Communications Society (IEEE),Boston, Mass.. Proceedings (10/13-15/82)

M KantApplying EMC Technology to the NextGeneration of Navy Ships -1982 IEEEInternational Symposium on Electromag-netic Compatability, Santa Clara, Calif..Symposium Proceedings (9/8-10/82)

M H PlofkerCombat System Availability-Summer

Course (Guest Lecturer): Combat SystemsEngineering and Ship Design, MIT, Cam-bridge. Mass. (8/82)

R.J. RenfrowTopside Design-Radar; and TopsideDesign-Weapons-Summer Course(Guest Lecturer): Combat Systems Engi-neering and Ship Design, MIT, Cambridge.Mass. (8/82)

E.E. Roberts, Jr.Error Budgeting Alignment and Ships'Flexure-Summer Course (Guest Lecturer):Combat Systems Engineering and ShipDesign. MIT, Cambridge, Mass. (8/82)

P.S. Sawkar 'T.J. ForquerE.J. ScherneckelH. LiA Multi -Port Memory Organization for Usein Distributed Computing Systems-Intl.Conference on Distributed Computing Sys-tems, Ft. Lauderdale, Fla. (10/18/82)

R.L. SchelhornUniversal Test Fixture for Testing ChipCarriers-Technical Note (10/12/82)

R L SchelhornTest Fixture for Testing Chip Carrier Devi-ces Assembled in Larger Circuits-Tech-nical Note (10/12/82)

R L SchelhornHigh Density Metal Core Substrate forChip Carrier Circuits-IEEE ComputerPackaging Conference, New York, N.Y.(9/9/82)

S.A. SteeleSystem Software Design Trade -Offs forReal -Time Data Measurement and ControlSystems-International Journal of Mini andMicrocomputers, Vol. 4, No. 2 (1982)

J.T. ThrestonSystem Functional Analysis and Func-tional Allocation: Performance TradeoffAnalysis: Anti -Air Warfare: Detect, Control,Engage; Computers/Digital Technology;and Contemporary Combat SystemDesigns-Summer Course (Guest Lec-turer): Combat Systems Engineering andShip Design, MIT. Cambridge. Mass. (8/82)

H Urkowitz K AbendSome Sensor Design Considerations foran Anti -Ship Cruise Missile-Proceedingsof the 1982 To -Service Workshop on Mis-

site Ship Targeting, Naval ResearchLaboratory, Washington, D.C. (8/10-12/82)

H. Urkowitz IMSR)1M.M. Rea (ATL)Some Sensor Design Considerations foran Anti -Ship Cruise Missile -1982 Tri-Ser-vice Workshop on Missile Ship Targeting,Naval Research Laboratory, Washington.D.C. (8/10-12/82)

National BroadcastingCompany (NBC)

C.E. SpicerNTSC Color Field Identification-SMPTEJournal, Vol. 91. No. 7 (July. 1982): pres-ented at the SMPTE 122nd Technical Con-ference. New York, N.Y. (11/14/82)

RCA Ltd.

P.P. Webb (Electro-Optics, RCA Inc.,Montreal) G.H. Olsen (Labs)Large Area and Visible Response VPEInGaAs Photodiodes-IEEE SpecialistConference on LEDs and Photoconductors(9/15-16/82)

RCA Service Company

A.B. JonesComputers Are Their Thing-Apple Users'Educational Software Convention(10/27/82)

J.C. PhillipsResponding to Competition -1982 IEEEConference of Professional Communica-tion Society. Boston. Mass. (10/14/82)

Solid State Division

H Veloric R DenningReliability of Radiation -Hardened CMOS/-SOS RAMs in Spacecraft Applications-GOMAC/82. Orlando. Fla. (11/2/82)

H Veloric R. Denning G. Schnable 'J YehReliability of Silicon -on -Sapphire ICs -1982 SOS/S01 Technology Workshop,Provincetown. Mass (10/5-7/82)

96 RCA Engineer 27-6 Nov./Dec. 1982

Engineering News and Highlights

Staff announcements

Thornton F. Bradshaw, Chairman of theBoard and Chief Executive Officer, an-nounced that RCA Corporation electedRobert R. Frederick, President and ChiefOperating Officer. Mr. Frederick will haveresponsibility for all Divisions and Subsid-iary Companies of the Corporation exceptthe National Broadcasting Company, Inc.,which will continue to report to the Chair-man of the Board and Chief Executive Of-ficer. Messrs. William C. Hittinger, Frank A.Olson, Roy H. Pollack, and Herbert S.Schlosser will report to Mr. Frederick. RCAStaff executives Eugene E. Beyer, Jr., Ken-neth W. Bilby, George H. Fuchs, Rocco M.Laginestra, Richard W. Miller, and ThomasB. Ross will continue to report to the Chair-man of the Board and Chief Executive Officer.

Laboratories

Emil V. Fitzke, Head, Technological Servi-ces, announces his organization as follows:Austin J. Kelley, Jr., Manager, TechnicalSupport Services; Jack F. Otto, Manager,Device Technology: William J. Schnelli, Ad-ministrator, Display Device Deve opment:and Donald J. Tamutus, Manager, ProcessTechnology.

Carmen A. Catanese, Director, Picture TubeSystems Research Laboratory, announcesthe appointment of Harry E. McCandlessas Manager, Advanced Development-Elec-tron Guns, Transfer Technology Labo-ratories.

National BroadcastingCompany (NBC)

Jeffrey P. Meadows, Vice -President, Engi-neering and Technical Services, Operationsand Technical Services. announces the ap-pointment of Donald R. Musson as Director,Technical Development Operations and Tech-nical Services. NBC, and the appointment ofSteven Bonica as Director, Broadcast Sys-tems, Engineering, NBC.

Patents

John V. Regan, Vice -President, Patent Oper-ations. announces the appointment ofJoseph S. Tripoli as Director, Patents-Electronic Systems.

Solid State Division

Stephen L. Pletcher, Division Vice -Presi-dent, Marketing, announces the appointmentof Michel Musso as Director of Marketing,Europe.

Erich Burlefinger, Division Vice -President,Electro-Optics and Power Devices, an-nounces the appointment of S. Paul Davisas Director, Power Operations.

Carl R. Turner, Division Vice -President, Prod-uct Assurance and Planning, announcesthe appointment of John E. Mainzer as Di-rector, Division Planning and OperationsSupport.

David S. Jacobson, Director, Custom LargeScale Integration, announces the appoint-ment of Peter Ferlita as Manager, CustomLSI Production and Production Control.

James W. Hively, Director, Semicustom De-vice Operations, Integrated Circuits, an-

Protessional activities

A tribute to Otto Schade

Otto Schade, Sr., devoted most of his pro-fessional career to research on televisionand the image sciences at RCA. Perhapsmore than anyone else, he was respons-ible for unifying our understanding of theevaluat on of optical, photographic, and elec-tronic imaging systems, as well as the hu-man visual system. In honor of his numer-ous contributions to the imaging sciences,the Optical Society of Ameica held a Sym-posium in Tribute to Otto Schade at itsAnnual Meeting in Tucson, Arizona, on Oc-lobe- 21, 1982. The Symposium began withthree invited papers that reviewed Schade'scontributions.

Dr. Albert Rose, retired Fellow at RCALaboratories, presented a paper entitled:"Otto Schade: A Personal Retrospective."Prof John Robson of Cambridge Univer-sity (U K.) then gave a paper summarizingSchade's contributions to the visual scien-ces, followed by a paper by Dean Brian

nounces the appointment of Henry S. Milleras Manager, Design Systems Engineering,and Walter Clauhs as Manager, ProductEngineering, Semicustom DeviceOperations.

RCA Cylix CommunicationsNetwork, Inc.

Eugene F. Murphy, Chairman of the Board,,RCA Cylix Communications Network, Inc.,announces the organization as follows:James C. Ziegler, Vice -Chairman; Ralph R.Johnson, President and Chief Executive Of--icer. The organization of the President andChief Executive Officer will be as follows:Bryan M. Eagle, Vice -President, Financeand Administration; Richard C. Furnival, Vice -

President, Operations; Floyd H. Jean, Vice-president, Engineering Development; andRonald L. Young, Vice -President, Marketing.

Otto Schade, Sr.

Thompson, of the University of Rochester,which reviewed Schade's research on im-age quality and evaluation. Other papers,on related topics, were given by Ted Adel-son, Jim Bergen, and Albert Pica. Thesescientists are all members of the ImageQuality and Human Perception group atRCA Laboratories. Curtis Carlson, also ofRCA Laboratories, presided at the Symposium.

RCA Engineer a 27-6 a Nov./Dec. 1982 97

Dr. Woll honored byUniversity of Pennsylvania

Dr. Harry J. Woll, Staff Vice -President andChief Engineer of RCA Electronic Products,Systems and Services, has received the1982 Yarnell Award from the University ofPennsylvania's Engineering Alumni Society.The award is given annually to a graduateof the School for outstanding contributions.

Dr. Woll received his Ph.D. degree fromthe University of Pennsylvania in 1953. Onthe 50th anniversary of the University'sMoore School of Electrical Engineering in1973, he was awarded the School's GoldMedal as a distinguished alumnus. Dr. Wollcurrently serves as Chairman of the Trust-ees for the Moore School and is a memberof the Board of Overseers for the School ofEngineering and Applied Science.

During his 41 -year carreer with RCA, Dr.Woll has advanced through a number ofengineering and management positions.Prior to being named to his current post in1981, he was Division Vice -President andGeneral Manager, RCA Automated Systems,located in Burlington. Mass.

Dr. Woll holds 20 patents in various fieldsof electronics. His activities and responsi-bilities at RCA have included the develop-ment of circuitry, micro -electronics, lasers,computers. electro-optics, automatic testequipment and air traffic control systems.In addition, he has been responsible forthe design of the rendezvous radar, atti-tude control electronics, and descent enginecontrol electronics for the Apollo Lunar Mod-ule spacecraft.

New TIMS President -Elect

Dr. H. Newton Garber, Director of Opera-tions Research for the RCA Corporation,has become President -Elect of the Instituteof Management Sciences (TIMS). His termas President of TIMS will begin in Sep-tember, 1983. The organization's goals areto identify, extend and unify scientific knowl-edge pertaining to management, throughpublication of journals and through local,regional, national and internationalmeetings.

Astro scientist cited atComputer Society conference

Kamal N. Karna, Astro-Electronics, servedas the tutorial chairman and an executivecommittee member for The CompCon Fall1982 conference held in Washington, D.C..in September, 1982. As a tutorial chairman,he organized four tutorials: Integrated voiceand data PBXs; Data Communications: Tech-niques and Approaches; Local Networks; Intro-duction and Equipment and Computer Com-munications-Protocols.

RCA Labs celebrates its 40th Anniversary

Aerial view of RCA Laboratories in Princeton, N.J. Established in 1942, the Princeton facilitywas 40 years old on September 27, 1982. In 1951, the facility was dedicated as the DavidSarnoff Research Center.

