radar (raidar). n. the presence and location of an …...dem. tot ,eterrntnnv.$15,,et by n. return 1,0711u 1144.1 MU, frelthe, J(eurtifyr) Our cover combines radar's past and present

IOTA IC

)AtE 14 ItCHRIFT- WON -

won 74i

sa. DONEL DCA rOra eilliros ailmemio *Raw daildirier

Amp 00estar w arcsamirteptromosi irrr WOW VII Ms

radar (raidar). n. Electronics. a device for determiningthe presence and location of an object by measuring the

m14, , L_

time for the echo of a radio wavelo return from it andthe direction from which ittretu [ra(dio) d (etecting)

- - _

a (nd) r(a nging)]

dem. tot ,eterrntnnv.$15,,et by

n. return 1,0711u 1144.1

MU, frelthe, J(eurtifyr)

Our cover combines radar's past and present. The large circle is a plan -position indicator showing a number of targets. Starting at the top center ofthe cover, we have: the trace of the first radar echo bounced off the moon; aface of the AN/SPY-1 phased -array radar system; the AN/FPQ-6 instrumen-tation radar; Dr. Irving Wolff performing radar experiments at RCA's Camdenplant in the 1930s; Hulsmeyer's German and British radar patents from theearly 1900s (surrounding modern electronic hybrid microcircuits);Guglielmo Marconi, inventor of the "wireless," shown in two views; a three-dimensional plot of a radar antenna pattern; Heinrich Hertz, who showed thatelectromagnetic energy could be reflected; an integrated circuit; and NicolaTesla, who first proposed the practical use of radio waves for detectingobjects.

Cliff Winner, from Missile and Surface Radar, Moorestown, N.J., painted thewatercolor for our cover.

[ram EngineerA technical journal published byRCA Research and EngineeringBldg. 204-2Cherry Hill, N.J. 08101Tel. PY-4254 (609-779-4254)Indexed annually in the Apr/May issue.

John PhillipsBill LaufferJoan ToothillFrank StroblPat Gibson

Joyce Davis

Harry Anderson

Jay Brandinger

John Christopher

Bill Hartzell

Jim Hepburn

Hans Jenny

Arch Luther

Howie RosenthalCarl Turner

Joe Volpe

Bill Underwood

Bill Webster

Ed Burke

Walt Dennen

Charlie Foster

RCA Engineer Staff

EditorAssociate EditorArt EditorContributing EditorCompositionEditorial Secretary

Editorial Advisory Board

Div. VP, Manufacturing,Consumer ElectronicsDiv. VP, Engineering,Consumer ElectronicsVP, Tech. Operations,RCA AmericomDiv. VP, EngineeringPicture Tube DivisionVP and Technical Director,GlobcomManager, TechnicalInformation ProgramsChief Engineer,Broadcast SystemsStaff VP, EngineeringDiv. VP, Integrated CircuitsSolid State Division

Chief Engineer,Missile and Surface Radar

Director, EngineeringProfessional Programs

VP, Laboratories

Consulting Editors

Ldr., Presentation Services,Missile and Surface RadarMgr., News and Information,Solid State DivisionMgr., Scientific Publications,Laboratories

To disseminate to RCA engineers technical information ofprofessional value To publish in an appropriate manner important technicaldevelopments at RCA. and the role of the engineer To serve as a medium of interchange of technical informationbetween various groups at RCA To create a community of engineering interest within the companyby stressing the interrelated nature of all technical contributions To help publicize engineering achievements in a manner that willpromote the interests and reputation of RCA in the engineering fieldTo provide a convenient means by which the RCA engineer mayreview his professional work before associates and engineeringmanagement To announce outstanding and unusual achievements of RCAengineers in a manner most likely to enhance their prestige andprofessional status.

The radar business 25 years later

Twenty-five years ago, RCA dedicated its new plant in Moorestown on what had formerly been a highlyproductive asparagus field. This expansion of facilities represented RCA management's long-term investmentin the vital field of radar engineering. In his dedicating speech, W. Walter Watts then Vice President, RCATechnical Products, stated:

"Radar is but one fast-moving frontier in the amazing science of electronics, which daily brings us newwonders in such fields as television, 'electronic brain' computers, communications, industrial productsand controls, and sound recording and reproduction. As in these other fields, we have only scratched thesurface of radar's potential services."

That perception has proved to be well-founded. Initially, straightforward applications of traditional radartechnology yielded a solid business base in neatly packaged user areas (e.g., the range instrumentation radar"product line"). With the passage of time, technological advances blurred the earlier distinctions betweenradars, computers, and communications. More importantly, the emphasis shifted from separate electronic"black boxes" to integrated electronic systems. In many cases, radar became the heart of such systems.

At the same time that our customers were demanding ever-increasing levels of performance from radarsystems, they have had less money to spend. Thus, "Design to Price" has becomea necessity rather than just aslogan.

Today we stand astride a radar technology with capabilities that were beyond corception when Moorestownopened its doors in 1953. Now we can cover the skies from horizon to horizon, detecting, tracking, identifying,and countering everything that moves. Non-military applications also abound, for we can measure ocean windcurrents, predict crop yields, and measure soil conditions and potential drought areas all over the world-allwith radar. Truly there is reason for pride in accomplishment among radar developers.

And uneasiness, as well. Radar users are becoming increasingly aware of how diverse technologies couldconverge to extend and expand radar's capabilities. We can be certain that some user, somewhere, is alreadymaking far-reaching assumptions on future radar applications. Almost as certainly, some industrial oracademic laboratory is nurturing a germ of an idea that will ultimately make these assumptions a reality.

So today the business outlook for radar is as robust as it was 25 years ago. The pattern that has characterizedradar development in recent years seems certain to continue. Our challenge is to assure that RCA remains aleader by stimulating and supporting the groundwork of today that will anticipateand satisfy the requirementsof the future. Clearly it is a challenge of major proportions.

But it's more fun than growing asparagus.

Max LehrerDivision Vice President and General ManagerMissile and Surface RadarMoorestown, N.J.

Radar it's changed

to this

How do RCA's information sources rate?

and this

from this

Sarnoff Award winners

46

coming up

Our next issue (Apr/May) covers the software explosion and how it is affecting thetraditional hardware -oriented engineer. Our anniversary issue (Jun/Jul) traditionallycovers the years most significant technological events at RCA-digital television, SOStechnology, optical videodisc memory, and more.

D.R. Higgs

T.G. Greene

I. Wolff

B. Fell

A.S. Robinson

W.W. Weinstock

D.L. PruittR.L. CamisaIJ.Goel

H.J. Wolksteinl R.L. Ernst

D.E. HutchisonI.J.C. PhillipsF.J. Strobl

MEM EngineerVol. 231No. 5 Febl Ma 1978

radar background4 Radar technology today --perplexity in Wonderland

6 The early days of radar

11 Radio Vision-the early days of radar at RCA14 An introduction to radar concepts25 Radar weapon system sensors-a system/technology perspective

radar technology articles32 Radar processing architectures38 Solid state for super -power radars

42 Pulsed GaAs FET microwave power amplifiersfor phased -array radars

Aires1G.A. Lucchi

R.M. Scudder

H.C. Johnsonl R.W. PaglioneJ.P. Hoffman

B.D. BuchIS.L. ClapperR.J. Smith

M.G. HeroldIM.C. Timken 76

J. ListonIG.M. Sparks

C.E. Profera

W.S. Pike

Copyright ©1978 RCA CorporationAll rights reserved

82

technical excellenceEngineering Information Survey results: Part 3

more radar technology articlesColor displays for airborne weather radar

Advanced antenna design reduces electronic countermeasures threatA short-range radar for measuring blast -furnaceburden height

Microcomputers for radar systems

Programmable processors and radar signal processing-an applications overview

Digital computer simulation of radar systems89 Enhancing antenna performance through

multimode feeds and distribution networks

on the job/off the job96 Electronic speed control for model railroad realism

departments100 The 1978 David Sarnoff Awards for Outstanding Technical Achievement102 Could you use a reprint?

104 Dates and Deadlines

105 Pen and Podium

106 Patents107 News and Highlights

about this issue...

Radar technologytodayperplexity inWonderlandPart of the wonder and fascination of radar is that it can doso much. In something like 40 years, defense es-tablishments all over the world have come to depend on itmore and more-originally as a sometimes -reliable sensor,and more recently as a key controlling element in majordefensive systems. Today, space-age radars routinelyidentify, track, and analyze virtually everything in earthorbit.

But the non-military applications of radar are even moreexciting. Radar is everywhere-as a bulwark of air trafficcontrol, in marine navigation (including more and morepleasure boats), in assessing soil conditions and cropharvests on an international scale, measuring and plottingterrain features and sea states, and even (as detailed herein)as a tool in the basic steel -production process. Perhaps thebest evidence of acceptance is the introduction of the word"radar" into the vocabulary of the consumer productsmarket.

This kind of application explosion didn't just happen. Itrepresents, instead, a studied and imaginative applicationof modern technology to exploit the intrinsic potential ofradar. Practically everything has been tossed into the pot-a broader slice of the electromagnetic spectrum, thewonders of solid-state devices and circuit miniaturization,and the entire range of computer elements. In essence, thecontinuing growth of radar derives directly fromtechnological diversity.

This kind of diversity, of course, begets complexity. Andanother name for complexity is perplexity-for the laymanwho is struggling to understand, and even to a degree forthe trained engineer who knows the basics but is uncertainabout how all the peripherals work. The pace of radardevelopment in recent years has opened a very real chasmbetween the conceptual simplicity and elegance of radartechnology and the ability of the average engineer to feelcomfortable with it.

A stated goal of the RCA Engineer is to help bridge this gap.And this issue represents an honest, if modest, attempt tomake some of the technology of modern radar under-standable both to those who are interested but not directlyinvolved and to those who are involved, but on the fringes.

Accordingly, the papers in this issue are not about RCAproducts for the most part. Neither do they describeeverything that is new and good in the field, either inapplications or in the technology.

What this issue does offer is a thorough -going primer onradar operation (Barry Fell, p. 14), some interestinghistorical background and personal reminiscences (TomGreene, p. 6, and Iry Wolff, p. 11), a summary of where westand today technologically (Art Robinson, p. 25), and aworking -level treatise on the explosion in signal and dataprocessing (Walt Weinstock, p. 32). The remaining papersform a montage of some of the work going on in RCAtoday-from feeds and antennas, to solid-state switchingand amplification, to the dark arts of balancing analog anddigital processing techniques, to a few special examples ofradar applications.

In summary, although this issue is certainly not com-prehensive, it gives a feeling for what radar is all abouttoday. If you see something that interests you and wouldlike to learn more about it, please call one of the authors.They are all articulate and enthusiastic. If you're looking forsomething and don't see it here, several of the papersinclude fairly extensive references on specific subjects. Andif you're still curious, the following brief bibliography lists anumber of recent papers on radar by RCA authors.

Don HiggsTechnical Publications AdministratorMissile and Surface RadarMoorestown, NJ

Ed. Note: Don Higgs presented the idea for the radar issue to theRCA Engineer Planning Board in June 1976. He then guided itthrough several levels of planning and pushed it to its final steps-working personally with most of the authors through rough drafts,rewrites, and approvals. The quality of the result speaks for itself.

Ed Burke, our Consulting Editor at MSR, provided the graphictreatment for Don's Editorial Input.

iftfilliii

Ii1011111,011mligoo, 'PH

4

Practical Phased Array Radar Systems (Apr 1975).

Ho, P.T.; Schwarzmann, A.; and Swartz, G.A.; "Low loss PIN diodefor high power MIC phase shifter," ISCC Digest (Feb 1977).

Scudder, R.M.; "AN/SPY-1 phased array antenna," MicrowaveJournal (May 1974).

Transmitters

Gross, S.D. and Liston, J.; "Wideband solid state transceivermodule evaluation for radar applications," IEEE InternationalRadar Conference Record (Apr 1975).

Mikenas, V.A.; "Advances in high -power S -band TRAPATT diodeamplifier design," Microwave Journal (Feb 1974).

Smith, W.I.; "CFA tube enables new generation coherent radar,"Microwave Journal (Aug 1973).

Signal/Data Processing

Gaffney, BP.; Martinson, L.W.; and Mayer, G.J.; "Model andsimulation of charge -coupled devices for signal processinganalysis," (Toronto, Sep. 1977).

Kalata, P.R.; "On implementation of the Kalman filter for radartarget tracking," Proceedings 7th Annual Pittsburgh Conferenceon Modeling and Simulation (Apr 1976).

Lunsford, J.A., and Martinson, L.W ; "A CMOS/SOS pipeline FFTprocessor-construction, performance, and applications,"NAECON '77 Record (May 1977); EASCON '77 Proceedings (Sep1977).

Mayer, G.J., and Upton, L.0.; "Charge -coupled devices and radarsignal processing," RCA Engineer (Jun/Jul 1975).

Perry, R.P., and Martinson, L.W.; "Radar matched filtering," RadarTechnology (Artech, Oct 1977).

Shapiro, L.; "A short course in digital electronics," RCA Engineer(Aug -Sep 1976, Oct -Nov 1976, Dec 1976 -Jan 1977, Feb -Mar 1977).

Smith, R.J.; "A bit -slice module set for microcomputing," COMP-CON '77 Proceedings (Sep 1977).

BibliographyGeneral

Bornholdt, J.W. and Hammond, V.W.; "Range instrumentationradar systems," Military Electronics Defence Expo '77 Proceedings(Sep 1977).

Nessmith, J.T.; "Range instrumentation radars," ELECTRO '76Proceedings (May 1976).

Volpe, J.C.; "RCA and radar-evolving technology in the realworld," Signal (Oct 1976).

Computer Control

Baugh, R.A.; Computer Control of Modern Radars (Jul 1973).

Mehling, T.H.; "Computer program architectural design forweapon system radar control," RCA Engineer (Jun -Jul 1974).

Weinberg, L.; "Scheduling multifunction radar systems," EASCON'77 Proceedings (Sep 1977).

Weinstock, W.W.; "Computer control of a multifunction radar,"Microwave Journal, Intensive Lecture Series (Apr 1975).

Antennas/Feeds

Eogner, B.F.; "Conformal arrays come of age," Microwave Journal(Nov 1973).

DiFelice, R. and Drenik, J.; "Mechanical design and integration ofphased arrays," Mechanical Engineering in Radar SymposiumFroceedings (Nov 1977).

Goodrich, H.C. and Landry, N.R.; "Practical aspects of phaseshifter and driver design for a tactical multi -function phased arrayradar system," IEEE G-MTT Transactions (Jun 1974).

Patton, W.T.; "Array feeds and selection of phase control,"Microwave Journal, Intensive Lecture Series on Practical PhasedArray Radar Systems (Apr 1975).

Patton, W.T.; "Compact constrained feed phased array forAN/SPY-I," Microwave Journal, Intensive Lecture Series on

,,,itatta((

5

The early days of radar

T.G. Greene Although the first radar patents were filed in the early 1900sand working systems existed in the 1920s, World War II wasthe real driving force behind radar development.

Ed. Note: This short history does not attempt to be all-inclusive; it is part of a more complete radar history,consisting of about fifty wall panels being prepared byTom Greene for display in the corridors of RCA'sMoorestown plant. Space constraints have excludedcredit to many radar workers within and outside of RCA;we have also stopped our history just after the end ofWorld War II, when large-scale radar -based defensesystems entered the scene.

Tom Greene's role here has been that of a collector andeditor, rather than as an author. He is deeply indebted tothe good memories and collecting habits of the manyGSD engineers who helped on this project, as well asthose engineers who long ago had the foresight to writethorough final engineering reports on their projects.

1886 Heinrich Hertz

Professor Heinrich Hertz in Germanydemonstrated experimentally in 1885 that66 -centimeter radio waves could be formedinto beams and that solid objects wouldreflect them. When the identity betweenlight and radio waves was established, it

became clear that a radio wave reflectedback on itself would create a wave -

interference pattern, and that this patternwould in itself be evidence of the reflectingobject.

This wave -interference detection method,the forerunner of the pulse method, wasreported by various groups of workers inwidely different applications, in the early1920s, both in the United States andabroad.

1900 Nicola Tesla 1903 First practical application

When Hertz demonstrated thatelectromagnetic waves could be reflectedby solid objects, he was far ahead of histime in the science of radio -location. It

remained for Nicola Tesla to recognize andto point out the practical application of theradio "echo." so vital in the Second WorldWar. Describing his 1889 method andtransmission of wireless energy in CenturyMagazine (June, 1900) he wrote:

"Exactly as with sound, so an electricalwave is reflected, and the same evidencewhich is afforded by an echo is offered byan electrical phenomenon known as a "stationary" wave -that is, a wave with fixed nodal and ventral regions. Insteadof sending sound -vibrations toward a distant wall, I havesent electrical vibrations toward the remote boundaries ofthe earth, and instead of the wall the earth has replied. Inplace of an echo I have obtained a stationary electrical wave- a wave reflected from afcr.

Stationary waves . . . mean something more thantelegraphy without wires to any distance . .. For instance, bytheir use we may produce at will, from a sending station, anelectrical effect in any particular region of the globe; we maydetermine the relative position or course of a moving object,such as a vessel at sea, the distance traversed by the same,or its speed."

I'LICHES PAT EN TA M1

PtITENTSCHRIFTAt 165546MASSE 74 d.

(flk 1101.SMIEYEP MSS1 LLX,P F.

VIVIVINA, on .tterral netaillftas Cepotellatia eiektrtseber WM*,

*Yea Ikelmbler salloa

Pau.. ks ow*. Mr. nom NM. holos r.

1,,

In 1903 Christian Hulsmeyer experimented with radiowaves reflected from ships. He obtained patents in severalcountries the following year for his obstacle detector andship navigational device. His scheme was demonstrated tothe German Navy but failed to develop interest because themaximum range of detection was only about one mile usingthe technology available at the time.

Halsmeyer recognized the problems of stabilizing a highlydirectional beam on a rolling, pitching naval vessel andprovided some correction for this in his British patententitled "Hertzian-wave Projecting and Receiving ApparatusAdapted to Indicate or Give Warning of the Presence of aMetallic Body, such as a Ship or a Train, in the Line ofProjection of such Waves."

6

1922 Shit detection by interference

At he Naval Aircraft Radio Laboratory at Anacostia, D. C.,Dr. A. Hoyt Taylor and Leo C. Young were conducting radiofield strength measurements when they observed that a shippassing through a high -frequency field affected theperformance of the receiver. In September 1922, theysuggested the use of this interference effect for thedetection of ships.

1924 Watson -Watt and the RAF

Beginning studies on mechanical direction finders in 1919,Robert Watson -Watt developed a cathode-ray directionfinder in 1928 capable of locating thunderstorms out to 400miles and in azimuth to 1°. With his wife Lady MargaretRobertson Watt as his principal laboratory assistant, hebegan major research on airplane radio -location in 1935. In1940-41 the Germans, continually encounteringconcentrated RAF fighter opposition to their bombing raids,assumed the British had large numbers of fighter planes. Atthat time woefully weak in fighter craft, Britain was able tovector those they had to the German attackers by usingWatson -Watt's equipment.

When Robert Watson -Watt was knighted on King GeorgeVI's birthday in 1942, he was identified merely as "a pioneerin radio location." When war security restrictions were laterlifted, he finally received credit for his principal role indeveloping Britain's radar, credited equally with the RAF inwinning the Battle of Britain.

1925 Measuring the ionosphere

The first reported use of pulsed radio energy to measuredistance was that of Gregory Breit and Merle Tuve of theCarnegie Institute in Washington, D.C. Reporting on theirbasic scientific investigation in the Physical Review in 1926,they described successful efforts to measure the height ofthe conducting layers in the ionosphere, using interruptedtrains of waves (ICW) and an oscillograph to record theechos. Several organizations and amateurs assisted in theirexperiment. Arrangements were made to transmit from

NRL's station NKF, Bellevue, Anacostia, D.C..Westinghouse's station KDKA in Pittsburgh, RCA's stationWSC in Tuckerton, NJ, and the Bureau of Standards' WWVin Washington, D.C. Their superheterodyne receiver, locatedat NRL, used type 201A tubes produced by RCA. Thesummary of the detailed 21 -page paper cited abovereported that the hypothesis of an ionized upper layer of theatmosphere was correct and that its height varies from 50 to130 miles.

1930 Young and Hyland

Leo C. Young and L. A. Hyland, engineers at the Naval Research Laboratory(successor to the Naval Aircraft Radio Laboratory), were experimenting withshort-wave direction finding. Hyland noted trouble with the performance of theequipment in the form of occasional violent signal fluctuations. He was ready toreturn his "balky" receiver to the laboratory for overhaul when he observed thatthe signal fluctuations occurred only when an airplane flew overhead. Hylandwrote a memo describing this method of detecting aircraft to Dr. Taylor(mentioned earlier), who reported it in a letter to the Chief, Bureau ofEngineering, on November 5, 1930. A development program startedimmediately.

Taylor. Young and Hyland received a patent on a "System for Detecting Objectsby Radio" for their NRL work on cw wave -interference radar.

x0.000.(000/

7

1932 RCA enters the field

Radar research and development were carried on exclusively by the military servicesuntil RCA entered the field in 1932, when a group of engineers under the direction ofDr. Irving Wolff began work in the microwave field. Although RCA began radar researchand development with marine and aircraft navigation and collision preventionapplications in mind, the military applications of radar soon became evident and workwith both the Army and Navy followed. (His accompanying article describes this earlywork more completely.)

By 1937 RCA had developed pulse ranging equipment which could determine therange of a target with considerable accuracy over short distances and obtain reflectionsfrom targets at greater distances. A radio detection and ranging equipment installed onthe roof of one of the RCA buildings in Camden, New Jersey, was the first microwavepulse radar system in the United States, and probably the world. The antenna wasdesigned with directional characteristics and mounted so that it could be rotated. Acathode-ray tube was employed as an indicator in the receiver output circuit. Onecoordinate on the indicator tube screen showed distance and a second coordinateshowed angle. The skyline of Philadelphia was plotted, and vessels plying the riverabout two miles distant were located.

1937 Col. Blair demonstrates radar

In 1937, the first equipment developed by the Signal Corps was demonstrated atFt. Monmouth to the Army Chief of Staff. "General Malin Craig was sold whenhe saw we were able to keep a plane flying overhead in the beam of a searchlightdirected by the radar position finder."

Colonel William Blair did not receive a patent covering the invention until 1957,after special legislation was passed permitting the filing after the legal time limithad expired. Col. Blair could not file until 1945 due to wartime secrecy. His pulseecho system used a single transmitter and receiver to determine distance anddirection. The system credited to Sir Robert Watson -Watt of England requiredtwo receivers.

1937 Shipborne radar

In 1937 RCA began work on thedevelopment of the CXZ, a shipborneequipment for detection and ranging,operating at 475 MHz. A single cabinethoused the transmitter, modulator, andpulse generator, and also served as apedestal for the antenna array. This wasthe first shipborne radar designed by acommercial firm to be installed on aNavy ship, the USS Texas.

At the same time a radar operating at200 MHz was developed at the NavalResearch Laboratories and placed onthe USS New York. Comparative testsof the two equipments showed that theRCA unit had superior definition, whilethe NRL version gave longer range. Amodel incorporating the best points ofeach was designed and RCA received aproduction contract for six unitsdesignated Model CXAM.

8

1939 Commercial radar for the Navy

RCA's production contract for six CXAM radars wasfollowed by a contract for fourteen additional equipmentsknown as Model CXAM-1.

The Model CXAM was the first radar produced for the Navyby a commercial firm. It was an air search instrumentproviding range and bearing information, and was designedfor installation on aircraft carriers, battleships, and cruisers.The first of these equipments was installed on the flagshipUSS Augusta in June, 1941. At the time of our entry intoWorld War II, the twenty sets installed on the mostimportant ships of the fleet were the only radars in use bythe Navy.

An officer serving on the USS California reported usingCXAM 1 equipment successfully for navigational purposeson a fogbound trip from Seattle to San Francisco.

1941 High -production shipborne radar

a..vol

The SA (Shipborne, Surface and Air Search Radar) wasdeveloped to provide early warning and keep track of surfacevessels and aircraft targets. It also proved useful as anavigation aid. The Model SA was intended for installationon destroyers and destroyer escorts. Development started inApril 1941 and the first model was delivered in September.Production started immediately after the Pearl Harborattack; 1565 sets of the Model SA series were produced byRCA at Camden. This was the largest quantity of this type ofequipment produced by any one manufacturer during thewar.

RCA engineers cooperated with the U.S. Navy'sBureau of Ordnance and the Office of ScientificResearch and Development in the development ofthe proximity fuze for use in rotating projectilessuch as field artillery or antiaircraft shells. Thisfuze was designed to burst in the vicinity of thetarget, within the fragmentation area of the shells,making it as effective as a direct hit.

Known officially as the VT fuze, it was unknown toall but a few military officials, scientists, andengineers until the war was over. To the factoryworkers, it was a mysterious project known asMadam X. Five and one-half million fuzes (morethan half those supplied to the armed forcesbetween October, 1942 and war's end) wereassembled at RCA's Camden, New Jersey andBloomington, Indiana plants.

The VT fuze was enclosed in the nose of theprojectile. A conical metal cap on the tip acted as

111FIrrillr,1111111.111.11.9

ANTENNA

ASSIMITAT

UHF LINE I IFF

rl TRANSMISSIONLINE

Suggw10 11 Mrs

TRANSMITTER

IC (TIM 11

ASSIMILY

56 SYNCITRO STROMAMPLIFIER CAPACITORS

MODULATOR

UNIT

VOLTAGET -REGULATING

TRANSFORMER

ORANGE

CALIBRATOR

DISTRIBUTION

TRANSFORM!!TRANSFORM!!

ALLAALL LL 74 ALL,

sgw ISAMMT

IN

CONSOLE

'111111111

an antenna. It continuously radiated high -frequency energy in a beam roughly matchingthe fragmentation pattern of the projectile. Whena portion of the radiated field encountered atarget, the antenna loading varied, which in turnchanged the plate current of the oscillator. Thischange was detected and amplified enough tostart a thyratron tube conducting and so actuatedthe detonator in the projectile. A special wetbattery, in which the electrolyte was contained ina glass ampule, supplied the electrical energy. Theshock of firing broke the ampule and the spinningof the projectile distributed the electrolytethrough the battery cells. The fuze required thedesigning of miniature tubes and associated partssturdy enough to withstand not only the terrificimpact when the gun was discharged, but also thecentrifugal force of the shell's high-speedspinning.

11111111111111111

The Model SR -2 shipborne surface and air surveillance radarwas designed to incorporate the experience gained on allprevious search radars to provide long-range warning forlarge ships.

Coordinated design began in October 1943 and the firstproduction model was shipped to NRL for test in April 1945.Two sets were delivered and installed aboard the USSMidway and the USS Franklin D. Roosevelt before WorldWar II ended. A total of 18 sets were produced by RCAbefore the program was cancelled at war's end. Operating ata higher frequency than the SA, the SR -2 had a longer rangeand supplied more accurate range and bearing data thanearlier shipborne search equipment. The antenna structurewas made of stainless steel to avoid corrosion problemsexperienced with former search arrays.

9

1946 Radar echoes from the moon

Late in 1945 the U.S. Army Signal Corps began a program to determine whetherradar signals could be reflected from the moon and what use might be made ofthem.

Two antenna "mattresses" of the type used on the Army SCR -270-271 radarwere assembled together at Evans Signal Laboratory in Belmar, NJ. Theresulting array of 64 dipoles, about 40 feet square, was supported on a 100 -foottower. Because the antenna could rotate in azimuth only, observations wererestricted to a relatively short time near moonset and moonrise.

On January 10, 1946, the first echoes from the moon were obtained atmoonrise. One of the earliest photographs from these experiments, that of anecho at moonrise on January 22, 1946, is shown on the sweep of a conventionaltype -A radar oscilloscope. At about 2-1/2 seconds after the first pulse wastransmitted, a vertical deflection of the trace occurred - the pulse returned fromthe moon had been received. This became the most widely published cathode-ray tube trace in history.

Vht

1I I I

0 100,000 000,000

030,000MILES

300,000

The Bumblebee contract was one of the earliest examples of an integrated radarsystem. The system did far more than detect targets. The contract was initiatedto have RCA support the U.S. Navy and their systems contractor, the AppliedPhysics Laboratory of the Johns Hopkins University. RCA's task included:"Research and development work . . . carried on in connection with electronicguidance equipment for guided missiles .... Special emphasis shall be placed onthe development of radar equipment to track enemy targets and to guide these(Bumblebee) missiles to the targets."

The significant feature of this contract for RCA was that it led into the entireinstrumentation radar business for the corporation: AN/FPS-16, AN/MPS-25,AN/FPQ-4, AN/FPQ-6, AN/TPQ-18, CAPRI, and AN/MPS-36.

Reprint RE -23-5-2Final manuscript received January 18. 1978

Author Greene holding aprototype four -horn monopulsefeed for the Apollo LM rendez-vous radar.

Tom Greene's long association with radar began in 1942,when he tested SA, SD, and Mark 11 radars for RCA inCamden. He worked on the AN/FPS-16, Talos, BMEWS,and DAMP programs at MSR and later served as Managerof Proposals and Contract Reports there. He is now aninstructor for PRICE, RCA's cost -predicting model, and isalso responsible for PRICE documentation.Contact him at:PRICE SystemsGovernment Systems DivisionCherry Hill, N.J.Ext. PY-5782

I0

Radio Visionthe early days of radar at RCA

RCA's involvement in radar started as an attempt to find anapplication for microwaves.

Finding work for microwaves

When radar development was first undertaken, mostlyby various military establishments, in the middle 1930s,the carrier frequency used for the transmission wasseveral hundred MHz at the high end. It wasn't untilcompactness and greater directivity indicated the use ofa basically higher carrier frequency, and a powerfulpulsed magnetron was developed in England, that workon microwave radar started in the early 1940s at M.I.T.Radiation Laboratory.

Irving Wolff initiated the program of microwave research at RCAdescribed in this article that eventually led to the production ofradar equipment. He became Director of the Radio Tube ResearchLaboratory in 1946 and was Vice President, Research, RCALaboratories from 1954 until his retirement in 1959.Contact him at:111 Red Hill Rd.Princeton, N.J. 08540609-924-3252

Author Wolff checking early radar equipment on the roof ofBuilding 5 in Camden during the 1930s.

I. Wolff

In RCA, one might say that we had the cart before thehorse. In the early 1930s radar, we called it RadioVision, was initially developed to find an application formicrowaves some seven years before microwavesbecame the backbone of radar.

About 1932 publications were appearing, originating inGermany and Japan, relating to developments ofmicrowave magnetrons. Research was also reported atthe University of Michigan. This seemed to be a fruitfularea for RCA research, and with the approval of Dr. E.W. Engstrom. Manager of Research, we initiated a pro-gram to develop a 3000 -MHz system. Dr. E. G. Linderwas recruited from Cornell to undertake a magnetrondevelopment. The receiver we used first was an old-style silicon -crystal type.

By 1934, work had progressed sufficiently to have anoperating transmitter -receiver system and RCAdemonstrated the equipment at several IRE meetingsduring that year. At that time, the possibility of reflectingsharp beams of microwaves from metal objects andionized gases was shown.

The Signal Corps was excited by these demonstrationsand invited us to bring the equipment to Navesink Lightat Sandy Hook, N.J. to test the range of the apparatus asa communication set. We were very thankful for thisinvitation because the location for distance tests wasmuch better than any we had in Camden. Thetransmitter site was at an altitude of 250 feetoverlooking New York Bay and there was a clear rangeover New York harbor of more than 15 miles. Thetransmission tests were quite successful, giving a rangeas great as line -of -sight.

From our standpoint the most significant experimentwas an attempt to obtain a reflection from a boatentering New York harbor. The channel was about one-half mile from the transmitting -receiving location andwe were elated to receive reflected signals from a shippassing through the channel. Some water tanks onshore at about the same distance were also goodtargets.

Pulse radar

Our equipment at this stage used audio modulation ofthe transmitter; three-foot parabolic dishes on bothtransmitter and receiver obtained directivity. Thus, onlythe azimuth of the reflecting object was determined. Thesuccess of the reflection experiments at the SignalCorps Laboratory changed the orientation of ourresearch from the possible application for radio relayingto use as a navigation instrument. For most effectiveapplication in navigation, however, distance as well asazimuth information would be required, leading us tostart a project for pulse modulation of the transmitterand pulse amplification in the receiver. We aimed for apulse of less than one microsecond duration to obtainsatisfactory distance resolution.

RCA's Camden plant was the site of early radar experiments.

Apparatus with such pulse modulation was constructedin 1938 and the equipment was placed on the roof ofCamden's Building 5 for tests. The receiving equipmentwas later modified to substitute a superregenerativemagnetron for the crystal or detector. We were able topick up and follow the ferry boats and other shipping inthe Delaware River as well as the trains on the elevatedline on the Philadelphia side of the river.

At first, we had only a distance trace on the scope with avertical blip showing the distance of the reflectingobject. It was not long, however, before we tied theposition of the antenna to the horizontal trace, the timeafter pulsing of the transmitter to a vertical trace, andthe pulse signal to the grid, giving what is now called Bscan.

In considering this development, it is interesting to thinkabout the debt we owed television in what we weredoing. If it had not been for the television research it isdoubtful that we would have had the tubes to amplifythe short pulses and the cathode ray tubes on which toshow the return signals. These pieces of equipment areso commonplace now that it is easy to forget that theywere new developments at that time.

How our radio -vision equipment could have beendeveloped to have a commercial application is anyone'sguess. We had in mind using it as a navigation aidinstalled on ships, and with additional refinement it

might well have gone in that direction.

A series of plane crashes gave radar development a push.

However, a seemingly more important commercial needtook precedence and changed the direction of ourresearch sharply. Shortly before the time when thesuccessful tests described above were made, there had

been a series of very bad airplane crashes into moun-tains. At a high-level meeting held to discuss the lackof progress that RCA had made toward securingleadership in aviation radio, it was suggested thatadopting our radiovision equipment to airborne use as acollision preventive could give us a big boost with theairline industry. This was the kind of imaginativetechnological advance using radio techniques thatappealed to General Sarnoff, and we were directed tosee what we could do.

The project, initiated in the spring of 1937, was orientedto the development of airborne equipment which wouldwarn pilots of the approach to mountains and otheraircraft. Owing to the instability, research nature, andantenna bulk at that time, a practical airborne systemseemed doubtful until microwave components had gonethrough a development phase. Hence, the highestfrequency which could be used with existingcommercially available components, 500 MHz, waschosen for the radio frequency of the airborneequipment. We did not have manpower available tocontinue research on both the existing 3000 -MHzproject and the new airborne unit, so the former wastemporarily set aside.

In the latter part of 1937, the new equipment wasinstalled in RCA's Ford Trimotor airplane and numerousflight tests were made during the ensuing year. Targetswere the Catskill Mountains and the Alleghenies ofPennsylvania. Two antenna systems were employed. Toget a signal from potential obstacles in front of theairplane, we used an inverted "V" type antenna installedalong the length of the Ford airplane. A dipole antennaunder the plane obtained the altitude signal. With theairplane in level flight at the height of the mountain top,signals were received at a distance of about 5 miles. Ifthe airplane were 1500 to 2000 feet above the moun-tain no signal was visible. The signal picked up fromthe ground reflection on the dipole altitude antenna wasalways visible. Other airplanes flying up to one-half miledirectly in front of our radar -equipped plane could bedetected. This was probably the first successful airborneradar equipment flown anywhere.

It had been our intention to give public demonstrationsof our apparatus, as well as a modification of it forshipboard use to guard against collisions with otherships and obstacles such as icebergs. But once more,our plans had to suddenly change. By the late 1930s,war clouds were thickening in Europe and the militarypersonnel who were shown the equipment and resultsimmediately clamped a lid of secrecy on it and gave uscontracts to develop apparatus for their use. Thereafterwe became essentially a government contractor.

In summary, some seven years after it had been firstundertaken, our microwave research led indirectly to

12

first development, and later manufacturing, contracts forvarious types of radar apparatus for the military. Directlyrelated to the airborne equipment which wasdemonstrated was a search radar using a Yagi antennamounted on the wing of PBR flying boats and a high -altitude altimeter to aid in accurate high -altitudebombing.

It was entirely fortuitous that we just happened to havesomething very much needed by the military at the righttime and it is interesting to speculate as to what mighthave happened if war had not been on the horizon.Assuredly, military application had not been ourobjective when we started the microwave research, oreven when "Radio Vision" was well along. One cancertainly say that progress in development andmanufacture would have been much slower without theimpetus of R&D funds and the urgency of equipment fora war.

FM radar

This, so far, has been the story of our early pulse radar.However, "circumstance of location" led us into anotheruse of reflected radio waves using frequency modulationcf the carrier (now known as fm radar).

At about the time we were making the flight tests in theFord Trimotor, a young man contacted our patentdepartment and said that he had some equipment todemonstrate. C. D. Tuska and I went to look. This wastne "circumstance of location" referred to above. If theyoung man, Royden Sanders, had not happened to liveclose to Camden, it is very doubtful that RCA wouldhave undertaken any fm radar research at that time.

While a sophomore at engineering college, and unawareof the Bell Laboratories fm radar work, Sandersindependently developed the concept of fm radar andleft college to work on his idea. We were familiar withthe Bell Laboratories work and noted that heappreciated some of the factors which limited accuracyand had taken steps to make needed improvements inhis development. We were sufficiently impressed tooffer to buy whatever improvement patents he mightobtain and to give him a chance to proceed with hisdevelopment in RCA. This proved to be a wise decision.Sanders turned out to be a prolific and determinedinventor as well as a sound engineer with a greatdedication to fm radar.

RCA's fm -radar altimeter was very accurate at low altitudes.

In due course, a development model of an altimeter wasbuilt and demonstrated to the military services. Thisaltimeter was not competitive with our high -altitudebombing pulse altimeter, since the pulse unit was mostuseful at then high altitudes such as ten to twentythousand feet, where the ambiguity caused by pulse

length was not of primary omportance. On the otherhand, the fm unit was most useful at low altitudes evendown to fifty feet or less. When the fm altimeter was putinto production, some tens of thousands were producedand, most unusually, the same apparatus wasstandardized for the US Navy, the Air Force, and theBritish. The production demands and timing were toogreat for RCA to handle alone, so two additionalmanufacturers were recruited to build the same unit.

The fm -radar altimeter proved to be most useful overwater, where its exceptional low -altitude performancecould be well put to use. As a next step, the output wastied in with the aircraft altitude control to set the altitudeautomatically. Since the output of the fm altimeter ismost readily a current proportional to altitude, this stepwas not too involved, at least to the extent of gettingadequate control signal from the altimeter.

The next step was a single radar that provided altitude, distance,and speed of approach.

Although it was obvious that the fm radar signalcontained the information for determining the distanceto the reflecting object, it was not generally appreciatedthat it also contained information on the speed ofapproach to the reflector. Sanders appreciated thatwhere there was relative motion between the radar andthe reflector, the sum of the up -sweep and down -sweepoutput frequencies gave a signal proportional todistance, whereas the difference was proportional tospeed of approach to the reflector. Thus, the informationrequired to drop a bomb in level flight over water,namely altitude, distance, and speed of approach totarget, were all available from one instrument. Oneantenna pointed forward to get the target informationand a second antenna pointed downward to obtainaltitude measurement.

Equipment for dropping bombs automatically wasconstructed and numerous flight tests were madedropping water bombs against the lighthouses inDelaware Bay. Fortunately, none of the numerousfishermen in small boats in the river was ever hit.

A final step was to adopt the equipment to automaticbomb release in other than level flight. Before the moresophisticated development could be completed and putinto production, though, the war was over.

This completes the story of the early days of radar(Radio Vision) in RCA, a project which was initiated todevelop equipment and study applications ofmicrowaves, and ended by supplying search radar, radioaltimeters, and automatic bomb -dropping research andapparatus to a military at war.

Reprint RE-23-5-21Final manuscript received September 8, 1977.

13

An introduction to radar concepts

B. Fell Radar is basically simple-transmitting a pulse of microwaveenergy at a target and timing the reflected return. But theradar designer must produce equipment that can also detectthe valid signals in a noisy environment, separate targetsignals from clutter, and avoid antenna -induced errors.

The origins of radar (RAdio Detection And Ranging) science areobscure, since early attempts to detect objects with radio wavesended either in failure or uncertainty as to the usefulness of thisphenomenon. As discussed in Skolnik,' although pulsed radarwas used as early as 1925 to investigate the ionosphere,2 theearliest radars used to detect aircraft were bistatic continuouswave (cw) systems.

Early radars were used to detect the presence of targets but, aspulsed radar techniques were developed, radars were able todetermine target slant range, angle, and rate of change of slantrange with respect to the radar.

With the advent of short -wavelength radar 1 cm to 3 cm) andminiaturization of electronic components, radar was placedaboard aircraft. Present-day airborne radar is used for avoidingcollisions with other aircraft, determining aircraft altitude,locating areas of extreme turbulence, such as weather fronts andthunderstorms, and controlling onboard weapon systems, such asguns or missiles.

The primary military use of radar during the 1950s was to detectand help identify aircraft. Two North American air -defensesystems built at that time were the U.S.-Canadian DEW (distantearly warning) Line along the arctic circle and the U.S. SAGE(semi -automatic ground environment) radar net within thecontinental United States.

The Soviet Union's launching of Sputnik 1 in October 1957 calledfor the development of high -power, land -based, long-range radarthat could probe the environs of outer space. The U.S. BallisticMissile Early Warning System (BMEWS) and the Space Detec-tion and Tracking Network (SPADATS) were developed anddeployed over the last 20 years to answer ballistic -missile andsatellite -defense problems.

Pulsed -radar functions are also being performed by opticalsystems, made possible by the development of the pulsed laserover the past 15 years. Optical radar is sometime referred to asladar (LAser Detection And Ranging):

Table I indicates the standard frequency bands allocated tomicrowave radar. Long-range (greater than 100 nautical miles)radars normally operate within vhf, uhf, or L -band. S- and C -band radars operate over medium ranges (50 to 100 nauticalmiles). X- and K -band systems are usually limited to short ranges

Table IRadar frequency bands are allocated internationally by the Inter-national Telecommunication Union (ITU). Atmospheric attenua-tion increases with frequency, so low -frequency radars are used forlong-range operation and high -frequency radars for short-rangemissions.