RCA Laboratories formally opened on Sep-tember 27, 1942. Research accomplish-ments during the past 40 years havebeen many and versatile. RCA scientistshave contributed importantly to the fieldsof optics, acoustics, and communicationsand have developed valuable tools andsystems for our national defense.

Dr. William M. Webster, Vice -Presidentof the Laboratories, in a message to em-ployees said, "We are a little more thanmidway between our 25th and 50th anni-versaries, so we might take time to lookback to the 60s and 70s and ahead tothe next decade. We have learned thatour strengths lie in our traditional busi-nesses-communications, electronics,and entertainment.

"We can take pride in an impressivelist of achievements over the years; per-haps the most notable one was the devel-opment of the all -electronic color televi-

sion system in use today. And a yearago, we introduced our new capacitanceelectronic disc VideoDisc system. Oursis an industry in which our eyes areusually on the future, but occasionally itis pleasant to look back on the past."

RCA Laboratories has carried out pio-neering work in computer memories andsystems, high-fidelity stereo recording,lasers, solid-state materials and devices,satellite communications, and numerousother achievements covering the entireelectronics spectrum. (See box, opposite.)

"Ten years from now, on our 50th birth-day, our research projects will be at leastas different as today's are from 1972,"Dr. Webster predicted. "RCA Laborato-ries has been outstanding in developingand applying technology throughout itshistory. By the diligence, dedication, andcreativity from all of us, let's prove thatlife begins at 40."

Mr. Karna was given a special citationfor his contribution to the success of theconference.

Degrees awarded

Two individuals at SSD, Somerville, recentlycompleted their undergraduate education.Alan S. Gutwillig received his Bachelor ofScience degree in Electrical Engineeringfrom Rutgers University. Louis Pennisi re-ceived his degree in Electrical Engineeringfrom Fairleigh Dickinson University.

New IEEE Senior Member

Richard M. Dombrosky, RCA Service Com-pany, Cherry Hill, has been elected a seniormember of the Institute of Electrical & Elec-tronics Engineers. Senior member, the high-est professional grade for which applica-tion may be made, requires experience reflect-ing professional maturity.

Licensed engineer"SelectaVision"VideoDisc Operations:T.J. Dudziak, IN 19887

98 RCA Engineer 27-6 Nov./Dec. 1982

Some milestones in1942 Dedication of RCA Laborato-

ries, on September 27. The techni-cal staff in Princeton numbered125 scientists and engineers.

1943 Development of the ImageOrthicon with sensitivity 1,000times greater than that of the Icon-oscope. For postwar television, itprovided flexible operation in thestudio or in the field.

1946 All -electronic color televisionsystem demonstrated publicly.... Development of aluminizedpicture tube which doubled bright-ness with no increase in power.

1949 Development of Vidicon, aminiature pickup tube with a pho-toconductive surface; important forclosed-circuit TV in industrial andeducational applications.

1950 Development and demonstra-tion to the FCC of three -gun shad-ow -mask tube-the tricolor kine-scope that became standard in theindustry.

1953 Magnetic tape recording sys-tem for both color and black -and -white TV programs developed anddemonstrated.

1954 "Electrofax," a high-speedelectrostatic printing processdeveloped; later licensed to pho-tocopier manufacturers.

Laboratories research1955 Electronic music synthesizer,

first of two models, built.... Alloy -junction "drift" transistordeveloped.

1958 Intensifier Orthicon cameratube, which could "see" in sur-roundings completely dark tohuman eye, developed.

1959 Direct-coupled-unipolar-tran-sistor-logic circuit developed.

1960 N -on -P radiation -resistantsolar cells developed for U.S. Sig-nal Corps.

1961 Cadmium -sulfide thin-filmtrans stor, made entirely by evapo-ration technique.

1962 First sun -pumped laserdeveloped.

1963 Development of the metaloxide semiconductor (MOS)transistor.

1966 First practical-techno ogy,vapor -phase growth processdeveioped for using gallium arse-nide in high-performance elec-tronic devices.

1967 S licon-on-sapphire fabricationtechnique developed for produc-ing large arrays of silicon field-ef-fect transistors Developmentof an experimental system

(Homefax) for broadcasting printedcopy into the home along withstandard TV programming.

1968 Liquid crystal techndogy pro-viding new type of electronic dis-play for print, pictures and movingimages; read by reflected light.

1973 First linear CMOS circuitsintroduced.

1974 Development of solid-stateimage sensor (CCD) containingmore than 120,000 electronic ele-ments on a silicon sensor chip thesize of a nickel-the forerunner ofthe CCD camera.

1975 Introduction of first COS/MOSmicroprocessor.

1976 Development of microwavehyperthermia units for treatment ofdeep-seated tumors.

1977 First SOS memory introduced.

1979 Development of solid-statelaser for high -bit -rate fiber opticcommunications.

1980 Completion of 15 -year devel-opment of "SelectaVision" Video -Disc system.... Demonstration ofconcept for a 50 -inch, flat -panelcolor TV display for wall mounting.

Technician awards at Lancaster plant

Each year outstanding technicians and drafts-men at Lancaster are chosen for Techni-cian Recognition awards to recognize thevery important contributions made by them.

The Picture Tube Division recipients ofthe 1981 Technician of the Year award areas follows:

C. Gerald Berrier-for outstanding contri-butions in picture tube life test expansionand modernization.

Robert P. Bitzer-for outstanding contribu-tions in picture tube electron gun shuntand enhancer modifications.

David E. Booth-for outstanding contribu-tions in PTD and SSD process equipmentmechanical coordination and redesign.

PTD's 1981 Technician of the Year award recipients, left to right: R. William Collins, William H.Shelton, Gerald R. Long, Robert P. Bitzer, Richard E. Roland, Kevin M. Rapp, James W. Hauer,C. Gerald Berrier, Dennis J. Urban, and Dennis L. Miller. (Missing from the photo are David E.Booth, Glenn W. Brunner, and John H. Ries.)

RCA Engineer 27-6 Nov./Dec. 1982 99

SSD's 1981 Technician of the Year award recipients, left to right: Scott K. Kurtz, Thadeus J.Jaworski, James P. Irvin, Kenneth J. Altmanshofer, Mark W. Howard, James E. Horst, Galen E.Shaud, Jr., Mary F. Kross, and Robert A. Barnes. (Missing from the photo is John S. Ritchey.)

Glenn W. Brunner-for outstanding contri-butions in color display tube pilot produc-tion matrix screening.

R. William Collins-for outstanding contri-butions in the picture tube high -voltage sta-bility computer -controlled data acquisitionsystems.

James W. Hauer-for outstanding contribu-tions in high -resolution color display tubequality standards and product demonstrations.

Gerald R. Long-for outstanding contribu-tions in picture tube process -equipment mech-anical design, development and manufac-turing start-up.

Dennis L. Miller-for outstanding contribu-tions in picture tube process -equipment com-puter -aided design systems.

Kevin M. Rapp-for outstanding contribu-tions in high -resolution color -display -tube

shadow -mask materials and etching devel-opments.

John H. Ries-for outstanding contributionsin the study of picture tube deflection com-ponent relationship on beam register.

Richard E. Roland-for outstanding contri-butions in start-up of the Defect AnalysisCenter.

William H. Shelton-for outstanding contri-butions in the elevation of POT to a highlevel of competence and respect.

Dennis J. Urban-for outstanding contribu-tions in the detection and analysis of plantwaste water effluents.

The Solid State Division recipients of the1981 Technician of the Year awards are asfollows:

Kenneth J. Altmanshofer-for outstanding

contributions in neutral beam source andpower tube product improvements.

Robert A. Barnes-for outstanding contri-butions in solid-state systems products, 9 -inch black -and -white monitor development.

James E. Horst-for outstanding contribu-tions in CCD imager design, layout andtesting.

Mark W. Howard-for outstanding contri-butions in closed-circuit video equipmentdevelopment utilizing computer -aided de-sign techniques.

James P. Irvin-for outstanding contribu-tions in vistacon yield and process -controlimprovement.

Thadeus J. Jaworski-for outstanding con-tributions in conversion tube SIT and vid-icon exhaust processing improvements.

Mary F. Kross-for outstanding contribu-tions in silicon wafer processing efficiencyand yield improvements.

Scott K. Kurtz-for outstanding contribu-tions in gallium arsenide material and laserfabrication yield and process -control im-provements.

John S. Ritchey-for outstanding contribu-tions in small power and pencil -tube pro-cess -control development.

Galen E. Shaud, Jr.-for outstanding con-tributions in closed-circuit video equipment,quality and reliability yield improvements.

Technical excellence

Frank Denny, Chairman of the new Technical Excellence Committee, addressing Americom'sengineers and engineering managers. To the lett is John Christopher, Vice -President of Tech-nical Operations, Americom. TEC members at the right are Carlton Barnes, Jim Colonna, LenDerkach, Fred Hoedl, and Doreen Jakubcak.

Americom launches its TEC

RCA American Communications formally in-troduced a Technical Excellence programto its engineers and engineering manage-ment on October 6, in a meeting at theHilton Conference Center in Princeton.

The meeting began with brief state-ments by Americom Vice Presidents JohnChristopher and Allen Cook and by Amen-com's President. Andrew F. Inglis, in whichthey voiced their support of the TechnicalExcellence Program. The chairman of Amer-icom's Technical Excellence Committee,Frank Denny, introduced the committee mem-bers and spoke about the purpose of theprogram.

A social hour after the meeting providedengineers, managers, and committee mem-bers with an excellent opportunity to ex-change ideas regarding the technical excel-lence program.

100 RCA Engineer 27-6 Nov./Dec. 1982

Five receive Consumer Electronics Division quarterly awards

Gobush

Mc Vety

Price

Lord

Nortrup

The Consumer Electronics Division TEChas selected their second quarter 1982 Tech-nical Excellence Award winners. Based on

managers' recommendations, the recipients'work was researched by the Technical Ex-cellence Committee. The nominating man-agers were then interviewed by the com-mittee.

The award winners listed below will re-ceive a plaque and a reference book oftheir choice. They are also eligible to re-ceive an annual award for their work.

Raymond Gobush-for the developmentand implementation of electrical, mechani-cal, and plumbing plans to convert the print-ed circuit board etching process from usingferric chloride to cupric chloride, resultingin considerable monetary savings.

Jill Lord-for the development and imple-mentation of chemical processes and safetyprocedures to convert the printed circutboard etching process from using ferricchloride 10 cupric chloride, resulting in con-siderable monetary savings.

Ronalc McVety-for the invention of a chip(leadless resistor or capacitor) removal tool,and for the factory implementation of themethod which prevents damaging printedcircuit boards.

Kevin Nortrup-for significant personal ded-ication for the use of innovative techniquesto maximize feature content and cost effec-tiveness in J -Line (1983) remote control TVsystems.

Tim Price-for significant contributions tothe area of computer assisted engineeringcharacterization of radio frequency (RF) de-vices and systems.

Second-quarter 1982 MSR award -winners announced

Bernie Matulis, Chief Engineer, Missile andSurface Radar, announced that the follow-ing award -winners will receive a commem-orative desk plaque and a current text orreference book, and will be honored at a func-tion next year.