LetterDesignation

Frequency

VHF 137 - 144 MHz216 - 225

UHF 420 - 450 MHz890 - 940

L Band 1215 - 1400 MHz

S Band 2300 - 2550 MHz2700 - 3700

C Band 5255 - 5925 MHz

X Band 8.5 - 10.7 GHz

K, Band 13.4 - 14.4 GHz15.7 - 17.7

K Band 23 - 24.25 GHz

Ka Band 33.4 - 36 GHz

(less than 30 nautical miles) since their transmissions experiencehigh attenuation as they propagate through the earth's at-mosphere. The microwave absorption is mainly caused by oxygenand water vapor in the atmosphere.

This paper deals exclusively with radar that operates in themicrowave region of the frequency spectrum. It should provide abrief answer to the questions:

1) What is a radar?

2) How does a radar work?

3) What are the analytical tools used in radar system design?

Types of radarRadars perform two primary functions: surveillance and tracking.

Surveillance (or search) radars, such as the RCA AN/ UPS -1 andthe GE AN/ FPS -24, search a volume of space and report thedetection of targets within the volume. Most airport radars are ofthe surveillance type. Tracking radars determine a time -history ofand Tracking System (SPADATS), in the late 1960s, doubled theUnited States space -tracking ability.

14

and the Sperry AN/ SPG-55, must be provided with initialpointing data in order to acquire and lock onto a target. Thetarget is tracked until it leaves the radar coverage, maneuvers outof the radar tracking "gate," or until the target is no longer ofnterest. Tracking radars have been used at such facilities as Cape

Canaveral, Wallops Island, and the Pacific Missile Range.

Many radars perform both surveillance and tracking. Included inthis group are: 1) track -while -scan radars; 2) track radars thatlave a search capability; and 3) agile -beam radars.

A track -while -scan radar produces track data by correlatingdetection reports as the radar scans continuously in angle. Thecorrelation procedure can be accomplished by a sophisticatedcomputer or an operator seated at a display.

Tracking radars, such as the RCA AN/ FPS -49, which can opentheir tracking gates to detect targets over a wide range interval, areable to search a volume of space. Once a detection is made, theradar can position its antenna beam in the direction of the targetand execute a tracking algorithm. This type of system is usuallyunable to track more than one target.

Agile -beam (or inertialess beam -steering) radars have been madepossible with the introduction of electronic, as opposed tomechanical, beam steering. Although electronic beam -steeringtechniques were used to a limited extent during World War II, thefull advantages of inertialess beam steering were not realized untilthe early 1960s. Agile -beam radars have the flexibility to schedulesearch or track functions in any direction as needed, since theirantennas are not constrained to rotate at a fixed rate (if indeedthey rotate at all).

The RCA shipboard AEGIS AN/SPY-1 radar system is anexample of an agile -beam radar. AN/ SPY -1 uses a two-dimensional array of phase shifters to steer the radar beam. Henceit is called a phased -array radar. The AN/SPY-1 maintains avolume search and, upon detecting a target, tracks that targetwhile still maintaining the volume search. The AN/SPY-1 canthus track many targets anywhere in its detection volume and stillperform surveillance of the radar coverage.

Since time is a fixed entity, an agile -beam radar must trade offavailable time between search and track functions. However, thedata -handling capabilities associated with an agile -beam systemcan be enormous. For instance, the addition of one BendixAN/ FPS -85 phased -array radar to the USAF Space Detectionand Tracking Systems (SPADATS), in the late 1960s, doubledthe United States space -tracking capability.4

Radars can also be characterized by the types of data they collect.Five common types of radar are continuous -wave, pulsed, pulsed-doppler, multistatic, and synthetic -aperture systems.

The continuous -wave (cw) radar transmits a continuous signal. Ifthis signal is transmitted at a fixed frequency, target doppler (ameasure of the target speed along the line connecting the radarlocation with the . target position) can be determined veryaccurately. Many police departments use this type of radar tomonitor the speed of automobiles. In addition to target doppler,frequency modulation provides a method of determining targetrange with cw radars.

Pulsed radars transmit a fixed duration pulse of energy atrepeated intervals called the pulse repetition interval (PRI).Pulsed radars measure target range and angle.

Pulsed doppler radars measure target position in the samemanner as pulsed radars but, in addition, extract dopplerinformation from a received train of pulses. The doppler andposition information is used not only to detect and track targets,but also to discriminate between moving targets and radar clutter.The remainder of this paper concentrates on describing theproperties of pulsed doppler systems unless stated otherwise.

Multistatic and synthetic aperture radar systems will be discussedbriefly for the sake of completeness.

Multistatic radar uses one or more transmitters and a set of radarreceivers over a long baseline to provide highly accurate angle, aswell as range, information. If the elements of the multistaticsystem are coherent, a multistatic radar becomes a microwaveinterferometer. The multistatic concept has been used extensivelyin the field of radar -astronomy.

Synthetic -aperture systems provide high angular resolution withsmall antenna apertures. Interferometric in nature, these radarsare placed on moving platforms such as aircraft or satellites anduse the platform motion to simulate a set of receive antennas.

Barry Fell is a physicist who has worked in the areas of phased -array radar and radar electronic countermeasues. Since joining theSystems Department of RCA Missile and Surface Radar in 1976, hehas been engaged in the development of ultra -high reliability, low -prime -power -consumption, unattended radars for use in remotelocations.

Contact him at:Systems DepartmentMissile and Surface RadarMoorestown, N.J.Ext. PM -3584

15

ANTENNALi\f"\-/"S'

TRANSMITTER

WAVEFORM GENERATOR

REMOTE RADARS,

COMPUTERS,ETC.

TIMINGAND CONTROL

DATA PROCESSOR

n-11VMOMm.

RECEIVER

SIGNAL PROCESSOR

111,6A,OEIrib

ANTENNA POSITIONSERVOS, ETC.

Fig. 1A radar consists of a number of subsystems that affect overallsystem performance, but are all tied to a central timing/controlsubsystem. The waveform generator produces a pulse that isamplified by the transmitter and radiated by the antenna. If thepulse strikes a target, a return (dependent on the target's radarcross section) will be reflected back to the antenna. This signal isamplified in the radar receiver; the S/N is maximized by a techniqueknown as matched filtering. The signal processor takes thisinformation and filters out unwanted returns. A data processorprovides for the radar's long-term timing and control, sendingsignals to display units, the antenna positioners, to other radars orcomputers, etc.

Radar functionsRadars can detect targets, estimate target position and velocity,and identify targets.

Target detection consists of searching a volume of space andreporting any targets which appear in that volume. Targetposition and velocity estimation includes the determination ofrange, angle (either azimuth, elevation, or both) and range rate(doppler) with respect to the radar. Target velocity cannot bemeasured instantaneously by a single radar, but must be derivedfrom a time history of target position data.

Radars that measure target range, azimuth, and elevation arecalled three-dimensional; radars that measure range and only oneangle are called two-dimensional. A two-dimensional systemwhich measures range and elevation is sometimes called a heightfinder, since it provides an accurate determination of targetaltitude.

There are two general approaches to target identification:Selective Identification Friend/ Identification Friend or Foe(SIF/ IFF); and Space Object Identification (SOI). SIF/ IFF is atechnique in which a radar sends out a coded series of pulses thatin turn trigger a transponder carried aboard the radar target. Thetransponder reply is directed toward the radar and contains targetinformation such as target identification, range, and altitude. Air-traffic controllers at major airports use such systems to im-mediately identify aircraft in the vicinity of the airport.

SOI is a radar identification technique that uses the "skin return"from a radar target. Over the past 20 years, much progress has

been made in identifying orbiting satellites by illuminating theseradar targets at high data rates. The resulting amplitude vs timehistory of the target returns is analyzed to determine targetcharacteristics such as size, shape, and tumbling rate.

Radar system operationAs shown in Fig. 1, a radar is composed of a series of subsystems:a waveform generator, a transmitter, an antenna, a receiver, asignal processor, a data processor, control equipment, anddisplays.

The waveform -generator produces a pulse of electromagneticenergy.

This pulse, the beginning of the radar cycle, can be a simplesinewave at a constant frequency or a complex waveform.Typically, complex radar waveforms may include a series ofsubpulses at the same or varying frequencies, a single phase -codedpulse, or a single frequency modulated (fm) -coded pulse.

The transmitter is an amplifier that increases the amplitude of thewaveform generator output to the desired level.

This amplification is accomplished at radio frequencies (rf) byeither a magnetron, klystron, traveling wave tube, conventionalmicrowave tube, or solid-state power amplifier. The transmitter ischaracterized by its power amplification (gain), peak poweroutput, pulsewidth, and duty factor (product of transmitted pulseduration and the pulse repetition frequency). The output of thetransmitter is fed to the antenna subsystem.

The antenna matches the impedance between the guided -waveoutput of the radar transmitter and the free -space propagation ofthe radar pulse.

Antennas are of many types and sizes. The antenna determines thetwo-dimensional beamshape and beamwidth of the transmitterradar energy. Since the antenna concentrates the transmittedenergy in a particular solid angle, passing energy through anantenna amplifies the total radar energy in a particular directionas opposed to transmitting radar energy equally in all directions.This characteristic of an antenna is called the directive gain.Antenna gain and beamshape are discussed below.

If a radar target is passive (i.e., if it does not actively produce radiotransmissions that are directed back to the radar), the targetmerely acts as reflector that intercepts a portion of the transmittedradar power and reradiates it in various directions. Thesereflection characteristics are described by the target radar crosssection (a).

A radar target can modify the radio frequency of the transmittedradar waveform.

The component of target velocity along the propagation directionof the radar transmission (vr) shifts the frequency of thetransmitted pulse, because of the doppler effect, according to therelation

Cif = 2 vr/ X (1)

where Af is the magnitude of the frequency shift and A is thewavelength of the transmitted pulse.

The radar mission determines the radar operation after transmis-sion of a pulse. Surveillance radars continually scan a volume of

16

space. The antenna beamwidth, antenna scan rate, and pulse-repetition frequency of such a radar determine the number ofpulses transmitted and hence received by the radar. A typicalsearch radar may transmit 20 pulses during the time it takes theantenna beam to sweep across a target. Tracking radars, on theother hand, know the target position relative to the beam axis,and by means of feedback they can continually steer the antennaso it is always directed towards the target.

The radar energy reflected from a target and captured by the radarantenna is sent to the radar receiver.

The receiver subsystem converts the frequency of received energyrrom radio frequency (rf) to an intermediate frequency (if), which_s usually around 50 MHz. The receiver amplifies the receivedenergy and maximizes the signal-to-noise ratio of individualpulses through a technique called matched filtering. This informa-tion is sent to the signal processor, which interprets the content ofthe received energy.

The signal -processing subsystem filters out unwanted returns.

Such unwanted "clutter," is typically energy reflected by obstacleson land, the land itself, the sea, or precipitation in the form of rainor snow. The signal processor can perform coherent or non -coherent pulse integration and presents its output to an indicatorscreen or an automatic target -detection subsystem.

A timing and control subsystem oversees overall operation of a

radar during the period from pulse formation to pulse formation.

Overall radar operation is controlled by an operator or, inmodern high -data -rate automatic systems, by a data processor. Inaddition, the data processor performs data smoothing andprediction for targets under track. The data processor alsocompensates for platform motion associated with radars locatedaboard aircraft or ships.

Radar output goes to a display or as control information for othersystems.

The output of the radar can be presented in various ways, such as:) a plan position indicator (PPI) scope, which plots target range

versus antenna azimuth position; 2) an "A" -scope, which displaysreceiver output amplitude as a function of time (i.e., range) for aparticular azimuth or elevation direction; or 3) a cathode-ray tube(CRT) display updated by a data processor in real time. The radarsystem can use its output data to position its own antenna or drivealso be sent to distant command and control posts, othercomputers, or to other radars. For example, a multistatic radarnetwork requires communication among the related sensors toenable maximum use of the information obtained from theextended radar baseline.

Radar detection of targetsDetection can be broken down to transmitting a pulse, receiving areturn, and separating signals from noise.

A radar determines the basic target -position parameters byproducing a high-powered pulse of microwave energy (Fig. 2).This pulse energy is concentrated into a beam and directed by theradar antenna into a solid angle centered along a line in space. If atarget lies in the general direction of this beam, the target willintercept and reradiate a portion of the transmitted radar energy.

R = SLANT RANGE = CAT/2

fl = RANGE RATE(DOPPLER) = XAf/2

BE = ELEVATION

OA= AZIMUTH

TRANS-MITTER

TRANSMITTEDPULSE

RECEIVERAND

SIGNAL PROCESSOR

REFLECTEDPULSE

OA

Fig. 2Target positions can be determined in terms of slant range,azimuth, and elevation (for a "3-D" radar) or range and only oneangle ("2-D" radar). Range is determined by timing the interval ittakes for a pulse traveling at the speed of light to travel to the targetand back. Range rate is derived from the doppler-induced frequen-cy difference between the transmitted and received pulses. Coarseangle measurements are obtained from the antenna pointingdirection.

H rTARGET ANGLE WITH RESPECT TOANTENNA BROADSIDE OR AVERAGEPOINTING ANGLE

Fig. 3 +e

Fine -angle measurements are obtained by comparing the returnedtarget amplitude (and/or phase) in two adjacent beam positions.These beams are generated either electronically or by physicalmovement of the antenna. The ratio of the received targetamplitudes in the left and right beams is used to measure the targetposition with respect to the antenna broadside or average steeringangle.

The radar antenna captures a portion of this reradiated energyand sends it to the radar receiver/ signal processor where thepresence of the target is detected, assuming the returned signalstrength can be differentiated from the system noise and clutter.

The slant range from the radar to the target is determined bymeasuring the time delay between the transmission of the pulseand the detection of the echo. The rate of change of this slantrange with respect to time (which is called the range rate ordoppler of the target) is determined by measuring the frequencydifference betwen the transmitted and received pulse.

The target angular position is determined from the antennapointing direction. Fine -angle information can be determined inmany ways, all of which essentially compare the target amplitudeand/or phase in two adjacent beam positions. These beampositions are generated either simultaneously (monopulse) orsequentially (sequential lobing). See Fig. 3.

17

RECEIVER

OUTPUT

TRANSMITTEDPULSE

OPEN RECEIVER GATE

TRANSMITTEDPULSE

Il

PULSE REPETITION INTERVAL =PRF

SEARCH PRI

CLOSE

RECEIVERGATE

OPEN RECEIVER CLOSE RECEIVER

GATE GATE

PULSE REPETITION INTERVAL =PRF

TRACK PRI

Fig. 4Radar pulses are transmitted at a fixed rate-the pulse repetitioninterval (PRI); its reciprocal is the pulse repetition frequency (PRF).A search radar (top) is interested in the entire volume it issearching, and so monitors returns throughout the radar's entirerange. A tracking radar (bottom), however, is interested only in thetarget it is tracking and so only monitors returns about the target'sestimated position.

NOISE SPIKE EXCEEDING DETECTION THRESHOLD (FALSE ALARM)

TARGET PRESENT AND DETECTED(VALID DETECTION)TARGET

DECLARED

TIME

TARGET PRESENT AND NOT DETECTED(MISSED TARGET)

DETECTION THRESHOLD

- RMS NOISE LEVEL

Fig. 5Detection threshold is a variable. It is set to produce the bestcompromise between avoiding false alarms and avoiding missedtargets.

Pte= PROBABILITYOF FALSE ALARM

SIGNAL-TO-NOISE RATIO (SA)

Fig. 6Detection probability and false -alarm probability vary with S/N.Threshold setting must take this into account.

As mentioned previously, target detection is performed in theradar's receiver and signal -processor subsystems. The radarreceiver must differentiate the reflected radar signal from thesystem noise background. The received signal strength dependson the target range and reflection characteristics, the radartransmitting power, and antenna gain. System noise is caused byunwanted electromagnetic energy that either enters the radarantenna from the radar environment or is produced by brownianmotion of electrons within the radar receiver itself.

The radar transmits pulses of electromagnetic energy at a fixedrate called the pulse repetition frequency (PRF). As illustrated inFig. 4, the time interval between two successive pulses is called thepulse repetition interval (PRI), which is the reciprocal of the PRF.

The receiver in a search radar monitors energy which has enteredthe radar receive antenna throughout the radar's entire range. SeeFig. 4(top). The receiver in a tracking radar, however, monitorsreceived energy only over the range extent which brackets thetarget's estimated position. See Fig. 4(bottom).

Target detection consists of establishing a threshold and declaringa target when a returned pulse of energy exceeds that threshold.

The detection threshold (Fig. 5) can be manually controlled froma gain control on the operator display, or it can be setautomatically either at a fixed level or at a variable level based ona constant -false -alarm -rate (CFAR) logic network. The presenceof noise makes target detection a statistical problem. Theprobabilities associated with target detection in a noise environ-ment are:

()probability of false alarm, which is the probability that anoise spike will exceed the preset threshold;

2)probability of detection, which is the probability that a targetreturn, if present, will exceed the threshold; and

3)probability of a missed target, which is the probability that atarget return will not exceed the threshold.

For a mathematical introduction to the detection of signals innoise, the reader is referred to section 2.5 of Ref. 5.

Fig. 6 plots detection probability versus signal-to-noise ratio forvarious detection threshold settings (i.e., for various probabilitiesof false alarm).6 For a given threshold (probability of false alarm),the greater the signal-to-noise ratio, the greater the probability ofdetection. Similarly, for a given detection probability, the higherthe threshold setting (i.e., the lower the probability of false alarm)the greater is the required signal-to-noise ratio.

The radar range equationThe signal-to-noise ratio (SI N) and system false -alarm probabili-ty (Pja) determine the probabilty of detecting a radar target (Po),i.e.,

PD = PD (S1 N, Pfa) (2)

In addition, the signal-to-noise ratio developed on a target alongwith an associated resolution factor determines the estimationaccuracy for the target range, range rate (doppler), and angle withrespect to the radar. The estimation accuracy standard deviations,OR, 01, and oe, are given by the relations:

OR = ARI[kR (SI Mil (3a)

18

aR = (S/ N)'1

co =001[103 (S/N)'1

(3b)

(3c)

AR, the radar range resolution, is determined by the reciprocal ofthe effective bandwidth of the transmitted pulse. AR, the radardoppler resolution, is determined by the reciprocal of the effectivetime the target is illuminated by the radar. AO, the angleresolution, is determined by the antenna effective beamwidth. Theconstant factors kR, la, and ko are determined by the relationbetween the transmitted waveform, the filter response of thereceiver, and the antenna beam characteristics.

The radar range equation represents the relationship between thereceived radar signal-to-noise power ratio and a series ofparameters which characterize the radar, the target, and theenvironment.

The radar range equation can be written as

SI N = (Ppeak/kTB NF) (GT) (c I 47R2) (A,/ 4ir R2) (1/ L) (4)

where(S/ N) is the signal-to-noise ratio at the output of the radar

signal processor;

is the peak power transmitted by the radar;

is Boltzman's constant;

T is the ambient temperature (degrees absolute);

B is the bandwidth of the radar receiver;

NF is the noise figure of the receiver;

GT is the transmit antenna gain;

R is the slant range from the radar to the target;

a is the radar cross section of the target;

A, is the effective area of the receive antenna; and

L is the total system losses.

Ppeak

k

The radar range equation provides an analytical tool foranalyzing and predicting radar performance. As shown in Eq. 4,the radar range equation determines the instantaneous signal -to -average noise power as a function of the radar and targetparameters.

As shown in Eq. 5, the signal power present at the output of theradar receiver (Sow) is determined by the peak power transmittedby the radar, the antenna gain provided by the radar upontransmission of the signal (Gr), the ratio of the effective targetradar cross section area (a) to the area over which the transmittedenergy has spread (47rR2), the ratio of the effective receiveantenna aperture area (Ae) to the area over which the reflectedenergy has spread (also 47r -R2), times the gain of the receiver (GA),divided by the system losses (L).

Saw = Ppeak Gr (a/471 -R2) (A,./47rR2) GA (1/L) (5)

The environmentally -caused noise within the receiver bandwidthB at absolute temperature T is

PN= kTB (6)

where k is Boltzman's constant, a universal constant that relatestemperature to energy. The noise figure of the receiver (NF) is a

measure of the noise produced by the receiver itself. It is definedby the relation

Now = kTBGA NF (7)

where N., is the total noise output generated by the receiver andGA is the gain of the receiver. Therefore, the signal-to-noise ratiopresent at the receiver output, the ratio of Eqs. 5 to 7, is describedby Eq. 4 above.

The radar antenna receive power gain, GR, is related to theeffective area of the radar antenna, Ae, through Eq. 8,

GR = 47rA,/ X2 (8)

where X is the free -space wavelength of the transmitted radarsignal.'

The antenna effective area, Ae, is related to the physical area of theantenna, A, through the expression

A, = p A (9)

p is called the antenna efficiency factor. Its numerical value isbetween zero and one. The efficiency is a function of: I) theefficiency with which the antenna aperture is illuminated; 2) theamount of energy which "spills over" the edges of the aperture(and is wasted); and 3) the efficiency of the antenna feed.

Solving Eq. 8 for Ae, substituting this result into Eq. 4, andcombining terms gives the following form of the radar rangeequation:

SI N = Ppeak GT GR X2/ [(4703 R4 (kT) B (NF)L] (10)

The radar range equation shows which system parameters theradar designer can control.

The peak transmit power (Pp,k), antenna gain (Gr,GR), transmitwavelength (X) and receiver bandwidth (B) are parameters whichare under the control of the radar systems designer.

The system designer has limited control over receiver noise figure(NF) and system losses (L). Low noise figures can be achieved byusing cryogenic techniques or, more practically, by usingparametric amplifiers as the receiver first stage. Principal systemlosses include: 1) losses in the radar waveguide on transmit andreceive; 2) propagation losses caused by atmospheric attenuation;and 3) signal -processing losses. The atmospheric noisetemperature (T) is dictated by the radar environment. The radarcross section of the target (a, which is not necessarily equal to thephysical cross-sectional area of the target) is specified whenstating the radar performance requirements for target detection.However, the actual radar cross section of a target depends on thetarget size, shape, and aspect angle with respect to the radar line ofsight.

The peak power, Ppeak, refers to the power produced when thetransmitter is turned on for the duration of the pulse, r. Theaverage transmitter power during a pulse, Pa g, is given by therelation

Pmq = Ppeak (r/ 7) (11)

19

COMPLEX

WAVEFORM

MATCHED FILTER

TRANSMITTED WAVEFORM IMPULSE RESPONSE

- -101VC-AA1'_ -

RECTANGULAR jPULSE

_ _

T -44 T

MATCHED FILTEROUTPUT

2T

Fig.7Matched filtering between transmitted waveform and receiver output produces the maximum S/N. It does not preserve the shapeof the original transmitted pulse as seen with the example of the rectangular pulse.

SIMPLE PULSE

CARRIER

LINEAR FM(CHIRP)

INCREASING FREQUENCY

180° PHASE REVERSAL CODE

ENVELOPE

Fig. 8Typical radar waveforms. Simple pulse is formed by rectangularmodulation of the rf carrier. "Chirp" waveform, used to achieve highrange resolution at moderate transmission power, is actually a longpulse generated by delaying different frequency components of ashort pulse. Phase -reversal codes are used to obtain pulsecompression for low as well as high time -bandwidth products.

where T is the pulse repetition period. Since T = 11 PRF, therelation between average and peak transmitter power can also bewritten as

Pavg = Ppeak (r) (PRF) (12)

The transmitter duty factor (DF) is defined to be the quotient ofthe average transmitted power to the peak transmitted power,hence

DF = Pavg I Ppeak = T (PRF) (13)

Continuous wave (cw) radars transmit a continuous "pulse" at afixed frequency. Under these conditions r -co, PRF-0, Pavg=

Ppeak, and thus DF = 1.

Signal-to-noise enhancementThe matched filter is the first step in S/N enhancement.

The signal-to-noise ratio at the receiver output is maximized whenthe receiver is a matched filter with respect to the transmitted

radar waveform.8 The impulse response of a matched filter is thetime reversal of the transmitted waveform, as illustrated in Fig. 7.The matched -filter output is the convolution of the transmittedwaveform with the matched -filter impulse response. Matchedfiltering does not preserve the original transmitted pulse shape.For example, the matched -filter output associated with arectangular -pulse transmitted waveform is a triangular pulse, asshown in Fig. 7.

The selection of the transmitted waveform used by a radar systemis based on factors such as required range and doppler resolution,peak power constraints, and clutter environment. Fig. 8 illustratesa few typical radar waveforms. A simple pulse transmitted by aradar is formed by rectangular modulation of the radio -frequencycarrier. The pulse shown has a pulse length r.

Transmitted waveforms can have internal codes. For example, along pulse generated by delaying different frequency componentsof a short pulse (pulse expansion) before transmission anddelaying the frequency components in an inverse manner uponreception (pulse compression) is called a "chirp" waveform. Thechirp pulse provides the equivalent of a high -power illuminationof the radar target using only a moderate peak transmissionpower because of the length of the transmitted waveform. At thesame time, the chirp pulse provides high range resolution, which isdetermined by the effective bandwidth of the pulse.

Pulses can also be coded by periodically shifting the transmittedradio frequency phase 180 degrees, as illustrated in Fig. 8. Thematched filter for this type of phase code has an impulse responseequal to the time reversal of the transmitted phase code.

Pulse trains are used to determine target doppler as well as targetrange. Since doppler resolution is determined by the reciprocal ofthe effective time the target is illuminated by the radar, multiplepulses transmitted in a single pulse repetition interval can be usedto make a gross doppler determination during that period. Finedoppler estimates can be made by processing a series of pulsestransmitted over a sequence of pulse repetition intervals. A seriesof pulse repetition intervals that are processed as a group toenhance target detectability constitute a dwell of the radar beam.See Fig. 9.

20

PULSE REPETITION INTERVAL

7 A In A In A

DWELL PERIOD

ri TRANSMITTED PULSE

A RECEIVED MATCHED FILTERED PULSE

Fig. 9Radar "dwell" enhances target detectability by processing a seriesof pulses.

Table IIRadar returns for a target usually vary from pulse to pulse becauseof the complicated reflection process taking place. The fluctuatingradar cross section for different target types was initiallycategorized by Swerling. His Models I and III usually apply to jetaircraft or missiles, while Models II and IV apply to propeller -drivenaircraft.

CROSS-SECTIONDISTRIBUTION

SWERLINGMODEL

PULSE -TO -PULSE

EXPONENTIAL-RANDOM ASSEMBLYOF REFLECTORS

I COHERENT

II NON -COHERENT

RICE -ONE LARGE STEADYAND NUMEROUSRANDOM REFLECTORS

III COHERENT

IV NON -COHERENT

The signal-to-noise output of a radar system can also beenchanced by integrating a series of radar returns.

The integration, which is performed in the signal processor ordata processor, effectively multiplies the single -pulse signal-to-noise ratio by a factor Ile. Hence,

(57 /integrated = ne (S N)single pulse (14)

where rz., is equal to the number of pulses integrated (n) if theintegration process is coherent over the integration period. If thiscoherence is disturbed or not present within the radar system, ne isa number between n1/2 and n, depending on the system factors thataffect the integration process.

Coherent integration depends on the coherence of the radartransmitter, the target reflections, and the receiver characteristicsfrom pulse to pulse over the duration of the beam dwell. Thetransmitted and received signals are both referred to a stable, andthus coherent, reference signal produced in a local oscillator.

Since a radar target is a complicated reflector or collection ofreflectors, the signal-to-noise ratio returned by a target usuallyvaries from pulse to pulse. This variation in target cross sectionwas accounted for in a systematic manner by Swerling9 in terms offour models summarized in Table II. Scan -to -scan fluctuation(Swerling Models I and III) usually applies to jet aircraft ormissiles, whereas pulse -to -pulse fluctuation (Swerling Models IIand IV) is representative of propeller -driven aircraft. A pointtarget results in a constant or steady radar cross section and iscalled a Marcum target model.

The effect target cross section behavior has on the detectability ofthese various target models is illustrated in Fig. 10. These curves

099 -a98-

095

-0 BO -

0 70 -

0.60 -050-040-030-020-

1

1-10 -5 0 '0

PER PULSE SIGNAL TO NOISE

25 30

FALSE ALARM PROBABILITY 10-7

NUMBER OF PULSES INTEGRATED 10

CURVE TARGET MODEL

1 SWRL I

2 SWRL II

3 SWRL III

4 SWRL IV

5 STEADY

Fig. 10Detectability varies with type of target model. Example hereassumes false -alarm probability of one in ten million and that 10successive pulses have been integrated. Note that detection -probability curves for fluctuating targets and steady targets crossas S/N increases.

TRANSMITTED

PULSE

RADAR

L:E-%/\PREFLECTEDPULSE

TARGET- A

Fig. 11Clutter is radar energy returned from objects other than the desiredtarget.

represent the signal-to-noise ratio needed per pulse to achieve aspecified detection probability, given a false alarm probability of10-7 and assuming ten successive target returns have beenintegrated.19

Note that for large detection probabilities, a lower signal-to-noiseratio is required to detect steady targets than to detect fluctuatingtargets. On the other hand, for small detection probabilities, thereverse is true. Fluctuating targets require lower signal-to-noiseratios at small detection probabilities than constant targets, sincefluctuating targets will occasionally scintillate in such a mannerthat a strong target signal will be returned to the radar.

Discriminating targets from clutter:signal processing

In addition to thermal noise, the detection of radar targets iscomplicated by the presence of clutter. As illustrated in Fig. 11,clutter is radar energy returned by objects in the radar environ-ment other than radar targets. Sources of clutter are stationaryobjects, land, large bodies of water, and precipitation in the formof rain or snow.

LANDCLUTTER

21

LAND CLUTTER

SEA

CLUTIER RAIN CLUTTER/ WHITE

NOISE

2V,-x (DOPPLER)

Fig. 12Clutter can be identified by plotting return amplitude against thedoppler frequency. Land is essentially stationary, but movingleaves, branches, etc. produce a spread centered around zero. Thesea has more motion associated with it, and so has a wider spread,but is still centered close to zero. Rain, however, is associated withwind velocities, and so has a non -zero mean and a wide distribu-tion.

LAND

SEA RAIN

0

CIRCUIT DIAGRAM

MTI TWO PULSE

CANCELER RESPONSE

DOPPLER FREQUENCY

FILTER RESPONSE

TARGET

PRE

Fig. 13Moving -target indicator canceler is one method of separatingclutter from targets. It works by subtracting successive radarreturns from one another and so attenuates signals that havedoppler frequencies centered at zero (land and sea clutter) or atmultiples of the system pulse -repetition frequency.

DOPPLER

FILTERSTARGET

DOPPLER FREQUENCY PRE

Fig. 14Doppler filters can be used to separate target returns from clutter infrequency space. Each doppler filter is monitored by CFARcircuitry to differentiate targets from distributed clutter in range.

Unlike noise, clutter exhibits properties that are at least partiallycorrelated from pulse to pulse. Target motion (doppler) can also beused to distinguish a target from clutter.

Land -clutter amplitude is characterized by a reflectivity per unitarea. The total energy reflected by land back to the radar dependson the radar pulsewidth, antenna azimuth beamwidth, antennapointing angle, and the type of terrain. The spectral (i.e., doppler)distribution of land clutter has a zero mean and a small spectralspread. The spread is caused by the rms velocity of the scatteringelements such as bushes, tree limbs, and leaves.

Sea -clutter amplitude is similar to land -clutter amplitude in thatits reflectivity is specified per unit area. The primary differencebetween land and sea clutter is the non -zero mean and greaterspectral spread of sea clutter. The sea -clutter mean velocity andspectral spread depend on the wind velocity near sea level, waterwave height, and water wave velocity.

Rain clutter is a volume effect; its reflectivity is specified per unitvolume. The energy reflected by rain back to the radar depends onthe radar pulsewidth and the antenna azimuth and elevationbeamwidths. Because of high -altitude (up to 20,000 feet) windshear, rain clutter exhibits a non -zero spectral mean and a largespectral spread.

Fig. 12 illustrates representative land, sea, and rain clutteramplitude as a function of doppler (clutter motion). As is shown,white noise is distributed uniformly over all doppler frequencies.

Targets can be separated from clutter with a Moving TargetIndicator (MTI) canceler or a doppler filter bank.

As illustrated in Fig. 13, an MTI canceler subtracts successiveradar returns from one another. This results in the synthesis of adoppler filter which, as shown in the figure, attenuates receiversignals that have doppler frequencies centered at zero or atmultiples of the system pulse repetition frequency (PRF). Targetsthat have non -zero doppler will be detected at the signal processoroutput, whereas most clutter will be suppressed.

Another method of differentiating targets from clutter is tosynthesize a doppler filter bank, shown in Fig. 14. This can bedone by taking a fourier transform of a train of radar pulses orsynthesizing a series of doppler filters from a train of radar pulses.Filter synthesis can take place in either the signal -processingsubsystem hardware or in the data -processor subsystem software.

The output of each filter is monitored by CFAR (constant false -alarm rate) logic. The CFAR logic detects threshold crossingsthat occur over only a few adjacent range cells. This differentiatestargets from clutter and suppresses clutter that is distributed inrange.

The doppler resolution of a doppler filter bank is determined bythe width of the individual doppler filters. Therefore, the filterwidth is inversely proportional to the effective time the radarilluminates the target.

Radar antenna theoryMost radar antennas are of two general types: reflector or array.

A reflector produces a radar beamshape from a continuousdistribution of energy across the antenna aperture. Its optical

22

analog is an aperture, which, when illuminated, results in adiffraction pattern. An array antenna consists of a discretedistribution of radiators across an aperture. Its optical analogwould be a one- or two-dimensional diffraction grating.

A radar aperture can be sized and illuminated to produce differentbeamshapes.

A fan beam is narrow in one dimension and broad in theorthogonal direction. This type of beam can be shaped toconcentrate radar energy near the radar horizon and graduallydecrease the energy transmitted at higher elevations. Ellipticalbeams are two-dimensional beams that have an elliptical crosssection. Pencil beams exhibit a circular cross section and are usedto simultaneously determine target azimuth and elevation.

As mentioned earlier, antenna beam steering is accomplishedeither mechanically or electronically. A mechanical scan entailsmoving the pedestal that supports the antenna reflector. Elec-tronic steering can be performed by varying the relative phase ofadjacent radiators (either in the feed mechanism or on the face ofthe antenna). Some radars use a mixture of these two steeringmethods in which, for instance, the radar beam is steeredmechanically in azimuth and electronically in elevation.

Fig. 15 illustrates the parameters that characterize antennaperformance. The antenna aperture is the active area of theantenna surface. The power gain of the antenna is determined bythe physical area of the antenna, the efficiency with which thisarea is illuminated, and the transmission wavelength. The half -power beamwidth of the antenna (in radians) is approximatelyequal to the ratio of transmission wavelength to the aperturedimension. Uniform illumination of the aperture shown wouldresult in a beam that is narrow in azimuth and broad in elevation.

If the aperture shown in Fig. 15 is uniformly illuminated, the far -field antenna pattern in each coordinate would be a (sinx / x)2

APERTUREILLUMINATION

FAR FIELDPATTERN

0

2

APERTURE COORDINATES

MAINLOBE

ANTENNA APERTURE

a

GAIN4TrA

G = P

Xi

BEAMWIOTH 03 dB a,

Fig. 15Antenna parameters depend on surface area, physical dimensions,transmission wavelength, and antenna efficiency.

power distribution consisting of a mainlobe and a series ofsidelobes. The 3 -dB beamwidth would be A/ a and Alb in thehorizontal and vertical coordinates, respectively. The firstsidelobe amplitude would be 13 dB below the mainlobe, as shownin Fig. 16.

From a radar standpoint, these sidelobes are high. They couldresult in increased clutter entering the system through the antennasidelobes and could also result in large -angle errors if the antennasidelobes illuminate large targets.

The microwave antenna designer can control sidelobes by varyingthe illumination distribution across the antenna aperture.

Taylor' "2 has shown that illumination of an antenna aperturewith a series of functions that approximate Chebyshevpolynomials decreases the antenna sidelobe level markedly-typical sidelobe levels of 30 dB below the mainlobe can beobtained. As shown in Fig. 17, these lower sidelobes areaccompanied by a slight broadening of the antenna mainlobe witha slight loss in angular resolution. The antenna gain decreases

(1ST SIDELOBESARE 13 dB BELOWMAIN LOBE)

11.1 IIA SIDELOBES

3 dB BEAMWIDTH

a=cc

_O

ANGLE FROM ANTENNA BROADSIDE

Fig. 16Antenna pattern for uniform illumination produces a mainlobe anda series of sidelobes. Even though the sidelobes are 13 dB belowthe main lobe, they are undesirable. Fig. 17 shows one method ofdecreasing sidelobe amplitude.

APERTURE COORDINATES

3 dBBEAMWIOTH

(1ST SIDELOBES ARETYPICALLY 30 dBBELOW MAINLOBE)

SIDELOBES

Aa IkaL/h. Ak-

ANGLE FROM ANTENNA BROADSIDE

Fig. 17Sidelobes are reduced by weighting the illumination across theantenna aperture. This weighting is accompanied by a slightbroading of the main lobe and a subsequent decrease in antennagain.

23

because of the decrease in the antenna illumination efficiency ascompared to uniform aperture illumination (i.e., the antennaefficiency factor decreases). Taylor weightings are but one of anumber of types of aperture illumination that have been used inradar design over the past thirty years.

ConclusionsRadars measure target echo characteristics (delay, amplitude,phase, etc.) to determine such parameters as target:

1) range;

2) range rate (doppler);

3) angle;

4) velocity (derived from a time history of range and anglemeasurements); and

5) signature (amplitude time history).

The probability of target detection and the accuracy of targetrange, range -rate, and angle determination depend on the signal-to-noise ratio returned by the target to the radar. Therefore, amatched -filter receiver is used to maximize the system signal-to-noise rate at the receiver output.

Since radar targets can be obscured by land, sea, and rain clutter,target doppler information is used to discriminate targets ofinterest from environmental clutter. This is accomplished throughthe use of MTI cancelers or doppler filter banks.

Radar antenna patterns are similar to the patterns developed byfar -field optical diffraction theory. The sidelobes that a radarantenna produces can be controlled and reduced through properaperture illumination weighting.

A thorough treatment of such an extensive subject as radar isdifficult in such a short paper. It is hoped that this generaldiscussion has set the stage for the informative papers that follow.For those interested in a deeper understanding of radar anddetection theory, a short bibliography of significant books andpapers devoted to radar is given below.

AcknowledgmentsMany valuable suggestions were received from Dr. Sam Sher-man, Frank Papasso, and Dr. Josh Nessmith of the RCA Missileand Surface Radar (MSR) Systems Department during thepreparation of this paper. Joyce Walters typed the originalmanuscript. The author thanks these persons and especially hiswife, Gretchen. Their support and assistance made this paperpossible.

BibliographyRadar systems

Barton, David K.; Radar System Analysis, Prentice -Hall (1964).Berkowitz, Raymond S.; (editor) Modern Radar: analysis, evaluation,

and system design, Wiley (1965).Blake, Lamont V.; A Guide to Basic Pulse Radar, Parts I and II, Naval

Research Laboratory Reports 6930 and 7010 (1969).Ridenour, Louis N. (editor); Radar System Engineering, MIT Radiation

Laboratory, Vol. 1, McGraw-Hill (1947).Skolnik, Merrill I.; Introduction to Radar Systems, McGraw-Hill (1962).Skolnik, Merrill I. (editor); Radar Handbook, McGraw-Hill (1970).

Antenna theoryCollin, Robert E. and Zucker, Francis J.; Antenna Theory, Parts 1 and 2,McGraw-Hill (1969).Hansen, Robert Clinton (editor); Microwave Scanning Antennas, Vol. 1,2, and 3, Academic Press (1964-66).Silver, Samuel (editor); Microwave Antenna Theory and Design, MITRadiation Laboratory Series, Vol. 12, McGraw-Hill (1949).

Signal processing/ detection theoryCook, Charles E. and Bernfeld, Marvin; Radar Signals; An Introduction

to Theory and Application, Academic Press (1967).Davenport, Wilbur B., Jr. and Root, William L.; An Introduction to the

Theory of Random Signals and Noise, McGraw-Hill (1958).DiFranco, J.V. and Rubin, W.L.; Radar Detection, Prentice -Hall (1968).Helstrom, Carl W., Statistical Theory of Signal Detection, Pergamon

Press (1968).Lawson, James Llewellyn and Uhlenbeck, George E.; Threshold Signals,

MIT Radiation Laboratory Series, Vol. -24, McGraw-Hill (1950).Marcum, J.I.; A Statistical Theory of Target Detection by Pulsed Radar,

RAND Research Memorandum RM-754 (Dec 1, 1947). Reprinted inIRE Trans. on Information Theory, Vol. IT -6 (Apr 1960), pp. 59-144.

Marcum, J.I.; A Statistical Theory of Target Detection by Pulsed Radar:Mathematical Appendix, RAND Research Memorandum RM-753(Jul 1, 1948). Reprinted in IRE Trans. on Information Theory, Vol. IT -6 (Apr 1960), pp. 145-267).

Meyer, Daniel P. and Mayer, Herbert A.; Radar Target Detection,Handbook of Theory and Practice, Academic Press, (1973).

Nathanson, Fred E.; Radar Design Principles, Signal Processing and theEnvironment, McGraw-Hill (1969).

Papoulis, Athanasios; Probability, Random Variables, and StochasticProcesses, McGraw-Hill (1965).

Rihaczek, August W.; Principles of High -Resolution Radar, McGraw-Hill (1969).

Swerling, Peter; Probability of Detection for Fluctuating Targets, RANDResearch Memorandum RM-1217 (Mar 17, 1954). Reprinted in IRETrans. on Information Theory, Vol. IT -6 (Apr 1960), pp. 269-308.

Woodward, P.M.; Probability and Information Theory withApplications to Radar, McGraw-Hill (1955).

References1. Skolnik, Merrill I.; Introduction to Radar Systems, McGraw-Hill

(1962), pp. 8-14.2. Tuve, M.A., and Breit, G.; "Terrestrial magnetism and atmospheric

electricity," Physical Review, Vol. 28, (1926) pp. 554-575.3. Proc. of the Society of Photo -Optical Instrumentation Engineers, Vol.

128 Effective Utilization of Optics in Radar, Symposium Proc. (Sep27-29, 1977). Huntsville, Alabama.

4. "U.S. Space Tracking Capacity to Double," Aviation Week, Jan 1,1968. Cover & pp. 64-67.

5. Skolnik, Merrill I.; Introduction to Radar Systems, McGraw-Hill(1962), pp. 29-35.

6. Barton, David K.; Radar System Analysis, Prentice -Hall (1964), p. 17.7. Silver, Samuel (editor); Microwave Antenna Theory and Design, MIT

Radiation Laboratory Series, Vol. 12, McGraw-Hill (1949), pp. 169-177.

8. Cook, Charles E., and Bernfeld, Marvin; Radar Signals, An In-troduction to Theory and Applications, Academic Press (1967).Chapters 1 and 2.

9. Swerling, Peter; Probability of Detection for Fluctuating Targets,RAND Corporation Research Memorandum RM-1217 (Mar 17,1954).

10. Skolnik, Merrill, I.; Introduction to Radar Systems, McGraw-Hill(1962) p. 52.

11. Taylor, T.T.; "Design of line -source antennas for narrow beamwidthand low sidelobes," IRE Trans. on Antennas and Propagation, Vol.AP -3 (1955), pp. 16-28.

12. Taylor, T.T.; "Design of circular apertures for narrow beamwidth andlow sidelobes," IRE Trans. on Antennas and Propagation, Vol. AP -8(1960), pp. 17-22.

Reprint RE -23-5-101 Final manuscript received January 30, 1978.

24

a system/technology perspective-

Radar weapon system sensorsA.S. Robinson Radar weapon system sensors have evolved from relatively simple

electromechanically scanned search radars to multifunctionelectronic scanning designs that simultaneously search,

discriminate, and track multiple targets and weapons in real time-all in an intense and growing radar countermeasures environment.

Early radar technologyRadar sensors have played a critical role in both offensiveand defensive weapon systems since they first maturedunder fire during World War II. Driven by a pressing need forearly warning of air attack and for effective deployment oftheir interceptor aircraft, British scientists and engineers, ina very short period of time, performed the prodigious feat ofconverting prewar laboratory research in electromagneticpropagation into a successful wartime operational airdefense system. It is interesting to review the architecture ofthis system, and of the environment in which it operated,since it established the basic framework from whichmodern radar systems have grown.

Art Robinson is Manager of Advanced Technology Programs atMissile and Surface Radar. This group is responsible fortechnological advances in three areas-advanced electronictechnology, advanced packaging and design automation, andadvanced microwave and antenna technology.Contact him at:Advanced Technology ProgramsMissile and Surface RadarMoorestown, N.J.Ext. PM -2296

Reprint RE-23-5-31Final manuscript received February 23. 1978.

Several tactical factors, matched with radar's abilities, contributedto radar's early success.

First, the threat to be detected consisted of the buildup oflarge numbers of aircraft into attack formations, and was atime-consuming process, involving sequential takeoffs andclimbing to loitering altitudes that were well within the fieldof view of defensive radars. Second, the radar -reflectingcross sections of individual aircraft were large, so thatsingle aircraft could be detected, and the approximate sizesof assembling forces could be approximated from their netcross sections and distributions. Finally, once turnedtoward their targets, these formations were limited to thespeed of their slowest aircraft, while the speed and climbingcapabilities of defensive interceptors, the distribution oftheir airfields, and the speed of communication from thedistributed radar network to centralized operations control,and then on to the interceptors themselves, was fast enoughto make it possible to achieve initial interceptions wellbefore target areas were reached.

From a technology perspective, communication equipmentdevelopments had already established a technology basefor transmitters, receivers, and antennas which, while atlower than optimum frequencies for radar, could be useddirectly in radar designs of limited angular resolution thatwere nonetheless effective for their primary "trip -wire"function. This "trip -wire" mode was needed to conservelimited defensive interception resources, making it possibleto initiate interceptions only when they were required, withcentralized control of tactical reserves, and the ability tomatch the number of interceptor aircraft to the estimatednumber of attackers. Once interceptors were airborne,relative, rather than absolute, positions were of importance.Radar tracks of both offensive and defensive formationsmade it possible for ground controllers to vector intercep-tors into advantageous attack positions to local control bythe interceptors based on visual (and, later, airborne radar)contact.

Initially, radar's critical role in air defense was a well kept secret.

Radar's outstanding success during the Battle of Britainresulted both from this fortuitous matching between air -defense needs and the performance that even early radartechnology could provide, and from a lack of appreciationby the offense of the capabilities of these new sensors andof the degree to which they had been integrated into anoverall defensive weapon system. As a result, attacks onboth radars and command centers were limited, and whilesignificant levels of initial damage were achieved, these

25

attacks were not pursued, so that both radars and intercep-tor command links remained effective throughout the war.This was the first and last time that a radar -based defensiveweapon system was in such a fortunate tactical position.Thereafter, all military commands have understood themajor role that radar sensors play, and so have reflected thisunderstanding in the design of both their offensive anddefensive systems.

The basic scanning radar designs developed during World War IIwere based primarily on mechanical scanning.

They included "2-D" search radars-antennas rotating inazimuth, using "fan beam" patterns-narrow in azimuth andbroad in elevation; height -finding radars-antennasslewable to a commanded azimuth position, with antennapatterns narrow in elevation and broad in azimuth so thatthey could be mechanically scanned in elevation("nodded") to provide height data; and tracking radars-antennas with patterns narrow in both elevation andazimuth, forming a "pencil beam" that could track a targetby nutating in a narrow circle around it ("conical scan").

Emergence of electronic scanningSince these early pioneering developments, mechanicallyscanning, single -function radars have evolved throughseveral generations of technology improvements andrepresent, today, the majority of deployed operationalradars. Electronic scanning technology has matured duringthis same time period and is being used increasingly eitherto augment mechanical scanning or to completely replaceit. High -precision, high -data -rate, single -target trackingradars continue to use mechanical scanning in bothazimuth and elevation, but conical scanning of the pencilbeam about the target has been largely replaced by"monopulse" designs. This approach is based on clusters offeeds that form multiple antenna beams, slightly skewed offthe precision antenna pointing axis. (See the paper byProfera.) These beams are processed and subtracted toprovide three signals-one proportional to target amplitudethat is used to measure range, one proportional to elevationangular deviation, and the other proportional to azimuthangular deviation from the antenna pointing axis. The rangemeasurement and the antenna pointing angle, corrected bythe two deviation signals, provide a precise measurement oftarget position, obtainable during a single radar pulsetransmission.

Search radars have evolved to "3-D" designs, with antennabeams narrowed in elevation as well as azimuth. Thesebeams scan electronically in elevation as the antennarotates mechanically in azimuth. For many applications thisrepresents an effective approach. It provides hemisphericcoverage with a single rotating antenna, detects all targetsin the radar field of view, and, by processing the series ofreturns received from these targets as the beam scans pastthem ("track while scan"), obtains target positions in threedimensions on each rotation (range, azimuth, and elevationangle).

For applications such as unjammed air traffic control, thequality of the resultant target tracks is adequate to maintain

required levels of air -space control. For many otherapplications, however, the data rate and beam -pointingflexibility obtainable with this approach is either marginal orinadequate. The problem arises from the lack of flexibility inthe mechanical azimuth scanning technique. As the anten-na sweeps by a target or groups of targets, the number ofdata samples obtained may not be adequate for targetdiscrimination, particularly if the target is immersed inclutter or jamming. The number of returns obtained from agiven target on each revolution can be increased bydecreasing antenna rotation rate, but such a decreaseincreases the delay before a new set of samples is obtainedon the next revolution. When targets are maneuveringrapidly, the quality of track data deteriorates rapidlybetween data samples, and is unsatisfactory for manyapplications. The expansion of electronic scanning in 3-Dradars from elevation scanning alone to combined elec-tronic scanning in elevation and azimuth greatly increasesradar flexibility and is a clear trend for future 3-D radars.There are, however, many applications such as multiple -target discrimination combined with the control of intercep-tion weaponry, in which both search and continuous,precision, high -data -rate, track data is required. In thesemulti -function systems, a multiplicity of fixed antennas, inwhich all scanning is electronic, is necessary. This situationfirst arose after World War II, when ballistic intercontinentalmissiles were developed and deployed.

The first reaction to this ballistic -missile threat was todevelop and deploy a radar network capable of providingearly warning of a ballistic missile attack. This network,based on mechanical searching and tracking radar designs,was intended to provide early warning to both militaryretaliatory forces and to the civilian community. Whilerelatively conventional radar technology was adequate forproviding early warning, it proved to be totally inadequatefor the problem of providing an active defense againstballistic missiles.

Active defense systemsMultifunction radar capabilities are necessary for active ballistic -missile defense.

The large number of missiles involved, the masses ofaccompanying decoys and jammers, and the reflectingwakes formed by the reentry of these bodies into theatmosphere combined to create a vast swarm of targets,immersed in noise and clutter, and to impose a newdimension on the radar performance required in search,discrimination, tracking, and fire control. The key problemswere and are discrimination and fire control-searchingthrough the target swarm using sophisticated waveformsand signal processing to separate out targets of potentialinterest, then tracking and performing further discrimina-tion tests on each of these targets at a high data rate toprovide precision inputs for fire -control computations.Conventional electromechanically scanned tracking radarscan achieve the required data rates only by using one radarfor each target, an approach that is not economicallyfeasible for the large number of simultaneous target tracksinvolved. In response to this threat, phased -array radars

26

technology perspectiv'

Phased -array antennasPhased -array antenna principles are il-lustrated in Figs. 1 through 3. The antenna inFig. 1 rotates mechanically in azimuth, whilescanning electronically in elevation. Theantenna face is divided into horizontalsegments, with rf energy from the radartransmitter distributed to each row throughindividual phase shifters. The energy fromeach phase shifter is then distributed to all ofthe radiating elements in its row. When thephase shifters all receive identical settings,energy emanating from each row adds inphase only along the antenna axis, and apencil beam is generated in a directionperpendicular to the antenna face. In allother directions, the energy adds withdifferent phases and therefore tends tocancel, with the lack of perfect cancellationleading to antenna sidelobes. As in all anten-na designs, the amount of energy lost tothese sidelobes can be controlled by taper-ing the level of rf energy across the antennaaperture.

At the cost of a substantial increase in thenumber of phase shifters (e.g., to 4900 in-stead of 70), the antenna beam can bedesigned to scan electronically in bothazimuth and elevation, as illustrated in Fig. 2.The rf energy from the transmitter nowreaches each radiating element through anindividual phase shifter. By introducing ap-propriate variations in phase in both thevertical and horizontal directions, the anten-na beam can be scanned electronically,typically over angles of ±60°. Typical phaseshifters use either magnetic materials ordiodes in the rf path, with these devices, inturn, actuated by digital control registers.While the antenna is receiving returns fromits last transmission, digital commands aredistributed to buffer registers associatedwith each control register so that, at the endof the listening interval, a single commandcan transfer the contents of all bufferregisters to all control registers, therebyimmediately switching the antenna beam toits new pointing direction. A photograph ofone face of modern four -face Navy phasedarray radar-the AN/SPY-1A-is shown inFig. 3. The AN/SPY-1A phased -array anten-na is comprised of approximately 4500waveguide radiating elements and ferritephase shifters, and is capable of electronical-ly positioning its narrow beam anywherewithin an octant of a sphere.

PHASE SHIFTERS'

POWER DIVIDERS

DIPOLES

Fig. 1Elevation -scanning phased -array anten-na works by transmitting rf energythrough individual phase shifters to allthe radiating elements in a row.

tq

PHASE SHIFTERS

GjIt.4111h11-11014mth

41(It

%L"

/11POWER DIVIDERS

DIPOLES

Fig. 2Elevation- and azimuth -scanningphased -array radar has many morephase shifters than radar in Fig. 1.

Fig. 3Modern phased -array radar, the AN/SPY-1 A, is used in air -defensesystem on Navy ships, four antenna faces per ship. This face is onthe developmental site at Missile and Surface Radar, Moorestown,

27

began to come of age. While initial designs were large andexpensive, they provided the first radar systems in whichthe pointing direction of antennas could be changed in afew millionths of a second, thereby making it possible tomultiplex search, discrimination, tracking, and guidancefunctions at high speed. (See technology perspective 1.)

Under the impetus of ballistic -missile defense, phased -array radar concepts evolved over the years into increasing-ly effective hardware designs in a number of frequencybands. Simultaneously, major advances were made in thelevel of sophistication of radar waveform generation andassociated signal processing, in the generation and dis-tribution of high levels of coherent rf energy, and in both thehardware and software required for massive real-timedigital data processing. These major radar advances werecoordinated with the development of quick -reaction, high-speed interceptor missiles and their associated command,control, and guidance. (The paper by Liston and Sparksshows how radar engineers simulate these sysems as an aidto the design and checkout processes.) The net result of allof these efforts was the establishment of a technology basefor ballistic -missile defense that would make possibledeployment of a defensive system which, against a full-scale nuclear attack, would have a high probability ofintercepting and destroying a significant number ofoffensive weapons, but a low probability of interceptingthem all.

ABM defense deployment has arguments pro and con.

It is possible to reason that deploying such a system wouldreduce the probable level of attack damage. Alternatively, itcan be argued that deploying a strong ABM defense wouldtrigger the production and deployment of a more -than -offsetting number of offensive missiles and so increase theprobable level of damage. Starting from these positions,more and more complex arguments for and against ABMdefense have evolved. At present, ballistic -missile defensehas halted short of full-scale deployment, with substantialongoing funding directed toward probing the potentialimpact of new technologies that might upset the presentstate of nuclear offensive dominance. If current efforts failto limit the number of nations with both nuclear weaponsand the capability to deliver them, it is not unreasonable toenvision the eventual deployment of some form of ballistic -missile defense as protection against the relatively small-scale (but highly damaging) attack that such nations will beable to mount.

Radar sensors must be designed as part of an overall system designprocess.

Lessons learned in attempting to solve the problems ofballistic missile defense are being translated rapidly into thedesign of offensive and defensive weapons systems fornon-nuclear warfare. In the non-nuclear domain, offensiveand defensive systems are continuously growing incapabilities in a state of "restless imbalance." The speedand maneuverability of both offensive and defensiveweapons is steadily increasing, as is the sophistication ofthe sensing, guidance, and warhead options that they carry.

World War II "time on target" tactics of multiple weaponfiring, so timed that all weapons reach the target together,continue in this new weapon environment, so that defensiveweapon systems have to be designed to cope with thesimultaneous arrival of a large number of targets. Further,the time available between target detection and requiredweapon commitment has steadily decreased, to the pointthat time delays inherent in human intervention have had tobe minimized and in some situations eliminated. These andrelated factors have made it essential that effective weaponsystems be designed as total entities, including the platformthat carries them, the radar and electro-optical sensors thatare their "eyes," the signal and data processing, commandand control, and weapon guidance that are their "brains,"and the weaponry that is their "striking arm." The periodicexercising of elements of these systems in actual combathas taught system designers a number of importantlessons, not the least of which is the central role to be playedby systems countermeasures.

Growing importance of countermeasuresOffensive and defensive weapons system designers seek todesign sensors that can see approaching weapons plat-forms (aircraft, ships, tanks, etc.), and the weapons thatthey launch (missiles, shells, etc.), determine theirlocations, velocities and accelerations, and assist in theguidance of weaponry against them. They also attempt todesign into their systems techniques to blind, confuse, anddestroy the sensors that will probably be deployed againstthem, and to overcome attempts to blind, confuse, anddestroy their own sensors.

In essence, countermeasures has become the name of the game.

The crucial role of sensors-both radar and electro-optical-in modern warfare is now fully recognized, and anyand all means for countering their effectiveness is receivinga high level of priority. A substantial percentage of themilitary intelligence efforts of all nations is directed towardevaluating the capabilities and limitations of oppositionweapons and of the sensor/computing techniques that willbe used to aim, launch, and guide them. Dual efforts tend toflow from these evaluations-the development and testingof tactics to take advantage of projected performancelimitations in opposition systems, and the development ofnew and/or modernized countermeasure technologies andequipments to further enhance these tactics. Typical anti -radar techniques include low -altitude, terrain -masked at-tack trajectories; high -power noise jammers to radiateenergy matched to radar transmissions, thereby maskingintermediate- and long-range target returns; repeaterjammers to receive radar interrogations and repeat back amultiplicity of false target returns; decoys designed tosimulate high -threat targets in order to induce heavyweapon expenditures from limited weapon inventories;passive reflector dispensers capable of sowing masses ofhigh radar cross section material over large areas, therebygenerating large numbers of false targets and clutter; andAnti -Radiation Missiles (ARMs) to receive, home on, anddestroy the radar itself, often requiring only radar sidelobeemanations to carry out their mission.

28

Technology implicationsAs these new tactics and equipments evolve, their degrees ofeffectiveness become known through the military intelligenceprocess, and technologies, equipments, and tactics are furthermodified to counter them.

The impact of all of these factors on radar technology hasbeen profound. One result has been increasing use ofphased -array radars, with one radar taking on a number offunctions, (e.g., search and multiple -target track), which inearlier systems required the time-consuming sequentialuse of a number of sensors. Radar bandwidths are beingbroadened and bandwidth diversity features, such aschanging transmission frequency from pulse to pulsethroughout this band, will be used to make the task ofjammers and deception repeaters more difficult. Wherefeasible, individual radars will be designed to operate inmultiple frequency bands, and within a given battle area, amultiplicity of radars, operating in concert, each in adifferent part of the frequency spectrum, will be used toimprove the probability of discriminating true target fromfalse, of defeating jamming, and of detecting missilesnoming on the radiation of a given radar in time to turn that-adar off before it is destroyed. Substantial efforts will beexpended on reducing radar sidelobes, in order to reducehe signal levels received from radar jammers, as well aseasing the task of designing deception transmitters asdecoys for the ARMs designed to home on radar sidelobeenergy. (See the paper by Scudder.) In general, the linebetween radar and electronic countermeasure subsystemscan be expected to blend and possibly disappear as thesesubsystems unite in their common task of sensing incomingtargets, overcoming jamming and deception, and jammingand confusing sensors deployed against them.

The computation speeds already required by weapons -systemsphased -array radars far exceed the wildest blue-sky thinking of theWorld War II era.

Radars are now required to search volumes that may behundreds of miles in diameter with resolutions in the orderof tens of feet, and to discriminate and track large numbersof targets with still higher resolutions. Limitations in thepeak power available with realistic transmitter designsmake it necessary to code transmitter waveforms so thatpulses of relatively long duration can be reduced by signalprocessing to the resolution dimension required by thesystem. (See the paper by Weinstock.) Rain, noise, fixedand movable clutter, chaff, noise jamming, repeater decep-tion jamming, deception targets, and real targets all giverise to radar detections that have to be processed in reallime in order to provide the discrimination and targetinginformation on which the entire system depends.

Pipeline processing, distributed data processing, and low-costdigital LSI make the high-speed computation possible.

Directly following the antenna, required computation ratesare often so high that they can only be achieved by"pipeline" computing architectures (see technologyperspective 2), in which signals are passed through a seriesof computing elements in a "pipeline" configuration, with all

elements operating simultaneously at high speed to achievethe required overall signal -processing throughput rates.Typically, this part of the signal processing retains a limited,but important, level of programmability in terms of suchfactors as pipeline configuration control, weighting valuesto be applied to the signals, etc. (See the paper by Timkenand Herold.) Typical competing implementation ap-proaches in this area include high-speed digital LSI, surfaceacoustic wave (SAW) devices, and sampled analog charge-coupled device (CCD) processing. Only after initial detec-tion processing do signal rates usually drop sufficiently tomake the use of general-purpose digital data processingfeasible. Digital data processing itself is going through animportant era of transition. Tne use of large, high-speedmachines with highly complex software, once the onlypractical design approach due to digital device costs, isgiving way to distributed data-processing architectures,(see technology perspective 3) now made economicallyfeasible by the advent of low-cost digital LSI devices(technology perspective 4). A strong drive is developing touse the opportunity presented by this new architecturalflexibility to improve the entire process of system softwaredesign, development and test-an area that has proved tobe a major cost and performarce stumbling block in manyrecent system developments. The expectation is that sub-stantial simplifications will result from distributing softwarewith hardware in separately testable entities of limitedprogram complexity, while retaining centralized"housekeeping" control of this array of processors, and ofdata -base functions that need to be shared between them.(See the paper by Buch, Clapper, and Smith.)

Two additional "drivers" underlie present trends in radartechnology-one a matter of national style and policy, the secondan economic overlay constraining all of the technologicallypossible alternatives.

Years ago U.S. scientists and engineers used to view theefforts of their U.S.S.R. counterparts with condescension.Each system designed by the Soviets has usuallyrepresented a relatively small incremental progression fromthe prior system designed for that function. The fact thateach of these systems has usually been produced inquantity, and distributed to both their armed forces andtheir allies in a timely manner, wasn't considered in ourthinking. U.S. programs have been characterized by use ofthe highest level of technology currently available, or soonto be available. Resulting systems have been substantiallymore sophisticated than their U.S.S.R. counterparts.However, these systems have taken longer and longer todevelop and have been deployed in limited quantities overrelatively long periods of time. The pattern now emerging isthat U.S.S.R. systems, having gone through substantialimprovements and upgradings over many years, nowrepresent very significant levels of performance and arewidely deployed. U.S. systems continue to demonstrate atechnological edge, but the margin is simply not as great asbefore and, in terms of deployed systems, we are clearlybehind in many situations.

Hand in hand with this problem is the issue of economicconstraints. Our high-technology military systems are

29

increasing in cost at a rate of approximately 400% perdecade. As a nation, we simply cannot afford the substantialincrease in defense expenditures and/or the decrease infielded systems that this cost growth implies. In 1978,approximately $124 billion will be budgeted for defense andapproximately 10% of these resources will be devoted toresearch, development, test, and evaluation. It is probablethat in terms of constant dollars, the budget segmentallocated to the design, development, and production ofweapons systems will remain essentially constant. Theclear challenge in the years ahead, in every area of defense,will be to identify the essential characteristics required ofour weapons systems, and to then focus our technologieson reducing the cost of these systems, so that they can be

-technology perspective 2 -

High-speed pipeline processingDirectly following the antenna,processes such as matched filtering,convolution, correlation, and spectrumanalysis must be performed. Digital im-plementation of these functions com-bines the advantages of high perfor-mance, stability, noise immunity, andprogrammability. However, even thelargest general-purpose digital com-puters cannot achieve the processingspeeds required for real-time signal -processing applications.

Pipeline processing permits thedesigner to achieve the desired datathroughput at the expense of signal timedelay. As illustrated in Fig. 1, signals arepassed through a series of computingelements, with all elements operatingsimultaneously at high speed, to achievethe overall throughput rates required.

Pipeline architectures simplify the con-trol and data storage problems whichwould occur if the throughput were to beachieved by paralleling large numbers ofprocessors. Furthermore, the FastFourier Transform, an algorithm whichgreatly reduces the number of com-putations associated with convolution,spectrum analysis, etc., can be easilyconfigured to fit pipeline architecture.

deployed in the quantities required to establish a realdefense in being, rather than a defensive potential. All signspoint to the probability that, exclusive of guerrilla warfare,when non-nuclear wars occur, they will be of relatively shortduration and will involve very substantial attrition. It is

unlikely that there will be time to tool up, produce, anddeploy systems that are still in research and development atthe outbreak of hostilities.

Each of the papers presented in this issue addresses someaspect of this technology/cost balance problem, and themeans for driving toward deployed technological ex-cellence at a price that we can afford.

Fig. 2 presents the block diagram of a high-speed digital pipeline pulsecompressor developed under Air Force sponsorship using silicon -on -sapphire LSI technology. This processor performs 60 multiplications and78 additions every 0.1 microseconds, for an effective computing speed of600,000,000 multiplications plus 780,000,000 additions per second!

INSTRUCTIONSET 1,

DELAY 1

EVENT 2,EVENT 1 -11'

OPERATION

INSTRUCTION INSTRUCTIONSET 12 SET 13

DELAY 2

1

OPERATIONII

TIME DELAY

N

DELAYS

DELAY 3

H OPERATION H. RESULTS 2RESULT 1

I I, 1

THRUPUT ISDEFINED BYMINIMUM TIMEBETWEENRESULTS

Fig. 1Pipeline processing achieves high data throughput at the expense of time delay.

INPUTSIGNAL

a

INPUT 0BUFFER )

DER AMPINGIN WEIGHTINGMEMORYITTL PROMSI

COMPLEXMULTIPLIER

6 STAGEPIPELINEFF T

0

PHASECORRECTION6 WEIGHTINGail PROW

REORDERMEMORY

.40

6 STAGEPIPELINEFFT 1

CONTROL SVSTEM,TTL PROMS

SOS COMPONENTS

0 9 9BIT MULTIPLIER

RANGEORDERED OUTPUT

)

JOUTPUT g0

2,02CUSTOM

0 6 BIT ADDER SOS

0 DUAL B BIT SHIFTER CUSTOM ICS

RETIMING REGISTER

U FLOATING POINT LOGICVARIABLE SHIFT REGISTER (UNIVERSAL ARRA VI

3 256.1 SOS RAM

AMPLI UDECORR CT IONITTLI

Fig. 2High-speed pipeline processor performs 60 multiplications and 78 additions every100 ns.

30

r technology persctivc

I

Distributed processingwith microcomputersThe block diagram at the right illustrates one architecturalapproach to the design of a distributed microcomputersystem for multifunction radar control. The controlprocessor acts as a control distributor for radar data. Radarreturn data from the signal processor is correlated withtrack files residing in the memories of the distributedmicrocomputers. This is accomplished by the centralmicrocomputer via direct memory access to the distributedmicrocomputer memories. After correlating old and newdata, the control microcomputer provides new data to theappropriate microcomputers, and assigns any new tracksto one or more of the distributed microcomputers, witheach microcomputer executing its tracking algorithmsindependently. Other functions such as scheduling,coordinate conversion and search -pattern generation arealso updated concurrently. Depending on the particularradar design, all data-processing functions can be dis-tributed in this manner, or the distributed microcomputerscan be used to decrease the processing load of a cen-tralized data processor.

GYRO DATASCHEDULING SEARCH COORDINATEPROCESSOR PATTERN

GENERATIONCONVERSION

TRACKCORRELATIONRETURN AND CONTROL BUSOUTPUTFORMATTING

TRACKPROCESSOR

TRACKPROCESSOR2

DOUBLE BUFFERED MEMORY

SIGNAL PROCESSOR

TRACKPROCESSOR3

In this approach to distributed microcomputing, thecontrol processor acts as a control distributor for radardata.

High -density electronic packagingThe increased speed and complexity of radar signal anddata processing has been matched by major advances indigital logic circuits and packaging.

Following World War II, the invention of transistorsprovided active logic elements requiring two orders ofmagnitude less volume than tubes. Logic circuitryshrunk even more in size as numbers of interconnectedtransistors were placed on the same chip by a processknown then as integration-today as small-scale in-tegration. Over the years, we have passed through thedevelopment of medium -scale and large-scale integra-tion and are now entering the era of very -large-scaleintegration (VLSI), with 10,000 to 100,000 activeelements interconnected on a single chip-a furtherincrease in packing density of 4-5 orders of magnitude.

The photograph at the right shows a multi -layer thick -film ceramic substrate interconnecting 16 chips on eachside of a board that is approximately 3" by 4". Using VLSItechnology, this packaging approach will provide circuitdensities exceeding 80,000 gates per square inch.

Mirror photograph of thick -film substrateinterconnecting 16 chips on each side of theboard. VLSI packaging densities will exceed80,000 gates per square inch.

31

Radar processing architecturesW.W. Weinstock

Contemporary radar systems use a widerange of minicomputers, microcomputers,programmable signal processors, andspecial purpose logic for flexible andreliable operations at reasonable cost. Theradar system requirements actuallydetermine which processing architecture touse for a given application.

This paper examines a relatively simpleradar system that uses a wide range oftypical processing functions. This exampleillustrates the types of processing re-quirements that must be handled inmodern radars and provides the basis forevaluating the various architectural alter-natives available.

A genericradar system example

A track -while -scan (TWS) radar is a searchradar whose output data is used to developtracks.* As a system, the TWS radar (Fig.1) is relatively simple-a continuouslyrotating search beam gathers targetazimuth and range data. Targets are il-luminated on each scan, and their

*Barry Fell describes the track -while -scan radar in his tutorialpaper in this issue.

System requirements determine the processing structure;flexibility and cost drive the mechanization.

detections are used by the system to formtarget tracks. It is a good example becauseit uses a wide range of radar -processingoperations: search, detection, acquisition,multiple -target tracking.

The signal processor extracts the targetfrom noise and clutter.

To detect a target, the system must separateit from a background of noise and clutter.Noise may be due to external interferenceor the radar receiver itself. Clutter may bedue to backscattering from the earth'ssurface, or from rain or clouds. Theprimary function of the signal processor isto reject noise and clutter by suitablefiltering, so that the target will be detectedevery time the beam scans by it.

This calls for the examination of everyrange -azimuth cell where a target can bepresent. The total number of cells that mustbe processed is the number of range -azimuth cells viewed in a single scan. Forexample, a radar scanning 360° with abeamwidth of 1°, a pulse length of 1 As(about 500 ft in range), and a maximumrange of 500,000 ft, will look at 360,000cells each scan. With a representative scan

ANTENNASYSTEM

7...1-'..

RECEIVER --." SIGNAL PROCESSOR 111.. DATA PROCESSOR

EXTRACTS TARGETFROM NOISE AND

DETECTSTARGETS BASEDI

I CLUTTER ON MULTIPLEI SCAN CRITERIA

TRANSMITTER ...-.1 REJECTS USELESSINFORMATION BY

a BLANKSRESIDUAL CLUTTER

THRESHOLDDETECTION ACQUIRES ANO

WAVEFORMGENERATOR

-I a TAGS TARGETSTRACKS TARGETS

WITH POSITIONCOORDINATES

PROVIDES SYSTEMCONTROL

Lfa SYSTEM

#CONTROL i

Fig. 1Generic two-dimensional track -while -scan radar provides aconsidering typical requirements of signal and data processing.

point

TO-"'USERS/

OPERATORS

of departure for

period of 3.6 seconds, this means that100,000 cells must be processed each se-cond.

The number of targets that can be seenduring a scan depends on the traffic. Atypical value for air -search applications isseveral hundred, or (nominally) one targetper thousand cells. Consequently, the rateof information flow drops by three ordersof magnitude after targets are extracted bythreshold detection. For radars with betterrange resolution (say 0.1 As pulse length, or50 ft in range) the number of range cells canincrease by an order of magnitude but thereduction in the rate of information flowcan be four orders of magnitude or greater.

In summary then, the signal processorperforms high speed filtering operations,repeated on a range -cell -by -range -cellbasis, to extract the signal from abackground of noise and clutter. Thisfiltering is followed by amplitude detectionand thresholding. To have real-time opera-tion, the processing time must not exceedthe time extent of the signal return. For thecases just cited, this is between 0.1 and 1µs.As we shall see, an extensive sequence ofoperations may have to be performedduring this very brief period.

The data processor correlates the signalprocessor outputs.

The signal processor output is an irregularflow of target -like returns. These can in-clude occasional noise spikes and residualclutter in addition to the targets. The dataprocessor now must correlate these returnson a scan -to -scan basis to develop targettracks.

This correlation process must do severalthings. First, it must associate successivereturns from the same target under avariety of conditions. There must be noconfusion because of target maneuver orthe presence of other returns nearby. Theidentity of targets on crossing or mergingflight paths must be retained. Clutter orinterference in the vicinity of the flight pathmust not be confused with target returndata.

12

In additon to maintaining tracks, thesystem must be capable of recognizing anew target; i.e. one which is not currentlyunder track. It must be able to do thiswithout being confused by transient in-terference or environmental returns. Noiseand clutter must also be excluded as soonas possible to lighten the processing load.

Different correlation methods are requiredto reject noise and clutter. Since noise israndom, it will not correlate in position ona scan -to -scan basis. Consequently, someform of multiple -scan correlation processmust be used to establish the presence of anew target. Clutter, on the other hand, isstrongly correlated scan -to -scan. This factcan be used to map and blank out theresidues of strong returns.

In addition to performing its target -handling functions, the data processorserves as the focal point for controlling thesystem and for disseminating target data.Using its target files, the processor drivesoperator displays and provides digital datafor any higher level system processingrequired.

Radar system designers today can choose;rom a variety of processing approaches.

Many current radars employ primarilyanalog signal processing, while othersdigitize all processing functions after thethe receiver (i.e., after baseband conver-sion). Although the two approaches existside -by -side today, this paper addressesonly digital implementations since theseoffer advantages in processing capacity andflexibility-clearly making them favoredcandidates for the radar system applica-tion.

Signal processing functionsThe series of operations required to extracta target from a background of noise andclutter is shown in Fig. 2.

All of the returns reflected from the targetduring a single scan are used in making thedetection decision. For the case of the radarwith a 1° beamwidth, a 3.6 second scanperiod, and range of 80 nmi (i.e., 1 msbetween pulses), about ten returns arereceived from a single target during thescan period. Numbers like a few dozenreturns are typical. Processing starts on asingle pulse with matched filtering. Subse-quent operations involve multiple pulsereturns.

RECEIVEDSIGNAL

MATCHEDFILTERING

CLUTTERREJECTION/MTI

DOPPLERPROCESSING

ENVELOPEEXTRACTION

NON -COHERENTINTEGRATION

THRESHOLDDETECTION

TARGETREPORTGENERATION

SINGLESCANREPORTS

Fig. 2Signal processing operations involved inextracting target information from abackground of clutter and noise.

Extracting a signal from noise is a two-stepprocess that makes use of the differences inspectra.

White noise has a flat spectrum, while thesignal spectrum is dominated by that of thesingle pulse. The multiple -pulse (target)spectrum has a fine structure of narrowlines separated by the pulse -repetition fre-quency. The specific nature of these linesdepends on the time that the target isilluminated by the radar beam and by thetarget's behavior during that time. Formaximum signal-to-noise ratio, we need afilter whose transfer characteristic ismatched to the target spectrum.

This matching can be achieved bycascading two filters-a single pulse -matched filter followed by a bank ofnarrowband (doppler processing) filters tocoherently integrate a number of successivepulses from the target. The bandwidth ofthese filters is a measure of the coherentintegration time. In cases where radarinstabilities or target fluctuations producedecorrelating effects, it may not be advan-tageous to coherently integrate all thereturns from one pass. In this case, thedoppler processing integrates as many asfeasible and non -coherent integration (us-ing amplitude information only) completesthe job.

Clutter rejection also takes place in twosteps.

Since the received signal frequency shiftswith target motion, the processor rejects

targets whose frequency shift (doppler) issmall-signifying small radial velocitiestypical of clutter. This process, calledmoving target indication (or MTI), isaccomplished by comparing the phasechange between successive returns. Inpractice, this usually involves from two tofour successive samples.

The rejection characteristic of an MTI filteris limited by the number of pulsesemployed. In general, clutter may havesome frequency -shift components dueeither to internal fluctuations (such asthose caused by wind blowing the leaves oftrees) or to radar instabilities. MTI will notreject these components completely.However, the doppler processing employedfor coherent integration has a narrowbandpass which provides the second step ofclutter rejection.

Non-linear operations eliminate unneededphase and amplitude information.

The four filtering operations that were justdiscussed-matched filtering, MTI,doppler processing, and non -coherentintegration-account for the bulk of thesignal -processing burden and all involvelinear filters. Two non-linear operationsare present as well; they destroy informa-tion. Envelope detection discards phaseinformation at the point where it is nolonger useful. Threshold detection discardslow amplitude targets.

The final step in signal processing is theconsolidation of all the data into single -scan reports. This requires an estimate oftarget range and azimuth, and possiblyamplitude as well. Such estimates involvecomputing weighted averages, and thiscomputation is similar to that involved inlinear filtering. Following parameter es-timation, the data is formatted as neededfor the data processing to follow.

In summary then, signal processing in-volves a sequence of linear filtering andnon-linear operations, performed on everyrange cell of interest. These operations arebasically arithmetic-multiplication, addi-tion, and time delay. The computations aresummarized in the appendix.

Data processing functionsData processing involves operations on thesignal processor output-a succession ofsingle -scan reports of target position andamplitude. The purpose is to establishtarget tracks and to reject any remaining

33

SINGLESCAN-.REPORTS

CLUTTERBLANKING

CORRELATIONWITHTRACK FILE

TRACKINGDATA

CANDIDATEDETECTIONS

MULT -SCANDETECTIONPROCESSING

TARGETOETECTIONS

TARGET USER/TRACKING

OPERATORDATA

MODE/OPERATION SYSTEM REQUESTS

CONTROLS CONTROL FROMUSER/

OPERATOR

Fig. 3Data processing involves operations on asuccession of single scan reports from thesignal processor to provide complete targettrack information.

undesirable returns. Fig. 3 shows themultiple -scan processing needed for detec-tion, acquisition, tracking, noise andclutter rejection, and system controlprocesses.

The processing burden caused by un-desired returns is minimized if these signalsare rejected early in the processing se-quence.

Clutter blanking removes those residualclutter reports that were strong enough topass the detection threshold-even afterMTI and doppler processing. This makesuse of a radar generated "map" of theenvironment. The map is, in effect, a file ofstationary targets that appear on arepetitive basis. Correlation with the maprejects returns in regions where strongclutter will mask desired targets.

Rejection of noise, on the other hand, isbased on its non -repetitive character. Noisereports are unlikely to correlategeometrically on a scan -to -scan basis. Ac-cordingly, the data processor requires aseries of single -scan reports before declar-ing a detection, thereby rejecting noisebefore the acquisition and trackingprocesses are started.

Target data are developed from severalsingle -scan reports.

The only targets that are candidates fordetection are those not currently in anyactive file. Consequently, all single -scanreports must be correlated with the systemtrack file. A target that fails to correlatewith an active track is a candidate for themultiple -scan detection processing.

After a succession of reports has beenreceived on a new target, the system can

establish velocity information that is goodenough to localize the likely position on thenext scan. This predicted position is thenused as the standard against which the newscan data are correlated. The uncertainty inthis position is a function of the quality ofprevious observations and deviations dueto possible target maneuvers. The design ofthe tracking and prediction filter thenbecomes a compromise between thenarrow bandwidth required for good noiserejection and the wide bandwidth needed tominimize lag errors in the presence of atarget maneuver.

The uncertainty in predicted target posi-tion for the next scan is the basis forcorrelation problems in a multiple -targetenvironment. Flight paths can cross; merg-ing or splitting formations can be en-countered; targets can fly near strongclutter regions. Each case can produce aproblem of maintaining track continuity,and the problem is aggravated by thepossibility of a target maneuver at anytime.

In addition to track maintenance, thetracking function must also discontinuetracks that show a run of poor quality dataor no data at all. If no returns have beenreceived for the memory span of the trackfilter, then there is no basis for coasting andthe track should be dropped. Failure todrop poor tracks will lead to unnecessaryprocessing burdens and correlationproblems.

The system control problem is one of chang-ing waveforms and exercising options tomatch changes in the environment.

The false reporting rate, in particular, mustbe monitored and controlled so thatspurious single -scan reports do notsaturate the system. Sector adjustments ofthreshold levels and blanking may berequired to control this situation.

The data processing functions center on theuse and management of target files.

The key files hold the information ontargets under track, the clutter map, andthe list of candidate new detections. The fileprocessing operations involve a series ofcorrelations or data associations in order tomake the basic decisions:

Is the return clutter? Is it from a target already under track? Is it from a new detection which has notyet passed the multiple -scan detectioncriterion?

The number of possible branches in thislogic is in marked contrast to the signalprocessing problem involving a fixed se-quence of operations on every cell. Thesequence of data processing operations ona single -scan report depends on the type ofreport. A fairly elaborate logic is requiredto ensure that all the possibilities areconsidered.

Another major difference between dataprocessing and signal processing is in thedegree of regularity of input data.

The data processor receives an input forevery threshold detection by the signalprocessor. The time between reportsdepends on the geographic distribution oftargets and the antenna scan period. Ingeneral, the inputs will be irregularlyspaced in time. Concentrations of targetscan produce peak rates considerably abovethe average. The signal processor, on theother hand, processes one resolution cellafter the other, on a regular basis.

How it's done-implementingthe processing functionsThe basic processing requirements we havediscussed are summarized in Table I. Theenormous differences in processing loadand speed requirements, as well as incomplexity of operations, clearlydemonstrate the need for diversity insystem development.

The evolving data processing technologyhas given rise to distributed processingsystems whose architectures are tailored tothe problem at hand.

Data processing requirements are typicalof those imposed on general-purpose com-puters. In the 1950s and 1960s, dataprocessor designs were based primarily oncentralized, large-scale machines. This ap-proach, of course, has been alteredsignificantly with the advent of theminicomputer and the microcomputer,which are based on different technologies.

The impetus toward the use of smallermachines has come from a second directionas well-software development. Smallmachines mean distributed processing withdedicated software. Monolithic softwaredevelopments are giving way to modularsets, offering a promise of easing theproblems of software management andsoftware change.

The architectural challenge of distributedsystems is to partition the problem into

34

Table IComparison of processing requirements demonstrates the need for a wide varietyprocessing approaches.

of

CharacteristicRepresentative capability required

Signal processing Data Processing

Processing load

Types of processingrequired

Processing speeds

Data span

Acceptable process-ing delays

Millions of inputs/ second

Regular, well defined,arithmetically intensesequence

Tens of million ofoperations per second

Tens of milliseconds

Hundreds of inputs/ second

Variable sequences of logic,control, and arithmeticoperations

Hundreds of thousands ofof operations per second

seconds

Microseconds Tens of milliseconds

elements with simple interface and com-munication requirements. Highlyrepetitive processing involving a limitednumber of different operations can bebased in dedicated smaller machines. Thisis why operations such as "smart terminal"drivers and coordinate converters havebecome distributed elements. Themidi/ mini/ microcomputer hierarchiesthat are appearing are one aspect of dis-tributed operations.*

However, certain functions by their naturerequire a strong centralized processingsystem capability. For example, if anumber of processing functions require theuse of an extensive set of files, they shouldshare them directly. The alternative to thisis duplicating these files for each processorand communicating any changes as theyoccur. In our generic TWS system, thetrack files were used in detection, acquisi-tion, and tracking. Since they might con-tain hundreds of targets whose coordinates

'Ruch, Clapper, and Smith discuss distributed processing at themicroprocessor level elsewhere in this issue.

are continuously changing, the file usersmust be tightly tied to the point where thefiles are generated.

The world of signal processing is evolvingfrom the other direction, with highly struc-tured requirements and extremely higharithmetic content.

The magnitude of the arithmetic processingproblem is illustrated by digital matchedfiltering. A transmitted waveform can beinternally coded so that a long pulse can becompressed into a narrow pulse with goodrange resolution. In this way, high energycan be achieved in a short pulse withoutexceeding the peak power limitations of thetransmitter. Compression ratios of a fewhundred to a few thousand are common. Ifthe range resolution is on the order of ahundred feet, then independent cells will beseparated by 0.2 /As after compression. Thispresents a formidable processing burden,as shown in the Appendix.

A time -domain form of matched filteringmultiplies the received signal by a replica of

OATH ATH PROCESSOR OUTPUT

INSTRUCTIONSTREAM

a) Re-entrant processor-single data stream/single instruc-tion stream.