D.L. Matthews-for system configuration de-sign and the development of prototype hard-

ware and unique printed circuit designs forthe RCA Control Processor Microcompu-ter. Largely as a result of Mr. Matthews'innovative approach, the prototype designsproduced have met all microcomputer sys-tem requirements with an accuracy thatallowed debugging to be accomplished inless than one month.

D. Shaw-for developing the basic compu-ter architecture for the RCA Control Pro-cessor Microcomputer. In particular, his ap-proach to coding led to the microcoding ofmore than 200 instructions in 13 weeksinstead of the anticipated 40 weeks. Mr.Shaw's architecture, developed in what wasconsidered by many to be an unworkablyshort schedule, has been proved by theperformance of the microcomputer testedwith an Air Force acceptance program.

Custom LSI Symposiumheld at Princeton

The second corporate technical symposiumon custom LSI circuits was held on Novem-ber 18 at the David Sarnoff Research Cen-ter, with over 60 in attendance. Accordingto Jack Hilibrand, GSD Engineering, whoserved as chairman and organizer, the ob-jective of the symposium program was "toencourage the use of custom and semi -custom LSIs within RCA where that use isadvantageous to the RCA product beingbuilt." The morning session focused ontools available in RCA, while the afternoonwas devoted to case studies of programsnow underway.

The speakers and their topics are listedbelow. Copies of the viewgraphs used inthe presentations are available at RCA tech-nical libraries. For more information on aspecific talk, contact the respective author.The Custom LSI Symposium was video-taped and tapes will be available from theCorporate Engineering Education videotapelibrary by January 1983. To borrow thistape (at no charge), call Margaret Gilfillanat TACNET 222-5255. Order Tape No. 440.Available in the 3/4 -inch U-Matic and the1/2 -inch VHS formats.Opening Remarks, H.J. Woll, Staff Vice -

President, and Chief Engineer, RCA Elec-tronic Products, Systems and Services.

"Why Custom Designs Now-A Current Per-spective," J. Hively, Solid State Division.

"A System for the Automated Design ofComplex Integrated Circuits," R.P.Lydick, Solid State Division.

"Custom VLSI Design for GSD, by GSD," A.Feller, Advanced Technology Labo-ratories.

"How to Use CAD Tools to be Sure YourCustom LSI Part Will Work the First Time,"L. Rosenberg, Solid State TechnologyCenter.

"What is Involved in Building a Custom LSISkill Group to Support Your Require-ments," F.H. Tillwick, Missile and SurfaceRadar.

Panel Discussion: What Tools/Techniques/Technology Are We Missing? Strengthsand Weaknesses?

"The Military Computer Family-An Appli-cation of Custom VLSI and StandardParts," S.E. Ozga, Missile and SurfaceRadar.

"Planning A Chip Set for Digital Video," I.H.Kalish, RCA Laboratories.

"Planning Custom LSI Development in Con-sumer Electronics," H. Scalf, ConsumerElectronics.

"Obstacles to the Use of Custom LSI inLow -Volume Systems-Two Broadcast Cam-era LSI Circuits," T.R. Smith, RCA Labo-ratories, and A.H. Lind, Commercial Com-munications Systems Division.

RCA Engineer 27-6 Nov./Dec. 1982 101

Index to Volume 27Subject Index In this Subject Index, we asked the authors to classify their own articles

under three principal subject headings. Under these headings you willfind appropriate article titles, each followed by author name(s), volumeand issue numbers, issue date, and initial page number.

AlgorithmsMechanically scanned radar systems, dynamic

analysis and simulation of. Liston. J. and Sparks.

G. M., v27n6, Nov/Dec 1982, p24.

AmplifiersGaAs microwave monolithic integrated circuits, a

review of. Huang. H. C., Upadhyayula, L. C.and Kumar. M., v27n5. Sept/Oct 1982, p58.

Solid-state power amplifier replacing TWTs in C -band satellites. Wolkstein, H. J. and LaPrade. J.

N.. v27n5, Sept/Oct 1982, p7.

Analog -to -digital convertersGaAs microwave monolithic integrated circuits, a

review of. Huang, H. C., Upadhyayula, L. C.and Kumar, M., v27n5, Sept/Oct 1982. p58.

Analytical chemistryChemical and Physical Laboratory supports Video -

Disc. Hakala, D. F., v27n1. Jan/Feb 1982, p39.

AntennasLow-sidelobe antennas for tactical phased -array ra-

dars. Patton. W. T., v27n5, Sept/Oct 1982, p31.

AstronomyMaking telescopes is my hobby. Schneider, W..

v27n2, Mar/Apr 1982. p75.

Automated designComputer design and testing of a microwave an-

tenna -feed manifold. Bolger, G. P. and Holaday,V. D.. v27n5, Sept/Oct 1982, p87.

MIMIC logic simulator. Ashkinazy, A.. v27n6,Nov/Dec 1982. p10.

Automated testingATE, distributed system architectures for. Fay. J.

E., v27n4. July/Aug 1982, p53.Automated measurement of transistor I -V charac-

teristics. Snowden. M. P. and Amantea. R.,v27n2, Mar/Apr 1982, p69.

Automatic test equipment for VideoDisc andstylus. Kowalchik, J. J. and Selwa, A. P., v27n1,

Jan/Feb 1982. p30.Large system development, digital interface simu-

lation for. Suhy, G. W., v27n6, Nov/Dec 1982,p39.

AutomationAssembly mechanization program at RCA Solid

State (RAMP). Koskulitz, J. and Rosenfield. M..v27n4, July/Aug 1982, p57.

Automated assembly of lumped -element GaAsFET power amplifiers. Klatskin, J. B., Haggis.D., Camisa, R. L. and Joyce, B. T., v27n5. Sept/Oct 1982, p64.

AwardsDavid Sarnoff Awards for Outstanding Technical

Achievement (1982). RCA ENGINEER Staff,v27n4, July/Aug 1982. p4.

"SelectaVision" VideoDisc Awards. RCA ENGI-NEER Staff. v27n1, Jan/Feb 1982, p4.

BroadcastingDirect Broadcasting Satellite: Black hole or bonan-

za? Clark, J. F., v27n4, July/Aug 1982. p82.

Business operationsStrategic framework to strengthen RCA. Brad-

shaw. T. F., v27n4, July/Aug 1982, pl.

Charge coupled devices (CCDs)An intelligent missile seeker, design and simulation

of. Hayman, J., v27n6, Nov/Dec 1982, p43.Infrared sensor systems, design and performance

prediction of. Cantella, M. J., v27n3, May/June1982. p67.

Infrared -camera system developed to use theSchottky -barrier IR-CCD array. Klein, J. J.,Roberts, N. L. and Chin. G. D., v27n3. May/June 1982, p60.

Circuit analysisMIMIC logic simulator. Ashkinazy, A.. v27n6.

Nov/Dec 1982, p10.

Circuit boardsParylene-coating process at MSR. Philip, G. O..

v27n2. Mar/Apr 1982. p56.

Circuit designANA radio frequency circuit simulation and anal-

ysis. Perlow. S. M., v27n6, Nov/Dec 1982. p46.

Circuit devicesActive and passive microwave components. Den -

linger. E. J., v27n5, Sept/Oct 1982, p50.

Command and control systemsLarge system development, digital interface simu-

lation for. Suhy. G. W.. v27n6, Nov/Dec 1982.p39.

Communication componentsLinear integrated circuits arc going digital. Wittlin-

ger, H. A., v27n1, Jan/Feb 1982, p58.

Communication satellitesRCA Satcom system. computer modeling in the.

Guida, A. A., Jonnalagadda, K.. Raychaudhuri.D. and Schiff, L., v27n6, Nov/Dec 1982, p84.

Receivers, wideband, for communications satellites.Goldberg, H., v27n5. Sept/Oct 1982, p37.

Communication systemsMicrowave communications system for RCA

Globcom. Solomon. S. M., v27n4, July/Aug1982. p75.

CompoundsMaterial compounding of VideoDiscs demands the

right chemistry. Whipple. B. and Dunn, V. S.,v27n1, Jan/Feb 1982, p16.

Computer applicationsAutomated measurement of transistor I -V charac-

teristics. Snowden, M. P. and Amantea, R..v27n2, Mar/Apr 1982, p69.

Computer music, an illustrated history of. Mercuri,R. T., v27n4, July/Aug 1982, p65.

Computer programming and systems analysis inthe rapidly changing VideoDisc environment(application of FOCUS). Myers, W. W., v27n4,July/Aug 1982, p48.

Engineering environment, simulation in an. Pitts,K. A. and Barton, R. R., v27n6, Nov/Dec 1982;p4.

Computer graphicsEngineering Form and Function: Windows on

RCA Technology. RCA ENGINEER Staff,v27n4, July/Aug 1982. p33.

Computer programmingComplex weapon systems, discrete -event simula-

tions of. Golub, J., v27n6, Nov/Dec 1982, p32.Computer programming and systems analysis in

the rapidly changing VideoDisc environment(application of FOCUS). Myers, W. W., v27n4,July/Aug 1982. p48.

Software reliability, discovering how to ensure.Trachtenberg. M.. v27n I, Jan/Feb 1982, p53.

Computer programsComputer design and testing of a microwave an-

tenna -feed manifold. Bolger, G. P. and Holaday,V. D., v27n5, Sept/Oct 1982, p87.

102 RCA Engineer 27-6 Nov./Dec. 1982

Industrial engineering

Computer simulationAn intelligent missile seeker, design and simulation

of. Hayman, J.. v27n6, Nov/Dec 1982. p43.Business modeling and the engineer: An Astro-

Electronics application. Nigam, A. K., Hong, S.and Spence, S. R.. v27n6, Nov/Dec 1982. p65.

Complex weapon systems, discrete -event simula-tions of. Golub, J., v27n6, Nov/Dec 1982, p32.

Computer -aided design of metal -forming processes.Chen, C. C., v27n2, Mar/Apr 1982. p45.

Electromechanical motor simulation. Browne, R.G., v27n6, Nov/Dec 1982, p80.

Engineering environment, simulation in an. Pitts,K. A. and Barton, R. R., v27n6, Nov/Dec 1982,

P4 -

HE modem simulation. DeMaria, P. A. and Bodzi-och, K. J., v27n6, Nov/Dec 1982, p18.

Infrared sensor systems, design and performanceprediction of. Centella, M. J.. v27n3, May/June1982. p67.

Large system development, digital interface simu-lation for. Suhy. G. W., v27n6. Nov/Dec 1982.p39.

Mechanically scanned radar systems, dynamicanalysis and simulation of. Liston, J. and Sparks.G. M., v27n6, Nov/Dec 1982, p24.

MIMIC logic simulator. Ashkinazy, A., v27n6,Nov/Dec 1982, p10.

Modeling and simulation: Tools for the 80s. Teger,A. H., v27n6, Nov/Dec 1982, p1.

RCA Satcom system, computer modeling in the.Guide, A. A., Jonnalagadda, K., Raychaudhuri,D. and Schiff. L., v27n6, Nov/Dec 1982, p84.

Computer storageData recording on optical discs at 300 Mb/s. Am-

mon, G. J. and Reno. C. W.. v27n3, May/June1982. p36.

Computer systemsATE, distributed system architectures for. Fay. J.

E., v27n4. July/Aug 1982, p53.

Computer terminalsColor -data -display CRT: A product whose time

has come. Barbin, R. L., Simpson, T. F. andMarks, B. G.. v27n4, July/Aug 1982. p23.

Computer -aided analysisANA radio frequency circuit simulation and anal-

ysis. Perlow, S. M., v27n6, Nov/Dec 1982, p46.Automated assembly of lumped -element GaAs

FET power amplifiers. Klatskin, J. B., Haggis,D.. Camisa. R. L. and Joyce, B. T., v27n5, Sept/Oct 1982, p64.

Mechanical and thermal design, modeling andsimulation in. Enstrom, R. E., v27n6, Nov/Dec1982, p71.

Computer -aided design (CAD)Computer -aided design and testing of microwave

circuits. Perlman, B. S., Rhodes. D. L. andSchepps, J. L., v27n5, Sept/Oct 1982, p76.

Computer -aided design of metal -forming processes.Chen, C. C.. v27n2, Mar/Apr 1982, p45.

MIMIC logic simulator. Ashkinazy, A., v27n6,Nov/Dec 1982, p10.

Computer -aided manufacturing(CAM)Numerical control: The cornerstone of computers

in manufacturing. Young, C., v27n2, Mar/Apr1982, p41.

Computer -aided testing (CAT)Computer design and testing of.a microwave an-

tenna -feed manifold. Bolger, G. P. and Holaday,V. D., v27n5, Sept/Oct 1982, p87.

Computer -aided design and testing of microwavecircuits. Perlman, B. S., Rhodes, D. L. andSchepps, J. L., v27n5, Sept/Oct 1982, p76.

Computers, homeComputer music, an illustrated history of. Mercuri,

R. T., v27n4, July/Aug 1982, p65.

Computers, manufacturingComputer -aided manufacture of precision mi-

crowave components for phased -array radarGaAs MESFET power amplifiers. Breese. M. E.and Mason, R. J., v27n5, Sept/Oct 1982, p71.

Data communicationsHE modem simulation. DeMaria, P. A. and Bodzi-

och, K. J., v27n6, Nov/Dec 1982, p18.

Data processingComputer programming and systems analysis in

the rapidly changing VideoDisc environment(application of FOCUS). Myers, W. W., v27n4,July/Aug 1982, p48.

DesignEngineering Form and Function: Windows on

RCA Technology. RCA ENGINEER Staff,v27n4, July/Aug 1982, p33.

Digital electronicsLinear integrated circuits are going digital. Wittlin-

ger, H. A., v27n1, Jan/Feb 1982, p58.

Direct broadcasting satellites (DBS)Direct Broadcasting Satellite: Black hole or bonan-

za? Clark, J. F., v27n4, July/Aug 1982. p82.

Display systemsColor -data -display CRT: A product whose time

has come. Barbin, R. L., Simpson, T. F. andMarks, B. G., v27n4, July/Aug 1982, p23.

EditorialsEditorial iiput: Two new editorial features. King,

T. E., v27n5, Sept/Oct 1982, p30.

EducationContinuing education for manufacturing engineers.

RCA Corporate Engineering Education, v27n2,Mar/Apr 1982, p38.

Electro-optics, lasersLong -wavelength semiconductor diode lasers for

optical communications. Olsen. G. H. and Et-tenberg. M., v27n4, July/Aug 1982, p38.

Space applications of lidar. Altman, W., Hogan, D.B., Hinter. E. C. and Rosenberg, A., v27n3,May/June 1982, p44.

Electro-cptics, systems andtechniquesData recording on optical discs at 300 Mb/s. Am-

mon, G. J. and Reno, C. W., v27n3, May/June1982, p36.

Electro-Optics: Insight to the future. Decker, C.D.. v27n3, May/June 1982, pl.

Silicon avalanche photodiodes, recent develop-ments in. Webb, P. P. and McIntyre, R. J.,v27n3, May/June 1982, p96.

Space applications for RCA infrared technology.Warren, F. B., Tower, J. R., Gandolfo, D. A.,Elabd, H. A. and Villani, T. S., v27n3, May/lure 1982, p76.

TV camera for target -acquisition and laser -designa-tor systems, high performance. Arlan, L., v27n3,May/June 1982, p87.

Electrochemical energy storageCoal. upgrading, for more efficient electric -power

generation. Williams, R., v27n4, July/Aug 1982,

p45 -

Energy conversionCoal. upgrading, for more efficient electric -power

generation. Williams, R., v27n4, July/Aug 1982,

P45

Engineering tools and techniquesEngineering environment, simulation in an. Pius,

K. A. and Barton, R. R., v27n6, Nov/Dec 1982,

P4

Fiber opticsFiberoptic communications systems, design, deve-

lopment and installation of. Ainsbury, D. C.,v27n3, May/June 1982, p28.

Finite element analysisMechanical and thermal design, modeling and

simulation in. Enstrom, R. E., v27n6, Nov/Dec1982, p71.

Heat transferMechanical and thermal design, modeling and

simulation in. Enstrom, R. E., v27n6, Nov/Dec1.4512, p7I

Image encodingHuman visual system, modeling of the. Adelson,

E H., Pica, A. P. and Carlson, C. R.. v27n6,Nov/Dec 1982, p56.

Imaging systemsInfrared -camera system developed to use the

Schottky -barrier IR-CCD array. Klein, J. L.Roberts, N. L. and Chin, G. D., v27n3, May/June 1982, p60.

Schottky -barrier infrared image sensors. Kosonocky,W.F., Elabd. FLA., Erhardt. H.G., Shal[cross F.V.,Meray. G.M., Miller. R.. Villani, T.S., Groppe, J.V..Franc. VI.. and Tams. F.J.. v27n3. May/June1982. p49.

Space applications for RCA infrared technology.Warren, F. B., Tower, J. R., Gandolfo, D. A.,Elabd, H. A. and Villani, T. S., v27n3, May/June 1982, p76.

TV camera for target -acquisition and laser -designa-tor systems, high performance. Arlan, L.. v27n3,May/June 1982, p87.

Industrial engineeringMirror Fusion Test Facility, producing neutral

beam ion sources for the. Lynch, W. S. andReed, R. E., v27n2, Mar/Apr 1982, p62.

RCA Engineer 27-6 Nov./Dec. 1982 103

Information management

Information managementComputer programming and systems analysis in

the rapidly changing VideoDisc environment(application of FOCUS). Myers. W. W., v27n4,July/Aug 1982, p48.

Technical abstracts database (TAD): RCA data-base now an on-line system. Technical Informa-tion Systems, v27n2, Mar/Apr 1982, p22.

Information sciencesTechnical abstracts database (TAD): RCA data-

base now an on-line system. Technical Informa-tion Systems, v27n2. Mar/Apr 1982, p22.

Infrared equipmentInfrared sensor systems. design and performance

prediction of. Cantella, M. J.. v27n3, May/June1982, p67.

Infrared -camera system developed to use theSchottky -barrier IR-CCD array. Klein, J. J.,Roberts, N. L. and Chin, G. D., v27n3, May/June 1982, p60.

Schottky -barrier infrared image sensors. Kosonocky,W.F.. Elabd, ILA., Erhardt. H.G., Shallcross, F.V..Meray. G.M., Miller, R.. Villani, T.S.. Groppe, J.V.,Frantz, V.L. and Tams. F.J.. v27n3, May/June1982, p49.

Integrated circuitsAssembly mechanization program at RCA Solid

State (RAMP). Koskulitz, J. and Rosenfield. M..v27n4, July/Aug 1982. p57.

Schottky -barrier infrared image sensors. Kosonocky.W.F.. Eland. H.A., Erhardt. H.G.. Shallcross, F.V..Meray. G.M.. Miller. R., Villani, T.S., Groppe,Frantz. V.1., and Tams. F.J.. v27n3. May/June1982. p49.

Introductions (to issues)Electro-Optics: Insight to the future. Decker, C.

D., v27n3. May/June 1982. pl.Manufacturing programs at Consumer Electronics.

Borman, B. L., v27n2, Mar/Apr 1982, pl.Microwave technology at RCA. Sterzer, F., v27n5,

Sept/Oct 1982, pl.Modeling and simulation: Tools for the 80s. Teger,

A. H.. v27n6, Nov/Dec 1982, pl.Quality plus diversity plus availability equals suc-

cess. Schlosser, H. S.. v27nl, Jan/Feb 1982, p1.Strategic framework to strengthen RCA. Brad-

shaw. T. F.. v27n4, July/Aug 1982, pl.

Laboratory equipment andtechniquesAutomated measurement of transistor I -V charac-

teristics. Snowden. M. P. and Amantea, R..v27n2, Mar/Apr 1982, p69.

Laser applicationsLong -wavelength semiconductor diode lasers for

optical communications. Olsen. G. H. and Et-tenberg. M., v27n4, July/Aug 1982, p38.

Space applications of lidar. Altman, W., Hogan, D.B., Hutter, E. C. and Rosenberg, A., v27n3,May/June 1982. p44.

Laser trackingSpace applications of lidar. Altman. W., Hogan, D.

B., Flutter, E. C. and Rosenberg, A., v27n3,May/June 1982. p44.

Lasers, injectionHigh -density optical recording using laser diodes.

Bartolini. R. A., v27n3, May/June 1982, p20.Single -mode diode lasers for systems -oriented ap-

plications. Botez, D.. v27n3, May/June 1982,p8.

LensesMaking telescopes is my hobby. Schneider, W.,

v27n2, Mar/Apr 1982, p75.

ManagementStrategic framework to strengthen RCA. Brad-

shaw, T. F., v27n4, July/Aug 1982. p1.

ManufacturingComputer -aided design of metal -forming processes.

Chen, C. C., v27n2, Mar/Apr 1982, p45.Equipment decisions for VLSI -circuit manufacture.

Blumenfeld, M. A. and Shambelan, R. C.,v27n2, Mar/Apr 1982. p35.

Manufacturing the VideoDiscs: An overview.Weisberg. H., v27nl, Jan/Feb 1982, p6.

Manufacturing. Custom and Quantity Manufactur-ing: An Engineering Comparison. D'Arcy, J. A.and Miller, R., v27n2, Mar/Apr 1982, p4.

Mastering VideoDiscs: The software to hardwareconversion. Kell, F. D., John, G. and Stevens, J.,v27n1, Jan/Feb 1982. p11.

Player Technology Center: Key to innovation.Mishra. D., v27n2. Mar/Apr 1982, p24.

Manufacturing engineeringContinuing education for manufacturing engineers.

RCA Corporate Engineering Education. v27n2.Mar/Apr 1982, p38.

Equipment decisions for VLSI -circuit manufacture.Blumenfeld, M. A. and Shambelan, R. C.,v27n2, Mar/Apr 1982, p35.