-

DATASTREAM

the transmitted signal and then integratesthe result.2 For the case where thewaveform coding has the form of a thou-sand subpulses, there are 1000 mul-tiplications and additions required every0.2 /As. This is an unacceptably highprocessing burden, even for the fastest ofcurrent machines. It can be reduced by twoorders of magnitude if time -domain filter-ing is employed using the Fast FourierTransform and inverse FFT. However, thearithmetic capability required is still verydemanding.'

Processing requirements drop con-siderably after matched filtering has beencompleted. However, the complete se-quence (MTI, doppler processing,envelope detection, and thresholding) mustbe accomplished every 0.2 As if theprocessor is to keep up with the input data.Present-day systems commonly requiresignal processors to achieve 100 times thethroughput of a minicomputer with only 10times the parts count.4'5 The signal process-ing implementation problem, then, is oneof balancing processing speed and flexibili-ty against cost.

Practical processors typically rely on someform of parallel architecture for high-speedoperation.

Processing architectures can be categorizedby the number of data streams and thenumber of instruction streams;6.7 fourtypical approaches are shown in Fig. 4.With the exception of the re-entrantprocessor (Fig. 4a), each type has explicitparallelism in its structure. This parallelismis central to achieving high-speed opera-tion. The array processor (4b) has beenused where data is divided on a range basisand different processors perform identical

PROCESSORNo. 2

PROCESSORNo. N

I--. OUTPUT

NSTRUCTION NSTRUCTIONSTREAM So. 2 STREAM No. N

c) Pipeline processor-single data stream/multiple instruc-tion streams.

PROCESSORNo. 1

INSTRUCTIONSTREAM No. 1

DATASTREAM-.No. 1

PROCESSORNo.1

DATASTREAM-.No. 2

PROCESSORNo. 2

DATA PROCESSORPROCESSORNo. N

DATAPROCESSORSTREAM-.

No. 1No.1

STREAM

tNo. N

INSTRUCTIONCOMMON INSTRUCTION STREAM STREAM No. I

b) Array processor-multiple data streams/single instruc-tion stream.

DATAPROCESSOR DATA PROCESSORSTREAM-.

No. 2

H OUTPUT

No.2 STREAM No. N

NSTRUCTIONSTREAM No.2

No. N tNST RUCT ION

STREAM No. N

d) Parallel processor/multiprocessor-multiple data streams/multiple instruction streams.

Fig. 4Basic processing architectures illustrate differences in approach for handling single and multiple instruction streams.

35

Table II

Parallelism, in many forms, must be integrated for high throughputs.

Form of parallelism Implementation

Arithmetic

Memory

Overlap of instructionand data cycles

Parallel controlof units which mustoperate simultaneously

Four multipliers and adders will perform an FFT,butterfly, or second -order section

Separate instruction and data memories allowparallel instructions and data accessesTwo operand memories allow simultaneousfetching of operands for signal processing computationsMultiple buses (3) support simultaneousmemory accesses

Separate logic/ control, registers and buses for instructionexecution and instruction and data accesses

Long instruction word for parallel controlMultiple address instructions (3) for data retrievaland storage of results.

processing on cells at different ranges.8 Thepipeline structure (4c) typifies most currenthardwired processors. It is also applicablewhere special microprocessors are used foreach processing function, and where thefunction is established by firmware. That iswhere processors are converted, for exam -

Walt Weinstock's early work in extendingthe area of radar target modeling broughthim international recognition through his"Weinstock cases." Later, his leadership indefense system development brought himRCA's highest engineering honor, the 1972David Sarnoff Award, and recently the IEEEFellow Award.Contact him at:Systems EngineeringMissile and Surface RadarMoorestown, N.J.Ext. PM -3487

ple, into an MTI processor, dopplerprocessor, or non -coherent integrator byvirtue of the program used in thatelement.9'1° The parallel processor is themost generalized form of parallel con-figuration.

Parallelism can also be implicit in theinternal structure of each processing ele-ment. In fact, most high-performance reen-trant signal processors are built using thetechniques given in Table II. Current signalprocessing systems use both classes ofparallelism. The specific choice for anyparticular application is a function of theproblem and the amount of flexibilitydesired in the solution.

Flexibility can be achieved in a straight-forward manner using digital techniques.

For example, in the case of pulse compres-sion discussed earlier, long codes can begenerated by simply adding more sub -pulses to the waveform. An analogmechanization would require a separatefilter for each waveform. Where flexibilityis required, the frequently higher cost ofdigital elements is offset by the multi -modeor multi -function capability of the digitalhardware.

Many levels of flexibility are possible. Atthe lowest level, a hardwired processor maybe capable of changing parameters (e.g.,bandwidth) without altering its basic con-figuration. This amounts to changingstored constants within a specific con-figuration.

On the other hand, the processing systemcan be implemented using a programmablesignal processor, which is a general-

purpose re-entrant machine tailored to thesignal -processing problem. All of the radarfunctions can be executed by sequentialoperations. Changes can be implementedby modifying the machine's program. Thisapproach offers a high degree of flexibility,especially where system requirements areexpected to evolve with time. Changes canbe accommodated without a hardwareredesign. *

But flexibility has its price. General-purpose devices must have capabilities thatare not always used in each application.Furthermore, in the case of programmablemachines of any kind, efficiency is lostwhen flexible, high -order languages areemployed. The use of machine languageyields greater efficiency but makes thesoftware more difficult to change.

A third aspect of flexibility must be con-sidered. This relates to cost as a function ofquantity. A common design with wideapplication is required to obtain a costadvantage. This commonality can beachieved with general-purposemicroprocessors as well as with program-mable systems. The high-performancemicroprocessor is well suited to the arrayand pipeline architectures that were dis-cussed earlier. Herold and Timken discuss programmable signal processors in

this issue.

For further readingSeveral papers in this issue illustrate anumber of specific implementations of thearchitectures discussed here. Theydemonstrate some of the state-of-the-artalternatives that RCA is currently in-vestigating. They also show the significanceof programmability as applied to signalprocessing.

ReferencesI. Rabiner, L.R. and Gold, R.; Theory and Application of

Digital Signal Processing (Prentice -Hall; Englewood Cliffs,N.J.; 1975) pp. 356-390.

2. Skolnik, M.; Introduction to Radar Systems (McGraw-Hill;New York, N.Y.; 1962) pp. 409-412.

3. Pezaris, S.; "A 40 Nanosecond 17x17 Multiplier," IEEETrans. on Computers, Vol. C-20, No. 4 (Apr 1971) pp. 442-447.

4. Fisher, J.R.; "Architecture and Applications of the SPS-41and SPS-81 Programmable Digital Signal Processors,"EASCON '74 Record, pp. 674-678.

5. Allen, J.; "Computer Architecture for Signal Processing,"Proc. IEEE, Vol. 63, No. 4 (Apr 1975) pp. 624-633. See p. 624specifically.

6. COMTRE Corporation (P.H. Enslow, Jr., Ed.)Multiprocessors and Parallel Processing, (John Wiley andSons; New York; 1974) pp. 77-80.

7. Cole, E.L. and Beaver, E.W.; "Matrix Controlled SignalProcessing," EASCON '74 Record, pp. 592-599.

8. Couranz, G.R. et al, "Programmable Radar Signal Process-ing Using the RAP," Parallel Processing (Springer-Verlag;Berlin; 1975) pp. 39-52

9. Veeknant, R.L. and Hughes, J.M.; "Reduction of ClassicalSonar Processors to Microprocessor Elements," EASCON'76 Record (supplementary handout)

36

AppendixSignal processing operations that remove noise and clutter fromthe radar signal

Signal processing operations are primarily arithmeticin nature, involving a number of multiplications,additions, and time delays.

Linear processing

The linear filtering operations required in signal processing can beexamined by considering the filter transfer function

H(z) = Y(z)I X(z), (1)

where X(z) is the transform of the input signal and Y(z) is thetransform of the output. The general transfer function of anysystem that can be described by linear difference equations withreal coefficients (and this covers all the sampled data cases ofinterest) is of the form

Y(z) bo + f + + bNfNH(z)= -= (2)X(z) 1 + alf1 + a2f2 + + amfm

The factor is associated with the delay between successivesamples; the a's and b's are real coefficients. By cross -multiplyingand taking the inverse transform, this becomes

y(n) = box(n)+ bix(n-1)+ + bNx(n-M)(3)

-aly(n-1) -a2y(n-2) - -amy(n-M).

Thus the current output y(n) is the weighted sum of the presentinput x(n) and all past inputs to x(n-M), together with pastoutputs y(n-1) to y(n-M). The general filter is recursive since theoutput feeds back into the input. This produces an infinite -duration response to any input. For the special case where thedenominator of Eq. 2 is a constant, Eq. 3 reduces to a finite-duration impulsive response.

The mechanization of Eq. 3 requires a series of multiplications,additions, and time delays. A wide variety of implementations isavailable; the choice depends in part on practicalconsiderations-the quantization effects associated with finiteword length and the computational efficiency of the candidatearchitecture.

The computations can be decomposed into a series of simplercomputations. Since the denominator in Eq. 2 is a polynomialwith real coefficients, its roots are either real or complexconjugates. This means that Eq. 2 can be represented as theproduct of transfer functions, none of which is greater thansecond order. Thus, instead of computing Eq. 3 directly, a cascadeof second -order operations can be used. This suggestsmechanizations which sequentially employ simple second -ordersections of the form:

y(n) = Ay(n-1) + By(n-2) + Cx(n-2) + Dx(n-1) + x(n). (4)

Therefore, the capacity to do four real multiplications and fourreal additions per step is the basic capability required for time -domain filtering.

Eq. 1 suggests the frequency domain implementation of the linearfilter. It requires that X(z), the Fourier transform of the input

signal, be computed and multiplied by H(z), the filter transferfunction. The inverse transform of the product H(z) X(z) thengives the filtered output y(n).

The use of the fast Fourier transform (FFT) algorithm allows thediscrete Fourier transform and its inverse to be computedefficiently.' The computational burden of the frequency domainapproach can be considerably less than the time domainequivalent, even though it involves both a transform and aninverse transform.

The FFT algorithm can be done in steps, by cascading com-putations of the following type

iu(m+1) = u(m) + W'v(m)

v(m+1) = u(m) - Wiv(m)

where u(m), v(m) are complex representations of the inputs to thestage; u(m+l), v(m+l) are the outputs; and W' is a fixed complexmultiplication factor. Complex operations are performed byoperating on the real and imaginary parts of the quantity ofinterest. Complex multiplication is equivalent to four realmultiplications and two additions. Thus Eq. 5 calls for the samecapability needed to do the time domain filtering in Eq. 4.Therefore, linear filtering requires the capability to do foursimultaneous multiplications and additions, regardless of theapproach.

(5)

Non-linear processing

The non-linear operations required in signal processing areenvelope detection and threshold detection.

Envelope detection removes the phase information from thesignal amplitude returns. Prior to envelope detection, the signal iscoherent and is represented by two quadrature quantities-an in-phase component, I, and a quadrature component, Q. These aregenerated by phase detecting the received signals against areference oscillator and a 90° phase -shifted version of thisreference. The components (I,Q) are the elements of the complexsignal discussed previously. Amplitude detection requiresthe computation of (/2 + Q2)1/.2An exact value of amplitude is notrequired for the threshold detection that follows. The use of anymonotonic function, such as (/2 + Q2), will allow for rejection ofsignals below the critical amplitude level.

Threshold detection involves the simple comparison of theintegrated amplitude in a cell with the critical threshold value. Ifthe threshold is exceeded, the return has passed the single -scandetection criterion and is sent on for further processing. If thethreshold is not exceeded, the return is irreversibly discarded.Thresholding is the primary branching operation in signalprocessing.

This brief summary confirms that signal processing operationsare primarily arithmetic in nature, involving a number ofmultiplications, additions, and time delays. Although theseoperations are relatively straightforward, the number ofoperations required, in microseconds of time, makes high-speedarithmetic capability an absolute requirement for real-timeprocessing.

Reprint RE-23-5-51Final manuscript received January 9. 1978. 37

Solid state for super -power radars

The ideal high-speed switch for multi -megawatt pulsed radarmodulators may be a thyristor.

"Dewey" Pruitt joined RCA in 1949, and hasbeen with the RCA Radar Transmitter groupat Moorestown since 1953. He has con-tributed to the design of pulsed radartransmitters for many projects, includingparticularly the AN/UPS-1, AN/FPT-2,TRADEX, and AN/FPS-95 radar systems.For the past several years, he has been theprincipal engineer in a series of projectsaimed at developing high -power solid-statepulse switches.Contact him at:Equipment Design andDevelopment EngineeringMissile and Surface RadarMoorestown, N.J.Ext. PM -2153

One trend seems inevitable in the continu-ing radar evolution-a constant demandfor more power. In particular, pulsed radar(and laser) systems are already pushingmodulator output pulse power levels ofseveral billion watts peak-at relativelyhigh (up to 0.5%) duty cycles.* Typicalmodulator requirements are forpulsewidths of 10 As, with 2µs risetime, at200 to 300 pulses/ s. Continuous or in-termittent duty may be required,depending on the application.

The heart of any high -power pulsemodulator is the pulse -switching device.

Hydrogen-thyratron, spark -gap, andmercury -pool devices-these have been thetraditional choices for handling the powerlevels required. But their size and weightare usually as impressive as their power -handling capacity.

The ideal switch, of course, would be smalland lightweight, with long life and highreliability-features that suggest a solid-state approach. This paper describes thefive-year history of RCA's investigations ofsolid-state devices for high -power pulseswitching. Several developmental switchesare described, with performance ratingsthat show very real potential as practicalswitches for high -power pulse modulators.

Background of development

In a line -type pulse modulator, energy istransferred from a high voltage powersupply to a "pulse -forming -network"(PFN) capacitance, via a charging induc-tance and a charging switch. This chargetransfer takes place over a relatively longtime called the interpulse period. The pulseswitch is then triggered to dischargethis energy into the load, producing a short,high -peak -power load pulse. The PFNconfiguration causes the energy to bedelivered to the load as a rectangular pulse.

Duty cycle is the ratio of pulsewidth to interpulse period.

D.L. Pruitt

After a short interval to allow the pulseswitch to recover voltage blocking ability,the charging switch is triggered to initiateanother cycle.

Power transistors generally make poorswitches in super -power short -pulsemodulators, mainly because the peakcapability of a power transistor is littlehigher than its dc capability. Thischaracteristic leads to an inefficient use ofsilicon, resulting in a large and un-reasonably expensive switch. Also,transistors are more susceptible to damagefrom fault current than are thyristors.

Power thyristors tend to be limited by rmscurrent rather than peak current, and arethus more appropriate for design of arelatively small, low-cost pulse switch.Power thyristors are limited in pulse -current risetime because of slow spread ofcurrent across the silicon chip after the gateis triggered. This characteristic (slow rateof current spreading) can cause localizedchip overheat for rapidly rising currentpulses-the well known dil dt limit. Largehigh -power thyristers are more susceptible.to di dt limitations than are small, low -power thyristors. Several interdigitatedand regenerative gate designs have beendeveloped to enhance high frequency per-formance. However, at the present state ofthe art, it appears that relatively small chips(up to 0.1 in. diameter and 35 A rms rating)have a definite performance, edge overlarger devices for pulses with risetimes upto 2µs.

Device selectionDuring the early study and investigation,arrangements were made with the RCASolid State Division to make pulse dissipa-tion tests on several thyristor-devicefamilies: RCA S3700 M, RCA S2600M,and 2N3899. From these tests, tentativeratings were derived, thus allowing costand size tradeoffs to be made in order toselect the best device for further work. Thetests were made using 20 -ms pulses, with

38

ANODE PLATE

CATHODE TERMINAL

Fig. 1Basic SCR switch "module" with 10 thyristors paralleled on a commonanode plate. (Dotted lines indicate that the SCR circuit is repeated tenrimes.)

risetime up to 2µs, and a low repetitionrate.

The resulting data showed that one 2N3899can replace four S3700Ms or threeS2600Ms. The significantly higher partscount with the smaller devices seemedcertain to make a S3700M or a S2600Mmodulator more expensive than a 2N3899modulator. Further, a 2N3899 modulatorwould be smaller in overall size. Therefore,the 2N3899 (stud mount)/ 2N3873 (pressfit) family was selected for further work.

Developing the switchThe switch was fabricated in two steps:

1) A "module" was constructed byparalleling a number of individualdevices on a common anode plate.

2) A number of "modules" were con-nected in series to obtain the desiredvoltage capability.

A ten-thyristor module yielded 1.8 mega-watts of peak power.

After an evolutionary period of manymonths, a module using ten type -2N3873thyristors pressed into a common anodeheat sink was developed. (Ten is anarbitrary number; any convenient numberup to probably several dozen may beparalleled.) As shown in Fig. 1, a pilotthyristor (SCR1) provides gate drive to theten main thyristors by switching the voltageon Cl onto the common gate bus. Cl ischarged to the module anode -to -cathodevoltage. Gate -isolating resistors R2

. I

*MorZoirotigZo)4044)

71 8.

Fig. 2Parallel SCR switch, showing press -fit thyristors on a water-cooled aluminum plate.

through R11 ensure a proper overdrive (2A peak at the nominal 500 V operatinglevel) into each of the main thyristor gates.Diodes D2 through D 1 1 provide furthergate isolation, preventing feedback from aleaky or shorted thyristor to the gate bus.Inductors LI through LIO are straight No.18 bus wires, each approximately 4 cmlong, which connect the ten main cathodesto the next anode plate for a series stack.Fuses F I through F10 provide automaticdisconnect of an individual failed thyristor.

Pilot thyristor SCR1, an RCA S2600M, isnormally triggered by a nominal 1-A(peak), 5 -us pulse from currenttransformer Tl. The secondary of T1 has50 turns of magnet wire on a small (CF111-Q1) ferrite toroid. In a series stack ofmodules, a 50-A (peak) primary pulse isconducted through a single -turn primarywhich links all of the toroids in series. Theprimary turn is a silicon rubber -insulatedhigh -voltage wire.

Early versions of the module were con-structed using the stud -mounted 2N3899thyristors on a water-cooled copper plate.Later versions were constructed usingpress -fit 2N3873 thyristors on a water-cooled aluminum plate. Fig. 2 shows thislatter version.

These switch modules were extensivelytested in a conventional artificial line typemodulator circuit at 6000 A peak. Theyyielded 1.8 -MW peak power and 9 kWaverage power.

Using RCA2N3873 thyristors selected atrandom from a lot of 1000 units containing

two different date codes, current sharingamong ten paralleled thyristors typicallyfell within ±20% of the average value. Twocontinuous runs of 8 hrs each were made atthe given conditions. In destructive tests,the devices exhibited remarkabletoughness: a t least four, and sometimes fiveof the ten devices had to be removed (byclipping the cathode leads) before failureoccurred among the remainder in a 5 -minute test. Typically, soft solder melted atthe cathode connection before the devicefailed in these destructive tests.

Twenty modules were connected in series toproduce a 30 -megawatt switch.

On the basis of the encouraging earlyresults, the Air Force in 1974 awardedRCA a contract to construct and test a 30 -megawatt switch.

Twenty modules were constructed, usingten type 2N3873 press -fit thyristors inparallel in each module; the 20 units wereconnected in series to form a 10 -kV peak -voltage, 6 -kA peak -current switch (Fig. 3).A modulator was constructed to test thisswitch to 50 MW peak power and 150 kWaverage power.

This switch has been operated for shortperiods (up to 5 minutes) at the full designpower levels. (The thermal constant of thethyristor switch is short compared to 5minutes.) With water cooling provided, theswitch heat sink temperature rise was onlya few degrees Centigrade. Longeroperational periods were precluded by thedanger of overheating test set components,

39

11111M1 maw!

...-!.2jurnyniratir!nnYrnr4141,..ire.

F-71247 3 Sr

41,11.% or.n.1,1,,,P111WILUATr MrPly P, MR. WyWiraativ.417

.-2,-)",t.Tomniar,..or

e'vss-per.'213C6Mr=glwr-11

Fig. 330 -megawatt switch shows 20 modules inbreadboard configuration used to verifyhigh -power switching performance.

1111 1 1 1 1 -1 11 1 I 11

SWITCH CURRENT: 1 kA/DIVISIONSWITCH VOLTAGE: 5 kV/DIVISIONHORIZONTAL: 2 As/DIVISION

s'141411VISAVIWAYNmii.IF4..1.134141Pn-an.Wi 1.6411

Trilli."VirdreaCtiL.seif

AftifiVi'reiVi r...6.1T

"ff'WV' 4 e1wwl!

1:4,,Vi;',..:4,1r4r010;aortitZ1

1:"

1,044

Fig. 4Extension of switching voltage capabilityfrom 10 to 15 kV was demonstrated in thisdevelopmental switch.

H 1 I I 1 I I 1 III 1111

111111113w*m19:01.1.

EXPANDED TIME SCALE:

SWITCH CURRENT:SWITCH VOLTAGE:HORIZONTAL:

1 kA/DIVISION5 kV/DIVISION0.5µs/DIVISION

Fig. 515 -kV switch current and voltage waveforms show instantaneous switching capability onnormal 2 microsecond time scale (left) and on an expanded 0.5 microsecond time scale(right).

particularly the pulse -forming -networkcapacitors.

Levels achieved were:

Peak switch voltagePeak switch currentPeak load powerAverage load powerPulsewidthCurrent risetime

(10% to 90%)Repetition rate

10 kV7000 A32 MW160 kW20 As

1.7 us

250 pulses/ s

Adding ten more modules produced anadditional 15 megawatts of peak power.

Under a contract extension from the AirForce, the 10 -kV, 30 -MW switch described

above was extended to a 15 -kV switch byadding 10 additional modules in series for atotal of 30 series modules. (See Fig. 4.)

Because of test -set limitations, this 15 -kVswitch could not be tested to its fullinherent capabilities of 45 -MW peak and225 -kW average power. Therefore, theswitch was tested for the followingparameters:

PulsewidthCurrent risetimePeak PFN votagePeak load currentPeak load powerRepetition rateAverage load power

10 pts0.8 As15 kV4700 A30.9 MW285 pulses/ s88.1 kW

The switch was repeatedly snapped on andoff at full power, demonstrating instant fullpower availability without warmup. Con-tinuous runs were again limited to about 5minutes by component (PFN-capacitor)heating. Fig. 5 shows switch current andvoltage waveforms during the pulse.

Hybrid circuitthyristor switchEarly in this program, the use of hybridcircuit techniques was recognized as offer-ing a potential dramatic reduction in sizeand weight for the parallel SCR switchmodule. Glass-passivated chips of an ap-propriate size became available in 1975(from Unitrode Corporation), and a hybridcircuit development program was started in1976.

A 40-A (rms rating) 600-V chip (R044060)was chosen; this chip is 5 -mm (0.2 in.)square. The concept involves attaching, byreflow soldering, 20 main switch SCR chipsin a ten parallel/ two series configuration,plus two trigger (or pilot) SCR chips, on aberyllia substrate. Beryllia was chosen forits excellent thermal properties-high con-ductivity and high specific heat.

circuit (Fig. 6) is essentiallytwo series circuits, much like the configura-tion shown in Fig. 1, but with an importantsimplification. The electrical isolation ofthe SCR anodes, provided by the berylliaheat sink, allows placement of the fuses (F1to F20) in the anode circuits; this, in turn,eliminates the requirement for gate -

isolating diodes (D2 to D11 in Fig. 1).

Otherwise, circuit operation is exactly asdescribed earlier.

Several trial layouts were generated anddiscarded in arriving at the hybrid circuitlayout. Because of the high rms currentsinvolved (40 A rms/ chip), the SCR chipsare not soldered directly to the substratemetallization, but rather to copper contactpads which are in turn reflow soldered tothe metallization pattern. The printed -circuit resistors (gate resistors and bleederresistors) add negligible weight to thehybrid circuit. The trigger transformers(T1 and T2) and the circuit capacitors (C1and C2) contribute significant (but notmajor) weight to the hybrid module.

The completed module weighs 190 gramswithout cooling fins, and 235 grams withcooling fins attached. Fig. 7 is aphotograph of a completed hybrid -circuitthyristor-switch module.

40

The performance objective, as an aircooled switch, was as follows:

Peak voltagePeak currentPulsewidthCurrent risetimeRepetition rateDuty cyclePeak powerAverage power

1 kV3 kA10 As

1 As100 pulses/ s0.0011.5 MW1.5 kW

Ten hybrid SCR modules were connectedin series to obtain a 10 kV (nominal) air-cooled switch as shown in Fig. 8, and theswitch was installed in the test modulator.In initial testing, some difficulty was ex-perienced with module voltage sharing,resulting in loss of devices. After repairs,maximum voltage was limited, resulting inthe following maximum operation:

Peak voltage 8.9 kVPeak current 3 kAPulsewidth 10 psCurrent risetime 1µsRepetition rate 100 pulses/ sDuty cycle 0.001Load resistance 1.2 Cl

Peak load power 10.8 MWAverage load power 10.8 kW

ConclusionsThe work performed on thyristor switchinghas shown that series/parallel arrange-ments of relatively small thyristors can beused effectively in high -power artificialline -type pulse modulators. A 30 -MW(peak) 150 -kW (average) power modulatorwas demonstrated using packaged 2N3873devices with water cooled heat sinks. Alightweight, compact, air-cooled hybridSCR switch attained 10 MW peak and 10kW average in an artificial line -typemodulator. These results show that solid-state switching has arrived as a practicalreality for advanced super -power pulsedmodulator applications.

AcknowledgmentsPart of the material contained in this paperwas published in the Record of 1976Twelfth Modulator Symposium sponsoredby IEEE. Significant portions of this solid-state modulator development work weresponsored by the Air Force Systems Com-mand's Rome Air Development Center,Griffiss AFB, N.Y. The following RCA

ANODE BUS

CATHODE BUS

Fig. 6Hybrid SCR chip switch circuit provides 20 switching chips without the requirement forgate -isolating diodes. (Dotted lines indicate that the SCR circuit is repeated ten times.)

RCAFig. 7Hybrid switch module in completed form shows compactness attainable in final design.

Fig 8Air-cooled hybrid switch incorporates ten SCR modules in series for high -voltage testing.

personnel made significant contributions:E.O. Johnson, RCA Laboratories; E.M.Leyton (deceased) formerly of the RCALaboratories; and G. Albrecht, formerly ofthe RCA Solid State Division.

BibliographyI. Gime, G.N., and Lebacqz, J.V.; Pulse Generators, Vol. 5,

MIT Radiation Laboratory Series (New York, McGraw-HillBook Company, 1948).

2. Pruitt, D.L.; "Multi -Megawatt Solid State Switch," IEEEConf. Record of 1976 Twelfth Modulator Symposium (Feb1976) pp. 62-66.

3. Pruitt, D.L.; "Thyristor Switches for Super Power Intermit-tent Duty Operation," Pulsed Power Systems WorkshopProc., Silver Spring, Md (Sep 1976).

4. RADC TR-75-73, "Multi -Megawatt Solid State Switch,"Final Technical Report (Mar 1975) and Supplement (Sep1975), RCA Government and Commercial Systems.

5. RCA 6T0062A I., "Multiple Chip SCR," Final TechnicalReport for RADC Contract F30602 -76-C-0197 (June 1977).

Reprint RE -23-5-10Final manuscript received January 19, 1978.

41

Pulsed GaAs FET microwave power amplifiersfor phased -array radars

R.L. CamisaJ. GoelH.J. WolksteinR.L. Ernst

The successful realization of I -band (8-10GHz) airborne phased -array radarsdepends upon the development of high-performance economical power -amplifiermodules. IMPATT diode amplifiers andtransistor amplifier -multiplier approacheshave been unsuccessfully tried at thesefrequencies. At lower frequencies, three -terminal silicon bipolar transistors are usedexclusively. Recent advancements in GaAstechnology indicate that efficient poweramplification is possible, with manyresearchers1'2 achieving output power in

This paper summarizes the performance ofa 9-10 GHz, 0.5-W pulsed GaAs FETamplifier developed as part of an Air Forceexploratory technology program.' Theamplifier reported here is considered adriver for higher -power stages yet to bedeveloped, with an eventual output powergoal of 5 W. In array -radar applications,the amplifiers must be operated in a pulsemode such that the devices are activatedonly when a radar pulse is beingtransmitted.

Fig. 1

Typical RCA GaAs FET chip before flip -chipmounting.

Phased arrays require pulsed operation at high frequencies,but at the beginning of this project, nothing had beenpublished on the pulse characteristics of GaAs FETamplifiers.

At the inception of this program, there wasno published prior art concerning the pulsecharacteristics of GaAs FET amplifiers.Therefore, the major goal of this effort wasto compare the cw and pulse performanceof microwave amplifiers using thesedevices. The evaluation of phase variationswithin a pulse was of particular interest forthis phased -array application, since lowphase variations make amplifier -to -amplifier tracking problems less difficult.

The distinctive features of RCA GaAsFETs used in this amplifier program arebriefly summarized below. Flip -chipbonding procedures, first applied to GaAspower FETs by RCA, presently differ-entiate our approach from others in thefield and will be described. The amplifierdesign approach discussed here emphasizespulse techniques and associated tradeoffs.

Device technologyFlip -chip devices have distinct advantages.

The distinctive features of RCA's GaAsFETs used in this amplifier program are: a)

Fig. 2"Flipped" device mounted on a gold-platedcopper carrier. This technique providesgood heatsinking and low source induc-tance.

flip -chip bonding, for its heatsinkingqualities and reduced source inductance; b)self -aligned gate processing, for itssimplicity; and c) multiple -layer epitaxy,for its non -degrading ohmic contacts. Thedetails of the device processing have beenpreviously published1.4 and only thehighlights will be briefly reviewed here. Ourfabrication processes use conventionalphotolithographic techniques and avoiddifficult alignment problems. Modernmicrofabrication techniques such as ionbeam milling are used, avoiding under-cutting and resulting in an almost 1:1 ratioof photoresist pattern to actual devicegeometry.

Flip -chip bonding of devices allowsoptimal heatsinking of units in a common -source configuration, while at the sametime optimizing gain by reducing sourceinductance to a minimum. This technologyuniquely lends itself to high-performance,reliable devices which can ultimately beadapted to large-scale production. Fig. 1

shows a typical RCA GaAs FET chip withplated source posts. Fig. 2 is a photographof a "flipped" device mounted on a gold-plated copper carrier. In the flip -chipbonding, ribbons or bond wires are at-tached on gate and drain pads and thedevice is then flipped down onto a coppercarrier, thereby contacting all the sourcesat the same time. The gate and drainconnections are then tacked down on aceramic ring having metallization pads.Inductance is minimized by grounding allsources directly, without the use ofwirebonds. The copper pedestals extractheat directly away from the surface of thedevice, where the heat is generated. Theflip -chip technology appears difficult, butthis technique has proved very practicalwith the advent of commercially -availableflip -chip bonding machines.

42

Amplifier developmentThe overall multistage amplifier wasdesigned in modular blocks.

This design allowed individual sections tobe separately optimized and cascaded easi-ly with other stages. A modular construc-tion is also desirable because it is

tractable-if an individual amplifier fails, itcan be easily located, reworked, or replacedby an equivalent unit. The most commontype of modular design is the quadrature-coupled balanced amplifier.' In this con-struction, each amplifier stage requires twoquadrature couplers and two single -endedcircuits. The main disadvantage of an all -

balanced approach is that two devices per

Reprint RE -23-5-9Final manuscript received December 21, 1977.

Ray Camisa was project engineer for theGaAs amplifier described here. His work attie Microwave Technology Center has alsoincluded GaAs FET devices and linearamplifiers. He has also published papers onlow -noise parametric amplifiers, microwaveintegrated circuits, and MIS varactors.Contact him at:Microwave Technology CenterRCA LaboratoriesPrinceton, N.J.Ext. 3136

Jitendra Goel is presently working cn GaAsMESFETs and amplifier circuit design. Hiswork has included the first X -band amplifierwith an output above 1 W and the first dual-cate high -gain octave -bandwidth (4-8 GHz)amplifier.Contact him at:Microwave Technology CenterRCA LaboratoriesPrinceton, N.J.Ext. 3166

Why do we need high -frequency pulsed amplifiers?Airborne radars are requiring higher and higher resolution, and high -frequency operation (up to 12 GHz) can provide it Higher frequencies alsomean smaller antennas and less aircraft weight. However, standardamplifiers do not work well at such high frequencies, and GaAs MESFETamplifiers seem to fill this gap.

It is also efficient to have an airborne radar work in both the air-to-air andground -to -ground modes. (The driving force here is for military tacticalaircraft, but future commercial applications saem possible-collisionavoidance with the air and ground are both important.) Thus, the radar mustbe able to "look" in a number of different directions in a very short time.Phased -array radars do this well, but they pulse beams in differentdirections by combining a large number of different -phased beams. Byvarying the individulal phases, the combined beams will cancel each otherout in different directions, resulting in a beam that can change its directionquite rapidly. The requirement for rapid pulsing is thus passed on to theamplifier.

Herb Wolkstein is now concentrating on thedevelopment and applications of solid-statemicrowave devices. He was previouslyManager of Advanced Programs andApplication Engineering for RCA'smicrowave activity, where he made manycontributions to TWT and solid-statetechnology.Contact him at:Microwave Technology CenterRCA LaboratoriesPrinceton, N.J.Ext. 2710

Bob Ernst has developed and supervised theconstruction of hundreds of microwave in-tegrated circuit modules operating at fre-quencies ranging from 70 MHz to 40 GHz.He is now working on power -combiningtechniques and the pulsed -microwavesystem described here.Contact him at:Microwave Technology CenterRCA LaboratoriesPrinceton, N.J.Ext. 2510

Authors Wol (stein, Goel, Ernst, and Camisa,along with the microwave amplifierdescribed in this article.

4.3

amplifier stage are required. An alternateapproach uses directly cascaded amplifierswithout couplers. The disadvantage of thisamplifier type is that individual amplifierstages interact with each other, whichmakes tuning the complete amplifierdifficult. The advantage of the directlycascaded amplifier, however, is that aminimum number of devices are used.

The final version of the driver amplifierused a mixed single -ended and balanceddesign as a compromise between thenumber of devices used, alignment difficul-ty, and overall volume. This was done byseparating the amplifier into three distinctmodules. The two output modules arebalanced stages. The input module is acirculator -coupled, three -stage single -

ended amplifier. Fig. 3 is a block diagranof the overall amplifier showing stage -by -stage performance. At band center, theoutput power of the overall amplifier was500 mW with an associated gain of 30 dB.The small -signal gain was 33 dB and thenoise figure was 12.5 dB.

The amplifier video circuitry design and rf-pulse operation differentiate this amplifierfrom all previous FET amplifier designs.

In an airborne -array application, in orderto conserve power, the FET amplifiersshould be off when no rf pulses are beingtransmitted. Two methods of pulsing theamplifiers were considered: pulsed -gateand pulsed -drain.

In a pulsed -gate configuration, the device iscut off by putting a reference voltage largerthan the pinchoff voltage on the device. Inorder to turn the device on, a positive -goingpulse decreases the effective negative gate -to -source voltage such that the FET candraw its normal operating current. In thesecond approach, the drain voltage is

pulsed from 0 V to the normal operatingdrain potential and current. The drain -biasing scheme requires a fast currentdriver and the gate -biasing scheme doesnot. However, a disadvantage of the gate -pulsing scheme is that it puts heavydemands on the maximum voltage that thedevice must tolerate without damage-thedevice breakdown voltage must be greaterthan the algebraic sum of drain voltage,gate voltage, and rf voltages. Our amplifierused the gate -biasing scheme because of itssimplicity and ultimate lower cost. Also,the amplifier was pulsed only in the last twobalanced stages, where most of the power isbeing dissipated.

STAGE I

SSGA,,, 7

(dB)

OUTPUT POWER

2 3

7 8

OVERALL NF 12.5dB

40

6

160

5

1_

TOTAL33dB

500

Fig. 3Stage -by -stage performance of the GaAs FET amplifier. Pulsed operation takes place in thelast two stages, where most of the power is dissipated.

4.-r-t

Fig. 4Experimental amplifier was produced in a microwave integrated circuit format. Note themodular approach taken and compare the modules with the blocks of Fig. 3.

The amplifier performs well, especially interms of its AM/PM conversion.

Fig. 4 is a top view of the amplifierdelivered to the Air Force AvionicsLaboratory. The overall package is ap-proximately 15X6X2.9 cm. The input andoutput connectors are the hermeticallysealable type with excellent rf performancethrough 18 GHz. The amplifier requirestwo dc inputs (Vv = 8 V, V,ef = -4 V) andone video pulse (0 to -4 V). With -4 V, thelast two stages are cut off, and with 0 V, theamplifier is turned on. The amplifierpackage is divided into rf and dc (or video)compartments.

The amplifier rf circuits were fabricated ina microwave integrated circuit (MIC) for-mat. All circuits were made on 0.635 -mm(0.025 -in.) A1203 microstrip transmissionlines with chrome -copper -gold metalliza-tion. The substrate thickness is dictatedmainly by the type of coupler used in the

balanced design. Actually, at I -band, athinner substrate is preferred to minimizedispersion and radiation effects.

Table I summarizes the design goals of theamplifier and the experimental results; Fig.5 shows the amplifier in pulsed operation.The amplifier met all room -temperaturedesign goals except power -added efficien-cy. The poor efficiency resulted from hav-ing to select devices with high breakdownvoltages so that the FETs could be pulsed.This should not be a fundamental problemfor the reasons stated in the article's conclu-sion. The AM/ PM (amplitude -modulation/ phase -modulation) conver-sion performance of this amplifier wasexcellent. At frequencies within the desiredband, the AM/ PM conversion was so lowthat it was hardly measurable. With furtheroptimization, AM/ PM could be furtherreduced over the entire band to ap-proximately 2° / dB.

44

ConclusionsA five -stage GaAs FET amplifier with 29.7±0.4 dB gain at 500 mW output power overthe 9-10 GHz band was designed,fabricated and tested. Extensivecharacterization of the amplifier perfor-mance was carried out under cw, pulsed-rf,and pulsed-rf pulsed -bias conditions toassess its suitability for airborne phased -array applications. The amplifier perfor-mance under cw and pulsed conditions wasalmost identical. To obtain pulsedamplifier stages not sensitive to the dutycycle, the gate of each FET was pulsed frompinchoff to its operating potential. Thistechnique eliminated the need for fastcurrent drivers, but put stringent con-straints on the dc characteristics of the rf

- devices. If this gate -biasing technique is tobe used, all the rf devices must have similarpinchoff values and their breakdownvoltages must exceed the algebraic sum ofpinchoff voltage, applied drain voltage,and the total rf voltage swing at maximumpower output. In the limited time availablefor developing this amplifier, it wasdifficult to meet these dc requirements andsimultaneously obtain good rf perfor-mance. This problem is not a fundamentalone, but it underscores the need for furtheroptimization of circuit and/or FETgeometries specifically designed for pulsed-rf applications.

AcknowledgmentsThe authors gratefully acknowledge thesupport of the Air Force AvionicsLaboratory, Wright -Patterson Air ForceBase, for their encouragement and supportof this work. All power devices used in thisprogram were supplied as part of anotherX -band power FET program. Dr. H.Huang was the project scientist for thedevice -development effort. The coopera-tion of W. Reichert, P. Pelka and J.Klatskin in handling the devices is grateful-ly acknowledged. All the FET amplifiercircuits were assembled and optimized byM. Kunz.

References

I. Drukier, I.; Camisa, R.L.; Jolly, S.T.; Huang, H.C.; andNaryan, S.Y.; "Medium -power GaAs field-effect transistors,"Electronics Letters, Vol. II, No. 5 (Mar 6, 1975).

2. Camisa, R.L.; Goel, J.; and Drukier, I.; "GaAs MESFETlinear power -amplifier stage giving 1 W," Electronics Letters,Vol. I1, No. 24 (Nov 27, 1975) pp. 104-105.

3. Air Force Avionics Laboratory, Contract F33615 -76 -C -I122.

4. Napoli, L.S.; Hughes, J.J.; Reichert, W.; and Jolly, S.; "GaAsFET for high -current power amplifiers at microwave fre-quencies," RCA Review, 1973, 24, pp. 608-615.

5. Englebrecht, R.S. and Kurokawa, K.A.; wideband lownoise L -band balanced transistor amplifier," Proc. IEEE 53,236 (1965).

Table IAmplifier performance met all room -temperature design goals except power -addedefficiency. Low efficiency came from having to select devices with high breakdown voltagesso that the FETs could be pulsed.

Gain and frequencyresponse

Output power

Efficiency

PulsewidthRepetition rate

Rise/ fall time

AM/ PM* conversion

Pulse -amplitude droop

Intrapulse phase shift

Unit -unit gain/phase tracking

Design goal Experimental results

30 ± 0.3 dB over 9- to 29.7 ± 0.4 dB at10-GHz band at 500 500 mW output powermW output power

500 mW 500 mW

20% 8%

0.2 to 65 IAS Both requirements met300 to 0.3 kHz

50 ns (max) < 30 ns

3° / dB (max) 3.5° / dB (max)

5% (max) < 5%

5° (max) < 5°

0.6 dB in gain No data obtained in7° in phase program time frame

AMA PM = amplitude -modulation/ phase -modulation

-2 V

-5 V

1111.11111.11111111

40 nanoseconds/div

mountimmiimmummi41, iimmerminumIE.riarem

40 nanoseconds/div

rt pulsed on

.41- gate pulsed on

Fig. 5In pulsed -gate operation (top) a positive -going pulse decreases the effective gate -to -sourcevoltage and so allows the FET to draw its operating current and turn on. Resulting phaseimbalance (bottom) is within specifications.

45

RCA Engineer-ranks second (after discussions withassociates) as a source of technical information aboutRCA.

TREND-ranks second (after the grapevine) as a source ofnon -technical information about RCA.

RCA Libraries-rank fifth as a source of all information-rated right after: Your own files (some of which have beenaccumulated with the help of the library), the engineers inyour own group, books, and handbooks (many providedby the library).

RCA Technical Abstracts-ranks lowest of the four, partlybecause of its low visibility.

Engineering InformationSurvey results

Part 3D.E. Hutchison) J.C. Phillips F.J. Strobl

A closer look at four information sources-how important are they to you?

What value do RCA engineers place on the RCA Engineer?TREND? RCA Technical Abstracts? The RCA libraries? Therecent Engineering Information survey answered thesequestions and several others* related to these four RCA -sponsored communication channels:

How accessible are they?How are they used?What should be done to improve them?

*An earlier paper focused on the general results of the Engineering InformationSurvey.' A second paper compared the information needs and use patterns of high andlow achievers.' This paper reviews four specific technical information sourcesavailable to RCA engineers.