Manufacturing programs at Consumer Electronics.Borman, B. L., v27n2, Mar/Apr 1982, pl.

Manufacturing. Custom and Quantity Manufactur-ing: An Engineering Comparison. D'Arcy, J. A.and Miller. R., v27n2, Mar/Apr 1982. p4.

Mirror Fusion Test Facility, producing neutralbeam ion sources for the. Lynch, W. S. andReed. R. E.. v27n2, Mar/Apr 1982, p62.

Numerical control: The cornerstone of computersin manufacturing. Young, C.. v27n2, Mar/Apr1982, p41.

Player Technology Center: Key to innovation.Mishra. D.. v27n2. Mar/Apr 1982. p24.

Manufacturing equipmentEquipment decisions for VLSI -circuit manufacture.

Blumenfeld, M. A. and Shambelan, R. C.,v27n2, Mar/Apr 1982, p35.

Manufacturing, Custom and Quantity Manufactur-ing: An Engineering Comparison. D'Arcy, J. A.and Miller. R.. v27n2, Mar/Apr 1982, p4.

Numerical control: The cornerstone of computersin manufacturing. Young, C., v27n2. Mar/Apr1982, p41.

MaterialsChemical and Physical Laboratory supports Video -

Disc. Hakala, D. F., v27n1, Jan/Feb 1982, p39.

Medical electronicsLocalized hyperthermia treatment of cancer. Sterz-

er, F.. v27n1, Jan/Feb 1982. p43.Medical applications of microwave energy. Pagl-

ione. R. W., v27n5, Sept/Oct 1982, p17.

MicroelectronicsParylene-coating process at MSR. Philip. G. 0.,

v27n2, Mar/Apr 1982. p56.

Microwave antennasLow-sidelobe antennas for tactical phased -array ra-

dars. Patton, W. T., v27n5, Sept/Oct 1982, p31.Medical applications of microwave energy. Pagl-

ione, R. W.. v27n5, Sept/Oct 1982, p17.Shaped -beam reflector antennas for communica-

tions satellites. Profera, C. E.. Rosol, G. L.,Soule, H. H. and Kara. J., v27n5, Sept/Oct1982, p43.

Microwave relay systemsMicrowave communications system for RCA

Globcom. Solomon, S. M., v27n4, July/Aug1982, p75.

Microwave technologyActive and passive microwave components. Den -

linger, E. J., v27n5, Sept/Oct 1982, p50.ANA radio frequency circuit simulation and anal-

ysis. Perlow, S. M., v27n6, Nov/Dec 1982. p46.Automated assembly of lumped -element GaAs

FET power amplifiers. Klatskin, J. B., Haggis.D.. Camisa, R. L. and Joyce, B. T., v27n5, Sept/Oct 1982, p64.

Computer design and testing of a microwave an-tenna -feed manifold. Bolger, G. P. and Holaday,V. D., v27n5, Sept/Oct 1982, p87.

Computer -aided design and testing of microwavecircuits. Perlman, B. S., Rhodes. D. L. andSchepps, J. L.. v27n5, Sept/Oct 1982. p76.

Computer -aided manufacture of precision mi-crowave components for phased -array radarGaAs MESFET power amplifiers. Breese, M. E.and Mason, R. J., v27n5, Sept/Oct 1982, p71.

Localized hyperthermia treatment of cancer. Sterz-er, F., v27n1, Jan/Feb 1982, p43.

Medical applications of microwave energy. Pagl-ione, R. W., v27n5, Sept/Oct 1982. p17.

Microwave components research at RCA. Nowo-grodzki, M., v27n5, Sept/Oct 1982. p4.

Microwave technology at RCA. Sterzer, F., v27n5,Sept/Oct 1982. pl.

Radar instruments: Sensors for industrial applica-tions. Nowogrodzki, M., Mawhinney, D. D. andKipp, R. W., v27n5, Sept/Oct 1982. p23.

Receivers. wideband, for communications satellites.Goldberg. H., v27n5, Sept/Oct 1982, p37.

Solid-state power amplifier replacing TWTs in C -band satellites. Wolkstein, H. J. and LaPrade. J.N.. v27n5, Sept/Oct 1982, p7.

MicrowavesGaAs microwave monolithic integrated circuits, a

review of. Huang. H. C., Upadhyayula, L. C.and Kumar. M., v27n5, Sept/Oct 1982, p58.

Microwave communications system for RCAGlobcom. Solomon. S. M.. v27n4, July/Aug1982, p75.

Shaped -beam reflector antennas for communica-tions satellites. Profera, C. E., Rosol, G. L.,Soule. H. H. and Kara. J.. v27n5, Sept/Oct1982. p43.

Military communicationsPin diode, high -power, the Engineer's Notebook.

Stabile. P. J., v27n5, Sept/Oct 1982. p21.

104 RCA Engineer 27-6 Nov /Dec 1982

RCA Solid State Division

Missile systemsAn intelligent missile seeker, design and simulation

of. Hayman. J., v27n6, Nov/Dec 1982. p43.

Molding, phonograph recordsMicro -molding is a VidcoDisc requirement.

McNeely, M. and Bock, M., v27n1, Jan/Feb1982. p19.

MotorsElectromechanical motor simulation. Browne. R.

G.. v27n6. Nov/Dec 1982. p80.

MusicComputer music, an illustrated history of. Mercuri,

R. T., v27n4, July/Aug 1982, p65.

Navigation satellitesThe second -generation Navy Navigation Satellite.

NOVA -1, expectations achieved. Staniszewski. J.R., v27n4. July/Aug 1982. p89.

Nuclear energyMirror Fusion Test Facility, producing neutral

beam ion sources for the. Lynch. W. S. andReed, R. E., v27n2, Mar/Apr 1982, p62.

On the job/off the jobMaking telescopes is my hobby. Schneider. W..

v27n2, Mar/Apr 1982. p75.

Operations researchBusiness modeling and the engineer: An Astro-

Electronics application. Nigam, A. K., Hong, S.and Spence. S. R., v27n6, Nov/Dec 1982, p65.

Optical communicationsElectro-Optics: Insight to the future. Decker. C.

D.. v27n3, May/June 1982. pl.Fiber-optic communications systems, design, deve-

lopment and installation of. Ainsbury, D. C..v27n3, May/June 1982. p28.

Long -wavelength semiconductor diode lasers foroptical communications. Olsen. G. H. and Et-tenberg. M., v27n4, July/Aug 1982, p38.

Optoelectronic systems, some of the new systems.Ettenberg, M., v27n3, May/June 1982. p4.

Silicon avalanche photodiodes, recent develop-ments in. Webb, P. P. and McIntyre, R. J.,v27n3, May/June 1982. p96.

Single -mode diode lasers for systems -oriented ap-plications. Botez. D.. v27n3. May/June 1982.p8.

Optical systemsHigh -density optical recording using laser diodes.

Bartolini, R. A., v27n3. May/June 1982, p20.Optoelectronic systems, some of the new systems.

Ettenberg. M., v27n3, May/June 1982. p4.

Package engineeringPackage engineering for the Picture Tube Division.

Haggerty. G. L., v27n2, Mar/Apr 1982, p50.

PhotoconductivitySaticon photoconductor: The latest in color -cam-

era -tube photoconductors. Neuhauser, R. G..v27n3, May/June 1982, p81.

PhotodiodesSilicon avalanche photodiodes, recent develop-

ments in. Webb, P. P. and McIntyre, R. J..v27n3, May/June 1982. p96.

Physical chemistryCoal, upgrading, for more efficient electric -power

generation Williams, R., v27n4, July 'Aug 1982,P45.

Picture tubesColor -data -display CRT: A product whose time

has come. Barbin, R. L., Simpson, T. F. andMarks, B. G., v27n4, July/Aug 1982, p23.

Package engineering for the Picture Tube Division.Haggerty. G. L.. v27n2, Mar/Apr 1982, p50.

PlasticsCaddy design and manufacturing of the VideoDisc

caddy. Torrington, L. A., v27n1, Jan Feb 1982.p23.

Material compounding of VideoDiscs demands theright chemistry. Whipple. B. and Dunn, V. S.,v27n1, Jan/Feb 1982, p16.

Printed circuit boardsParylene-coating process at MSR. Philip, G. 0.,

v27n2. Mar/Apr 1982, p56.

Process controlStatistics applications in manufacturing. Gunter, B.

H. and Rayl. M., v27n2, Mar/Apr 1982, p14.

ProductivityQuality Circles at RCA Scranton. Mecca, S. J.,

v27n2, Mar/Apr 1982. p30.

Propulsion, electronic and ionicThe second -generation Navy Navigation Satellite,

NOVA -1, expectations achieved. Staniszewski. J.R., v27n4. July/Aug 1982, p89.

PublishingEditorial input: Two new editorial features. King,

T. E., v27n5, Sept/Oct 1982, p30.Editorial Representatives, meet your. RCA ENGI-

NEER Staff. v27n1, Jan/Feb 1982. p35.Technical abstracts database (TAD): RCA data-

base now an on-line system. Technical Informa-tion Systems, v27n2, Mar/Apr 1982, p22.

Quality controlProduct Assurance of VideoDisc : VideoDisc per-

formance in the real world. Huck, R. H., v27n1,Jan/Feb 1982, p27.

Quality Circles at RCA Scranton. Mecca, S. J..v27n2, Mar/Apr 1982, p30.

Statistics applications in manufacturing. Gunter, B.H. and Ray', M., v27n2, Mar/Apr 1982, p14.

Radar antennasLow-sidelobe antennas for tactical phased -array ra-

dars. Patton, W. T., v27n5, Sept/Oct 1982, p31.

Radar applicationsRadar instruments: Sensors for industrial applica-

tions. Nowogrodzki, M., Mawhinney, D. D. andKipp, R. W , v27n5, Sept/Oct 1982. p23.

Radar systemsMechanically scanned radar systems, dynamic

analysis and simulation of. Liston, J. and Sparks,G. M., v27n6, Nov/Dec 1982, p24.

Radar, DopplerRadar instruments: Sensors for industrial applica-

tions Nowogrodzki. M., Mawhinney, D. D. andKipp. R. W.. v27n5. Sept/Oct 1982, p23.

Radar, phased arraysComputer -aided manufacture of precision mi-

crowave components for phased -array radarGaAs MESFET power amplifiers. Breese, M. E.and Mason. R. J.. v27n5. Sept/Oct 1982, p7I

RadiometersMedical applications of microwave energy. Pagl-

ione. R. W., v27n5. Sept/Oct 1982, p17.

RCA Astro-ElectronicsBusiness modeling and the engineer: An Astro-

Electronics application. Nigam, A. K., Hong, S.and Spence, S. R., v27n6, Nov/Dec 1982, p65.

RCA Consumer ElectronicsManufacturing programs at Consumer Electronics.

Borman. B. L., v27n2, Mar/Apr 1982. pl.

RCA Electron Tube DivisionMicrowave components research at RCA. Nowo-

grodzki. M.. v27n5, Sept/Oct 1982, p4.

RCA ENGINEEREditorial Representatives, meet your. RCA ENGI-

NEER Staff, v27n1, Jan/Feb 1982, p35.

RCA LaboratoriesMicrowave components research at RCA. Nowo-

grodzki. M.. v27n5, Sept/Oct 1982, p4.

RCA Picture Tube DivisionPackage engineering for the Picture Tube Division.

Haggerty, G. L., v27n2, Mar/Apr 1982, p50.Quality Circles at RCA Scranton. Mecca. S. J.,

v27n2, Mar/Apr 1982. p30.

RCA "SelectaVision" VideoDiscOperationsAutomatic test equipment for VideoDisc and

stylus. Kowalchik. J. J. and Selwa. A. P.. v27n1,Jan 'Feb 1982, p30.

Caddy design and manufacturing of the VideoDisccaddy. Torrington, L. A.. v27n1, Jan/Feb 1982,p23.

Manufacturing the VideoDiscs: An overview.Weisberg, H.. v27n1. Jan/Feb 1982, p6.

Micro -molding is a VideoDisc requirement.McNeely, M. and Bock, M., v27n1, Jan/Feb1982, p19.

Product Assurance of VideoDisc : VideoDisc per-formance in the real world. Huck, R. H., v27n1.Jan/Feb 1982. p27.

RCA Solid State DivisionLinear integrated circuits are going digital. Wittlin-

ger. H. A., v27n1, Jan/Feb 1982. p58.

RCA Engineer 27-6 Nov./Dec. 1962 105

Recording systems, laser

Recording systems, laserData recording on optical discs at 300 Mb/s. Am-

mon, G. J. and Reno, C. W., v27n3, May/June1982, p36.

High -density optical recording using laser diodes.Bartolini, R. A., v27n3, May/June 1982, p20.

Optoelectronic systems, some of the new systems.

Ettenberg, M., v27n3, May/June 1982, p4.Single -mode diode lasers for systems -oriented ap-

plications. Botez, D., v27n3, May/June 1982,p8.

ReliabilitySoftware reliability, discovering how to ensure.

Trachtenberg, M., v27n1, Jan/Feb 1982. p53.

Satellite communicationDirect Broadcasting Satellite: Black hole or bonan-

za? Clark. J. F., v27n4, July/Aug 1982, p82.RCA Satcom system, computer modeling in the.

Guida, A. A., Jonnalagadda, K., Raychaudhuri,D. and Schiff, L., v27n6, Nov/Dec 1982, p84.

Receivers, wideband, for communications satellites.Goldberg. H.. v27n5, Sept/Oct 1982, p37.

Shaped -beam reflector antennas for communica-tions satellites. Profera, C. E., Rosol, G. L.,Soule. H. H. and Kara, J., v27n5, Sept/Oct1982, p43

Solid-state power amplifier replacing TWTs in C -band satellites. Wolkstein, H. J. and LaPrade, J.N., v27n5, Sept/Oct 1982. p7.

Satellite technologyThe second -generation Navy Navigation Satellite,

NOVA -1. expectations achieved. Staniszewski, J.R.. v27n4, July/Aug 1982, p89.

Signal processingAutomatic test equipment for VideoDisc and

stylus. Kowalchik, J. J. and Selwa, A. P.. v27n1,Jan/Feb 1982, p30.

Solid state devicesActive and passive microwave components. Den -

linger. E. J., v27n5, Sept/Oct 1982, p50.Pin diode, high -power, the Engineer's Notebook.

Stabile, P. J.. v27n5, Sept/Oct 1982, p21.

Solid state materialsSpace applications for RCA infrared technology.

Warren, F. B., Tower, J. R., Gandolfo, D. A.,Elabd. H. A. and Villani, T. S.. v27n3, May/June 1982. p76.

Solid state processingAssembly mechanization program at RCA Solid

State (RAMP). Koskulitz, J. and Rosenfield, M.,v27n4, July/Aug 1982, p57.

StatisticsStatistics applications in manufacturing. Gunter, B.

H. and Rayl, M., v27n2, Mar/Apr 1982. p14.

Stereophonic systemsStereo CED VideoDisc system. Freeman, E. J.,

Mehrotra, G. N. and Repka, C. P., v27n4, July/Aug 1982, p12.

Switching circuitsPin diode, high -power, the Engineer's Notebook.

Stabile, P. J.. v27n5, Sept/Oct 1982. p21.

Systems engineeringATE, distributed system architectures for. Fay, J.

E.. v27n4, July/Aug 1982. p53.Fiber-optic communications systems, design, deve-

lopment and installation of. Ainsbury, D. C.,v27n3, May/June 1982, p28.

HF modem simulation. DeMaria, P. A. and Bodzi-och, K. J., v27n6, Nov/Dec 1982. p18.

Tape recordingHawkeye camera. Bendell, S. L., Bazin, L. J. and

Clarke. J. J., v27n4, July/Aug 1982, p18.

Television camerasHawkeye camera. Bendell, S. L.. Bazin, L. J. and

Clarke, J. J.. v27n4, July/Aug 1982, p18.Saticon photoconductor: in color -cam-

era -tube photoconductors. Neuhauser. R. G..v27n3, May/June 1982, p81.

TV camera for target -acquisition and laser -designa-tor systems, high performance. Arlan, L., v27n3,May/June 1982, p87.

Television display systemsHuman visual system, modeling of the. Adelson,

E. H., Pica, A. P. and Carlson, C. R., v27n6,Nov/Dec 1982, p56.

Television, audioStereo CED VideoDisc system. Freeman, E. J.,

Mehrotra. G. N. and Repka, C. P., v27n4. July/Aug 1982, p12.

Television, mobile unitsHawkeye camera. Bendell, S. L., Bazin, L. J. and

Clarke, J. J., v27n4, July/Aug 1982. p18.

TestingComputer -aided design and testing of microwave

circuits. Perlman, B. S., Rhodes, D. L. and

Schepps, J. L., v27n5, Sept/Oct 1982, p76.Software reliability. discovering how to ensure.

Trachtenberg, M., v27n1, Jan/Feb 1982, p53.

TypesettingEditorial input: Two new editorial features. King,

T. E., v27n5, Sept/Oct 1982, p30.

Value engineeringMirror Fusion Test Facility, producing neutral

beam ion sources for the. Lynch, W. S. andReed. R. E., v27n2, Mar/Apr 1982, p62.

VideoDisc, Capacitance electronicdisc systemCaddy design and manufacturing of the VideoDisc

caddy. Torrington. L. A., v27n1, Jan/Feb 1982.p23.

Chemical and Physical Laboratory supports Video -Disc. Hakala, D. F., v27n1, Jan/Feb 1982, p39.

Manufacturing the VideoDiscs: An overview.Weisberg, H., v27n1, Jan/Feb 1982, p6.

Mastering VideoDiscs: The software to hardwareconversion. Kell, F. D., John, G. and Stevens, J.,v27n1. Jan/Feb 1982, p11.

Material compounding of VideoDiscs demands theright chemistry. Whipple, B. and Dunn, V. S.,v27n1, Jan/Feb 1982. p16.

Micro -molding is a VideoDisc requirement.McNeely, M. and Bock, M., v27n1, Jan/Feb1982, p19.

Player Technology Center: Key to innovation.Mishra, D., v27n2, Mar/Apr 1982. p24.

Product Assurance of VideoDisc : VideoDisc per-formance in the real world. Huck, R. H., v27n1,Jan/Feb 1982, p27.

Quality plus equals suc-cess. Schlosser. H. S., v27n1, Jan/Feb 1982, pl.

"SelectaVision" VideoDisc Awards. RCA ENGI-NEER Staff, v27n1, Jan/Feb 1982. p4.

Stereo CED VideoDisc system. Freeman, E. J.,Mehrotra, G. N. and Repka, C. P., v27n4. July/Aug 1982, p12.

VidiconsSaticon photoconductor: The latest in color -cam-

era -tube photoconductors. Neuhauser, R. G.,v27n3. May/June 1982, p81.

VisionHuman visual system, modeling of the. Adelson,

E. H., Pica. A. P. and Carlson. C. R., v27n6,Nov/Dec 1982, p56.

Weapon systemsComplex weapon systems, discrete -event simula-

tions of. Golub, J., v27n6. Nov/Dec 1982, p32.

106 RCA Engineer 27-6 Nov./Dec. 1982

Author IndexAdelson, E. H., Pica, A. P. and Carlson, C.

Human visual system, modeling of the. v27n6.Nov/Dec 1982, p56.

Ainsbury, D. C., Fiber-optic communications sys-tems. design, development and installation of.

v27n3, May/June 1982, p28.Altman, W., Hogan, D. B., Flutter, E. C. and

Rosenberg, A., Space applications of lidar.v27n3, May/June 1982. p44.

Amantea, R. with Snowden, M. P., Automatedmeasurement of transistor I -V characteristics.v27n2, Mar/Apr 1982, p69.

Ammon, G. J. and Reno, C. W., Data recording onoptical discs at 300 Mb/s. v27n3, May/June 1982,p36.

Arlan, L, TV camera for target -acquisition and las-er -designator systems, high performance. v27n3,May/June 1982, p87.

Ashkinazy, A., MIMIC logic simulator. v27n6,Nov/Dec 1982. p10.

Barbin, R. L, Simpson, T. F. and Marks, B. G.,Color -data -display CRT: A product whose timehas come. v27n4, July/Aug 1982, p23.

Bartolin', R. A., High -density optical recording us-ing laser diodes. v27n3, May/June 1982, p20.

Barton, R. R. with Pitts, K. A., Engineering envi-ronment, simulation in an. v27n6, Nov/Dec 1982,p4

Bazin, L J. with Bendell, S. L et al., Hawkeyeamera. v27n4, July/Aug 1982. p18.

Bendell, S. L, Bazln, L J. and Clarke, J. J., Haw-keye camera. v27n4, July/Aug 1982, p18.

Blumenfeld, M. A. and Shambelan, R. C., Equip-ment decisions for VLSI -circuit manufacture.v27n2, Mar/Apr 1982, p35.

Bock, M. with McNeely, M., Micro -molding is aVideoDisc requirement. v27n1, Jan/Feb 1982,p19.

Bodzioch, K. J. with DeMaria, P. A., HF modemsimulation. v27n6. Nov/Dec 1982. p18.

Bolger, G. P. and Holaday, V. D., Computer designand testing of a microwave antenna -feed manifold.v27n5, Sept/Oct 1982, p87.

Borman, B. L, Manufacturing programs at Con-,umer Electronics. v27n2, Mar/Apr 1982, p:.

Botez, D., Single -mode diode lasers for systems -ori-ented applications. v27n3, May/June 1982. p8.

Bradshaw, T. F., Strategic framework to strengthenRCA. v27n4. July/Aug 1982, pl.

Breese, M. E. and Mason, R. J., Computer -aidedmanufacture of precision microwave componentsfor phased -array radar GaAs MESFET power am-plifiers. v27n5. Sept/Oct 1982. p71.

Browne, R. G., Electromechanical motor simula-tion. v27n6, Nov/Dec 1982. p80.