How accessible are they?Most engineers follow the path of least resistance.

Research on the use of information sources shows thataccessibility often determines frequency of use.v*Engineers frequently turn first to the information sourcethat is most accessible; perceived technical quality in-fluences his decision to a lesser extent. This implies thatimproving the quality of an information source may not leadto increased use of that source-unless it is accessible.

"See, for example, Allen, T.: Managing the flow of technology (MIT Press; 1977) p. 184.

46

RCAEngineer

How accessible is the RCA Engineer? Trend? RCATechnical Abstracts?

% of respondents having accessRCA

TechnicalTREND Abstracts

Do you use the RCA Engineer, TREND, and TechnicalAbstracts fo find personal contacts with experts with whomyou can discuss technical matters?

RCA

Engineer

RCA

TechnicalTREND Abstracts

Direct access 73% 67% 9%Source has provided

Indirect accessNo access

13%14%

23%10%

36%55% contacts with experts. 39% 28% 18%

Direct access means that the information is distributeddirectly to the engineer's office or home. Sources identifiedas indirect access are available through circulation orborrowing. No access means these sources are notavailable or not used.

RCA Engineer is distributed on a company -paid subscrip-tion basis and is generally sent to an engineer's home.According to the Survey, coverage varies substantially bylocation (100% in some to virtually none in others).Research and development engineers get the most com-plete coverage; manufacturing and service engineers, theleast.

TREND is sent in bulk quantities to each engineeringlocation within RCA. A distributor in the location routessufficient copies to engineering groups for distribution toeach engineer. The survey indicated that about 10% of therespondents did not receive TREND.

RCA Technical Abstracts is distributed to libraries, toengineering management, and to those who feel they have aneed for direct access. Many survey respondents (55%)have no access (do not know what RCA Technical Abstractsis or have no access to it.)

What is your access to an RCA Library?

Library at my location 77%

Remotely located library 7%

No access 16%

Let's analyze what these data tell us about the value of eachof these technical information sources.

RCA Engineer:

The above data confirmed:

-that the RCA Engineer's major value is as a source forgeneral technical information. (Tech info-other.)-that it is a valuable source of business informationabout RCA.-that in some cases it can be very valuable as a source ofjob -related information, but-that it has only moderate value as a source for industry -related business information.-that another important measure of the RCA Engineer'susefulness is its value in fostering intracompany, in-terdivisional contacts.

Reprint RE -23-5-17Final manuscript received February 28,1978.

Contact him at:Advanced WidgetDevelopmentWidget SystemsExt. 1234

Library services are available to most survey respondents;77% have direct access and 7% can take advantage of thecentral library of their major operating unit.

How are they used?-a question of value

How would you rate RCA Engineer, TREND, RCA TechnicalAbstracts, and the RCA libraries as sources for the follow-ing types of information?

RCARCA Technical

Engineer TREND Abstracts Library

Tech. info-job related 60% 42% 37% 90%Tech. info-other 90% 68% 37% 93%Business info -RCA 70% 95% N/A 82%Business info-Other 30% 47% N/A 81%Professional 61% 56% N/A 85%

This table summarizes the percent of respondents who saidthese sources are valuable, very valuable, or somewhatvaluable.

41

00 0

Thirty-nine percent of RCA Engineer readers have used the journalto locate experts on technical matters.

47

TREND

The Research and Engineering News Digest, is valued as aprimary source for business information about RCA, hasfairly high value as a source of general technical informa-tion about RCA, and is of moderate value for the remainingthree categories of information (professional, business -other, and technical -job related).

RCA Technical Abstracts

Although Technical Abstracts is valued by 37% of therespondents, this constitutes the lowest ranking of the foursources considered. One of the reasons for this low rankingis that 55% of the respondents did not have access to RCATechnical Abstracts.

Library

The libraries rate very high and maintain an image ofproviding material in all five areas of information to 81-93%of the respondents. There is little question that their valueranks highest among the sources covered here.

How much of most issues of the RCA Engineer and TRENDdo you read and scan?

RCA Engineer TREND

Read 20% 410/o

Scan 52% 46%

The RCA Engineer publishes over 500 pages a year; thus,each year the average survey respondent reads over 100pages-a considerable reading volume. An interestingfinding was that the percentage read was similar for alltypes of engineers (research, design, manufacturing) butvaried greatly with professional achievement (discussedlater).

TREND presents short capsules of diverse items:respondents read what interests them and scan the rest.The almost 50/50 read/scan relationship suggests a goodmix of content.

How often do you obtain copies of reports or papersreferenced in RCA Technical Abstracts?


Use more than twice a month 4%

Occasionally 29%

Never use 67%

RCA Technical Abstracts is used by 33% of the surveyrespondents, 4% at least once a month, and 29% use itoccasionally. Recalling that only 45% have access (9%direct, 36% indirect), many of those who have access followthrough by obtaining copies of the documents abstracted.

How often do you use the Library?

Libraries accessible at a location are used by 77% of therespondents; 48% use it more than twice a month; 29%occasionally. The heaviest use is for literature searches,followed by reading current journals and proceedings.

Other uses of the library range from checking MilitarySpecifications and vendor information to seeing what's newand finding a quiet place to work. 9% are too busy to use thelibrary, and 4% say that their management discouragesthem from using the library.

How does achievement level relate to accessand use of these sources?An earlier paper on the Engineering Information Survey2compared the differences in information use between highand low achievers.* It showed that high achievers useappreciably more initiative in seeking out information andmake the effort required to be well informed beyond theimmediate job. Generally, the same pattern shows up in thedata for the four information sources reported below.

Access

% of respondents answering positivelyHigh Low

achiever achiever

Receive RCA Engineer at homeNo access to RCA EngineerReceive TRENDDirect receipt of RCA Technical AbstractsLibrary at my location

Use (by those who have access)

% of RCA Engineer read% of TREND readUse RCA Technical Abstracts

twice a monthNever use RCA Technical AbstractsUse library more than twice a monthNever use library

84 46

5 33

78 56

63 28

86 60

21 16

42 37

12 4

20 44

70 45

2 13

'Achievement was measured by six criteria: perceived technical currentness,effectiveness as an information source, and the number of papers, presentations, patents,and awards.

20

10

WELL BELOW AVERAGE ABOVE WELLBELOW AVERAGE AVERAGE ABOVEAVERAGE AVERAGE

TECHNICAL CURRENCY

Fig. 1RCA Engineer is read more by engineers who perceive themselvesto be more up to date. Respondents were asked to rate themselvesin terms of being up to date with the current state of the art in theirtechnical field; these responses are plotted against percentage ofthe RCA Engineer that is read for each category.

48

We were able to derive several other interestingrelationships from this comparison and other surveyanalyses:

RCA Engineer

Engineers who read more of the RCA Engineer regardthemselves as being more up to date (see Fig. 1).

Engineers who have more access to the RCA Engineerperceive that their management places more emphasis onstaying up to date.

Engineers who have more access to the RCA Engineer areused more frequently as technical information sources.

Engineers with more professional achievements havebetter access to the RCA Engineer and use it more to findpersonal contacts with whom they can discuss technicalmatters.


Those who find RCA Technical Abstracts most useful arethose who receive it directly.

Those who regard their management as placing strongemphasis on keeping up to date had easy access to RCATechnical Abstracts.

Those receiving RCA Technical Abstracts directly ratedthemselves near the top in keeping up to date.

Engineers who have more access to RCA TechnicalAbstracts have more patents, papers, presentations, andawards (Fig. 2).

Library

Those who put more emphasis on keeping technicallycurrent make more use of, and have better access to,libraries.

30

20

10

DON'T LIBRARYKNOW NOTWHAT AVAILABLEIT IS

ACCESS TO TAB

DIRECTRECEIPT

CIRCULATED

Fig. 2Access to RCA Technical Abstracts relates to accomplishments(papers, patents, presentations, awards).

Those who use the library more also publish more papers,make more presentations, file more patents, and winmore awards (Fig. 3).

Those who have used computer -assisted searching havemore professional accomplishments.

What can we do to improve?

The survey results contained many recommendations for enhanc-ing the value of the information sources covered.

What could make the RCA Engineer more important?

By far, the most common response to this open-endedquestion was for "more directly job -related material." Otherwrite-in suggestions repeated most often were:

-include more tutorial or self -study material.

-include more state-of-the-art technology.

-make it easier to understand-less math, less technical.

-go into more technical depth.

-make it more applications oriented.

Following up on this open-ended question, the surveyattempted to delve more specifically into recommendationsfor change.

When asked about state-of-the-art reviews, for example,more than 79% of the engineers responding (about 2100)considered it important to publish more of this type ofarticle in the RCA Engineer.

Respondents were asked to write-in state-of-the-art topicsand other topics that should be in the Engineer, along with

20-

15-

10-

5

NOT AT ALL <2 2-5 6-10 >10

TIMES/MONTHHOW FREQUENTLY DO YOU USE LIBRARY?

Fig. 3Library use correlates strongly with professionalaccomplishments. Engineers who visit their libraries frequently arethose who are writing more papers, making more presentations,obtaining more patents, and receiving more awards.

49

cu heard wKir-From"the deuarcir opera/6r

TREND is second only to the grapevine as a source of nontechnical information about RCA.

suggested possible sources or authors. Table I lists themost frequent recommendations by general topic.

Table IGeneral topics that survey respondents suggested for theRCA Engineer. (The list is in order of the popularity of thetopics based on write-in answers to the question "Whatwould you suggest as some of the more important topics tobe covered in the RCA Engineer").

Computers/digital technology

Solid state technology and applications

Manufacturing

Communications

Business/policy

Profiles of other activities

Electro-optics

Circuits

Mechanical engineering and packaging

Television

Heat transfer

Career information

Devices

Materials

The survey also provided a listing of types of informationthat might be useful and asked respondents to check asmany as apply. The results are given in Table II.

Table IIMore information about competitive technology and moreeducational presentations were the two strongest surveyrequests. (These are responses to the statement "The RCAEngineer should have more information about.")

of respondents answering0/0

Competitive technology 64

Educational presentations-for self -study 52

Profiles of divisions and activities 40

Various aspects of engineering as a profession 33

Educational presentations-for career guidance 28

Technical reference file (technical bibliographiesof various topics compiled by experts) 28

Blue-sky look into the future 27

Entertainment (electronics is fun and educational;hobbies, etc.) 23

Book reviews 20

Two-way communications/letters to editor 13

Profiles of engineers 12

Picture features (photo essays) 12

Interviews 10

50

What subjects would you like emphasized more in TREND?What are the subjects you would like emphasized less?

The results, in descending order of importance, show thepercentage of readers desiring more information in aparticular area, minus the percentage of those desiring lessemphasis:

New products, developmentsNew marketsBusinessScientific advancesPolicyAvailable services (Library, MIS,

Information Services, etc.)

65%52%51%48%35%

24%

What services could the library provide that would behelpful to you that it does not now provide?

In the answer to this question, there seemed to be a generalfeeling that our libraries are not being kept up to datebecause of budget limitations. Requests were made formore up-to-date books, broader coverage, more copies ofpopular books so there wouldn't be such a long wait forthem, and a larger selection of journals and conferenceproceedings.

Library users would also like their libraries to have morespace for older material, which now must be discarded.They would like on-line computer searching, audio-visualaids and materials, video tapes of RCA courses, catalogs ofholdings in other RCA libraries, bulletins from their librarylisting new acquisitions, faster response time on documentsrequested, and better copying facilities.

Reactions to the survey-what are we doing with these results?

RCA EngineerJohn Phillips, EditorContact him at:Corporate Engineering, Cherry Hill, N.J.Ext. PY-4254

As I reported in the August -September issue,3 we havealready started publishing more review and survey articles,more tutorial information, and more about RCA'sbusinesses. We plan to continue these efforts. Further, wefeel that most papers we publish-whether product-,technology-, or application-oriented-can teach, can makesome mention of the business climate, and can reviewrelated developments. To that end, we have been en-couraging authors to include additional broader -interestmaterial in each paper. This may be a perspective statementadded as a "box" to the article, or it may simply be anexpanded introduction, further explanation of the theory,stronger conclusion, or clarifying illustration.

Most important, we are making the needs of our broadreadership known to potential authors as we invite papersfor future issues; we are also insisting that those needs be

met as a condition of publication. Many engineers wantedmore information on the competition and on competitivetechnologies. Here we hit a real obstacle. Including infor-mation on competition is difficult, partly because suchinformation is difficult to get, and partly because the RCAEngineer is published by RCA. While we cannot evaluateour competitor's products in a company -sponsored journal,we will stress competing technologies in any technologyreview we publish. For example, in the August -September1977 issue, the paper on CMOS Reliability, by LarryGallace, et al, also treats bipolar device reliability. In thesame issue, Jerry Bouchard and John Bauer reviewed thecompeting hybrid technologies, not just the ones used attheir locations.

Because access and use of the Engineer relate so stronglyto the survey's measures of engineering achievement, weare continually working with management to broaden thedistribution, and have directed much of our editorialresources toward improving readability. Direct distributionto all engineers remains a primary objective.

We have also taken steps to build upon the journal's value asa source of personal contacts for technical information.This shows up in several ways:

Direct statements within the body of each article, lettingreaders know where more information is available fromthe author or from others at RCA.

Useful references and bibliographies.

Specific contact information for each author.

Reviews of work being done around the company (e.g.,"Where the electro-optics action is at RCA" by Haynes,Apr - May 1977)

We plan, to use the many write-in suggestions for articlesand features. As the mechanism for follow-up, all thesuggestions from a location will be given to the EditorialRepresentative for that locatoon. If we obtain half of them,we will fill the pages of the Engineer for at least the nextdecade.

When we come to the question of including material relateddirectly to every engineer's job, we have a real problem.Many of us would like to solve our next major problem bysimply picking up a piece of literature. But no informationsource is that good. And, although the request for directlyjob -related information was a major recurring theme of thissurvey, the best we, as information gatherers, can do ismake it easier for readers to determine what is relevant totheir jobs and provide them efficient tools and methods forfurther literature or expert searching. We are emphasizingthe importance of useful references and bibliographies ineach article; and, as reported, if an article has some bearingon a particular job, author contact is simple and straight-forward.

In summary, then, the survey results set a major goal for theRCA Engineer staff, Editorial Representatives, and authors:to produce a journal that meets the needs of thereadership-said another way, to produce a journal "by and

51

Many readers of RCA Technical Abstracts follow through byobtaining copies of the documents it abstracts.

for the RCA Engineer." This survey was a large step in acontinuing process of trying to better understand ourreaders' needs and interests, recognizing, of course, thatthese needs and interests change continually. We willcontinue to solicit your ideas and try to be responsive toyour needs.

TREND-The Research and Engineering News DigestFrank Strobl, EditorContact him at:Corporate Engineering, Cherry Hill, N.J.Ext. PY-4220

The survey results have provided an additional perspectiveon what TREND readers want to see in RCA's technicalnews digest.

A main concern that generally underlies all the survey datais how TREND can be more useful to RCA engineers. One ofTREND's purposes is to point out to engineers where usefulinformation can be found-Continuing Engineering Educa-tion courses; valuable resource materials (such as RCALibrary programs, RCA -MIT Industrial Liaison Symposia);RCA technical symposia; and engineering or marketingcontacts who can provide in-depth information on technicalwork going on in other divisions.

This last item is receiving special emphasis because theSurvey indicated that 59% of RCA engineers "never useTREND to find personal contacts with whom they candiscuss technical matters." To remedy this, a contact nameis being included in every story covering a technicalachievement.

Business information about RCA fared well in the readers'ratings. TREND includes business information not only onRCA's technical operations but also on the contributions ofits diverse companies. We constantly try to get articleslooking at the business perspectives of a particular opera-tion, and more of these will be published in the future.

In response to the Survey question on "technical informa-tion not directly related to job duties but important tokeeping broadly abreast of the state of the engineering artand science in a particular field," TREND scored aboutaverage. TREND, as The Research and Engineering NewsDigest of RCA, has as another goal to make engineers andscientists aware of the latest technical achievements in theCorporation-the whole Corporation. We hope to includemore of the technical work going on in the Corporation thatis not highlighted elsewhere. How? By periodically askingour readers for "news tips" on developments they want toshare with the total technical community at RCA, and byutilizing the Editorial Representative network to help getinputs.

As for supplying technical information that is directlyrelevant to the performance of present job duties, TRENDscored below average and it should. We don't pretend tohelp an engineer do a specific job, but do attempt to pointout where a colleague may be doing comparable work andhow additional information can be obtained.

TREND was rated above average as an information sourceabout professional achievements, i.e., society work,awards, honors. We will continue to concentrate on majortechnical honors and achievements, but leave coverage ofthe bulk of professional accomplishments to the RCAEngineer journal and other media such as TechnicalExcellence newsletters.

The survey showed that 10% of the respondents did notreceive TREND. The news digest is not sent directly to eachengineer but is disseminated via a distributor network. Aperson at each location is charged with the responsibility ofbreaking down a bulk shipment and forwarding copies toengineering departments at that location. As engineersmove around in their divisions or change locations, theyshould advise the TREND office of their new addresses toensure uninterrupted service. Distributors are asked to pollperiodically the engineering departments they service tomake sure TREND distribution is adequate.

Indications of unsatisfactory distribution uncovered in theSurvey are being followed up. All Survey respondentsshould receive their own copy of TREND, as the engineer-ing community is the prime audience for the news digest.

Although content of a digest is primary, its readability isenhanced by its look. TREND readers scan 46% of thedigest and read 41%-not bad, but the ratio should beshifted to more "read." With Editorial Board and Staff help,we are attempting to make TREND more interesting and

52

lively by improving its graphics (typography, photos,cartoons), content (shorter stories to gain more variety incoverage of all RCA's operations), and by running specials(technology quizzes, guest editorials). A highly valuedfeature in TREND is the organization chart. Readers areinterested in following the progress of their colleagues.More charts are planned in upcoming issues.

In summary, then, these Survey results will serve as aguideline, along with suggestions from the TREND EditorialBoard, to ensure that readers get balanced and adequatecoverage of RCA's technical and related business activitiesin each issue. But perhaps more importantly, we can'tpublicize your achievements if we don't know about them.TREND is a "readers' digest" in the sense that it reflects itsreaders' work. Our output is your input. We look forward tohearing from you if you feel you have an achievement thatcan help others at RCA in their work.

RCA Technical AbstractsDoris Hutchison, EditorContact her at:Corporate Engineering, Cherry Hill, N.J.Ext. PY-5412

The survey showed that many of those reading RCATechnical Abstracts do follow through by obtaining copiesof documents abstracted therein. However, it was a surpriseto find that 21% of those replying have encountered somuch delay or difficulty in getting documents that they havegiven up trying. Copies of all documents abstracted in RCATechnical Abstracts are available, and there is no reasonwhy you should not be able to obtain them. The best source,of course, is your librarian, who should be able to get anyRCA document for you fairly quickly. If you experience anydifficulty in obtaining documents, we would like to knowabout it.

RCA Technical Abstracts is of value to 37% of therespondents; this constitutes the lowest ranking of the foursources considered. One reason for the low ranking is that55% of the respondents do not use RCA TechnicalAbstracts because they either do not know what it is (26%)or do not have access to it (29%). To help remedy thissituation, a reprint of an RCA Technical Abstracts coverwith a description of its purpose, coverage, how to use it,and where to find it was included with an RCA Engineermailing. Hopefully, some of the 55% mentioned above willnow look for RCA Technical Abstracts in their library, asktheir engineering management to put them on their circula-tion list, or, if not otherwise available, request to be put onthe distribution list. A number of such requests have beenreceived and honored.

RCA LibrariesContact your librarian, orDoris HutchisonCorporate Engineering, Cherry Hill, N.J.Ext. PY-5412

The libraries rate very high in this survey and maintain animage of providing business, professional, and technicalinformation to 81-93% of the respondents. There is littlequestion that their value ranks highest among the informa-tion sources covered in the survey.

Why does RCA provide these technicalinformation sources?RCA Engineer-A bimonthly technical journal thatpublishes state-of-the-art reviews, tutorials, and othertechnical, and business -oriented papers. It provides anin depth look at what's going on at RCA and helps youupdate your knowledge of the electronics field. It alsoprovides a forum in which you can publish to enhanceyour professional reputation. For more information, callJohn Phillips, Cherry Hill, Ext. PY-4254.

TREND-a monthly briefing of what's going ontechnically at RCA, what RCA's businesses are doing,and what services, products, and other aids are availableto support your work. For more information, call FrankStrobl, Cherry Hill, Ext. PY-4220.

RCA Technical Abstracts-A monthly corn pany-privateabstract bulletin of RCA's proprietary technical reportsand publications. It cites author, source, and document -number and includes RCA patent disclosures and MITresearch reports as well as the RCA papers and reports.Annual cumulative indices are also available, startingwith 1968. For more information, call Doris Hutchison,Cherry Hill, Ext. PY-5412.

RCA Libraries-The highly skilled librarians in RCA'stechnical libraries can get just about any piece oftechnical information you may need or desire. If yourlocation doesn't have a library, your divisional library iswilling to support you. For more information, call yourlibrarian or Doris Hutchison, Cherry Hill, Ext. PY-5412.

All locations employing large numbers of engineers areserviced by an RCA technical library operated by a highlyskilled professional librarian.' Many other locations havemodest libraries or technical information collections staffedon a part-time basis. This type of information facility couldalso be useful at several additional locations. Engineers inlocations lacking library facilities can use the major libraryof their operating unit; at present, about 7% of the surveyrespondents take advantage of a remote library.

Some of the needs expressed by survey respondents arenow being provided at several locations. For example, on-line computer searching is now available at Moorestown,Princeton (Laboratories and Astro-Electronics), Lancaster,Indianapolis, and Burlington; soon to be available inCamden, with other libraries considering this service.

References

1. Underwood, W.J. and Jenny, H.K.; "Engineering Information Survey Results." RCAEngineer, Vol. 22, No. 3 (Oct - Nov 1977).

2. Underwood, W.J. and Jenny, H.K.; "Engineering Information Survey Results, Part 2(High vs. low achievers-how are they different?)" RCA Engineer, Vol. 23, No. 4 (Dec1977 - Jan 1978).

3. Phillips. J.C.; "There've been some changes made," RCA Engineer, Vol 22, No. 2 (Aug -

Sep 1977).

4. Whitehead, O.F. and Hutchison, D.E.; "Technical Information: where to get it," RCAEngineer, Vol. 22, No. 4 (Dec 1976 - Jan 1977).

53

Color displays for airborne weather radar

R.H. Aires) G.A. Lucchi

In 1977 RCA introduced weather radarsystems with color indicators for bothcommercial air carriers and general -aircraft users. The use of color hassignificantly improved the pilot's ability torecognize potentially dangerous storms. Ithas also improved the radar's usefulness inthe ground -mapping mode, where the pilotlooks at the terrain below instead of theweather ahead. But before describing thisadvance, some background on weatherradar is necessary.

Relating rain densityand turbulenceWeather -radar systems used in the air -carrier and general -aviation industrymeasure the rainfall density that existswithin a storm cell-they are not able toassess turbulence directly.

Experience has shown, however, thatstorm cells containing rainfall rates of over11.5 millimeters per hour quite oftenproduce enough turbulence to cause air-craft structural damage or passenger in-jury. Therefore the FAA advises pilots tostay at least 20 miles away from storms withrainfall rates exceeding 11.5 mm/ hr.

The radar return from a storm depends uponthe backscatter from the raindrops.

Eq. 1 is the basic radar range equation usedto calculate the detectability of a specificradar target.

P, = P,G2X2cr/ [(4/r)3 R4/] (1)

wherePi = transmitter peak powerG = antenna gain used both for

receiving and transmitting= radar carrier wavelength

a = target radar cross section areaR = one-way range between

radar and targetP, = received power from the targetL = system losses, correlation gain, etc.

Reprint RE -23-5-13Final manuscript received February 14. 1978.

Pilots recognize potentially dangerous storms more readilywhen the storm locations are displayed in color.

This equation is derived from thetheoretical considerations with the radartarget consisting of a solid, highly con-ducting surface of generally sphericalshape. In order to apply this equation tometerological-type targets, it must bemodified to reflect a three-dimensionalfluid -impregnated volume representativeof a storm cell core.

The radar backscatter coefficient Z forspherical raindrops has been empiricallyestablished. The coefficient is a function ofthe sixth power of the raindrop diameterand the number of raindrops in a specificvolume. It is denoted by

Z= E D,6,=,

and is inversely proportional to the fourthpower of the radar carrier wavelength andproportional to the second power of ab-solute value of the dielectric constant, K, ofwater. Since the backscatter coefficient isalso proportional to rainfall rate, therelationship* Z = 200 r''6 is used to definethe reflectivity, where Z is in mm6/ m3 and ris the rainfall rate in mm/ hr. For thepencil -shaped radar antenna beam general-ly used on airborne weather radar systems,the resultant pulse volume can be definedas (n-18)R202cT, where R is the radar range,

is the antenna beamwidth, c is the speedof light, and T is the transmitterpulsewidth. The resultant radar targetbackscatter cross section a is the pulsevolume times the factors which relate to themeteorological targets described below.

',This relationship pertains to Strataform-type storms; Z = 55 r1'is sometimes used for convective storms.

This backscatter coefficient is shown in Eq.2.

a = (n-5/ X4) ZK2(r18)R202cT (2)

By substituting a from Eq. 2 into Eq. 1,converting the terms into appropriatevalues, and converting the equation into adB format, the weather range equation fora beam -filling target is shown in Eq. 3. Thisformat makes the equation suitable forrelatively simple graphing and so assiststradeoff analyses in design.

P, = 10logP, + 20log0 + 10log T+ 10logZ-20logA-20logR +2G-L-168.25

(dB) (3)

where P, is in dBm, 0 is in degrees, Z is inmm 6/ m3, R is in nautical miles, P, is in kW,T is in Ais, is in cm, G is in dB, and L is indB.

For a storm cell whose diameter does notfill the radar beam, Eq. 3 must be modifiedby substituting the non -beam -fillingvolume equation (r/ 8)dc T for the beam -filling volume equation (ir/ 8)/2202c T intoEq. 2, where d is the storm cell diameter.The same procedure used to establish Eq. 3yields Eq. 4 for non -beam -filling storms.

Pr= 10logP, + 20logd + 10log T + 10logZ-20logA - 40logR + 2G -L -133.08

(dB) (4)

All terms in Eq. 4 have the same units as inEq. 3 except d, which is in nautical miles.

Eqs. 3 and 4 are used to calculate thereceived power which appears at the input

54

to the receiver for the range and radarparameters selected for a specific storm celldensity.

Both range equations are plotted in Fig. 1using the system parameters for the RCAPriM US -400 ColoRadar system with anI8 -inch flat -plate antenna. It can be seenthat the received power decreases at 6 dBper octave of range at distances up to thepoint where the 3 -nautical -mile -diameterstorm -cell model changes from beam -filling to non -beam -filling. At greaterranges the received power decreases at 12dB per octave of range.

Pilots must avoid storms above a certainintensity, so radar displays must make thesestorms stand out.

The radar meteorology industry hasrelated rainfall rates to the storm intensities(Table I). The table also gives the colorsRCA uses to indicate these rainfall rates.

-50 -

- 60

-70

- 80

-90

-100

-110

1152 3 4 5

RCA's background in weather radarRCA has been producing airborne weather radar systems for commercialair carriers and general -aircraft users since 1955. The first system producedwas the well known C -band AVQ-10, which is still being used by many of theworld's airlines. Since that time, most of the airborne weather radar systemshave been designed to operate at X -band. So far, RCA has produced a totalof eleven different types of X -band and two types of C -band radars; they areused in aircraft ranging from wide-bodied jets down to the smallest twin -engine aircraft. RCA's most recent weather -radar advancement is thePriMUS-400 ColoRadar described here.

Table IRadar measures rainfall, which can be correlated to storm turbulence. (Radar cannotmeasure turbulence directly.) Radar meteorology industry has grouped rainfall rate (andthus turbulence) via this table. RCA has selected colors in the table to indicate the differentstorm levels.

Storm intensity

DrizzleLightModerateIndustry standard

pilot alertHeavy 16.0 mmi hr

or greater

Rainfall rate Display color Digital level

0.25 mm/ hr black 0

1.0 mm/hr green 1

4.0 mm' hr yellow 2

11.5 mm; hr red 3

*Storm levels with rainfall rates greater than 11.5 mm. hr are shown as the number 3 level regardless of theirintensity.

PRIMUS -400WEATHER DETECTIONI CHARACTERISTICS

BEAM FILLING RANGE

10 20 30 40 50 60 70

RANGE NAUTICAL MILES

100 200 300 400 500

Fig. 1

Weather detection characteristics of the PriMUS-400 ColoRadar. The four solid lines show the returned power from storms withdifferent rainfall rates. The 11.5 mm/hr. storm is potentially dangerous. System identifies levels of storm intensity by differentcolors on the display. The vertical "beam -filling" line, which defines 3 -nautical -mile -wide storms that co or do not fill up the radarbeam, also establishes the break point between the two radar range equations. The shaded areas show how the receiver gain isshaped.

55

I -

ANTENNA

FL F OAT ION

ROTARY10141

GEARREDUCTION

ELEVATIONDRIVE MOTOR

AZIMUTHMOTOR CONTROL

AZIMUTH DRIVESTEPPER MOTOR

GEAR

REDUCTION

REV TIONRESOLVER

SWEEP

RESOLVER

///

/

AZIMUTHPDT!DINT

AZIMUTHRESOLVER

PITCH AND ROLL SIGNALS

rRECEIVER-TRANSMITTER

SPAC

CIRCULATORt44to.j

COSINE PITCH AND SINE ROLL

POSITION TEED/AC

LEvATEON MOTOR °Rio

RECEIVED OR REFLECTED RI

TRANSMITTED RI

REFLECTSRF ONLY

MOIR. W HO -

FROM AIRCRAFTVERTICAL GYRO

MHz PITCH SIGNAL

AM H, ROLL SIGNAL

CO AXE ALMAGNETRON

9345 MOO I 0/1 SPULSE

STABILIZATION

R

LIMITER

1345 MHz RE PULSE

GISE 1,011IT ST ONLY

AIR STRIP MODULE

NOISE DIODE

ANO

BANOPASS FILTER

SINS MtLOCAL

OSCILLATORIVO)/

VALANCEDMIXER

,SANS MH/ CW

FILTER

AFC

MIXER

MODULATORDC CONTROL

VOLTAGEAFC

0OULATOR TRIGGER

PULSE WIDTH CONTROL

POWER SUPPLY

PROTom of AYMODULATOR TR,PUL SI WIDTH CON

AGE

Fig. 2PriMUS-400 system. Radar pulse (variable width for looking at ground or weather) goesthrough circulator to stabilized antenna, then returns through antenna and circulator. Aftersignal processing and amplification, the returned signal is digitized to one of four storm -intensity levels, scan -converted from polar to x -y coordinates, and displayed on the colorcrt.

Meteorologists have established thatstorms whose rainfall rate equals or ex-ceeds 11.5 mm/ hr (a Z factor of 104) shouldbe avoided by aircraft. As a result, inmonochrome displays, areas with this Zlevel are blinked between the brightest level(number 3) and then off so that the pilotcan easily distinguish the storm's position.In the color display system, this levelappears as red, with a pilot option to havethe area blinked between black and red. Inorder to determine where lesser rainfallrates are located in a storm cell, the 4mm/ hr signal level is detected and dis-played as the 2nd from the brightest(number 2) in the monochrome indicatorand in yellow for the color indicator. The 1mm/ hr storm is indicated as the lowestlevel (number I) on the monochromeindicator and green on the color indicator.In this manner, the pilot can easilydetermine where to fly to avoid storms ofvarying intensity levels. The "drizzle" levelis not shown on the display because itsturbulence intensity is insufficient to be ofconcern.

PriMUS-400 ColoRadar

On March 15, 1977 RCA announced thefirst general -aviation airborne weatherradar system with a coldr indicator. Theacceptance of this system by pilots has beenoverwhelming, primarily because the use ofcolor greatly simplifies the recognition ofdisplayed storm intensities. In addition tothe advantages provided by color, theresolution of the display used in thePriMUS-400 ColoRadar is four timesbetter than previously available digitalstorage type indicators. Another feature ofthe PriM US -400 ColoRadar is the use of adifferent set of colors (cyan [blue-green],yellow, and magenta) in the ground -mapping mode. In this mode the

transmitted pulsewidth and receiverbandwidth have been optimized to take fulladvantage of the increased display resolu-tion. The overall result is much betterground -mapping than previouslyavailable. Major design considerations forthe primary functions of the PriMUS-400ColoRadar system will now be described.

The major elements of the PriMUS-400system are the receiver/transmitter, the in-dicator, and the antenna.

Fig. 2 is a block diagram of the radarsystem. A magnetron generates microwavepulse energy at 9345 MHz and the cir-culator then directs it to the antenna, whereit is radiated to illuminate the target.Reflected energy from the target is in-tercepted by the antenna and directed bythe circulator to the microwave mixer. Themixer converts the microwave frequency toa first i.f. of 60 MHz. After amplification at60 MHz, a second conversion to 10.7 MHzoccurs within the i.f. amplifier to achievethe desired bandwidth prior to video detec-tion. An AFC adjusts the receivermicrowave local oscillator to track themagnetron frequency.

The detected video signal is digitized in theindicator to four discrete levels. Afterdigital integration of the returns from fourpulses, the data is scan -converted frompolar -coordinate (rho -theta) format to tv

56

F & VIDEOAMPLIFIER

rDIGITAL INDICATOR

NPUTS

RCVRVIDEO

A I I

GRAMMERSTC

GER NOISE GATEIROL FAULT DETECTORS

POWER SWITCHING

120 FMSYSTEMTRIGGER

1 97

DIGITIZER

DIG TIDEDVIDEO

97 VARIABLE PHASE

CORRE LA OR

RANGECLOCK

RANGCLOG

GENERATOR

I G ANDCONTROL

NTENN01111vELow

HNUMERICALPHA -

GENERATOR

ADORE SS

UPDATEADDRESS

ANTENNA FILUPIESIGNAL TO ALFA.,NUAERiC GENERATOR

A 'C LAMP 5 OR 28 VOLTSBUS

5 VOCLOW VOLTAGEPOWER SUPPLv

CONYERS!.

COLOR REFERENCBAR GENERATOR

CYC CONTOUR

CONTROLPANEL

RANGE 'AZMARK

GENERATOR

VIDEODECODER AND

AMPLIFIER

DEFLECTIONCIRCUITS

GREEN

PAGEDISPLAY

ORIZ IS TS.,

AT 60 H 0 Hz

HIGN VOLTAGEPOWER SUPPLY

L

(x -y) format for storage in a RAMmemory. Readout of the RAM memory isin standard tv format at a 60 -Hz field rate.A decoder turns on the appropriate red,green, or blue gun of a color CRT,depending on whether the data represents a0, 1, 2, or 3 level.

External video signals can be added to theradar data by video mixing at the videoamplifier. Range and azimuth marks andeven alphanumeric information can also beadded to the display by video mixing. Theindicator logic generates timing signals totrigger the transmitter and to drive theantenna azimuth stepper motor. Theantenna elevation motor is driven fromsignals derived from the aircraft attitudereference (vertical gyro) to stabilize theantenna against pitch and roll maneuversand permit manual control of the antennaelevation angle for ground mapping.

The transmitter has a variable pulsewidth.

A 10 -kW coaxial magnetron has beenselected for reliability and frequency -

stability characteristics. This magnetronhas been used in the top -of -the -line RCAgeneral -aviation weather radars since theintroduction of the AVQ-21 in 1970. Itmaintains nearly full power output untilend -of -life because of its conservativecathode loading. The large coaxial cavity ismuch less subject to particle build-up thanthe older strap -vane design and thereforehas less tendency to change frequency withage. A solid-state line -type modulator withan SCR switch provides the required 4.7 -kV pulse for the magnetron.

In the weather mode, the transmitterpulsewidth is 3.5 µs. For optimum perfor-mance this pulsewidth requires an i.f.bandwidth of approximately 350 kHz.From Eqs. 3 and 4 it can be seen that thereceived power (P,) is greater with a widerpulsewidth. A pulsewidth of 3.5 As

sacrifices little weather resolution, since theestablished model storm cell is 3 miles or 37As in diameter. The 3.5-µs pulsewidth istherefore a reasonable compromisebetween sensitivity and resolution.

1

In the ground -mapping mode, the resolu-tion of small targets is limited by thepulsewidth and receiver bandwidth. Theselection criteria here was a tradeoff withthe display resolution, which is determinedby the size of the digital memory. In thePri M US -400 Colo Radar the digitalmemory consists of two planes of 64k bitseach. The screen therefore can be dividedinto 256 X 256 cell locations, with the rangeresolution determined by dividing theselected range scale into 256 cells. On the25 -mile -range scale the smallest incrementstored is 25/ 256 miles or approximately 0.1mile, which is equivalent to 1.2 As. A 1-µspulsewidth is used in the ground -mappingmode for selected range scales of 50 milesor less. On the 100-, 200-, and 300 -mileranges the pulsewidth is increased to 3.5µsfor increased range performance andbecause the display is the limiting factor onresolution.

The pulse -repetition frequency (PRF) isdetermined by the desired display resolu-tion and the antenna scan rate. The

57

BLANKING

Fig. 3I.F. amplifier optimizes systemsensitivity by generating a secondi.f. center frequency of 10.7 MHz.Sensitivity time control (STC) ad-justs gain with range; manual gaincontrol allows pilot to vary color -level settings and see which part ofstorms are most intense.

Fig. 4AFC system essentially guarantees"lock -on."

Fig. 5Perceived-power/range slopecan be varied at the i.f.amplifier or video amplifier.Beyond 50 nautical milesrange, additional sensitivitycomes from varying thereference voltages of the com-parators.

60 MHzFROMMICROWAVEMIXER

AGC

PULSE WIDTHCONTROL

60 MHzLOW NOISEPRE AMP

60 MHz2ND STAGEAMPLIFIER

A

MAP -100

WEATHER. -4,

PILOT MANUAL

STCGENERATOR

DIODEMIXER

49.3 MHz2NDL.O.

10.7 MHzAMPLI Fl ER

4

SELECTABLEBANDPASSFILTER

HIGHLEVEL

DETECTOR

10.7 MHzAMPLIFIER

LOWLEVEL2NDDETECTOR

26 dBVIDEOAMPLIFIER

10.7 MHzAMPLIFIER

4- 10.7 MHzAMPLIFIER

INDICATOR:1-

FROMTRANSMITTER

WAVEGUIDE AFC 60 MHz CONVERTER 10.7 MHzPROBEIPO MIXER FILTER 60 TO 10.7 MHz AMPLIFIER11110

1 MW SIGNAL9345 ±30 MHz

49.3 MHzLOCALOSCILLATOR

GUNN DIODE

-h.DISCRIMINATOROSC-VARACTORFREQUENCYCONTROL

INTEGRATORSEARCHOSCILLATOR

LOW FREQ.LAG LEADNETWORK

VIDEOAMPLIFIER

+15

+10

+5

CC0

X

CC

O

-1020

00

zao

W0-ra

U cc -3oto toz

-20

-35

TO RECEIVERMIXER

VIDEO AMPLIFIERINCREASING SENSITIVITY

PRIMUS-400STC CHARACTERISTICS

EXTENDEDSTC RANGE

IF AMPLIFIERDECREASING SENSITIVITY

STC CONTRIBUTED BY IF(7 dB/OCTAVE)

STC CONTRIBUTED BY IF AND VIDEO(12 dB/OCTAVE)

II {11''STC CONTRIBUTED BY VIDEO(5 dB/OCTAVE)

STC CONTRIBUTED BY VIDEO(12 dB/OCTAVE)

2 3 4 5 6 10 20 30 40 50 60

RANGE NAUTICAL MILES

100 200

58

PriMUS-400 antenna scans 120 degrees in4.3 seconds. For an azimuth resolution ofone quarter of a degree, a PRF of 120 Hz isrequired.

The receiver has a very good noise figure.

The PriMUS-400 achieves a 7 -dB noisefigure by using a bandpass filter to reducethe noise at the image frequency along withlow -noise hot carrier diodes, a balancedmixer to reduce feedthrough of local os-cillator noise, and a low -loss circulator andTR device. An isolator between the Gunndiode local oscillator and the mixer reducesfrequency pulling. A separate AFC mixerpermits accurate adjustment of thesampled magnetron frequency amplitude.This configuration provides a system noisefigure at least 1 dB lower than competitivesystems and provides 40 -dB rejection ofsignals at the image frequency.

A 60 -MHz first i.f. amplifier center fre-quency permits the practical design of theX -band bandpass filters used to reduceimage -frequency noise mentioned in theprevious section. In order to achieve the350 -kHz predetection receiver bandwidth,a second local oscillator and mixergenerate a second i.f. center frequency of10.7 MHz to optimize the system sensitivi-ty when in the 3.5-µs-pulsewidth weathermode. Sensitivity time control (STC) isgenerated in the STC generator (Fig. 3) andinjected into the second stage of the 60 -MHz amplifier. The STC is adjusted forapproximately 35 -dB receiver gainattenuation at 3 nautical miles and has thegain increase at a rate of 7 dB per octave ofrange, reaching maximum gain at ap-proximately 90 nautical miles. The adjust-ment allows for 3 dB of internal systemdegradation and 1 dB per octave of rangeto account for atmospheric attenuationfrom intervening rain up to 35 nauticalmiles, as shown in Fig. 1.

AGC, applied to the first 10.7 -MHzamplifier, is established by sampling thevideo output signal at a range beyond themaximum range of expected radar signalreturns. A pilot -operated manual gain -control voltage is applied in lieu of AGCwhen the gain is turned out of thecalibrated preset position. The manualgain control can determine which of the 3 -level (red) areas contain the most intenserain. While in this variable -gain position,an alphanumeric warning on the indicatorreminds the pilot that the system is notcalibrated for the standard rainfall rates.

The gain control is also very useful foroptimizing target recognition when in theground -mapping mode.

A separate X -band mixer diode provides asample from every magnetron pulse at 60MHz, which is then down -converted to a10.7 -MHz amplifier and discriminator.When not "locked -on," the control voltageautomatically sweeps the X -band localoscillator ±30 MHz until "lock -on" isestablished. The AFC servo loop isdesigned to achieve the highest stable gainpossible with a 120 -Hz sampling frequen-cy. More than a 10:1 range of a "-2" gainvs. frequency slope is used in the AFCservo loop is achieve a high effective K.The average search sweep rate of 10

MHz/ second is slow enough to essentiallyguarantee "lock -on" with this loopcharacteristic and a discriminator band-pass of 1.5 MHz. Fig. 4 is a block diagramof the AFC.

Three comparators determine the 0, 1, 2,

and 3 levels corresponding to rainfall lessthan 1 mm/hr, 1 mm/hr, 4 mm/hr, and 11.5mm/hr, respectively.

Fig. 1 shows that the power -received/ rangeslope is 7 dB/ octave at less than 50 nauticalmiles range and 12 dB/ octave beyond 50nautical miles range. Within the range ofSTC, 7 dB/ octave is achieved by varyingthe i.f. gain, as shown in Fig. 5. Beyond 50nautical miles range, additional sensitivityis provided by varying the referencevoltages, and thus the sensitivity, of the 2 -and 3 -level comparators to achieve a totalof 12 dB/ octave of range up to 150 nauticalmiles. Since the PriMUS-400 displays redfor a 3 -level signal strength, the pilot isaccurately warned of 3 -level storms up to arange of approximatley 150 miles (18minutes away at a jet speed of 500 knots).

A four -pulse digital integrator with a 3 -out -of -4 algorithm reduces the false -alarmrate, thus permitting the gain to be in-creased by 3.8 dB before the false -alarmrate is objectionable. A minimum discerni-ble signal (MDS) of -115 dBm can beachieved with the PriM US -400 whenreceiving a pulsewidth equal to a standard(37-µs) storm cell.

Range and azimuth data are collected inpolar (rho -theta) coordinate format, butdisplayed in x -y format.

Until recently, most airborne radar dis-plays were scanned in the rho -theta format(Fig. 6a), which has the disadvantage of

.%)

\ /\ /\ /\ /\ /\ /\./

Fig. 6Range and azimuth data is collected in polarcoordinates (top), but displayed in tv-scancoordinates (bottom). Digital storage makesthis display method practical and producesuniform brightness and resolution over theentire displayed area, something not possi-ble with the polar -coordinate display.

non -uniform resolution over the area ofdisplay. Near the origin of the display, thedata is crowded, and at the edge of thedisplay, it is spread out much as the spokesof a wheel. With the advent of digital datastorage in solid-state random-accessmemories, many advantages of x -y scanbecame practical. Some of these advan-tages are:

uniform brightness and resolution overthe displayed area; efficient deflection and high -voltagesystem; compatibility with a tv raster to com-bine data from other sensors andsystems; and compatibility with the requirements ofa line -screen color CRT.

A preprogrammed read-only memorysteers the received signals from eachtransmitted pulse to the appropriate loca-tion in a random-access memory cor-responding to the antenna angle and targetrange. Scan conversion is achieved byreading the random-access memory in x -yformat, as shown in Fig. 6b. The readout isperformed at standard tv rates, thusproviding a flicker -free display using aconventional tv crt.

59

305

0

0

RANGE OFF

50 200 1900TEST

510F MA cyc MAP AVM

0

iiti

._,5

Fig. 7High -resolution dot pattern for rangeazimuth markers is difficult to produce on a525 -line tv raster. Special LSI circuit makesthe dots appear as smooth lines.

The video display has high -resolutionazimuth marks; it also accepts externalinputs for pilot checklists and other data.

Very high video bandwidth is required todraw smooth arcs representing rangemarks and straight lines for azimuth markson a 525 -line tv raster. A custom large-scale -integrated circuit was developed withRCA's Solid State Technology Center.This IC provides 50-ns video pulses togenerate a high -resolution dot pattern thatappears as smooth lines on the display (Fig.7). The color of these lines and marks iscyan in the weather mode and green in themap mode for good color contrast.Alphanumerics generated by a charactergenerator identify the range marks,

selected mode, and special alerts. An exter-nal video input connector allows the in-dicator to be time-shared for otherfunctions such as checklist, navigationinformation, optimum flight profiles, andother tv-formatted data.

The antenna must have good stabilizationwith the high -resolution color display.

The antenna can be scanned through 120degrees in azimuth and 60 degrees inelevation. The normal pattern is a horizon-tal line scan. A two-phase stepper motordrives the azimuth axis, based on com-mand pulses from the indicator. The direc-tion of scan is controlled by the phaserelationship. A feedback signal determineswhen the antenna is passing the dead -ahead position and this signal is comparedwith the indicator command to verifysynchronization. If an error exists, theradar data is blanked and the letters ANTare flashed on the screen.

To maintain a scan pattern that is fixed inspace relative to the earth's surface, pitchand roll signals received from the aircraftattitude -reference system are applied to anelevation servo system. The use of two axesof freedom (typical in most general -aviation aircraft) means that the antennaelevation servo must continuously correctfor the aircraft's roll and pitch. Brushlessresolvers and a two-phase ac motor areused in the elevation axis servo for long life.A major improvement over previousstabilization designs has been achievedthrough the use of good servo design

Ray Aires is Chief Engineer of AvionicsSystems. Under his guidance, RCA hasintroduced the PriMUS-20 and -30 weatherradars, which were the first to employ x -yscanned indicators providing alphanumericinformation outside the radar presentation.Avionics Systems' PriMUS-21, -31, and -50were developed specifically for helicopterapplications. Most recently, he wasresponsible for RCA's introduction of thefirst weather radars with color indicators,including the PriMUS-90 for commercial aircarriers and the PriMUS-300 and -400 forgeneral aviation.

Contact him at:EngineeringAvionics SystemsVan Nuys. Cal.Ext. 3644

practices to achieve a high effective loopgain (KO, which became more importantwhen the high -resolution color display wasintroduced.

The waveguide slotted -array antenna usedobtains 70% aperture efficiency withsidelobe levels 25 dB or more below themainlobe. These two characteristics areextremely important to maximize therange performance of the radar and tominimize ground clutter when looking forweather.

SummaryThe PriMUS-400 ColoRadar systemdescribed in this paper has introducedmany new features that have led to itsimmediate popularity in the general -aviation marketplace. The most importantfeature was the introduction of a colorindicator with 4 times the display resolu-tion of previous digital storage systems.For a lightweight radar system, the 150nautical miles of storm intensity calibra-tion provides a capability not previouslyfound in the most sophisticated large andexpensive airline -type weather radars.

George Lucchl has over 20 years of radarexperience with RCA. His recent projectsinclude a number of distance -measuringequipments and transponders, plus con-tributions to a number of weather radarsystems-the AVQ-47, -21, and -30 systemsplus the PriMUS series.

Contact him at:EngineeringAvionics SystemsVan Nuys, Cal.Ext. 3638

60

Advanced antenna design reduceselectronic countermeasures threatThe expansion of electronic countermeasures has changedthe ground rules for air -surveillance radar design.Low-sidelobe radar is a good example of electroniccounter -countermeasures at work.

R.M. Scudder

The demands on radar systems are ever-changing, and nowhere is this more truethan in those made on high-performanceradar operations by the expandingenvironment of active countermeasures.The sophistication of advanced ECM (elec-tronic countermeasures) equipment, andthe recent addition of ARM (anti -radiationmissiles) can, in combination, severelycompromise the effectiveness of the stan-dard surveillance radar systems built in the1950s and 60s.

Countering the countermeasures involvestwo principal approaches:

1) Signal diversity-varying the frequen-cy, power, polarization, or waveform ofthe radar signal so that the necessarilylimited jamming capability of the ECMequipment receives more variety inreceived signals than it can counter.

2) Sidelobe suppression-maintainingthe sidelobes that inevitably accompanythe main radar beam at such low levelsthat the ECM equipment cannot gen-erate effective blanking countersignalson these lobes.

Implementing these approaches takes avariety of mechanizations, any or all ofwhich can be combined in advanced -design, sophisticated systems (e.g., mul-tipurpose phased -array radars). Suchsystems, of course, are costly, usually largeand therefore relatively immobile, andimpose special power and signal -processing requirements. These re-quirements combine to preclude such anapproch for tactical surveillance radars,which must be highly mobile and easy to setup and operate in forward areas.

But it is possible to apply ECCM (elec-tronic counter -counter-measure) tech-niques to tactical systems, using straight -

Fig. 1

Stepped -beam reflector antenna as it would be emplaced as an element of a tactical radarsystem. Typical rotation rate is 6 rev/m. Transportable radar shelter at left contains radartransmitter plus operator display and control functions.

forward modifications to standard radardesign to achieve really effective operationsin advanced ECM environments. Thispaper addresses such an approach anddescribes the preliminary development of aparaboloidal -reflector, stepped -beamantenna system that promises excellentradar performance in an environment ofmodern, advanced ECM.

Antenna design andthe environment

The antenna configuration described hereis compatible with the constraintsassociated with the operatingenvironments for today's tactical radars.Attributes of the new antenna include:

very low azimuth peak sidelobes, andlow average sidelobes;

polarization diversity necessary for all-weather operation;

electronically stepped beam in elevation,with mechanical rotation in azimuth, toprovide full search capability with asingle wideband signal -processingchannel;

aperture size compatible with small -target detection and height estimation atextended ranges;

simple structure to assure minimumacquisition and life -cycle costs;

high reliability and maintainability, foroperation in tactical environments;

ease of conversion into and out of thetransport configuration; and

light weight for air transport.

Antenna descriptionA standard tactical -radar pedestal canrotate the antenna for azimuth scan, whilethe antenna rapidly steps its beam elec-tronically up and down over 20 degrees ofelevation scan.

Suitable for a radar using a single -channel,wideband signal processor, the new anten-na achieves reduced ECM vulnerabilityprimarily through reducing the magnitudeand extent of its radiation sidelobes. Fig. 1is an artist's concept of the antenna inoperating condition. Major componentsinclude the reflector structure with edge -mounted shields (absorber panels), struc-tural supports, the vertical array of dual -polarized feed -horns, and the associatedwaveguide interfaces with the transmitterand rf receiving equipment.

Other equipments located on the rotatingstructure include the unit containing thetransmitter polarization switching andbeam steering controls, the rf power dis-tribution network, the circulator/ transmit -receive limiter group, and the rf / if

61

preamplifier group. All these items arelocated within the rf equipment housingstructure, which also supports the feed andreflector structures in both the operatingand transport configurations. A separateIFF (for identification friend or foe, aseparate radar return coming from atransponder on friendly aircraft) antennaassembly is mounted at the top edge of theradar reflector.

Features of the antenna configuration in-clude the following:

The reflector surface is 20 ft wide and 12ft high, a section of a paraboloid with a6.5 -ft focal length, having its vertex atthe bottom edge. The reflector -surfaceaccuracy required for low azimuthsidelobe performance in operatingenvironments is maintained by buildingthe reflector as a lightweight monocoquestructure. It is 10 in. deep, built in threesections, and joined by pin -aligned hookcatches.

Absorber panels, at the top and bottomedges of the reflector, are designed tosuppress wide-angle feedhorn spilloverand/or edge diffraction around theperiphery of the reflector surface.

The feedhorn cluster consists of sevencorrugated horns, one protruding slight-ly above the paraboloidal axis, with theremainder below that axis. The feeds arepositioned to provide secondaryoverlapping beams compatible withmeasurement of radar target -height datafrom the horizon to 20 degrees elevation.The crossover level between adjacentbeams is at approximately -3 dB foreach beam in the vertical cluster. Eleva-tion beamwidths range from 1.7 degreesfor the lower beams to 7.0 degrees for thehighest -elevation beam. The beamwidthin the azimuth direction is less than 1.4degrees for all beams. The feeds providedual orthogonal -mode excitation to im-plement radar polarization diversity.

The microwave components, located inthe rotating base structure, interface theantenna with its associated radar system.

The stepped -beam antenna system uses acombination of the orthogonal beammatrices, phase shifters, and power -distribution networks to provide both eleva-tion beam steering and independent controlof transmit and receive polarization modes.

Fig. 2 shows the system in block -diagramform. Stepped -beam steering in elevation isaccomplished by commanded phase tapers

that are identical for the orthogonalpolarization paths. The polarization modeis selected by command of a mean phasedifference between the two groups of eightphase shifters. Duplexing isolationbetween the transmit and receive paths isimplemented by a 180 -degree phasedifference between the two groups of phaseshifters, set immediately after the time ofthe transmit pulse to effectively "switch"the antenna from the transmitter channelto the receive path.

VERTICALREFERENCE

Lz0VUWQ

44,z

BEAM NO. 7

-4. BEAM NO. I

ORTHOGONALBEAM

MATRIX

The power -distribution network providesamplitude weighting on various horns inthe seven -horn group as a function ofelevation beam steering position. Twohorns are driven for the lowest beam, threefor each of the next -higher elevationbeams, and one for each of the three highestelevation beams.

Although the transmit and receive pathslargely share common circuitry, theybranch at a point where operation of the

EIGHT POWERPHASE DISTRIBUTE

SHIFTERS NETWORK

S

TRLIMITER

FORTHOGONAL EIGHT POWER.1BEAM PHASE -- DISTRIBUTE

MATRIX SHIFTERS NETWORK

RF/IFRECEIVER

HEIGHT DATA CORRECTION 4- AZMODE & BEAM CONTROL-. ROTARY

SIGNAL PROCESS.*-- COUPLERS

TRANSMITTER La. I F F

1

Fig. 2Antenna subsystem block diagram indicates the general nature of the equipment housed inthe rotating base assembly shown in Fig. 1. Components use rectangular waveguides,packaged to be compatible with field maintenance procedures.

-10

-12

-14

-16

-18

-20

-22

-24

-6 0 6 12

ELEVATION ANGLE (DEGREES)18 24 30

Fig. 3Computed elevation patterns for the seven beams are shown in angular positions optimizedfor combined radar target detection and heightfinding functions. The beams becomebroader at higher elevation angle because they are created by feed assemblies displacedfrom the geometric focus of the paraboloid reflector.

62

elevation beam -steering phase shifters canpermit those phasers to perform also atransmit -receive duplexing function. Theantenna group includes low -noise rf and ifamplification of signals for external single -channel wideband processing.

Complete receiver protection is providedby the transmit/ receiver -limiter assembly.The vertical reference unit is used as thecalibrated reference for measurement oftarget height.

For transport the antenna is disassembledand folded down. Two end sections of thereflector are unlatched from the centersection and stowed flat on the base. Thecenter section hinges down so that it, too,lies flat. The resulting transport configura-tion is a pallet -like assembly compatiblewith all modes of transportation, includingC-130 cargo aircraft.

Antenna performance

Immunity of the antenna to ECM is aprimary objective. The following discus-sion treats the measures taken to obtain agood ECCM design.

The antenna feed system was configured toprovide peak azimuth sidelobes of theorder of 40 dB down, while achieving theelevation beamshape required for precisetarget height estimation. The patterns ofthe antenna design were calculated using anRCA -developed computer program,DISH, a tool for observing and analyzingreflector -antenna pattern characteristics.

The elevation patterns have low sidelobesbecause the beams are formed by excitingadjacent feedhorns.

Elevation patterns obtained from the feedand reflector geometry were calculated andare shown in Fig. 3, which gives theelevation beam set (group) for verticalpolarization. The patterns were calculatedfor a frequency of 3.35 GHz. The fourlowest beams are formed by exciting sets ofthree adjacent feedhorns (the lowest beamis formed by exciting only two horns) inorder to keep the elevation sidelobe levelsbelow 24 dB and to obtain the beamwidthsnecessary for proper beam -crossover levelsfor target heightfinding functions. Forexample, to obtain the second -lowestbeams, horns 1, 2, and 3 (Fig. 4) are excitedsimultaneously, with horn 2 receivingabout 92 percent of the power and theremainder evenly split between horns 1 and

FEED ARRAY

TOP SHIELD

REFLECTOR

6 .4- BOTTOM SHIELD7

Fig. 4Feed -reflector assembly includes severalmeasures used to suppress wide-anglesidelobe responses in the elevation dimen-sion. The microwave absorptive shieldingreduces the radar's susceptibility to anti -radiation missiles and electronic in-telligence devices.

3. Both the beam locations and crossoverlevels are dicatated by the required ac-curacy of height estimation. Elevation -pattern performance showed that, as ex-pected, the maximum gain and minimumbeamwidth correspond to a beam radiatingalong the axis of the paraboloid.

The azimuth patterns have a constantbeamwidth and low sidelobes.

Whereas the elevation beamwidth in-creases as the beam is scanned in elevationfrom the axis of the paraboloid, theazimuth beamwidth is virtually indepen-dent of elevation beam position. This is theresult of positioning the feed along theazimuth focal locus of the reflector.

In addition, the illumination of the reflec-tor is tailored in azimuth to provide thevery -low -azimuth sidelobes required foroperation in an ECM environment. Thisconstant beamwidth and low sidelobe per-formance is demonstrated in Fig. 5, whichshows calculated azimuth patterns cor-responding to beams 2, 4, and 7.

For successful operation of a radar antennain an ECM environment, the antenna mustradiate very little extraneous energy.

In other words, the antenna patterns havevery low sidelobes. Low -azimuth sidelobesnear the main beam were obtained bydesigning the feed for peak center illumina-tion, with sharply tapered illuminationtoward the sides of the reflector. The edgeillumination is typically -20 dB or less,resulting in peak azimuth sidelobe levels of-40 dB or lower in the absence of reflectorerrors or surface deformation.

Since the reflector will be fabricated as aself-supporting monocoque or honeycombstructure, manufacturing -produced sur-face errors will be random across thesurface. The average sidelobe level causedby a 0.020 -in. (rms) surface tolerance will

0

-10

7.4 -20

-302

-40

rc -50

70

-10

-20

2>

cc -50

5

Beam 2

0 15 20 25 30 35 40 45AZIMUTH ANGLE (DEGREES)

Beam 4

15 20 25 30 35AZIMUTH ANGLE (DEGREES)

Beam 7

40

25 3510

AZIMUTH ANGLE (DEGREES)

45

Fig. 5Predicted azimuth patterns of elevationbeam positions 2, 4, and 7 respectively.Achievement of low azimuth sidelobes is aprimary objective of the antenna design.Computed patterns such as these, using theRCA simulation program DISH, guided thedesign to that end. Quantitatively, the ob-jective is to provide azimuth sideloberesponses below -35 dB within 5° of themain beam, and below -50 dB outside of thatazimuth angle region.

63

be lower than -70 dB. Wind -inducedreflector deflection will lead to slightlyincreased sidelobe levels in the vicinity ofthe main beam, but for the 0.18 -in. max-imum deflection specified in the design, theimpact on sidelobes is not significant.

Aperture blockage by the feed is often animportant source of close -in sidelobes inmost reflector antennas. Typically, the feedstructure of a multiple -elevation -beamantenna is located directly in front of thereflector, creating an inherent blockage. Asshown in Fig. 4, the feed system for thisantenna is offset substantially, therebyminimizing the blockage. In the case of thisantenna design, the maximum blockagearea is 6 X 8 in., which results in a sidelobelevel below -60 dB. The blockage contribu-tion to close -in sidelobes is thereforenegligible.

The primary contributor to wide-angleazimuth sidelobe energy is feed spilloverpast the edge of the reflector. In this design,spillover is controlled by a combination ofdirectivity of the feed and reflector in-tercept intensity. The spillover sidelobe willbe no greater than an acceptable -50 dBrelative to beam peak.

Extraneous energy radiated into the con-ical region 45 degrees from the zenith is alsocaused by feed spillover. This would resultin sidelobes of about -35 dB in the vicinityof the zenith. However, much of thisspillover energy can be eliminated by plac-ing an absorbing shield at the top of the

Fig. 6One -tenth scale model verified low-sideloberesponses predicted by computer simula-tion. Operating frequency covers 31 GHz to36 GHz. Test feed horns were designed toprovide either horizontal or vertical linearpolarization.

reflector, as shown in Fig. 4. Suppressionof spillover by this direct means has beenexperimentally demonstrated. Spilloverenergy below the bottom reflector is

prevented in similar fashion, as shown inthe sketch. A similar layer of absorbingmaterial will be applied to the surface of thebaseplate to make it nonreflective.

Since the reflector will have a solid, ratherthan a mesh, surface, no energy will

0 dB

-10 d131--

-20 dB

E5

g- -30 dBkg

. 0 -40 dB

-50 dB

108 72

0

-10 dB

-20 dB

c63 -30 Mt

¢O

-40 dB

-50 dBF

108 72 36

penetrate the surface and producebacklobes.

Experimental resultsA one -tenth scale model of the antennaverified that the desired sidelobe responsespredicted by computed pattern simulationsare attainable.

The model's reflector parabolic surface,with aperture dimensions of 2 ft width by

0 36-116111721/Ail1

08AIL-

ANGLE - DEGREES

ANGLE - DEGREES

Fig. 7Measured azimuth patterns of beam position number 4 at 33.5 GHz (top) and 36 GHz(scaled -up frequency taking model size into account) for the vertically polarized mode.These low sidelobes are typical of the key results of the development program, whichverified that an antenna structure compatible with mobile, tactical environments can also bemade to provide a high degree of immunity from ECM threats.

64

1.2 ft height, was machined from solidaluminum to a tolerance of better than0.0015 in. rms. Fig. 6 shows the assembly ofthe scaled reflector, a five -horn segment ofthe feed group, and the absorptiveshielding above and beneath the reflector,all mounted in test position.

An indoor anechoic test range was chosenas the pattern test site since it couldmeasure antenna sidelobe responsesreliably down to levels below -60 dB (thiscapability is difficult to obtain in mostoutdoor test ranges).

Fig. 7 shows vertically polarized azimuthpatterns measured on the scale -modelfeed/ reflector at two different operatingfrequencies, with peak near -in sidelobesbelow -40 dB diminishing rapidly to lowerthan -50 dB, in accordance with the designobjectives. In general, wide-angle sidelobesare below -60 dB. These measured resultscorrelate well with the computed pattern ofbeam No. 4 (shown in Fig. 5) in terms ofmaximum near -in sidelobe responses.

Fig. 8 shows how effective the absorptiveshielding is in suppressing wide-angleelevation lobes in the zone of interest inrelation to anti -radiation missiles. Thispattern spans the angular region from

0 dB-

-10 dB

-20 dB

rc

9,3

a.;-,( -30 dB

'.3tW2cc 0

-40 dB

-50 dB

"

A/

108

directly overhead (at left) to verticallydownward (at the right of the plot). Theeffectiveness of various widths of ab-sorptive shielding at the top of the reflectoris shown at the left, where upward radia-tion as a function of shield width isillustrated for widths equivalent to full-scale dimensions of 5 ft., 3 ft., and zero areoverlaid. A nominal three-foot shield limitsupward sidelobe radiation to less than -46dB.

Implication/applicationof results

The overall result of this effort is anantenna design for tactical radar with:1) wide instantaneous bandwidth (600MHz at S -band); 2) low azimuthalsidelobes; 3) low sidelobes in a cone of 45°half -angle centered at the zenith (below-44 dB in a small region and below-S0 dBelsewhere); 4) polarization diversity, in-cluding orthogonal, linear, and circular;5) electronically steered beam over 20° inelevation compatible with one wide -bandwidth signal -processing channel.

This ECCM antenna demonstrates that, byapplying design techniques specificallydirected at sidelobe suppression, excellent

72 36 0

ANGLE - DEGREES72 108

Fig. 8Measured wide-angle elevation patterns of scale model. Strong spurious high -angleradiation increases the susceptibility of the radar to anti -radiation missiles. Most of thishigh -angle response is due to direct radiation propagating around the upper edge of thereflector from the feedhorn assembly. A moderate amount of shielding provides adequatesuppression-dashed curves marked A and B show radiation pattern with small and noshield, respectively.

sidelobe performance can be achieved witha reflector antenna. These techniques in-clude: 1) much lower edge illuminationthan has been customary, at a modestsacrifice of aperture efficiency; 2) edgeshielding to reduce the sidelobes caused byspillover at the edge of the reflector; 3) feedoffset configurations that minimizeblockage of the aperture by the feed struc-ture; 4) shielding techniques that minimizehigh -angle radiation from the feedhornstructure.

This demonstrated capability shows thepotential for the continued effective use oftactical radars of modest complexity andsize-at modest cost-in an era of hostileenvironments for radar operations. Andthis potential, of course, implies evengreater pressures on antenna designers inthe future, to meet the inevitable demandsof ever-growing sophistication andcapability of ECM techniques.


Bob Scudder has had approximately 30years of professional experience at RCA inthe design and development of antenna andmicrowave systems applicable to surfaceradars. This has included antennasassociated with air -search and height -finding equipment, precision trackingradars, and phased -array antennasdesigned to support defense weaponssystems.Contact him at:Advanced Technology ProgramsMissile and Surface RadarMoorestown, N.J.Ext. PM -2420

65

A short-range radar for measuringblast -furnace burden height

H.C. JohnsonR.W. PaglioneJ.P. Hoffman

Radar's high reliability, ability to operate continuously, and lackof mechanical parts can find work in unusual locations.

To improve the control of iron -producingblast furnaces, key furnace variables mustbe measured more accurately and reliably.One of these variables is the height of theburden (iron-ore/coke/limestone mixture)in the furnace. Each furnace has anoptimum burden height determined byoperating and design considerations.Deviating from this optimum height lowersefficiency and, in an extreme case, cancause furnace damage.

Problems withelectromechanical systemsPresently, two electromechanical methodsare commonly used to measure burdenheight.

They use: 1) a bar known as a stockrod or2)a steel cable and heavy weight assembly.

MICROWAVE ELECTRONICSAND ANTENNA

Both methods measure burden height bymonitoring the distance that the rod orweight travels before contacting the burdensurface. These electromechanical methodsseem ideally suited for an environment assevere as the interior of a blast furnace, butproduction experience shows otherwise.With either method, burden heightmeasurement is inaccurate if the equipmentis not operating freely or if the rod orweight contacts a fluidized region of theburden and penetrates too deeply. Evenmore important, both systems have arecord of frequent breakdowns. Thus, themain reason for developing an alternativemethod of measuring burden height was toimprove measurement reliability.

System requirements

Reliability, maintainability, and accuracywere the major system requirements.

RCA and Bethlehem Steel worked togetherat producing a new measurement system.At the outset of the program, they es-tablished the following system re-quirements:

high system reliability, to be gained byminimizing the use of mechanical com-ponents;

easy equipment serviceability, to be

gained in particular by eliminating theneed for components that are installedinside the furnace;

Reprint RE-23-5-141Final manuscript received January 13, 1978.

SHUT-OFF VALVE

Fig. 1Microwave measuring system, placed at the top of the blastfurnace, avoids temperature extremes and problems withmechanical contacting systems. Blast -furnace burden (iron-ore/limestone/coke mixture) enters furnace through the tophopper, past charging bell and fills furnace as shown. Radarmeasures distance to burden surface within 6 inches.

QUARTZ WINDOW

ADAPTER FLANGENITROGEN PURGEOPENINGS

SHUT-OFF VALVE

Fig 2Microwave instruments are external to the furnace and protectedby the microwave -transparent quartz window and continuousnitrogen purge.

66

measurement accuracy of ±6 inches;

measurement range up to 35 feet fromthe furnace charging bell, as opposed to15 feet or less for the conventionalstockrods;

fully automatic operation; and

continuous operation (the mechanicalstockrods must be raised above thecharging bell while material is beingdumped into the furnace).

These requirements led to a radar -basedmeasuring system.

RCA and Bethlehem jointly developed aprototype measurement system; it wasinitially constructed and tested at RCALaboratories in 1976 and then installed andevaluated on the "D" blast furnace atBethlehem's Burns Harbor, Ind., plant.Initial startup problems were overcomeand the equipment continues to be used forroutine control of burden height. Based onthe successful operation of the prototype,RCA developed two improved productionsystems, which were installed on the "C"blast furnace at Burns Harbor. They arecurrently being used in production tocontrol furnace burden height. Designdetails of the production systems will nowbe described, in addition to results of theprototype evaluation program.

Mechanical equipment

The microwave instruments require a clean,dirt -free opening to the furnace interior anda protective enclosure for the microwavecircuits.

The furnace had to be modified slightly forthe microwave system. Fig. 1 is a simplified

exposed view of the upper section of "C"blast furnace. Four stockrod pipes arelocated at 90° intervals around the conicalsection of the furnace exterior. Onestockrod pipe houses a mechanicalstockrod, the pipe located 180° from themechanical rod is unused, and the remain-ing two pipes are used for the rangingradars.

While the furnace was down for a reline (acleaning and resurfacing procedure done tothe inside of the furnace approximatelyevery five years), the following items,shown in Fig. 2, were added to the existingstockline indicator pipes:

a furnace shut-off gate valve;

an adapter flange with a quartz window;

provisions for purging the window withnitrogen gas; and

an enclosure to protect the microwaveelectronics and antenna.

The quartz window is transparent tomicrowaves and is required to passmicrowave signals into the furnace interior.Nitrogen gas is introduced immediatelybelow the window, through ports in theadapter flange, to prevent the buildup ofcarbon and iron dust deposits on thewindow.

Electronics

The electronic equipment consists of threebasic subsystems;

1)microwave electronics and antenna,installed in a stainless steel enclosure ona rebuilt stockline indicator pipe;

Fig. 3System has three major components-microwave electronics/antenna enclosure (left) and signal -processing enclosure (right),which are installed on the furnace top, and display enclosure,which goes in furnace control room.

FREQUENCY

f2

fi

2)processing electronics, also located onthe furnace top and installed in astainless steel enclosure; and

3) display and computer interface elec-tronics, located in the furnace controlroom.

Fig. 3 shows the three electronic systemsbefore installation.

The cw-fm (continuous -wave frequency -modulated system) determines burdenrange by measuring the instantaneousdifference in frequency between thetransmitted and received signals.

Fig. 4 illustrates the principle underlyingthe method of measuring burden height. Asshown, a microwave oscillator ismodulated in a manner that produces aswept transmission of frequencies. At sometime, to, a frequency, fi, is transmitted. Thissignal travels through the space in thefurnace until it strikes the burden surfaceand is reflected back to the microwaveelectronics. This signal arrives at sometime, ti. However, now the transmittedsignal has changed to a different frequency,f2. The difference in frequency, fb, betweenf2 and fi is proportional to the rangetraversed by the signal and is given by:

fb = 4R Aff,lc (1)

where

fb = beat frequency or V2 - fil (Hz)R = range to the target (ft.)

of = swept frequency bandwidth (MHz)f,= modulation frequency (Hz)

c = 983.573 (ft. -MHz); speed of light

- TRANSMITTED SIGNAL--- REFLECTED SIGNAL

TIME

Fig. 4Continuous -wave frequency -modulated system determines rangeby measuring the instantaneous difference in frequency betweenthe transmitted and received signals. Because signal frequenciesvary with time, frequency offset of reflected signal will depend onthe time the reflected signal takes to return.

67

Fig. 5 is a block diagram of the microwavecircuitry that generates and acquires thedifference frequency signal. Themodulated microwave signal is passedthrough a directional coupler and cir-culator and is transmitted to the burdensurface through a 22-d B -gain waveguidehorn antenna. A single antenna is used toboth transmit and receive the signals. Themixer produces an output signal with afrequency, fb, equal to the difference infrequency between the instantaneoustransmitted and received signals.

Fig. 6 is a simplified block diagram of theentire radar system. All the electronics,

MICROWAVEINPUTFROMOSCI LLATOR

13 -dBDIRECTIONALCOUPLER

including the microwave oscillator, employhigh -reliability solid-state devices.

One difficulty often encountered in usinghigh -accuracy, close -range radar systemsis that any drift in the modulationparameters will produce an error in range.

To prevent such drift errors, calibrationcircuits, shown in Fig. 6, are incorporatedinto the production version. They aredesigned to maintain accurate rangemeasurements regardless of modulation -parameter (f,,, and LJ) drifts caused bycomponent aging or temperaturevariations. These circuits operate asfollows: a portion of the transmitted signal

BALANCED

MIXER

CIRCULATORrf

.0 OUTPUTTOANTENNA

Fig. 5Microwave circuitry sends and receives signals throughfrequency -difference output in mixer.

MICROWAVEOSCILLATOR

10dBCOUPLER

OUTPUTo TO

PREAMP

same antenna, then produces

MODULATION

ISOLATOR

COUPLER- MIXERCIRCULATORASSEMBLY

FOR CALIBRATION

COUPLER -MIXERCIRCULATORASSEMBLY

FOR TARGET

TARGETPREAMP

t DELAYLINE

CALIBRATIONPREAMP

DISPLAY

SIGNALPROCESSING

A ANTENNA

QUARTZWINDOW

COMPUTERINTERFACE

Fig. 6Entire system incorporates microwave circuitry of Fig. 5, calibration circuitry to avoidproblems with component drift, and separate signal -processing and display/interfaceelectronics.

is directed through a separate coupler -mixer -circulator and sent down a coaxialdelay line of fixed length that has beenshorted at the far end. The signal is

reflected at the shorted end of the cable andreturns to the mixer, where it is combinedwith a portion of the transmitted signal.This produces a beat frequency determinedby Of and fm. This "calibration" beatfrequency is divided down digitally by thesignal -processing circuits and used as agate to count the target beat frequency. Achange of the modulation parametersresults in a corresponding change of boththe calibration and the target beat fre-quencies. Since the change in the calibra-tion beat frequency produces a cor-responding change in the counter gateinterval, the target frequency measured bythe counter remains constant.

The calibration preamp includes a band-pass filter with 3 -dB gain points at 94 and109 kHz. The modulation parameters cantherefore drift as much as 15% withoutaffecting the measurement accuracy. Theactual delay for the shorted coaxial line is151.5 ns, which corresponds to a nominalfrequency of 102 kHz.

The target preamp includes a bandpassfilter with 3 -dB gain points at 12.5 and 40kHz. These frequencies correspond to arange between the large charging bell in itsopen position and a stockline of ap-proximately 35 feet. The signal -processingcircuits are shown schematically in Fig. 7.These circuits are contained in an enclosurethat is also mounted on the "furnace top"next to the microwave componentsenclosure.

The target and calibration preamp outputsare connected to tuned -gain -responseamplifiers in the electronics enclosure.

The target amplifier includes a 40 -kHz low-pass filter with a 6-dB/octave gainresponse in the passband. This shapedresponse compensates for the fall -off inreturn -signal amplitude that occurs withincreasing burden distance. The calibra-tion amplifier is a 90 -kHz high-pass filterwith a flat gain response in the passband.

A portion of the amplified signal from eachamplifier is used as a measure of return -signal amplitude. These signals are con-verted to a dc voltage and compared topreset threshold values by the signal -leveldetector circuits. The output of the calibra-tion signal -level detector is used as anoperational check for the system and the

68

output of the target signal -level detector,which varies considerably during normaloperation, is used as a squelch.

The target and calibration amplifiers con-nect to target and calibration phase -lockedloops (PLL). The calibration PLL is asingle loop with a long time constant totrack the slowly varying calibration fre-quency. The target PLL has two series -wired loops with shorter time constants.

The output of the calibration PLL and thetarget PLL are connected to the divider -counter circuit. The target frequency isdivided by three and counted over a gatetime interval determined by the calibrationfrequency. The calibration frequency isdivided by an adjustable constant rangingbetween 41160 and 49096. This produces agating rate between 1.9 and 2.7 Hz,depending on the modulation parameters.The total number of target pulses countedin this gate interval is equal to 100 times therange in feet from the microwave receiverto the burden surface.

The three most significant digits aretransmitted in BCD format to the control -room display.

Although these BCD digits represent therange in feet from the microwave receiverto the burden surface, the furnace operatorgenerally is interested in knowing thedistance in feet from the large bell in itsclosed position to the burden surface. Thisdistance reading is developed in the displayenclosure by digitally subtracting a user -specified fixed distance from the totalrange distance. This offset distance isdetermined by the furnace geometry andnormally is between 15 and 25 feet.

The display enclosure also contains astockline command circuit, which com-pares a user -selected burden level with themeasured level. When the burden level fallsbelow the user -selected level, the commandcircuit sends the furnace computer a signalrequesting that more burden materials beadded to the furnace. The furnace operatormay select burden levels between 5 and 10feet in steps of one foot.

Evaluation ofthe prototype systemThe six-inch accuracy goal was met, and theunits have operated reliably so far.

Following construction, temperature andmeasurement -accuracy tests were con -

TARGET-SIGNAL -

FROMTARGET

LEVELDETECTOR

0 SQUELCH

TO

TARGET TARGETPHASE.LOCKEDLOOP

DIVIDER -COUNTER

BCD

DIGITSPREAMPOUTPUT

AMPLIFIER

DISPLAYENCLOSURE

FROMCALIBRATION

0 CALIBRATIONCALIBRATION

AMPLIFIERPHASE -LOCKED

LOOP

PREAMP -OUTPUT

CALIBRATIONSIGNAL -LEVELDETECTOR

OPERATIONAL RADARCHECK WARNING

+15 TH V

Fig. 7Signal -processing circuits are also mounted at furnace top. Calibration works byautomatically scaling the indicated target range by the ratio of the true to indicated delayline range. Target preamp is tuned to compensate for lower returned -signal amplitude atgreater measuring distance.

35

30

25

20

15

10

5

0

MICROWAVESYSTEM

BELL -TO -BURDENRANGE

STOCKLINEINDICATOR PIPE

MOVABLEBURDEN

\MATERIAL

0 0\\\\\\\\ BELL \\\\\ \\\\I I I

5 10 15 20 25 30 35

ACTUAL BELL -TO- BURDEN RANGE -FEET

Fig. 8Experimental setup showed that microwaveaccuracy over the 35 -ft. range.

ducted on the prototype system at RCALaboratories before shipping the equip-ment to the Burns Harbor plant. Measure-ment accuracy was determined by con-structing a horizontal target range tosimulate the physical geometry inside thefurnace. Fig. 8 shows this experimentalsetup together with test results. For thesetrials, a pipe of the same size and shape asthe furnace stockline indicator guide -pipewas modified to accept the quartz window.Sheet metal, placed approximately 16 feetfrom the antenna, simulated the largecharging bell. Samples of burden were

system worked within the desired 6 -in.

attached to a movable 6 -foot -squarevertical surface. The trials conducted withthis setup showed that the microwavedistance measurements agreed with theactual distances to within the desired 6 -in.accuracy over a range up to 35 feet from thebell. For the temperature trials, the rangereadout was monitored while equipmenttemperatures were varied from -20 to+140° F. which is the expected range forthe externally mounted radar unit. Thevariations in range readings, at a range of30 feet from the bell, were less than 3 inchesover the entire temperature range.

69

Following the trials at RCA, the equipmentwas installed on the Burns Harbor "D"blast furnace. After initial equipmentdebugging, the equipment was calibratedand production evaluation was conducted.Although a rigorous determination of

8

6

4

2

0

8

6

4

2

0

MICROWAVE

burden -height measurement accuracy wasnot possible, because there was no accurateindependent reference in the furnace, apractical estimate of the accuracy wasobtained by comparing the microwavemeasurements with mechanical stockline

I

MECHANICAL

/-/IIft-100 SECONDS

cSLIP

Fig. 9Comparison of microwave and mechanical systems used in same furnace shows goodagreement. Burden height is taken as the distance from the large charging bell to the burdensurface. Burden "slipped" in fifth trial shown here.

Bob Paglione has ten years of experience inthe microwave field. He is now working withrf and microwave equipment for heatingbiological materials. Besides designingnumerous microwave components, he hasalso developed a computer -controlledvariable mismatch used in designing solid-state microwave amplifiers.Contact him at:Microwave Technology CenterRCA LaboratoriesPrinceton, N.J.Ext. 3167

Henry Johnson has 19 years of experiencewith the Microwave Technology Center. Hisrecent work has centered on commercialradar applications-besides the blast -furnace radar project described here, he hasalso worked on collision -avoidance radarsand a microwave speed sensor forlocomotives.Contact him atMicrowave Technology CenterRCA LaboratoriesPrinceton, N.J.Ext. 2549

Authors Paglione (left) and Johnson with the microwave electronics and antennasubsystem for the blast -furnace radar. Coil of coaxial cable at top is for calibration,rf/transmitter electronics are at center, and transmit/receive antenna is at bottom.

measurements averaged over many charg-ing cycles and under various furnaceoperating conditions. Fig. 9 shows typicalresults obtained from the mechanical in-dicator and the microwave equipment dur-ing eight consecutive burden charges. Thefirst three recordings show rapid burdenmovement, while the remaining five showslower movement. On the fifth recording, asmall burden "slip" of approximately onefoot in height occurred. Based on the datafrom the experimental trials at RCA andthe analysis of operating data of the typeshown in Fig. 9, we judge that the equip-ment indicates burden height within ap-proximately 6 inches of actual burdenheight.

AcknowledgmentsGrateful acknowledgment is given to L.M.Zappulla, E. Mykietyn and J.E. Brown ofRCA Laboratories and to G.R. Crossley,W.E. Swan and T. Koselke of BethlehemSteel for their many constructive con-tributions to this effort.

John Hoffman has a background in elec-trical engineering and 17 years of ex-perience with Bethlehem Steel's ResearchDepartment. His recent work has centeredon the development of instrumentation forblast -furnace applications. Examples in-clude the development of infrared imagingequipment to measure temperature dis-tributions, tests involving the use of neutrongages to measure coke moisture, andparticipation in the development of themicrowave equipment described here.

Contact him at:Research DepartmentBethlehem SteelBethlehem, Pa.215-694-7102

70

B.D. Buch S.L. Clapper

Microcomputers for radar systems

Radar designers are using microcomputers to distributecomputer -processing load throughout their systems forlower cost, improved thruput, greater versatility, and easiersystem development.

R.J. Smith

General-purpose digital computers havebeen used to control multifunction radarsfor the past ten years, but not withoutproblems. In particular, the need for ex-tensive software development and its con-trol have been very costly. One alternativeis to develop special-purpose hardware;however, this alternative is also costly withthe additional drawback of inflexibility.Microprocessors offer a third alternative.

Special-purpose programmable micro-processors have already been used in radarfor such functions as coordinate conver-sion. Antenna -array beam steering andtracking also have been prime candidatesfor microprocessor -based implementation.Removing such functions from the general-purpose computer and distributing thempromises several benefits:

Lower cost-a smaller general-purposecomputer can be used. Removing well-defined computational functions permitsmore computer resource for control anddecision processing. This also implies theuse of higher -level command languages inthe general-purpose computer forreduced software cost.

Higher traffic rates-total processingcapability is increased by adding morespecial-purpose processors.

Expandability-the system can be up-graded to higher traffic rates by addingmore special-purpose processors.

Software/ firmware management-thesystem partitioning forced by distributingthe processors reduces the size of the largere-entrant program resident in thegeneral-purpose processor. Each special-purpose processor is identified as a radarfunction, and design review techniquesfor firmware similar to those forhardware can be instituted.

associated with controlling a radar by ageneral-purpose computer, new problemareas arise because a gap exists between themicroprocessor as an LSI chip and itsactual realization as hardware andfirmware in a system. LSI microprocessorsrepresent complex subsystems, but theyrequire supporting hardware and softwareto become microprocessor -based systems.