Camisa, R. L with Klatskin, J. B. et al., Automat-ed assembly of lumped -element GaAs FET poweramplifiers. v27n5, Sept/Oct 1982. p64.

Cantella, M. J., Infrared sensor systems, design andperformance prediction of. v27n3, May/June1982, p67.

Carlson, C. R. with Adelson, E. H. et al., Humanvisual system, modeling of the. v27n6, Nov, Dec1982, p56.

Chen, C. C., Computer -aided design of metal -form-ing processes. v27n2, Mar/Apr 1982. p45.

Chin, G. D. with Klein, J. J. et al., Infrared -camerasystem developed to use the Schottky -barrier IR-CCD array. v27n3, May/June 1982, p60.

Clark, J. F., Direct Broadcasting Satellite: Blackhole or bonanza? v27n4, July/Aug 1982, p82.

Clarke, J. J. with Bendell, S. L et al., Hawkeyecamera. v27n4, July/Aug 1982. p18.

D'Arcy, J. A. and Miller, R., Manufacturing, Cus-tom and Quantity Manufacturing: An EngineeringComparison. v27n2, Mar/Apr 1982, p4.

Decker, C. D., Electro-Optics: Insight to the future.v27n3. May/June 1982, pl.

DeMaria, P. A. and Bodzioch, K. J., HF modemsimulation. v27n6, Nov/Dec 1982, p18.

Denlinger, E J., Active and passive microwave..omponents. v27n5, Sept/Oct 1982. p50.

Dunn, V. S with Whipple, B., Material compound.ing of VideoDiscs demands the right chemistry.v27n1, Jan,'Feb 1982. p16.

Elabd, H. A. with Kosonocky, W. F. et al.,Schottky -barrier infrared image sensors. v27n3,May/June 1982, p49.

with Warren, F. B. et al., Space applica-tions for RCA infrared technology. v27n3, May/June 1982, p76.

Enstrom, R. E., Mechanical and thermal design,modeling and simulation in. v27n6, Nov/Dec1982. p71.

Erhardt, H. G. with Kosonocky, W. F. et al.,Schottky -barrier infrared image sensors. v27n3,May/June 1982, p49.

Ettenberg, Optoelectronic systems, some of thenew systems. v27n3, May/June 1982. p4.

with Olsen, G. H., Long -wavelength semi-conductor diode lasers for optical communica-tions. v27n4, July/Aug 1982, p38.

Fay, J. E., ATE, distributed system architecturesfor. v27r.4. July/Aug 1982, p53.

Frantz, V.L., et al, Schottky -barrier infrared image.ensor. 27n3. May/June 1982, p49.

Freeman, E. J., Mehrotra, G. N. and Repka, C. P.,Stereo CED VideoDisc system. v27n4, July/Aug

1982, p12.

Gandolfo, D. A. with Warren, F. B. et al., Spaceapplications for RCA infrared technology. v27n3.May/June 1982. p76.

Goldberg, H., Receivers, wideband, for communi-cations satellites. v27n5, Sept/Oct 1982, p37.

Golub, J., Complex weapon systems, discrete -eventsimulations of. v27n6, Nov/Dec 1982, p32.

Groppe, J. V. with Kosonocky, W. F. et aL,Schottky -barrier infrared image sensors. v27n3,May/June 1982. p49.

Guido, A. A., Jonnalagadda, K., Raychaudhuri, D.and Schiff, L, RCA Satcom system, computermodeling in the. v27n6, Nov/Dec 1982, p84.

Gunter, B. H. and Rayl, kl., Statistics applicationsin manufacturing. v27n2, Mar/Apr 1982. p14.

Haggerty, G. L, Package engineering for the Pic-ture Tube Division. v27n2, Mar/Apr 1982, p50.

Haggis, D. with Klatskin, J. B. et al., Automatedassembly of lumped -element GaAs FET poweramplifiers. v27n5. Sept/Oct 1982, p64.

Hakala, D. F., Chemical and Physical Laboratorysupports VideoDisc. v27n1, Jan/Feb 1982, p39.

Hayman, J., An intelligent missile seeker, designand simulation of. v27n6, Nov/Dec 1982, p43.

Hogan, D. B. with Aftman, W. et al., Space ap-plications of lidar. v27n3, May/June 1982, p44.

Holaday, V. D. with Bolger, G. P., Computer de-

sign and testing of a microwave antenna -feedmanifold. v27n5, Sept/Oct 1982, p87.

Hong, S. with Nigam, A. K. et al., Business model-ing and the engineer: An Astro-Electronics ap-plication. v27n6, Nov/Dec 1982, p65.

Huang, H. C., Upadhyayula, L C. and Kumar, M.,GaAs microwave monolithic integrated circuits,

a review of. v27n5, Sept/Oct 1982, p58.Huck, R. H., Product Assurance of VideoDisc :

VideoDisc performance in the real world. v27n1.Jan 'Feb 1982, p27.

Huffer, E. C. with Altman, W. et al., Space applica-tions of lidar. v27n3, May/June 1982, p44.

John, G. with Kell, F. D. et aL, Mastering Video -Discs: The software to hardware conversion.v27n1. Jan/Feb 1982, p11.

Jonnalagadda, K. with Guide, A. A. et al., RCASatcom system, computer modeling in the. v27n6,Nov/Dec 1982, p84.

Joyce, B. T. with Klatskin, J. B. et al., Automatedassembly of lumped -element GaAs FET poweramplifiers. v27n5. Sept/Oct 1982, p64.

Kara, J. with Profera, C. E. et al., Shaped -beamreflector antennas for communications satellites.v27n5. Sept/Oct 1982, p43.

Kell, F. D., John, G. and Stevens, J., MasteringVideoDiscs: The software to hardware conversion.v27n1. Jan/Feb 1982, pl I.

King. T. E., Editorial input: Two new editorial fea-tures. v27n5, Sept/Oct 1982, p30.

Kipp, R. W. with Nowogrodzki, M. et al., Radarinstruments: Sensors for industrial applications.v27n5. Sept/Oct 1982, p23.

Klatskin, J. B., Haggis, D., Centime, R. L andJoyce, B. T., Automated assembly of lumped -element GaAs FET power amplifiers. v27n5,Sept/Oct 1982. p64.

Klein, J. J., Roberts, N. L and Chin, G. D., In-frated-camera system developed to use theSchottky -barrier IR-CCD array. v27n3, May/June 1982. p60.

Koskulitz, J. and Rosenfield, M., Assembly me-chanization program at RCA Solid State(RAMP). v27n4, July/Aug 1982, p57.

Kosonocky, W.F., Elabd, H.A., Erhardt, H.G.,Sballcross, F.V., Meray. G.M.. Miller, R., Vil-lani, T.S., Frantz, V.L., and Tams, F.J., Schot-tky -barrier infrared image sensors. v27n3. May/ -line 1982. p49.

Kowelchik, J. J. and Selwa, A. P., Automatictest equipment for VideoDisc and stylus. v27n1,Jan/Feb 1982, p30.

Kumar, M. with Huang, H. C. et al., GaAs mi-crowave monolithic integrated circuits, a reviewof. v27n5, Sept/Oct 1982, p58.

LaPrade, J. N. with Wolkstein, H. J., Solid-statepower amplifier replacing TWTs in C -band sa-tellites. v27n5. Sept/Oct 1982, p7.

Liston, J. and Sparks, G. M., Mechanicallyscanned radar systems, dynamic analysis andsimulation of. v27n6, Nov/Dec 1982, p24.

Lynch, W. S. and Reed, R. E., Mirror FusionTest Facility, producing neutral beam ionsources for the. v27n2, Mar/Apr 1982, p62.

Marks, B. G. with Barbin, R. L et al., Color -data -display CRT: A product whose time hascame. v27n4. July/Aug 1982. p23.

Mason, R. J. with Breese, M. E., Computer -aid -

RCA Engineer 27-6 Nov./Dec. 1982 107

Mawhinney, D.D. with Nowogrodzki. M. et al.

ed manufacture of precision microwave compo-nents for phased -array radar GaAs MESFETpower amplifiers. v27n5, Sept/Oct 1982, p71.

Mawhinney, D. D. with Nowogrodzki, M. et al.,Radar instruments: Sensors for industrial applica-

tions. v27n5. Sept/Oct 1982, p23.McIntyre, R. J. with Webb, P. P., Silicon ava-

lanche photodiodes, recent developments in.v27n3. May/June 1982. p96.

McNeely, M. and Bock, M., Micro -molding is aVideoDisc requirement. v27n1, Jan/Feb 1982,

p19.Mecca, S. J., Quality Circles at RCA Scranton.

v27n2, Mar/Apr 1982, p30.Mehrotra, G. N. with Freeman, E. J. et al.,

Stereo CED VideoDisc system. v27n4, July/Aug 1982, p12.

Meray, G. M. with Kosonocky, W. F. et al.,Schottky -barrier infrared image sensors. v27n3,May/June 1982. p49.

Mercuri, R. T., Computer music, an illustratedhistory of. v27n4, July/Aug 1982, p65.

Miller, R. with D'Arcy, J. A., Manufacturing,Custom and Quantity Manufacturing: An Engi-neering Comparison. v27n2, Mar/Apr 1982, p4.

- with Kosonocky, W. F. et al., Schottky -barrier infrared image sensors. v27n3, May/June1982. p49.

Mishit, D., Player Technology Center: Key to in-novation. v27n2, Mar/Apr 1982, p24.

Myers, W. W., Computer programming and sys-tems analysis in the rapidly changing VideoDiscenvironment (application of FOCUS). v27n4,July/Aug 1982, p48.

Neuhauser, R. G., Saticon photoconductor: Thelatest in color -camera -tube photoconductors.v27n3. May/June 1982, p81.

Nigam, A. K., Hong, S. and Spence, S. R., Busi-ness modeling and the engineer: An Astro-Elec-tronics application. v27n6, Nov/Dec 1982, p65.

Nowogrodzki, M., Microwave components re-search at RCA. v27n5, Sept/Oct 1982, p4.

- Mawhinney, D. D. and Kipp, R. W., Ra-dar instruments: Sensors for industrial applica-tions. v27n5. Sept/Oct 1982, p23.

Olsen, G. H. and Ettenberg, M., Long -wave-length semiconductor diode lasers for opticalcommunications. v27n4, July/Aug 1982, p38.

Pagllone, R. W., Medical applications of mi-crowave energy. v27n5, Sept/Oct 1982, p17.

Patton, W. T., Low-sidelobe antennas for tacticalphased -array radars. v27n5, Sept/Oct 1982. p31.

Perlman, B. S., Rhodes, D. L and Schepps, J.L, Computer -aided design and testing of mi-crowave circuits. v27n5, Sept/Oct 1982, p76.

Perlow, S. M., ANA radio frequency circuit simu-lation and analysis. v27n6, Nov/Dec 1982, p46.

Philip, G. 0., Parylene-coating process at MSR.v27n2, Mar/Apr 1982, p56.

Pica, A. P. with Adelson, E. H. et al., Humanvisual system, modeling of the. v27n6, Nov/Dec1982, p56.