In considering a microprocessor* for aparticular application, several factors mustbe evaluated:

Speed-Is the microprocessor fastenough for the application?

Word size-Does the microprocessorhave the required precision?

Architecture-How is the micro-processor system best configured for theapplication?

Interface-How is the processor inter-faced with other system components(memories, control registers, busses,etc.)?

Packaging-How can the micro-processor system functions be packagedto be consistent with existing packagingsystems yet be flexible enough for futuresystems?

Programming-Does sufficient supportexist so that the microprocessor systemprogramming can be performed?

'The terms microcomputer and microprocessor are often usedinterchangeably, yet there are important differences in theirmeanings. A microprocessor is the small central processing unit(CPU) that performs the logic operations in a microcomputersystem. In the microprocessor or CPU, instructions are decoded,arithmetic and logic operations are performed, and timingsignals are generated. However, much more hardware-such asmemory, a clock, and interface circuits-may be necessary forfull-scale "computing." This additional hardware, in conjunctionwith the microprocessor, is called a microcomputer.

Although a distributed processing ap- Reprint RE -23-57-7proach solves some of the problems Final manuscript received January 31, 1978.

MSR's approachto distributed processingRecent efforts at MSR have addressed thedevelopment of the operational elements(hardware and software) required for dis-tributed processing in radars.' The firststep in this development is the selection of amicroprocessor family; the second step ispartitioning the designs into functionalmodules that can be used with each other indifferent ways so that a given requirementcan be met with a microcomputer matchedto that need.

The questions of speed, word size andinterface have been considered versus theapplication requirements, and thehardware approach developed is based onbit -slice functional modularity. We use abit -slice bipolar microprocessor (Ad-vanced Micro Devices, Type 2901) forspeed, and other functional modules(RAM, PROM, bus interface, etc.) whichcan be interfaced and configured for wordsize in a variety of processing architecturesas a function of the application.

Six functional modules have been identifiedand implemented.

The microprocessor module uses fourAMD type 2901's in a 16 -bitmicroprocessor module. The module isdesigned to be a 16 -bit slice so that a 32 -bitprocessor can also be configured. Themodule uses seven other LSTTL (Lowpower Schottky TTL) MSI/ SSI devicesand dissipates approximately 4.0 W ofpower.

The microprogram PROM module is a 512word X 40 bit PROM (programmable read-only memory) to hold microprograms. Itconsists of 17 bipolar devices and dissipatesapproximately 4.3 W of power.

Buch, B.D.; Clapper, S.L.; Smith, R.L.; "Bit -slice module set formicrocomputing"-private correspondence.

71

The macroprogram PROM module is a 2kword X 16 bit PROM used to hold themacro -level program. It consists of 10bipolar devices and dissipates ap-proximately 5.1 watts of power.

The RAM module is a lk word X 16 bitRAM (random-access memory). It uses 16CMOS/ SOS RAMs for optimumspeed/ power tradeoff and three bipolardevices (19 devices total). The typicalpower dissipation is estimated at 0.2 W.

The bus interface module contains tri-stateregisters and drivers and is used to "hold"the various system configurations together.It consists of 14 LSTTL devices anddissipates approximately 1.0W.

The goal is to incorporate high -densitystate-of-the-art packaging techniques withno loss of generality.

This goal was met by implementing thefunctional modules on 3.8 X 1.78 -in.

Richard Smith has 15 years experience indesign and development of radar signalprocessing architectures. He was responsi-ble for development of microprocessor -based radar applications for beam -steeringcontrol, tracking, coordinate conversion,and servo loops.Contact him at:Advanced Technology ProgramsMissile and Surface RadarMoorestown, N.J.Ext. PM -3319

Bruce Buch has contributed to design anddevelopment of special purpose microcom-puter systems since 1975. He is presentlyengaged in developing signal and mul-

ceramic substrates with active devicesenclosed in leadless hermetic packages.

The large packages in Fig. 1 are the leadlesscarriers holding the 2901 microprocessors.The smaller packages are the MSI/ SSIdevices. Each module is implemented as alarge 72 -pin (36 pins/ side) dual -in -linepackage and is made to be mounted onstandard 0.3 -in. centers. This type ofmodule can be used on either a printed -circuit board or wire -wrap board.

Software/firmwaredevelopmentSince the hardware module set is subject tochange as new devices are introduced, themain challenge is to produce firmware thatdoes not become obsolete as the hardwarechanges. This has been accomplishedthrough the development of "architectureindependent" support software used inconjunction with a PDP 11/03 -basedmicrocomputer development system.

tiprocessor architectures for beam -steeringcontrol of phase array radars.Contact him at:Advanced Technology ProgramsMissile and Surface RadarMoorestown, N.J.Ext. PM -3587

Stephen Clapper has been involved indesign and development of special purposemicrocomputer systems since 1975. He hasalso contributed to various signal -processing designs since 1969.Contact him at:Advanced Technology ProgramsMissile and Surface RadarMoorestown, N.J.Ext. PM -3587

Authors (from left to right) Stephen Clapper, Richard Smith, and Bruce Buch.

72

Program development using the actualhardware is accomplished by using macro -ROM and micro -ROM simulator modulesas shown in Fig. 2. These modules areloaded from the PDP 11/03 microcom-puter but are functionally and mechanical-ly identical to the micro -ROM and macro -ROM modules used in an actual applica-tion. The difference between the simulatormodules and actual modules is speed. Thesimulator modules are slower since aninstruction decode requires that controlbits be obtained from the floppy disk.

Another approach to program develop-ment is to use a simulator. Simulators for16- and 32 -bit microcomputer systems havebeen developed to run on any PDP 11minicomputer.

Microcomputer systemAlthough the hardware modules developedat MSR allow for architectural variations,many applications in radar systems can besatisfied using a standard processingarchitecture. The 16 -bit microcomputer ofFig. 3 represents a general-purposearchitecture that can be tailored to aspecific application by using microcodeand/ or by adding special purpose process-ing hardware (e.g., multipliers). A 16 -bitmicrocomputer with 2k words ofmacroprogram and lk words of data RAMrequires 12 modules total. The architecturecan be extended to form a 32 -bitmicrocomputer by adding more modules.A 32 -bit microcomputer with the 2k wordsof macroprogram PROM and RAM maybe added if required. The microcycle timeof this 16/32 bit microcomputer is 250 nswith a typical register -to -register add timeof 500 ns at the macro (assembly language)level. Provision has been made for micro -programmed two's complement multiplyand divide. The 16 -bit multiply requires 19microcycles or 4.75 As and the 16 -bit dividerequires 29 microcycles or 7.25 As.

Although the microcomputer has beendesigned so that it can be tailored for aparticular application using microcode, ageneral-purpose assembly -level instructionset has been implemented for branching,memory/ memory moves, andmemory/ register moves. These in-structions are modeled (not exact emula-tion) after the PDP-11 instruction set;actually, they form a subset of the PDP 11instruction set and use the same assemblermnemonics so that a source program canbe written on a PDP 11 and run on themicrocomputer.

11 IISdT

11.11maas. otwil :

microprocessor

fa

****100Ur, llllllllllllllllll tilt tnnu uttuutuur uutatuutak,

data RAM

micro ROM

macro ROM

-..mnmvumrommnnmmrm,rnninvmynnowwm.,

111 II IISill

controller

011110R a ` At 2 2

1*bus interlace

Fig. 1Functional microcomputer modules. This approach permits use of advanced packagingtechnology to incorporate high density packaging without loss of generality.

PDP 11/03MICROCOMPUILSYSTEMWITHFLOPPY DISK

MICROCOMPUTER SYSTEMBEING DEVELOPED

Fig. 2Microcomputer development system. This system uses standard I/O ports to create a testbed for performing macro- and micro -level programming for the basic module architec-tures.

RESTART DMA IN

Fig. 3Basic microcomputer architecture. This drawing shows the module types and quantitiesneeded (in parenthesis) to implement a 16 -bit microcomputer. Note that the moduleslabeled BUS IN and BUS OUT are actually one module type, the BUS INTERFACE module.The major functions in the 16 -bit microcomputer are: 2901 16 -bit microprocessor (CPUfunction); an operand select function (OPSEL) used to select 2901 register file addressesfrom either micro or macro program; a macroprogram address register (MACADR); aMACRO -ROM module for macroprograms; a 2909 control register (RAMAR) and DATARAM module; two holding register/driver modules (DRAML) for RAM data (also DMAinput/output); and a register/driver module for transferring 16 -bit constants from themacroprogram ROM to the 2901 input (MACC).

Modular firmware concept

The basic microcomputer of Fig. 3 is thefocal point for using hardware modules asa building block to construct "computer-like" hardware designs. A similar conceptapplies to the firmware.

Many radar algorithms are common tomany radar systems. If these commonalgorithms can be identified andmicroprogrammed, elements of a radar -processing language can be evolved. Thefirmware concept entails development ofmany special-purpose microprogramsfound in radar systems, so that the systemdesigner can select those microprogramsuseful to his application. Once PROMs areburned in, he then essentially has a special-purpose computer for his unique applica-tion.

Applications

Several radar functions have beenidentified for microcomputer implementa-tion. As summarized in Table I, these fallinto three categories: off-loading maincomputers by processing well-definedfunctions in special purpose micro-processor systems; replacing functions nowperformed in a minicomputer where thegeneral-purpose nature of a minicomputerhas more processing power (and cost) thanrequired; and hardware reduction by usingmicroprocessors to perform functionstraditionally done with hardware. A fewspecific examples are described below.

Coordinate conversion can be performed inless than 3 ms using a microcomputer.

A basic computational task in a radarsystem where the antenna is located on amoving platform is to convert the variableplatform coordinates to a fixed referencecoordinate system. This task can bedelegated to one of the system computers.Since the coordinate conversion must beupdated at frequent intervals (i.e., 5 ms)and the computations involve generatingsines and cosines, considerable CPU time isconsumed. Furthermore, operation withlarge word size (i.e., > 20 bits) is required.If this task can be removed from the centralcomputer and done with a bit -slice micro-processor, the central computer loadingcan be reduced significantly.

In the coordinate conversion process, amatrix based on the variable platformcoordinates must be multiplied with areference matrix. Fig. 4 shows the typical

73

Table IMicroprocessor application to various radar system functions.

Application Function

Off-loading main computers Coordinate conversionTrack functionsRadar return processingCommand output processing

Minicomputer replacement DisplaysCheckout and monitoringServo computersFire control computers

Hardware reduction Beam steering controllerPost processing (monopulse, interpolation, etc.)Input/output buffer to signal processorData formatting (asynchronous data collection)Data bus controllers ("smart" busses)

PURPOSE: REDUCE THE LOADING OF CENTRAL COMPUTER

PROBLEM:

PLATFORMMOTION

GYRO DATACONVERTER

DATA SMOOTH ANDMATRIXGENERATION

CENTRALCOMPUTER

TWO 'A' MATRICES BASED ON TWOPREDICTIONS OF SHIP'S POSITION

3 ANGLES (ROLL PITCH,YAW ...C,Z El

DATA SMOOTHING (PERFORM FOR EACH ANGLE)

Bo- -A20 n, -A30 4-R4On_, +A50n)/A6

(B18n-4 B26n-3-A36n-2 -Been-1 +Ben1/B6

WHERE FIRST C, then Z, then E

A MATRIXICALCU LATE TWICE FOR Ti AND T2 )

a11= COS C COS Z + SIN C SIN E SIN Z

a21 = SIN C COS E

a31= SIN C SINE COS Z- COS Z SIN Z

a12 --SIN C COS Z + COS C SIN E SIN Z

a22 = COS C COS E

a32= COS C SIN E COS Z + SIN C SIN Z

a73= COS E. SIN Z

a23 =-SIN E

a33 - COS E COS Z

C = HEADING, E = PITCH, Z = ROLL

PREDICT 3 ANGLES AT Ti AND T2

g -1) = Bn +901T1 -To

60(721 =en+ 601T2 -T01

ENTIRE COMPUTATION MUST BE PERFORMEDIN 5 MSEC.

6COMPUTER SOLUTION PERFORMS COMPUTATIONIN 2.9 MSEC.

Fig. 4Coordinate conversion problem. Variable motion of the antenna'sconverted to fixed reference coordinates and updated every 5 ms.

PDP11/34

0

FFT CONSTANTSIN PROM

2901 MICRO-PROCESSORSYSTEM

i

},256 x 256 x 64BIT DYNAMIC RAMDATA ARRAY(INTEL 1600)

HARDWARE:MULTIPLIERS,SCALER,BIT REVERSAL

platform must be

32 BIT PROCESSOR

FLOATING POINT IMPLEMENTATION

250 NSEC OPERATIONS

RADIX -2 FFT

PROGRAMMABLE 16, 32, 64, 128, 256 2-0 FFT'S

Fig. 5Two-dimensional fast Fourier transform performed via microprocessor.

calculations required. Three positionangles (roll, pitch, yaw) are digitized, thensmoothed using a polynomial filter. Therate of change for each angle is alsodetermined so that the platform positioncan be predicted at some future time (T1and T2). The predicted position angles areused to calculate two 'A' matrices. Thesetwo matrices are then multiplied with thereference -coordinate matrix.

A bit -slice microprocessor system has beenused for this type of coordinate conversionprocess in the AEGIS Air Defense System.This processor represents an early bit -slicedesign-a 24 -bit implementation usingstandard AEGIS modules. Themicroprocessor is the AMD 2901 and has a380-ns clock time. This allows computationof a 24 -bit sine in 98 is. This processorrepresents a significant off-loading of oneof the AN/ UYK-7 system computers.

Data Formatting Unit (DFU) is another bit -slice microprocessor implementation inAEGIS.

The Data Formatting Unit processes theinterfacing data between the ship's datasensors and the command computers. Thefunction of this processor is to gatherasynchronous data (e.g., 1, 2, and 3 bits at atime) from the various ship sensors, placethe data into standard formats (e.g., 16- or24 -bit words), and , upon command,transfer the formatted data to the centralcommand and decision computer. Thisprocessor is the same basic AMD 2901implementation as the AEGIS coordinateconverter.

Microprocessors are being used to speed upthe two-dimensional fast Fourier transformfor antenna pattern measurements.

Testing and analysis of phased array anten-nas requires far -field pattern measurement.One technique for doing this is to take near -field measurements and transform thesenear -field samples to far -field patternsusing a two dimensional (2-D) fast Fouriertransform (FFT). In testing the phasedarray, a large amount of data is collectedand the requirement exists for many 2-DFFT computations. This requires con-siderable computer processing, and resultsin a long delay between measured data andprocessed data used for analysis. Oneapproach to reducing this time is theimplementation of a bit -slice micro-processor system, microprogrammed toperform the FFT algorithm. This FFTprocessor is used in conjunction with aminicomputer (PDP 11) as shown in Fig. 5.

74

PROGRAMSTORAGE

MICROCOMPUTER AND1

4 4 CONTROLf, SIN (a)

MICROCOMPUTER BUS SUB -ARRAY 1

. . . .

BLOCKS REPRESENTREGISTER FILE ANDPHASE SHIFTER

MICROCOMPUTER2

MICRO-PROCESSOR

1

141°Y 4.. I 1 4

SUB -ARRAY 2 SUB -ARRAY 1ARRAY

MICRO-PROCESSOR

2

SUB -ARRAY 2MICROCOMPUTERFig. 6

Phased -array -antenna beam steering viamicrocomputer.

Nr 4 4 4

MICRO-PROCESSOR

NSUB ARRAY N4 4 4

Since the processing is two dimensional, alarge random access memory is required.Furthermore, special-purpose hardwaresuch as TRW's 16X16 -bit multiplier isemployed to speed up the radix -2 FFTcalculations. Communication between theFFT processor and PDP 11 computer is viaa direct memory access (DMA) channel.This application is a good example of howthe basic microcomputer of Fig. 3 can betailored with special-purpose hardwareand microprogramming to implement avery specific hardware function.

An approach for microprocessor steering ofa phased array has been developed andproposed for future shipboard systems.

The technique used incorporates a "read- while write 'B' " type register file at

each phase shifter. This enables the array touse the phase words for transmit andreceive during radar dwell period N whilethe bit -slice microprocessor microcom-puter is loading the phase words for dwellperiod N+ 1 .

Since the loading of the phase -shifterstorage elements is done sequentially, themicrocomputer can interface to the phaseshifter through a single bus for data, phase-

shifter address, and control as shown inFig. 6. The dwell time or time to computenew beam positions can be speeded up bypartitioning the array into subarrays andusing a microcomputer to load each sub -array in Fig. 7a. This approach, however, iswasteful because of the redundancy ofhaving many microcomputers executingthe same routine. The redundancy occursin program storage hardware and can beeliminated by using a multi -processormicrocomputer. The multiple micro-processors execute a single instructionstream using constants which correspond

a. Muiticomputer beam steering control.

SUB-ARRAY N

b. Multiprocessor beam steering control.

Fig. 7

New beam positions can be computed more rapidly if the array is partitioned for separateloading.

to a particular subarray. This architectureis shown in Fig. 7b. This approach to beam -steering control, along with the hybridpackaging technology will result in aphysically small beam -steering controllerthat can be located near the antenna. Thiswill reduce the cabling required to interfacethe beam -steering controller to the array.

The beam -steering controller is yet anotherexample of how the module types describedearlier can be used to implement differentcomputing architectures.

Another proposed microprocessor applica-tion is checkout and monitoring.

This approach is based on distributingmicroprocessor checkout -and -monitoring(CAM) systems based on functional andcabinet -level partitioning. Each micro-processor CAM (µCAM) is tied intohardware test- and data -monitoring points.The µCA Ms sequence through testroutines that query the individualhardware points and format the data forblock transfers to the central processor.The central data processor manages thedistributed µCAM system on a functionalbasis.

Faults are detected at the subsystem levelthrough continuous monitoring by eachµCAM. When a fault occurs, the µCAMsenses this and notifies the central dataprocessor via interrupt. The central dataprocessor then queries the µCAM todetermine the nature and seriousness of thefault. Any further diagnostics or fault

isolation routines are then directed by thecentral data processor. Each µCAM willcontain "canned" diagnostics for faultisolation at the least -replaceable -unit levelof hardware and will report the failedhardware unit location back to the centraldata processor.

This approach to checkout and monitoringeliminates the large data base requirementsof a central computer approach since thedata required for fault detection and isola-tion (test -bit patterns) are distributedthroughout the hardware.

ConclusionsThe microprocessor in a radar system is acomponent to be sold as a part of a largerproduct, and can be used to implementthose functions which are more or lesscommon to all radars. The microprocessoras a device is not particularly significant inthese applications; however, the micro-processor as a system function is. Theimpact of the microprocessor in radarsystems is related to the degree to which itremains a flexible device that can be usedfor multiple functions.

The key issues, therefore, in applyingmicroprocessors to radar systems, are notthe device itself but the software andfirmware required to make it into a radarsubsystem component. The questions to beresolved are ones dealing with systemarchitectures and firmware development.The modular approach to hardware andfirmware described here is directed towardsmicroprocessor use in that context.

75

Programmable processors andradar signal processing-an applications overview

M.G. HeroldM.O. Timken

Marty Herold has been involved with thedesign of signal -processing systems sincejoining RCA in 1957. His experience in-cludes analog computing, servos, frequen-cy synthesis, receivers and digital circuitand system design. Since 1970 his primaryassignments have involved systems designof real-time programmable signalprocessors for communications and radarapplications. He is currently the systemsengineer responsible for design anddelivery of a programmable, real-time,digital communication receiver.

Maurice Timken joined MSR in 1957. He hasextensive background in signal processingtechniques and has designed anddeveloped these subsystems for applicationto radar systems such as TRADEX, BMEWS,ASFIR, and AN/FPS-95. In his presentassignment as a unit manager in the SystemEngineering Department, he is responsiblefor development of a programmable signalprocessor for communications application.

Contact them at :

Systems EngineeringMissile and Surface RadarCamden, N.J.Ext. PC -4514

Authors Marty Herold (left) and MauriceTimken with an SPS-81 ProgrammableSignal Processor in the background.

Developers of programmable signal processors are con-tinually attempting to outreach each other with greaterthroughput or programming ease, or both.

In radar, signal -processing functions areperformed on received signals so thatdesired target returns are enhanced andnoise, clutter, and other extraneous signalsare de-emphasized.* In early radars, signalprocessing was an analog -circuit function;later came hardwired digital logic circuits;the more recent trend is toward program-mable methods.

Today's radars place wide -bandwidthhigh-speed demands on the signalprocessors; these processors must also beadaptable to changing needs. Thus, speedand flexibility establish the architecturaltradeoffs for selecting the best signalprocessor. Speed can be achieved in avariety of ways, principally by means ofperforming functions simultaneously or inparallel; flexibility is achieved by software.

At one extreme of the tradeoff is thehardwired approach-maximum speed,little flexibility; at the other extreme is thegeneral-purpose computer-a lot ofprogramming flexibility with high-levellanguages, but orders of magnitude belowthe speed required for radar processing.Programmable processors fall in a wideregion between these two extremes, ap-proaching one or the other limit,depending on the degree of parallelism inthe architecture and hence the complexityin programming. No clear criteria can beset forth to identify an optimum level offlexibility (programmability) and speed(parallelism). The system design goals in-volved, however, are quite clear: maximumthroughput, with optimum use of availableresources, and without an overcomplicatedprogram.

" Walt Weinstock describes the radar signal -processing (unction

in some depth elsewhere in the issue.

Reprint RE -23-5-8Final manuscript received February 1, 1978.

What's a programmablesignal processor?All programmable signal processors are"standalone" units controlled from a hostminicomputer. The software package in thehost minicomputer provides programassembly, control, and debugging capabili-ty. Programs for signal -processor opera-tion can be stored on disk or tape andloaded under control of the host minicom-puter. Therefore, all peripherals availableto a computer (teletype, CRT terminal,printers, disk operating systems, magnetictapes, etc.) can be integrated into a signal -processing module through the hostminicomputer.

Programmable processors are usuallydelivered with a software library packageproviding such functions as array multiply,add, subtract, and combinations thereof,plus more complex functions such as com-plex multiply, FFT (fast Fouriertransform), recursive filter, and powerspectral density. The sophisticated user canadd functions of his own to the library. Theattractiveness of the array library softwareis that it provides a means of quicklyapplying a programmable processor to acomputation problem, since the functionsare Fortran -callable.

The major drawback to use of array librarysoftware for real-time signal -processing isthe efficiency lost in overhead. That is, eachlibrary function called requires the com-puter to perform extra operations (useextra memory and time) needed to linkthese various "canned" programs at theprocessor level. For real-time signal -

processing applications such as radar, theprocessor must be programmed inmicrocode to achieve the desiredthroughput. Programming at themicrocode level, called microprogramm-ing, requires that the user develop in -

76

structions practically at the logic designlevel.

What happens inthe real world?

Consider a typical situation that mightemploy a programmable signal processor.The usual case is a multiplicity of basebandsignal processing computations that wouldotherwise require many dedicated digitalprocessors. For instance, a typical radarmay require FFT, doppler filtering, rangeand monopulse angle estimation, inter-polation, etc. If a programmable processoris fast enough to perform all of thesefunctions, we have a prime applicationcandidate. (The programmable processoris misused if dedicated to a single heavy-load function such as complex translationand filtering for preprocessing-the samplerates alone could take all the resources ofthe processor and you're not able to use itsprogrammability.)

Two considerations are of prime impor-tance when designing a programmableprocessor into a system:

Partitioning of processing function forefficient "overlay" management.

Efficient flow of input and output data.

The functional partitioning of processingfunctions into overlays (smaller programsegments) is necessary since programmableprocessors have relatively small programmemories and thus cannot hold the totalsignal -processing repertoire. The programoverlays are generally kept in large bulkmemories, easily accessible to theprocessor. These program segments arethen called and placed in program memoryonly when the functions they perform areneeded in real time.

The dynamic swapping of programsegments in program memory imposes thefurther requirement for sequential batch-

mode processing of data queued up intoblocks. This mode requires the completionof each processing program segment (on alldata) before proceeding to the next signalsegment. Such a characteristic leads tomemory size tradeoffs to reduce theoverhead associated with program swapp-ing.

The input/ output flow of data and accom-panying control strategies must becoordinated. Some processors cannotreach out and obtain data. Instead, blocks

DIGITAL COMPLEXVCOINPUT

DATABUFFER

LOW PASSFCFILTER

-0/O-4.LOW PASS2

FILTER0

WEIGHTINGFILTER

Fig. 1Typical processing scenario. Hypotheticaltime, frequency, and display processing.

FFT

POSTPROCESSING

1-1

-- T:72

TOHOSTCPU

SPECTRALDISPLAY

multiple -function radar processing involving

_0ST

I

SPS-81 SIGNAL PROCESSORCOMPUTER

HARDWARECOMPUTER

CPU MULTIPLIERPERIPHERALS

CORE BULKMEMORY

I MEMORY

L__ INPUT-OUTPUT

FCI SPEEDPROCESSORBJFFER

DATA INPUT-. DUAL PORTMEMORY

ARITHMETICPROCESSOR

D/ACONVERTERIT

_

CRTDISPLAY

Fig. 2Signal -processing system. A typical application architecture involving host computercontrol, display, and data/signal processing flow from input to computational outputdeposited in the host computer memory.

of data and/ or control must be moved intoa processor memory before processing canbegin. Output is the reverse process. Inmost cases, processors must be stopped forexternal program changes by the hostCPU. If uninterrupted continuous process-ing is necessary in real-time controlsituations (e.g., operator inputs to add orsubtract processing functions or changeparameters), an indirect means of control isneeded.

Because these processors operate in-dependently of the host minicomputer, andbecause control is exercised indirectly via acontrol program in the mini's memory, anexecutive control program must bedesigned into the software system of theprocessor itself. These control programscan range from very rudimentary in-structions to high-level control programs.The form of control advocated here is acontrol program residing in the signal

processor which can fetch controlstatements from the host minicomputer'smemory, interpret the requirement, andsend to the proper control subroutine thatpart of the control program required toaccomplish the intent of the control in-struction.

A typical problem

Consider the typical processing segmentshown in Fig. 1. The process shown in-volves an AFC loop for doppler extraction,with complex translation to baseband. Thebaseband signal envelope is detected forvarious post -processing uses while the in-put spectrum is analyzed in parallel fordisplay and other automatic, spectrum -dependent post -processing functions.

Assume a typical programmable processorsuch as the S PS -8 I (SPS, Inc.) which has asystem architecture as shown in Fig. 2. This

77

system provides a natural partitioning offunctions. A dual -port memory is

provided: one port buffers input data froman external processor; the second port isused by the processor to fetch dataindependently. The SPS-81 has a bulkmemory to store program overlays, a high-speed buffer for scratch -pad and interimstorage, a hardware (hard -wired) mul-tiplier, and a DMA (direct memory access)interface to deposit results directly in hostminicomputer's memory. Results can alsobe output via the D/ A converter fordisplay.

By functional partitioning, we match theprocess to be computed with the internalprocessor.

Filtering, AFC feedback, complex transla-tion, FFT, and envelope detection areperformed in the arithmetic section. All ofthese processes involve various com-binations of multiplies and adds for whichthe arithmetic processor is designed. Postprocessing, which generally involvessearching, thresholding and branching(i.e., single -thread processing) is performedby the input-output processor, which is, infact, a highly efficient microprocessor. Thehardware multiplier speeds computationsand avoids the need to commit thearithmetic section (and its setup overhead)to randomly programmed, single mul-tiplications.

Functional partitioning of the coding forthe arithmetic processor depends verymuch on the small program memory of thearithmetic section. For example, amicrocoded instruction for the arithmeticsection has 96 bits per word, but only 16instructions can be stored at once. Theindex section, which drives the arithmeticsection, can store 64 program instructions;its instruction has 32 bits. Each of theseinstructions is very powerful, as evidencedby the number of bits in the instructionword, so that much can be accomplished ineach instruction.

Experience in programming the SPS-81indicates that our sample problem of Fig. 1can be partitioned into three overlays:translation filtering with AFC, filteringwith envelope detection, and FFT withweighting.

Next the partitioning of the process and themultiple channel capability of the input-output processor must be considered.

The highest -priority channel can beused to output data to the D/ A converter

at a clocked rate. The clock rate will bevery slow compared to the speed of theprocessor, and the high -priority channelis used to guarantee a steady output forflicker -free display.

The next -highest -priority channel canbe used to control and move data in andout of the arithmetic section.

The next -lower -priority channel wouldcontain the post -processing functions.

The lowest -priority channel wouldcontain data output to the host minicom-puter and the executive program.

This prioritized partitioning provides for anatural flow of control from the highest -priority processing channel to the executivein the lowest. When a processing segment iscomplete on a block of data, the processingof the higher -priority channels stops, caus-ing control to fall through to the next lowerpriority channel, etc.-finally revertingcontrol back to the executive program. Theexecutive then fetches the next controlinstruction from computer memory forappropriate action.

The architecture shown in Fig. 2 waschosen to facilitate optimum processing ina parallel pipeline fashion. The controlprogram is resident in the host computermemory where it can be modified by thehost minicomputer. Program overlays arestored in bulk memory from which high-speed overlaying of program memory ofany internal processor can quickly beaccomplished. Processing proceeds undercontrol of the executive program thatresides in the SPS-81.

The memory system is also partitioned anddistributed to facilitate efficient processingflow. Data is fetched from the dual -portmemory for input to the arithmetic section;interim results from the arithmetic sectionprocessing are then deposited in a high-speed buffer. Data volume at this point ismuch reduced, allowing use of the smallhigh-speed memory to hold interim results.The buffer is high speed in that its cycletime is the same as the input-outputprocessor clock.

The post -processing functions are thenperformed on these interim results. Theresults may be directly deposited in hostmemory or in bulk memory for later blocktransfer to the host minicomputer memory.The display data can be stored in any of thememories for continuous cycled output tothe D/ A converter.

Microcoding the arithmetic processor isakin to logic design.

Efficient coding of the arithmetic processorrequires a somewhat formal programmingorganization so that the programmer canstay abreast of events at each step of theprocess. Optimum coding requires max-imum use of the arithmetic section at eachcycle.The primary difficulty lies invisualization of all the assets and mul-tiplexer switched data paths whenattempting to fold in pipelined or parallelprocesses. The efficient programming andsubsequent debugging require carefulcharting of the processing flow through thearithmetic section.

For this purpose, a charting process hasbeen developed that is akin to logic design.In this technique, the sequential arithmeticsection cycles are charted out on aschematic format depicting the arithmeticelements, data paths, and memories. Thisprovides immediate visibility of the state ofthe arithmetic section at the end of eachcycle.

Once the program is mapped, program-ming requirements for addressing, func-tion selection, scaling, etc., are charted intoa table. This table is then analyzed to besure that arithmetic section waiting time isminimized. The coding plan is juggled tobest meet the waiting -time requirement.(At times, this juggling is unsuccessful andit may be necessary to modify the code.)

The input-output processor program isthen coded to support the arithmetic -section requirements as coded. Input-output processor constraints maynecessitate still another reshuffling of code.The microcoding of programmableprocessors, therefore, involves logic designwith cut -and -fit programming of thevarious internal processors for efficientthroughput. Processing is furtherenhanced via algorithm tradeoffs whichselect processing structures that best matchthe processor.

Speeding up the processFaster devices are an obvious means ofenhancing processing speed; however, theprimary challenge is in the selection ordesign of efficient signal -processingarchitectures and appropriate program-ming techniques. The sequential processingand high-level languages found in thegeneral-purpose machine have beenreplaced with highly parallel structures and

78

machine coding. Although programmingthese highly parallel processors is morecomplex, the potential for increasedthroughput is great. The aim is to providesufficient switching within the arithmetichardware to allow optimum programmingfor a wide variety of signal -processingfunctions. Both architecture andprogramming design are tightly coupledwith signal -processing techniques andthese skills are required for optimumdesign.

Architecture enhancements: the ultimategoal is to use 100% of the resources 100% ofthe time.

The most widely used and accepted conceptto improve processing speed is a multiple -processor architecture. This concept hasbeen introduced to enhance throughputwith a motive for efficient use of thearithmetic unit resources. By providing aseparate processor to perform the tasks ofinput, output, and control, the arithmeticunit can be fully dedicated to arithmeticcomputation. The approach is to providean arithmetic processor for computationonly and support it with another processorfor address computation, data movement,control, and other housekeeping chores.

The memory is in general the slowest devicein any signal processor computer structure.

Regardless of the degree of throughputcapability achieved with multipleprocessors and arithmetic unit resources,the cycle times of memories become thelimiting factor. Much engineering efforthas gone into the design of "smart"memory interfaces and memories to speedcomputer processing. This type of memoryinterface is programmed to allow fetchingor storing of data arrays at given addressincrements from a specified starting ad-dress, thereby reducing the load on thecontrolling processor for address com-putations.

One technique used to overcome thememory speed problem is to distribute thememories (e.g., separating program fromthe data storage to provide program in-struction fetches on a non -interfering basiswith retrieving and depositing results indata memories). Distributed datamemories and multiple bus structures arealso utilized to increase processing speed byminimizing memory wait times. Othertechniques applied to reduce memoryaccess time include interleaving memorymodules within a single memory on con-secutive addresses, and using a small, fast,

bipolar memory for scratchpad computa-tion.

Software enhancements: the tradeoffs in-volve programming techniques as well asthe language level employed.

A careful tradeoff must be made betweenthe required performance and com-putational burden. Both programming andsignal -processing ingenuity are essentialskills to make this tradeoff. But thepotential increase in throughput is high.

The primary effort in software develop-ment for computers has been towardhigher -level languages. This developmentproceeds from machine code, to assemblylanguage, to Fortran et. al. to even higherlanguages, such as those used for struc-tured programming. Each level abovemachine coding invokes some overhead,and thereby sacrifices processing speed forprogramming generality. Even assemblycode, when it is applied to a particularproblem, is not necessarily the mostefficient attainable for a particularproblem.

The microprogrammable computer breaksthis constraint by allowing the programmerto define new computational kernelsmatched to a particular problem, thusenhancing processing speed by moreefficient exploitation of the computerarchitecture. Microprogramming thereforerequires the partitioning of the process tobe coded into an efficient set of kernels,followed by microcoding of the kernel.

The coding of programmable signalprocessors has similar tradeoffs inlanguage level. All of the processormanufacturers provide a basic set ofprograms (commonly referred to as anarray processing library) which can becalled at the Fortran level. These librariesinclude a number of high-level processingfunctions so that a signal -processing func-tion can be programmed. Use of theselibraries, however, involves additionaloverhead and results in slower overallcomputational speed.

In general, the integration of the program-mable signal processor into a system forreal-time radar signal processing will nottolerate this loss in performance caused byextra overhead. It makes no sense to designa processor architecture for optimumefficiency and then throw away much of theperformance with setup overhead. To alarge extent, overhead will be proportional

to processor complexity. Integrating aprogrammable processor into a real-timesignal -processing system, then, imposes arequirement to microcode the applicationsoftware rather than use the librarybuilding blocks.

In summary, speed enhancement viasoftware has three main goals:

1) Selecting algorithms which best fit thearchitecture of the processor.

2) Partitioning processing functions tominimize overhead.

3) Microcoding the processor to obtainmaximum use of the programmableprocessor's capability.

ConclusionsThe programmable signal processor is oneof the most important developments insignal -processing applications today. Newprocessors providing greater processingspeed and/ or flexibility will continue to bedeveloped. The primary effort is in moreand more parallelism of one sort oranother. Efforts to provide higher -levelsoftware will also continue, to easeprogramming complexity while extractingprocessing performance. Developers arecontinually attempting to outreach eachother with greater throughput orprogramming ease, or both. Usually theease of programming is based on theavailability of a library of preprogrammedfunctions.

Although these libraries have their place,they extract a price in reduced efficiencybecause of their increased overhead. Eachlevel of parallelism adds, in effect, a level ofspecialization which makes it much moredifficult to develop a higher level languagethat takes advantage of the efficiencies ofthe parallel structure.

Application of this type of highly parallelstructure to maximize throughput, then,demands that the programming be equallyas efficient. The use of microcoding istherefore essential to overcome the in-efficiencies of high-level languages. Hencethe techniques employed are more akin tologic design. It is clear, then, that thepractical implementation of a program-mable processor into a system involves notonly tradeoffs of system functions,algorithms, and processor structures, butalso careful consideration of softwaredesign, encompassing the whole range oflanguage levels and associated assemblersand programming aids.

79

BibliographyI. Herold. M.G.; "Investigation of programmable signal

processors for processing communications type signals,"private correspondence.

2. Lucas, A.; "High level control of signal processing functions,"Proc. of 1977 IEEE International Conf. on Acoustics, Speechand Signal Processing.

3. Morris, L.R. and Mudge, J.C.; "Speed enhancement of digitalsignal processing software via microprogramming a generalpurpose minicomptuer." Conf. Record, 1977 InternationalConf. on ASSP, Hartford, Conn. (May 1977).

Comparison oftypical programmableprocessor architectures

The three programmable signal processors discussed hereare by no means the only commercially availableprocessors, but were selected because they depict a rangeof differing architectures and typify the considerationsdescribed in this paper.

Array Transform Processor (AP -120B, Floating PointSystems, Inc.). A multiple processor architecture featuringpipeline arithmetic elements, a single interleaved memory,and independent I/O processing.

I I

Some general comparisonsTable I compares the performance that can be achieved on typicalsignal -processing functions for these processors. Of additionalinterest, the MAP 300 and AP -120B are floating-point processorswhereas the SPS-81 is I6 -bit fixed point. Floating point providesgreater dynamic range and relieves the programmer of scalingworries.

For data I/O the AP -120B and MAP 300 feature independentDMA processors. The DMA processors are set up with a wordcount, source, and destination address, and increment by eitherthe processor or the host computer. They can then move blocks ofdata independent (and in the background) of on -going process-ing.

The SPS-81 has a primary difference in operation in this regard inthat all the memories are in effect on the IOP processor bus. Thereis no need to move blocks of data in and out since it can reach outand fetch data as needed and deposit results in memories ordevices requiring them when computed.

Table IPerformance comparison of three commercially available signalprocessors

Processing time, point

Function SPS-81 MAP 300 AP -120B

FFT (1024 pt. complex) 3.2 4.5 4.7Recursive filter (2 pole / 2 zero) 0.5 0.85 0.83

Complex mulitply 0.5 0.85 0.67

MULTIPLE38 -BIT DATA PATHS

TABLEMEMORY

HOSTCPU

I/O DMA

INTERFACE

1 I

DATAPAD

X

MULTIPLE

DATAPADY

16 BITS

sPAD I

MEMORY', -I 5 -PAD1

16 BITS

MAINDATAMEMORY

38-BIT DATA PATHS

Ml I M2PIPE LINED

Al A2 PIPE LINED

STAGE 1 STAGE 1FLOATING FLOATING POINT

POINT ADDERSTAG STAGE 2

MULTIPLIERSTAGE 3

MULTIPLE PATHS

The AP -120B does not function at nearly the multiple -processorlevel of the other two processors; rather it provides a parallelarchitecture of varying computational and memory elementscontrolled from a single program memory. The primaryarithmetic elements are a single multiplier and adder which can bepipelined to achieve speed.

Pipelining is a form of parallelism which, in the AP -120B case, isequivalent to three parallel multipliers and two parallel adders.That is, the multiplier of the AP -120B requires three cycles tocompute a result; however, additional multiplies can be startedbefore the result is obtained. The AP -120B multiply time is 0.5µs,but pipelining increases the effective multiply time to 167ns/ computation. The AP -120B adder is also 2:1 pipelined, with asingle multiply taking 333 ns and a pipeline rate of 167ns/ computation.

The architectural simplicity of the AP -120B arithmetic unit isdeceptive in terms of programming difficulties. To obtainprocessing efficiency, the programmer must constantly visualizethe time phasing of data in and out of memory and arithmeticelements. Also, the multiple processors of the AP -120B are nottotally independent, since they operate from a single programmemory.

80

Programmable Digital Signal Processor (SPS-81, SPS,Inc.). This processor architecture features a highly parallel,number -crunching arithmetic processor (IS and AS), sup-ported by a fast microcomputer -like I/O processor.

10P

TIMESHARELOGIC

PROG AMEM MEM

B

MEM

ALU

I/ODEVICES

IPERIPHERALMEMORY

HOSTCOMPUTER

Is_i_PROGMEM

AMEM

B

MEM

HALU

t 4

-FLAGS FCN COEF READ WRITE -

AS

PROGMEM -

COEFMEM

DATAMEM

-... ARITH

The three processors of the SPS-81 have independent programmemories and they operate independently. The machine issynchronous in that the clocks for each processor are derivedfrom a common source. The cycle time of the input-outputprocessor and index section is 167 ns. The SPS-81 can do a fullcomplex multiply (four real multiplies plus two adds) and two fullcomplex adds (four real adds) in 0.5 As. A hardware multiply inthe input-output processor can also speed processing when use ofthe complex arithmetic processor is inappropriate.

In addition to being one of three processors, the SPS-81 input-output processor is itself a four -channel processor, therebyfurther extending processor parallelism. A memory register file,program counters, and control are replicated four times. Opera-tion of the four channels is on a time-share basis. If a channel iswaiting for a memory cycle to complete, the next -lower -prioritychannel can perform a computation in the arithmetic logic unit,service a device, or fetch data from an alternate memory.

Macro Arithmetic Processor (MAP -300, CSPI, Inc.). Amultiple processor with a multiple -bus, multiple -memoryarchitecture and separate, independently programmableI/O processors.

HOSTPERIPHERALS

t t t HOSTCOMPUTER 0,,TORY

THE MAP PRODUCT LINEHOST INTERFACE SCROLL

IHISI

MEMORYBUS 1 MULTIPORTED

MEMORY16 PORTSMEMORY

BUS 2MEMORY

MEMORYBUS 3

MEMORY

CENTRALPROCESSOR

BI/O

SCROLL SCROLL

ARITHMETICPROCESSOR

(AP32 -BIT

FLOATING POINT

DIGITALDEVICES

ANALOGDEVICES

t t t

The MAP is also a federation of independently programmedparallel processors. The MAP is unusual in that is is a totallyasynchronous processor. Operations are not clocked as in otherprocessors; rather operations flow from one to the next as neededdata or operands become available. Processing proceeds at devicespeed, theoretically obtaining maximum computational speedavailable from given devices. The MAP -300 performs twomultiplies in 0.42 As plus two adds in 0.21 As (i.e., one in eacharithmetic element), for an effective rate of 0.21 µs and 0.105 Asrespectively.