Pitts, K. A. and Barton, R. R., Engineering envi-ronment, simulation in an. v27n6, Nov/Dec1982, p4.

Profera, C. E., Rosol, G. L, Soule, H. H. andKara, J., Shaped -beam reflector antennas forcommunications satellites. v27n5. Sept/Oct1982. p4.3

Raychaudhuri, D. with Guida, A. A. et al., RCA

Satcom system, computer modeling in the.v27n6, Nov/Dec 1982, p84.

Rayl, M. with Gunter, B. H., Statistics applica-tions in manufacturing. v27n2, Mar/Apr 1982,p14.

RCA Corporate Engineering Education, Con-tinuing education for manufacturing engineers.v27n2. Mar/Apr 1982, p38.

RCA ENGINEER Staff, David Sarnoff Awards fo-Outstanding Technical Achievement (19821v27n4, July/Aug 1982, p4.

Editorial Representatives, meet your.v27n1, Jan/Feb 1982, p35.

Engineering Form and Function: Win-dows on RCA Technology. v27n4, July/Aug1982, p33.

- "SelectaVision" VideoDisc Awards.v27n1. Jan/Feb 1982. p4.

Reed, R. E. with Lynch, W. S., Mirror FusionTest Facility, producing neutral beam ionsources for the. v27n2, Mar/Apr 1982, p62.

Reno, C. W. with Ammon, G. J., Data recordingon optical discs at 300 Mb/s. v27n3, May/June1982, p36.

Repko, C. P. with Freeman, E. J. et al., StereoCED VideoDisc system. v27n4, July/Aug 1982.

p12.Rhodes, D. L with Perlman, B. S. et al., Com-

puter -aided design and testing of microwave circuits. v27n5, Sept/Oct 1982, p76.

Roberts, N. L with Klein, J. J. et al., Infrared -camera system developed to use the Schottky -

barrier IR-CCD array. v27n3, May/June 1982.

P60.Rosenberg, A. with Altman, W. et al., Space ap-

plications of lidar. v27n3, May/June 1982, p44.Rosenfield, M. with Koskulitz, J., Assembly me-

chanization program at RCA Solid State(RAMP). v27n4. July/Aug 1982, p57.

Rosol, G. L with Profera, C. E. et al., Shaped -beam reflector antennas for communications sa-tellites. v27n5, Sept/Oct 1982. p43.

Schepps, J. L with Perlman, B. S. et al., Com-puter -aided design and testing of microwave circuits. v27n5, Sept/Oct 1982, p76.

Schiff, L with Guide, A. A. et al., RCA Satcomsystem, computer modeling in the. v27n6. Nov/Dec 1982, p84.

Schlosser, H. S., Quality plus diversity plusavailability equals success. v27n1, Jan/Feb1982, pl.

Schneider, W., Making telescopes is my hobby.v27n2, Mar/Apr 1982, p75.

Selwa, A. P. with Kowalchik, J. J., Automatictest equipment for VideoDisc and stylus. v27n1,Jan/Feb 1982, p30.

Shalicross, F. V. with Kosonocky, W. F. et al.,Schottky -barrier infrared image sensors. v27n3,May/June 1982, p49.

Shambelan, R. C. with Blumenfeld, M. A.,Equipment decisions for VLSI -circuit manufac-ture. v27n2. Mar/Apr 1982, p35.

Simpson, T. F. with Barbin, R. L et al., Color -data -display CRT: A product whose time hascome. v27n4, July/Aug 1982, p23.

Snowden, M. P. and Amantea, R., Automatedmeasurement of transistor I -V characteristics.v27n2, Mar/Apr 1982, p69.

Solomon, S. M., Microwave communications sys-tem for RCA Globcom. v27n4, July/Aug 1982.p75

Soule, H. H. with Profera, C. E. et al, Shaped -beam reflector antennas for communications sa-tellites. v27n5, Sept/Oct 1982, p43.

Sparks, G. M. with Uston, J., Mechanicallyscanned radar systems, dynamic analysis andsimulation of. v27n6. Nov/Dec 1982. p24.

Spence, S. R. with Nigam, A. K. et al., Businessmodeling and the engineer: An Astro-Electron-ics application. v27n6, Nov/Dec 1982, p65.

Stabile, P. J., Pin diode, high -power, the Engi-neer's Notebook. v27n5, Sept/Oct 1982, p21.

Staniszewski, J. R., The second -generation NavyNavigation Satellite, NOVA -1, expectationsachieved. v27n4, July/Aug 1982, p89.

Sterzer, F., Localized hyperthermia treatment ofcancer. v27n1, Jan/Feb 1982, p43.

- Microwave technology at RCA. v27n5,Sept/Oct 1982, pl.

Stevens, J. with Kell, F. D. et al., MasteringVideoDiscs: The software to hardware conver-sion. v27n1, Jan/Feb 1982, p11.

Suhy, G. W., Large system development, digitalinterface simulation for. v27n6, Nov/Dec 1982,p39.

Tams, F. J., et al, Schottky -barrier infrared imagesensors. v27n3, May June 1982, p50.

Technical Information Systems, Technical ab-stracts database (TAD): RCA database now anon-line system. v27n2, Mar/Apr 1982, p22.

Teger, A. H., Modeling and simulation: Tools forthe 80s. v27n6, Nov/Dec 1982, p1.

Torrington, L A., Caddy design and manufactur-ing of the VideoDisc caddy. v27n1, Jan/Feb1982, p23.

Tower, J. R. with Warren, F. B. et al., Space ap-plications for RCA infrared technology. v27n3,May/June 1982, p76.

Trachtenberg, It, Software reliability, discover-ing how to ensure. v27n1, Jan/Feb 1982, p53.

Upadhyayula, L C. with Huang, H. C. et al.,GaAs microwave monolithic integrated circuits,a review of. v27n5, Sept/Oct 1982, p58.

Villani, T. S. with Kosonocky, W. F. et al.,Schottky -barrier infrared image sensors. v27n3,May/June 1982, p49.

- with Warren, F. B. et al., Space applica-tions for RCA infrared technology. v27n3,May/June 1982, p76.

Warren, F. B., Tower, J. R., Gandolfo, D. A.,Elabd, H. A. and Villani, T. S., Space applica-tions for RCA infrared technology. v27n3,May/June 1982, p76.

Webb, P. P. and McIntyre, R. J., Silicon ava-lanche photodiodes, recent developments in.v27n3, May/June 1982, p96.

Weisberg, H., Manufacturing the VideoDiscs: Anoverview. v27n1, Jan/Feb 1982, p6.

Whipple, B. and Dunn, V. S., Material com-pounding of VideoDiscs demands the rightchemistry. v27n1, Jan/Feb 1982, p16.

Williams, R., Coal, upgrading, for more efficientelectric -power generation. v27n4, July/Aug1982. p45.

Wittlinger, H. A., Linear integrated circuits aregoing digital. v27n1, Jan/Feb 1982. p58.

Wolkstein, H. J. and LaPrade, J. N., Solid-statepower amplifier replacing TWTs in C -band sa-tellites. v27n5, Sept/Oct 1982, p7.

Young, C., Numerical control: The cornerstone ofcomputers in manufacturing. v27n2, Mar/Apr1982, p41.

108 RCA Engineer 27-6 Nov./Dec. 1982

Editorial RepresentativesContact your Editorial Representative at the TACNETnumbers listed here to schedule technical papersand announce your professional activities.

Cablevision Systems TACNET*John Ovnick Van Nuys, California 534-3011

Commercial CommunicationsSystems Division (CCSD)

*Bill SepichClinton Glenn

Camden, New Jersey 222-2156Meadowlands, Pennsylvania 228-6231

Consumer Electronics (CE)

*Eugene JansonFrancis HoltChuck LimbergDon WillisByron Taylor

Indianapolis, IndianaIndianapolis, IndianaIndianapolis, IndianaIndianapolis, IndianaIndianapolis, Indiana

Government Systems Division (GSD)

Advanced Technology Laboratories*Merle PietzEd Master

Astro-Electronics

Carol Klarmann

Automated Systems*Paul SeeleyDale Sherman

Camden, New JerseyCamden, New Jersey

New JerseyPrinceton, New Jersey

Burlington, MassachusettsBurlington, Massachusetts

Government Communications Systems*Dan Tannenbaum Camden, New Jersey

Harry Ketcham Camden, New JerseyGSD Staff*Ed Moore Cheery Hill, New JerseyMissile and Surface Radar*Don Higgs Moorestown, New JerseyJack Friedman Moorestown, New JerseyGraham Boose Moorestown, New Jersey

National Broadcasting Company (NBC)

*Bob Mausler New York, New York

422-5208422-5217422-5117422-5883 RCA Records426-3247

RCA Communications

American Communications

Murray RosenthalCarolyn Powell

Global Communications

*Dorothy Unger

RCA Limited (Canada)

222-2161222-2731

229-2544229-2919

326-3095326-2985

222-3081222-3913

222-5833

224-2836224-2112224-3680

324-4385

Patent Operations

George Haas Princeton, New Jersey 226-2491

Picture Tube Division (PTD)

*Ed MadenfordNick MeenaJack NubaniJ.R. Reece

Lancaster, PennsylvaniaCircleville, Ohio

Scranton, PennsylvaniaMarion, Indiana

227-3657432-1228329-1499427-5566

TACNET

Princeton, New Jersey 258-4192Princeton, New Jersey 258-4194

New York, New York 323-7348

Bob McIntyre Ste Anne de Bellevue 514-457-9000

*Greg Bogant; Indianapolis, Indiana 424-6141

RCA Service Company

*Joe SteogerRay MacWilliamsDick Dombrosky

Cherry Hill, New JerseyCherry Hill, New JerseyCherry Hill, New Jersey

Research and Engineering

Corperate Engineering Hans Jenny

LaboratoriesEva Duke,.

222-5547222-5986222-4414

Cherry Hill, New Jersey 222-4251

Princeton, New Jersey 226-2882

"SelectaVision" VideoDisc Operations*Nelson Crooks Indianapolis, Indiana 426-3164

Solid State Division (SSD)

*John SchoenPower Devices

Harold Ronan

Integrated CircuitsDick MoreySy SilversteinJohn Young

Somerville, New Jersey 325-6467

Mountaintop, Pennsylvania 327-1633or 327-1827

Palm Beach Gardens, Florida 722-1262Somerville, New Jersey 325-6168

Findlay, Ohio 425-1307Electro-Optics and Power DevicesJohn Grosh Lancaster, Pennsylvania 227-2077

Solid State Technology CenterJudy Yeast Somerville, New Jersey 325-6248

Technical Publications Administrators, responsible for review and approvalof papers and presentations. are indicated here with asterisks before their names.

RCA EngineerA technical journal published byCorporate Research and Engineering"by and for the RCA engineer"

Bldg 204-2. Cherry Hill. NJ 08358Forwarding and return postage guaranteed

rEM MUSSELMANNPU BOX 9STRASBURGPA 17579

Printed in the U.S.A.Form No. RE -27-6

LCTT

BULK RATEUS Postage

PAIDPhila., Pa.

Permit No. 2906