The two arithmetic processors proceed similarly. As operandsbecome available in the arithmetic -element input registers,computations proceed. If compuation results are required asinputs to the next computation, the next operation proceeds assoon as the previous result ripples into the operand register.Synchronism between processors is achieved on a programmedbasis with flags and status bits.

81

Digital computer simulation of radar systems

J. Liston G.M. Sparks

This paper provides an overview of thekinds of simulation models, together withtheir application to radar systems designand analysis. The simulations discussed inthis paper are broadly categorized as real-time and non -real-time simulations. A real-time simulation includes real-time cyclingconstraints associated with operation in arealistic scenario. An example is a real-timeoperational simulation, which facilitatesthe routine evaluation and parameter"trimming" of an illustrative aircraft vec-toring system without the necessity forflight support aircraft. A non -real-timesimulation, on the other hand, is notconstrained by the real computation time;however, such a simulation is a powerfultool for designing and evaluatingalgorithms and functional inter-relationships in a complex radar -centeredsystem.

Within each of the foregoing categories,the elements of the system to be simulatedare categorized as discrete or analogprocesses. Discrete processes are usuallydescribed by difference equations and canbe directly programmed on a digital com-puter. These processes, such as thoseemployed by digital signal processors, areusually simulated on a general-purposedigital computer by simply using the For-tran (or other high-level language)equivalent of the specific algorithmassociated with the process in question.Analog processes are usually described bydifferential equations, so to program theseequations it is necessary to convert them toa set of difference equations. The initialform of the process may be a set ofdifferential equations or a transfer func-tion, G(s), which represents a set ofdifferential equations. The approach toconverting the analog process to a discreteprocess may be to solve the set ofdifferential equations and to sample thesolution at discrete intervals or to usetransforms such as the bilinear transforma-tion to develop the difference equationsdirectly from the Laplace -transformtransfer function, G(s). Recursive algo-rithms can be used for either approach.

Simulation is a necessity for radar systems as complex asAEGIS. Without it, problems and optimizing would appeartoo late in the design process.

Discrete and analog process simulationscan be realistically integrated to representthe operation of a complex radar system orperhaps certain critical elements of thesystem. This paper describes how theeffects of noise, thresholds, nonlinearities,system errors, quantization, and variousaspects of the system environment(propagation, multipath, etc.) are intro-duced and how each affects the overallsystem. It also gives specific examples onthe simulation effort that RCA Missile andSurface Radar has used in the design andevaluation a number of radar systems.

Simulation approachesMost processes encountered in the studyand analysis of radar systems can beclassified as either analog or discrete.Analog processes have a continuous timeresponse to a given excitation function and,as such, are best described in terms ofdifferential equations or Laplacetransforms expressed as functions of thecomplex frequency variable; i.e., G(s). Dis-crete processes, on the other hand, haveprocess variables defined only at aparticular set of time values, implying thatthe independent variable (time) is

quantized. Such processes are typicallydescribed in terms of difference equationswhich can be directly programmed forsolution on a digital computer.

One way of simulating analog processes ona digital computer is by solving the definingdifferential equation directly.

For some analog processes described by aset of N simultaneous linear constant -coefficient differential equations with Mtime -varying inputs, it is more efficient tosolve the equations and store the results asan N XN and an NXMmatrix in a data file,rather than use substitution techniquessuch as the bilinear transform. This datafile is then available to the appropriatesimulation program and can be executedby simple multiplications and additions atthe system data rate. Substitution methodsor numerical -integration techniquessometimes require integration time inter-

vals much smaller than the computersystem update time interval.

To illustrate this approach consider thescalar differential equation:

X(t) = AX(t) + BA(t) (1)

The solution to this equation is:

X(t) = eA(")X(tk)

+ J;:eA(') BAL(a)da (2)

where X(tk) is the initial condition and gois a time -varying input.

This is the solution to an analog process.For a discrete process, the equivalentsolution is

X[(k+1)1] = eATXACkn+ [e -1](BI 24) (kT) (3)

This equation then solves for a sequenceX(kT) for k = 1, 2,... using an inputsequence )2(k T) for k = 1, 2,... . Thissolution can be used in a feedback systemwhere µ (k7) is a function of previousvalues of X.

Eq. 3 is the solution to a single differentialequation with one input. Consider now Ndifferential equations with M inputs. Thesolution to these equations can be writtenin matrix form similar to Eq. 3.

The resulting matrix equation is written inthe following general form:

X[k+l)T] = aX(kT) ± 13µ(kT) (4)

where a and /3 are respectively NX Nand NX M constant matrices.

The key to this solution is to determine thematrix AT in order to arrive at a and /3. Thematrix el'. is computed by calculating theeigenvalues and eigenvectors of A andperforming the appropriate transfor-mations. This was a very convenientmethod, since the time-shared computersystem available at RCA MSR had libraryroutines which calculated eigenvalues,

82

eigenvectors, and matrix inversions. Aseparate program used these libraryroutines to calculate and store a and 18 sothat they were then available in a data filefor use by a simulation program.

In substitution methods, the differentialequation or its Laplace transform isrepresented as a discrete process thatclosely approximates the time response ofthe analog process.

Fryer and Schultz' describe variousmethods of simulating system transferfunctions on a digital computer. Thesemethods yield solutions that are expressedin terms of a recursive or difference -equation form, through the substitution s =(2/ T)(1-A)/ (1+A) when T= time intervalbetween data points and A = delayoperator e-sT.

ao+al A+a2A2G(s)-G[f(A)]=

1+biA+b2A2 +. .+b,A'n(7)

The corresponding difference equation orrecursion formula is written by inspectionas

Y. = aox. + +....+ amxrm

(8)

which can immediately be programmed ona digital computer. Fig. 1 gives a flowchartfor such a program. Table I is a short tableof bilinear transforms which give the valuesof the coefficients (a's and b's) in terms ofthe parameters of the transfer function G(s)and time interval T. We note that thebilinear transform can be cascaded; that is,if G(s) can be factored into the form

G(s) = Gi(s)G2(s)

Gi(s) and G2(s) can be programmed intotwo independent blocks of code and theoutput of the GI(s) simulation used toprovide the input to the G2(s) simulation.The resulting output of the G2(s) block,under this condition, will simulate theoutput of G(s).

The difference equation that is the result ofthe bilinear transform process is easilyprogrammed; however, evaluating the con-stants ao, al b1, b2 ... can involve aconsiderable amount of tedious algebraicmanipulation, particularly in the case ofhigh -order polynomials ins. However, thisalgebraic manipulation can be carried outusing a digital computer, thus relieving the

Table I

Bilinear transforms (or Tustin transforms) convert Laplace transformsdifference equations that are easily programmed on a digital computer.

Network

Integrator

Lag network

Lead -lag network

INPUT STREAM OF X'SREPRESENT TIMESAMPLED EXCITATIONFUNCTION

into discrete

G(s) Yn

1/s (Ti 2)(xo - xo-i) - ye -1

11[(s1b)+ 1] [Tb/(2 + Tb)] x +[Tbl (2 + Tb)] x.-1 +[(2 Tb)/(2 + Tb)]

[(s /a) + I]I[(s I b) + I] [b(2 + Ta)]I [a(2 + Tb)][b(Ta -2)]/[a(Tb -2)] xn-i

+ 1(2 - Tb)I (2 + Tb)]

Fig. 1

Flowchart for program solving the difference equations of the type given in Table I.

Y = ai Kn-1

1=0

IS DELAYOPERATOR, i.e.

Yn = CTri CII Tri,1

Fig. 2Simulated digital filter is a result of using difference equations derived from knowledge oftransfer functions or by tracing the signal flow.

user of the burdensome algebraicmanipulation and reducing the possibilityof introducing errors in the process. Themethod assumes that the function G(s) tobe simulated is of the form of the ratio oftwo polynomials in s.

Simulation of discrete processes, such asdigital filters, on a general-purpose digitalcomputer is a straightforward procedure.

Fig. 2 shows illustrative digital filters andthe corresponding difference equation thatprovides a discrete time series of outputvalues, y,,, given a time series of input

values, x. Such difference equations canbe directly programmed on a general-purpose computer to exactly represent thediscrete process of interest. The differenceequations are derived from knowledge ofthe transfer functions appropriate to cer-tain classes of digital filters5; e.g., directform I, direct form 2, parallel form, or incertain simple cases, by tracing the signalflow as implied by the examples in Fig. 2.Noise, signal -amplitude quantizationeffects, and logical operations can also bereadily accommodated within theframework of a suitable high-level com-puter language such as Fortran.

83

In certain cases, the discrete process isdefined in terms of an algorithm, which canbe directly transcribed into simulationcode. For example, a specific fast-fourier-transform algorithm used in a radar digitalsignal processor can be directly coded intoa Fortran simulation of the processor.

Examples of small-scaleanalysis by simulationSimulation can help analyze the centralissue in many radar signal -processingsystems-detection performance.

In order to illustrate a generalmethodology for such cases, let us usedigital or Monte Carlo simulation todetermine the target detection probability,given a specific signal-to-noise ratio andfalse alarm probability. The theoreticalsolution to this problem is well known,6thus we can compare simulation -derivedand theoretical results to gain a feeling forwhat can be achieved in terms of accuracy.

Consider that the output from a lineardetector, which provides the envelope ofsteady sinusoidal signal plus noise, isstatistically governed by the modifiedRayleigh distribution. This process issimulated by forming the vector sum of asteady signal and independent in -phase andquadrature-phase noise samples from aGaussian distribution of zero mean andunity variance. A simple threshold detectorcan then be simulated by generatingvoltage samples in accordance with theabove procedure and comparing theresulting amplitudes with a threshold givenby

T = (-2 loge Pfr)lh (9)

where Pi -a is the false -alarm probability.

Some insight as to the accuracy of oursignal -plus -noise simulation can be ob-tained by calculating detection probabilityversus signal-to-noise ratio (S/ /V) forvarious values of false -alarm probabilityand comparing the results with theoreticalvalues computed with a highly accuratealgorithm.' This has been done in Fig. 3 for2000 Monte Carlo trials; correspondencewith theoretical values is excellent.

Simulation is not limited to cases in whichthe radar signal is steady.

It is possible to simulate amplitude samplesof the envelope of signal plus noise wherethe signal components are fluctuating.Such fluctuations arise due to changes in

99.99

990

998

99

98

95

90

80

70

50

ci 40

E, 30

'I' 20

010

5

2

1

0.5

0.2

0.1

0.05

one

)

//3 le,

43

E'"i-L8

SMOOTH CURVESVALUES GIVEN

PURPOSES

ARE -NOTE:THEORETICALFOR COMPARATIVE

2 4 6 8 10S/N (dB/

12 14 16

Fig. 3Good comparison exists between theoretical (smooth curves) and simulated probability ofdetection.

the relative phase relationships* among thevarious scattering centers which comprisethe target. The fluctuations can beclassified as either independent from sam-ple to sample (rapidly fluctuating) or total-ly correlated within a given stream ofsamples (slowly fluctuating). Here, thestream of samples might simulate thesequence of target returns which occur asthe radar beam scans past a target during asingle rotation (or scan) of the antenna.

Additional effects which influence theamplitude of the signal returns include theantenna pattern shape and reflections fromthe surface or other large objects in thevicinity of the radar. These are accountedfor in many simulation analyses, but willnot be discussed in detail here.

Radar system simuation is often called uponto provide a model of the dynamic effects ofa moving target of interest.

The simulation generates a realistic timesequence of radar measurements (obser-vables) in terms of range, azimuth, andelevation. These values are modeled to

Due to changes in signal frequency (frequency diversity) ortarget rotation relative to the radar line of sight.

represent the "true" position of the target.The resulting time sequence of observablesare then perturbed within the simulation torepresent the effect of radar -induced"noise" on the process under analysis.

Simulation showed how adding a sidelobeblanker affected system performance.

A sidelobe blanker inhibits receiverresponse to signals which enter the radarvia the antenna sidelobes. The gate intowhich the reference and blanker channelsrespectively feed passes the reference -channel signal -plus -noise only if theenvelope of the signal -plus -noise of thereference channel exceeds that of theblanker channel by a specified blankerthreshold, expressed in dB. If blankerthreshold is exceeded, a single pulse "detec-tion" or "hit" is declared if the envelope ofsignal -plus -noise exceeds the threshold es-tablished in a second gate. For linearenvelope detection, this threshold value isgiven in accordance with Eq. 9.

N successive beacon response samples,characterized by a given reference -channelsignal-to-noise ratio, are considered in thesimulation. The detection probability is

84

9929

99.9

99.8

99

98

95

90

80

70

L-60

0 5040

30

E20

10

5

2

0.5

0.2

0.1

0.15

0.01

BLANKER THRESHOLD2.632x103

RCVR

2000 MObITE-CARLO

GAINBELOWBEAM

GAIN

GAIN

= 7.5 dbI

= RCVR

TRIALS

PM=BLANKER

BLANKER

GAIN

20dbMAIN

OFFBLANKER

MAIN LOBE. BLANKER ON

I

EXCESSBLANKER

S/L GAINGAIN 1db)

BEYOND -

5.0

....,;<.--

4.0

Ilk3.5

3.0----__,--_ -

11111111416.

0

-.....N.s...

2 4 6 8 10S/N 1dB)

12 14 16 18

Fig. 4Sidelobe-blanker simulation showed the loss in detection (two curves)performance that occurs when sidelobes are blanked out as an electroniccounter -countermeasure. Set of 11 curves shows probability that sidelobe(S/L) signals will be falsely accepted as a mainlobe signal.

then computed as ratio of the accumulatednumber of "hits" to the number of trials, N.For each trial, independent samples of in -phase and quadrature-phase Gaussiannoise perturb the signal levels in therespective reference and blanker receiverchannels. The effect of the reference -pattern beam shape loss or sidelobeattenuation on the reference -channel signallevels is accounted for in establishingrespective signal-to-noise ratios in thereference and blanker receiver channels.Typical results obtained with such asimulation are shown in Fig. 4. Theseresults show the loss in detection perfor-mance which is introduced by the blankercircuit; they also indicate the probabilitythat a sidelobe signal will be falselyaccepted as a mainlobe signal for variousvalues of sidelobe gain greater than theblanker antenna gain.

Another program simulated a clutter filterand Fast Fourier Transform on an AN/MPS-36 radar modification.

This system problem was one ofautomatically detecting, acquiring, andtracking a high-speed artillery projectilewith a pulse doppler radar in the presence

COMPUTERDESIGNATE

*DIGITIZE I & Q

64 POINTS

11 10

CLUTTERNOTCH FILTER

32 POINTFFT

CLUTTERNOTCH FILTER

DETECT

YES

NO

ESTABLISHRANGE ANDDOPPLER TRACK

NO

ESTABLISHANGLE TRACK

Fig. 5Simulation of artillery -shell detection wasuseful in quantizing and scaling system duringdevelopment, then debugging in field. I and Qare in -phase and quadrature signals.

of ground clutter. The goal was to completethe detection and acquisition process in lessthan one second.

The signal -to -clutter ratio (S/ C) whenattempting to acquire the target was ap-proximately -30 dB. To acquire the targetit was necessary, via signal processing, toimprove this by 43 dB to an S/ C of 13 dB.

The signal processor had available in -phase (I) and quadrature (Q) signals. Eachof these signals was processed through athree -pole Cauer notch filter which gave astop -band rejection of 51 dB. The I and Qwere then processed with a 32 -point FastFourier Transform (FFT). The output ofthe FFT was 32 signals equally distributedover a radar pulse repetition interval(PRI). The target doppler frequency andamplitude was determined by examiningthese 32 signals and performing a sinx/ xinterpolation between the largest returns.A threshold was then examined and, if itwas exceeded, a target detection wasdeclared. A three -pole bandpass filter thenverified if this was indeed a target.

The entire detection process, shown in Fig.5, with a clutter signal, target signal, and

receive noise, was simulated on a digitalcomputer. The program was set up torandomly select both a PRF and a dopplerfrequency within the middle 50% of thePRI. Using an S/ C=-30 dB, 80 detectionswere made. The calculated doppler foreach detection was within ± 10 Hz of thetrue doppler. This simulation program wasalso very useful in investigating such thingsas the effect of quantizing and scaling. Itproved extremely useful in debuggingoperations in the field.

Examples of large-scalesystem simulations(non -real time)The AN/MPS-36 instrumentation trackingradar was the subject of a detailed computersimulation.

A data rate approximately equal to thepulse repetition frequency was used and allradar functions, modes, and conditionswere included. The basic modules of theprogram are:

Trajectory generator-target trajectory

Target effects-signal return from target

Power program and transmitter

85

Atmospheric effects-clutter and mul-tipath

Automatic gain control

Track servos-azimuth, elevation, rangeand doppler (range rate)

Systematic errors

Random errors

Executive control-controls system bymaking a priori and real-time decisions

In simulating missile targets, the launchpoint, impact point, and trajectory can beprogrammed and either a passive echotarget or a beacon -transponder target canbe selected. The passive target consists of acone -cylinder missile configuration withuser -selected dimensions and a resultingradar cross section varying realisticallywith viewing aspects. For the beacon targetan N -antenna -port system equally spacedabout the missile, including resultingpolarization and interferometer effects, issimulated. The simulation also includesrange delay and doppler shift for echo orcoherent -beacon targets.

The simulated electromagnetic environ-ment includes error contributions causedby atmospheric refraction, ground clutter,and multipath. The multipath effects con-sider the diffraction, intermediate, andinterference regions. The radar model in-cludes the effects of the radar dataprocessor so that the output data obtainedfrom either the real radar or the simulatedradar will have been identically processed.

All four of the radar tracking channels(azimuth, elevation, range, and range rate)were simulated using the same approach.First, detailed mathematical represen-tations of the track circuitry weredeveloped in the frequency domain. Theseexpressions were then transformed intodifference equations for digital simulation.

TPQSYM simulates the operation of theAN/TPQ-27 precision tracking radar, track-ing filters, and all the guidance algorithmsnecessary to guide a "simulated" aircraftmodel on a radar -directed bombing -missionflight path.

Ballistic drag tables are stored within theprogram for a variety of bomb types alongwith a fourth -order Runge-Kutta integra-tion routine to solve the ballistic equationsof motion for the falling bomb. Majorfeatures of TPQSYM are as follows:

All four TPQ-27 guidance modes areincorporated;

F4 aircraft flight dynamics is based onhigh -degree polynomial representationof aircraft using manufacturer's data;

lateral and pitch control of aircraft isprovided;

wind effect on the aircraft and bombtrajectory is accounted for;

earth's curvature between radar site andtarget is accounted for through a rota-tion matrix;

radar noise and bias effects can bespecified;

input data (R, A, E) can be obatinedfrom a recording of field -derived data(actual) or simulated on the basis ofinterval noise generators;

a wide variety of plots can be specifiedand made by the user; and

TPQSYM currently runs on MSR time-shared system in Fortran.

Fig. 6 shows a typical plot obtained fromTPQSYM. This plot represents thealtitude -control -channel response to aninitial altitude error of 200 ft, given an F4Baircraft at a nominal 2000 -ft. altitude and500 -knot airspeed.

A dynamic, single -target, weapon -systemsimulation measures AEGIS performanceagainst contemporary threats in hostile,uncooperative environments.

This simulation, known as SystemPerformance Evaluator Comprising

300

200

100

LL

0cc -

0

0

A-100

-200

-3000

Target, Radar, Missile (SPECTRM)models a target generator, the AN/ SPY -1phased -array radar, Mk91 slaved il-luminator, S M-2 missile, and ship's motionalong with the Mk19 gyro. Developed overthe past seven years, SPECTRM is stillundergoing changes which reflect fieldalterations, and additions which are in-cluded as new operating modes becomeviable.

To date, SPECTRM has had a variety ofuses, including:

initial verification that slaving an il-luminator to SPY -1 outputs was feasibleagainst a wide spectrum of targets;

error budget analysis (error sensitivityand allocation); design and verificationof a universal range -adaptive a, /3 -filterfor weapons -system users;

measuring AEGIS performance againstspecification targets; developing 'specialcase' system logic, such as low -E -modelogic and coast -mode logic;

determining system performance againstspecial threats;

assessing effect of face-to-face (array -to -array) handover errors on missile/ targettrack, verifying that single -face testingwas adequate;

assessing efficiency of various midcourseguidance laws for missile flight control;

comparing capabilities of severalproposed rocket motors;

200 FT INITIAL ALTITUDE ERROR

500 KNOTS TRUE AIRSPEED

COMMAND VERTICAL ERROR

ALTITUDE RATE

ACTUAL ALTITUDE ERROR

10 20 30 40TIME, SECONDS

50

30

20

10

0

-10

-20

60-30

Fig. 6Tracking -radar simulation is possible with TPQSYM program. Plot represents the radar'saltitude -control response to an initial altitude error of 200 ft.

86

developing missile `fly -out' curves; and

determining AEGIS performance in anECM environment.

SPECTRM contains about 5000 Fortranstatements, and requires approximately300,000 bytes of memory. The program isextensively modularized and, through theuse of control cards, only those equipmentsessential to the particular analysis need beexercised. In addition, the many systemnoise sources can be individually selected-or omitted entirely if a 'reference run' isdesired. When the entire weapons system isunder analysis, program running time isapproximately three times faster than realtime.

The MEDUSA simulation evaluates the per-formance of the AEGIS system againstmulti -target attacks.

MEDUSA (Multi -target EffectivenessDetermined Under Simulation for AEGIS)is written in Fortran and currently requiresapproximately 160k words of DEC20storage. It includes models of the SPY -1radar, Mk 26 launcher, Mk 91 illuminator,SM-2 missile Combat Air Patrol (CAP)and software logic to control theseequipments. Models to describe enemyscenarios are also provided.

The specific uses for MEDUSA are toprovide:

a test bed to aid the design and evalua-tion of AEGIS tactical algorithms

an operation analysis tool to assess thetactical performance of AEGIS

assistance for AEGIS test analysis

MEDUSA has been successfully used as atest bed to evaluate candidate SM-2scheduling algorithms and Threat Evalua-tion Weapon Selection (TEWS)algorithms. The modular structure ofMEDUSA allows the simulation of anunlimited number of candidates for eachAEGIS algorithm being evaluated. Thealgorithms for a particular run arespecified by input parameters. MEDUSAis also capable of acting as a test bed to aidthe design and evaluation of other WeaponControl Sys,tem (WCS) and Command andDecision (C&D) algorithms. MEDUSA isnot intended to be used to design AEGISequipment. Models of the SPY -1 radar,Mk26 launcher, Mk91 illuminator, and theSM-2 missile are part of the MEDUSA testbed.

Trajectory -generator subprograms tomodel threat targets are also included inthe test bed, as are default models for thealgorithms being evaluated. The test-bedmodels are intended to have sufficientfidelity to provide valid results in theevaluation of tactical algorithms or systemtactical performance. In general, this doesnot require that the models mimic in detailthe equipment or phenomena they repre-sent. Greatly simplified models are usedwherever possible in the interest of simula-tion computer program development andrunning time costs.

As system design continues, more specificrequirements for MEDUSA runs developand with them come concomitant re-quirements for additional sophistication insome of the models. The modular structureof MEDUSA permits ready modificationor replacement of the models (either thoserepresenting algorithms under test, orthose making up the test bed) as the needarises. In the meantime, evaluations requir-ing only simple models are not delayedpending availability of the "ultimate"MEDUSA.

MEDUSA is now available for both pre-test and post-test analysis for actualAEGIS tests at the Combat SystemsEngineering Development site. CurrentlyMEDUSA is being used to designscenarios before they are run on theAEGIS Interface System Simulator. Thisis an inexpensive way to uncover anomaliesin the scenarios.

Real-time simulationof large systemsThe develoment of radar command andcontrol computer programs frequently oc-curs parallel with hardware development.The computer systems now containfunctions which in the past were hard-wired. Many of these functions, such asclosure of angle servo loops, closure ofrange loops, target acquisition, targetdetection, and waveform selection, areimplemented in computer programs, mak-ing the computer system an integral part ofthe radar system design.

To test the system, therefore, it is necessaryto integrate the hardware and software. Asthe complexity of computer programs in-creases, a means of verifying the perfor-mance of the computer programs prior tohardware integration is needed to assurethat schedule and cost objectives can beobtained.

Real-time simulation, therefore, has animportant role to provide a real-time inter-face which can be used as a certificationand maintenance tool to:

reduce costs over the total program, asthe simulator

-does not require the use of a livehardware environment foroperational computer program testing

-allows for parallel testing ofhardwaPe and computer programprior to and after integration.

minimize risk in development by

-providing a means for early detec-tion of problems

-providing management withvisibility of the program -developmentcycle and allow for action to be takenin problem areas

-assuring that problems in the com-puter programs and/or hardware donot paralyze independent testing ofboth

provide a means of verifying computerprograms during acceptance testing andafter implementation of approved com-puter program change requests.

TPQSIM verifies the AN/TPQ-27operational program and conducts softwareacceptance tests in a controlled environ-ment; i.e., without the need for physicalradars and aircraft.

TPQSIM operates on the PDP-11/ 70computer. It simulates, in real time, an IFF(identify friend or foe) radar, a precisiontracking radar (PTR), a Tadil-C data link,and several aircraft. TPQSIM com-municates with a program running on anAN/ UYK-7 computer across a special-purpose hardware interface. The headingand attitude of each of the aircraft can bemodified by Tadil-C messages. Messagesconcerning IFF and PTR detections,beam -steering commands, weapon -releasesignals, etc., are sent between the twocomputers in formats and rates identical tothose used in the AN/ TPQ-27 operationalprogram.

TPQSIM is valuable because it:

exercises computer programs in realtime without requiring live aircraft;

provides simulated I/O channel inter-face data for IFF, PTR, and com-munications functions,

generates dynamic aircraft positionaldata on line;

87

includes functional tracking radarhardware simulation for angle -error andrange estimation;

includes functional IFF radar hardwaresimulation for antenna motion, azimuthestimation, and range estimation;

does not interfere with basic operationalprogram timing;

does not require modification of basicoperational computer programs;

synchronizes I/O using a real-timeclock; and

allows for test operator control.

The AEGIS Interlace Simulation Systemcomputer programs test the AEGIS tacticalcomputer programs.

The interface simulators consist of anintegrated, modular set of computerprograms that can be tailored to simulate

Glenn Sparks has been associated withMSR since 1954. During this period he hasbeen engaged in key radar system engineer-ing assignments emphasizing control -system analysis, computer simulation, com-mand and control, and estimation theory.Most recently he was involved in systemengineering of the HR -76 radar system.

Contact him at:Systems EngineeringMissile and Surface RadarMoorestown, N.J.Ext. PM -3037


Authors Sparks (left) and Liston

unavailable interfaces; e.g., comptuerprograms and/or equipment and externalstimuli; e.g., targets to the tactical com-puter programs. The interface simulatorsprovide a test bed to support the verifica-tion, integration, test, and acceptance ofthe tactical computer programs during thevarious stages of their development. Theinterface simulators are used to support:

the tactical computer program systemelement builds and Phase II tests at theComputer Program Test Site (CPTS);

multi -element computer -program in-tegration and test at the CPTS; and

Combat System integration and test atthe Combat Systems EngineeringDevelopment Center (CSED).

The ISS computer programs operate in setsof two -bay and three -bay AN/ UYK-7computers and use a common user

Jack Liston has held major systemsresponsibility for concept design, perfor-mance analysis, and design optimization onadvanced phased -array radar systemdevelopments. Some of the areas of hisprincipal contributions include system com-puter modeling for cost optimization andperformance tradeoffs, advanced self-duplexing receiver concepts, and systemperformance verification through simula-tion. He is now systems engineer on theAN/TPQ-27 radar.Contact him at:Systems EngineeringMissile and Surface RadarMoorestown, N.J.Ext. PM -3037

language (SCRIPT) and a common con-trol system.

The ISS computer programs includeseveral on-line programs which provideinput to exercise the following segments ofthe AEGIS combat sysem:

AN/ SPY -1A phased-aray radar-thebasic AEGIS sensor;

C&D (Command and Decision), whichdrives displays and reacts to operatorcommand;

ORTS (Operation Readiness TestSystem), which provides on-linemonitoring;

GDC (Gyro Data Converter), whichsimulates ship's motion;

WCS/ FCS (Weapons ControlSystem/ Fire Control System), whichcomputes launch parameters and aimingangles for the missile launcher; and

SYSTEM, which exercises all systemcomponents as a complete system.

Concluding remarksWe have presented a summary of the kindsof digital simulations which have been usedat MSR during the past few years to design,troubleshoot, and validate a variety ofradar systems. Specific examples have beendrawn from the AN/MPS-36, AN/ TPQ-27, and AEGIS programs. Suchsimulations have made a vital contributiontowards the development of these systems.

AcknowledgmentsWe are grateful to C. Falcon, D.M. Fuerle,J. Golub, and E.J. Hartnett for discussionand descriptive data on various AEGISsimulations. Helpful discussions with R.Lieber, N.A. Ricciardi, and P.J. Schickalso contributed to the preparation of thispaper.

ReferencesI. Fryer, W.D. and Schultz, W.C.; "A survey of methods for

digital simulation of control systems," Cornell AeronauticalLaboratory, Inc., Report XA-1681-E-I, (Jul 1964).

2. Boxer, R.; "A note on numerical transform calculus," Proc.I RE (Oct 1957).

3. Tustin, A.; "A method analyzing the behaviour of linearsystems in terms of time series," J. I. E. E. (Proc. of theConvention on Automatic Regulators and Servomechanisms)Vol. 94, Part 11-A, (May 1947).

4. Heizman, C.; Lambert, L.; and Millman, J.; "A comparison ofprincipal and secondary operator for the approximatetransform calculus," Columbia University ElectronicsResearch Lab. Tech. Report. T -6/C, (Aug 1955).

5. Rabiner, L.R. and Gold, B.; Theory and Application of DigitalSignal Processing, Prentice -Hall, Inc. (1975).

6. Barton, D.K. Radar System Analysis, Prentice -Hall, Inc.(1964) p. 17, Fig. 1.9.

7. Brennan, L.E., and Reed, I.S.; "A recursive method ofcomputing the Q -function," Trans. IEEE, Vol. IT -I1, No. 2(Apr 1965).

88

Enhancing antenna performance throughmultimode feeds and distribution networks

C.E. Profera

Probably the most widely used microwaveantenna is the paraboloid reflector, il-luminated by a source of electromagneticenergy called the feed, which is placed atthe focal point of the reflector. The feeddirects energy toward, and receives energyfrom, the reflector surface. The paraboloidreflector antenna radiates (or receives)energy in a prescribed pattern charact-erized by a single dominant narrow mainbeam or lobe and a plurality of lower -levelminor lobes or sidelobes.

Reflector antennas that are required toexhibit optimum performance (generallyimplying maximum efficiency of antennaillumination by the feed and minimumsensitivity to external noise) are oftenilluminated by a Cassegrainl feed system asshown in Fig. 1. Such a feed system consistsof a secondary hyperboloid subreflectorand feed. In this configuration the feeddirects and receives energy to and from thesubreflector surface. The subreflector, inturn, directs and receives energy to andfrom the main paraboloid reflector surface.


TRANSMITRECEIVENETWORK

Parabolic antennas have high efficiency and gain when theyare fed by electromagnetic energy that is radiating in morethan one waveguide mode.

The feed, for either the conventionalparaboloid reflector antenna or theCassegrain antenna, is generally an elec-tromagnetic waveguide horn connected tothe associated system transmit/ receivenetwork by waveguide transmission line.The simplest horn feed that may be used isone which propagates and radiates theelectromagnetic field characterizing thefundamental or dominant waveguidemode, denoted TEio mode for generalrectangular waveguide structures. Feeds ofthis type are always designed so that theonly electromagnetic wave propagated andradiated characterizes the fundamentalmode.

Horn antenna technology advancements inthe 1960s included the discovery of ahigher -order waveguide mode radiationproperty. When combined with fundamen-tal mode radiation, this property producesfeed radiation characterisitics leading tosubsequent antenna performanceenhancements.' The most notable im-provements obtained relate to antennaefficiency. The primary advantages of thisdual -mode radiation principle are reduced

FEEDHORN

PARABOLOID MAINREFLECTOR

HYPERBOLOIDSUB-REF LECTOR

FOCAL POINTOF PARABOLOIDREF LECTOR

CONJUGATEFOCI OFHYPERBOLOIDREFLECTOR

Fig. 1Cassegraln reflector antenna system cross section illustrates the relationships among thefeed horn, the confocal hyperboloid subreflector, and the paraboloid main reflector.

feed -pattern sidelobe levels and improvedfeed -pattern main -beam symmetry. Thesefeed radiation properties produce lowerspillover loss (i.e., feed radiation outsidethe boundaries of the collimating reflectorsystem) and a symmetric reflector illumina-tion characteristic, both of which increaseantenna gain.

Additional specific higher -orderwaveguide modes produce the radiation -pattern characteristics necessary to per-form monopulse radar tracking.3'4 Theability to independently generate and con-trol the multiple modes required for im-proved performance and monopulse track-ing, within a single waveguide horn feed,has led to the development of the mul-timode monopulse feed. This feed con-figuration is the high-performance stan-dard for tracking radar applicationsemploying Cassegrain reflector antennas.

The low-sidelobe illumination propertiesprovided by the multimode feeding tech-nique are also applicable to antenna con-figurations employing horn feeds forphased -array antennas.' Multiple -mode,low-sidelobe distribution networks forboth planar6 and circular -array' antennashave been developed using this technique.

This paper presents the fundamentals ofmultimode feed techniques and describesboth the reflector -antenna and phased -array applications where these techniqueshave been employed.

Multimode fundamentalsMonopulse tracking radar system re-quirements for reflector antennas withoptimum high -gain and low -noiseproperties are most readily satisfied by thedual -reflector Cassegrain antenna with amultimode, monopulse feed.

This application of multimode technologyis the one of most general interest at RCAMS R. Multimode fundamentals are

89

ly- POLARIZED FIELDS)

MODETYPE

E la,y)

ly- POLARIZED FIELD)

TRANSVERSE FIELD ILLUSTRATION

Iy- POLARIZEIYFIELD)

TE10

( 1)

COS Et

an V

-3/2- 1 a/2%

-a/2

LSE12

crE12+Tm121

(I)cos Ilk, cos 2p

'I

- 4

4

4I

I t

4 i

- ,,

LSE11

(TEWTMII)

i AEI

SIN 17

Y

-ir 1

t I t

-

TE20

(4)SIN 120

Y

-I

a

MULTIMODE MONOPULSE FEED

REFERENCE BLOCK DIAGRAM

CHANNEL - MODE MODEc.)MONOPULSEE) BRIDGE - GENERATING PHASING HORN

CEHRARONNLSR CATr NETWORKSECTION SECTION

b I

Fig. 2Trigonometric functions and transverse electric fields areillustrated (a) to describe the feed aperture distributionsproduced by each of the modes of a high-performancemultimode monopulse feed. The block diagram (b)shows the major functional sections of a typical mul-timode monopulse feed.

E FIELDAMPLITUDE

El° vs y

ITE10)

" COS2-0N-PEDESTAL" HORNILLUMINATION FUNCTION

Ei2YY

ILSEI2)

NiNk. (Eio*E121 vs y

CTEio + LSE12)

Fig. 3E -plane feed aperture "tapering" shown here results when the TE 10and LSE12 modes are summed in a co -phased sense. The resulting"cost -on -pedestal" distribution is produced by summing theuniform TEio mode and cosine -tapered LSE12 mode E -plane dis-tribution.

described here in terms of the monopulseapplication, but the techniques used areapplicable, in a non-monopulse sense, tohigh-performance reflector antennas inother radar and non -radar applications.

The single -polarization waveguide modeset employed in a high-performancemonopulse feed application contains thefundamental TEto, and higher order TE12 +TM12 (called LSE12), TE11 + TMI I (calledLSE11) and TER, rectangular waveguidemodes. The mathematical functionsdescribing the transverse electric fields ofthese modes (in square waveguide) aretabulated in Fig. 2, along with their fieldconfigurations and a block diagram of atypical feed system that produces them.

The electric field composed of a sum ofwaveguide modes is "tapered" to control thebeamwidth and sidelobe patterns of thefeed.

The reference or sum channel of a con-ventional four -element monopulse com-parator network connected to a multimodefeed produces, at the feed aperture, anelectric -field distribution composed of thesum of TElo and LSE12 waveguidemodes. Combining these modes produces aresultant electric -field distribution whosetransverse amplitude (x -y plane), in asquare feed aperture, is represented

Fig. 4E- and H -plane sum and difference pattern functionsof a well -designed multimode feed.

MULTIMODE FEED (TEio LSE12)

REFERENCE (£) PATTERN

E PLANE

- - H-PLANE

SINGLE MODE ITE101FEED PATTERN

Il (1

40

phr

Fig. 4aThe sum patterns are compared with the fundamentalmode (TEio) pattern of a single horn to illustrate itslower beamwidth and higher sidelobe levels.

90

mathematically as

EREF (x,y) = [1 + A21 cos (27ry/ a)]. (1)cos (7rx I a)

= 2 A21[T+ cos (277y/ a)]*cos (7rx I a)

where T= (1- A21)1 A21

when the modes are co -phased. The dis-tribution of EREF is "cosine tapered" alongthe x-axis (H -plane) of the horn apertureand "cosine -squared -on -pedestal tapered"along the y-axis (E -plane), as illustrated inFig. 3. The "cosine -squared -on -pedestal"amplitude tapering in the E -plane of theresultant distribution, which is the result ofLSE12 mode incorporation, gives rise to thelow-sidelobe and beamwidth equality(symmetry) property of dual -mode radia-tion. A21, the ratio of LSE12 to TEio modeamplitude, controls the E -planebeamwidth and sidelobe characteristics ofthe horn. For 10 -dB beamwidth equality ofE- and H -plane patterns, a typical designrequirement for efficient illumination of areflector antenna by a horn feed, A21 =0.67.

The mathematical expression of themultimode-feed radiation -pattern func-tion, for the reference -mode illuminationcharacteristic of Eq. 1, is

(u) = gio(u) + Al2g12(u) (2)

where

glo =sin(u sin 0)

u sin 0cos (u cos 0)

(7r/ 2)2 - (u cos 0)2

u sin4i sin(u sing))812 =

7r2 - (u sin0)2

cos (u cos 0)

(7r/ 2)2 - (u cos 0)2

U = (7ra/ X) sin 0= wavelength

0 = angle measured from feedaxis in the plane ofpattern calculation

ci) = 7r/ 2 for E -plane,0 for H -plane

Computed sum pattern functions of themultimode feed, with A21 = 0.67, are shownin Fig. 4a superimposed with the E -planepattern function of the simpler single -modefeed. The E -plane beam broadening andsidelobe-level reduction afforded by thedual -mode technique are evident in thisfigure. Maximum E- and H -planesidelobes (of the multimode feed patterns)are observed at levels of -30 dB and -23 dBrespectively, relative to the pattern max-imum. The 10 -dB beamwidth equality andlow sidelobe levels of the feed pattern

produce nominally optimum illuminationand spillover efficiencies for the Cassegrainreflector antenna, yielding an antenna gainthat approaches the maximum availablefrom a given aperture dimension. Reflectorshaping techniques may be incorporatedinto the antenna design to obtain a furthergain increase, but are beyond the scope ofthis paper. Well -designed multimodemonopulse feeds are capable of providingCassegrain reflector antenna referenceefficiencies in the 55% to 65% range.

The monopulse radar tracking capabilityof the multimode feed derives from itsability to independently couple themonopulse comparator network differenceports to the TED) and LSE11 waveguidemodes in the multimode horn. The TENand LSEI I modes, whose transverse elec-tric field configurations are also describedmathematically and illustrated in Fig. 2,provide difference or monopulse errorpatterns along the principal x- (H -plane)and y -(E -plane) axes respectively. The H -plane difference -pattern function of theTEN) mode and E -plane difference -patternfunctions of the LSE', mode, whichdescribe the feed error patterns, are:

(u) = sinu/(7r2 - u2) (3a)

gaE (u) = u cosu/[(7r/ 2)2 - u2] (3b)

ERROR (AE) PATTERN

E-PLANE

Fig. 4bE -plane difference (monopulse error) pattern func-tion for a multimode feed provides the radar trackingfunction in a single plane.

ERROR (OH) PATTERN

3.0

p/

4.0

Fig. 4cH -plane difference (monopulse error) pattern func-tion provides the radar tracking function in a second,orthogonal plane.

91

Fig. 5Computed patterns illustrate the effects of mode phase error ( ) and hornflare angle phase error ( /3 ) on the E -plane sum patterns of the multimodefeed. Note the increasing beamwidth and sidelobe levels accompanyingthese phase errors.

Computed E- and H -plane difference -pattern functions of the multimode feed areillustrated in Figs. 4b and 4c, and togetherwith the reference patterns complete themonopulse pattern set.

Antenna efficiencies associated with thedifference patterns, which directly affectradar tracking sensitivity, are better for themultimode feed than for conventional four -and five -horn monopulse feeds.

This improvement is the result of a reduceddifference -mode spillover loss charact-eristic afforded by the rotationallysymmetric, low-sidelobe reference il-luminations, permitting an increased angleintercept of feed radiation by theCassegrain subreflector.

The previous multimode analysis assumeda "co -phase" condition between the TEioand LSE12 reference modes. Since thesemodes do not constitute a degenerate pair(i.e., propogate with the same phase veloci-ty), this condition will exist at only a singlefrequency. The relative phase shift betweenthese modes, which is introduced by thewaveguide feed structure, is a function offrequency and is the major parameter thatdetermines the useful bandwidth of themultimode feed. Increased bandwidth isobtained with minimum relative phasedifference of the TEio and LSE12 modes,introduced by the section of feed betweenthe mode generator and horn aperture.Feed designs requiring a 27r -radian phaseshift between TEio and LSE12 modes havebeen developed, and provide up to 10%

Fig. 6RCA MSR's first high-performance multimodemonopulse feed was developed for the Apollo ShipsInstrumentation Radar (ASIR). This feed operatedover a nominal 5.4 to 5.9 GHz frequency band.

useful bandwidth capability. Increasedbandwidth capabilities in excess of 10%have been realized with feeds requiring aphase shift of only w radians betweenreference modes.

A second parameter influencing the perfor-mance capability of the multimode feed isthe flare angle of the waveguide horn.

This flare angle imparts a nominalquadratic phase distribution to thetransverse modal aperture illuminationfunction, and degrades the low-sidelobecharacteristics of the feed. The effect ofphase difference between the TElo andLSE12 modes at the horn aperture and thequadratic phase distribution produced bythe horn flare angle are illustrated in thecomputed E- and H -plane reference -pattern data of Fig. 5. This data illustratesthe E- and H -plane pattern functions of asquare -aperture horn with dual -mode sumillumination, where a relative -mode phasedifference and quadratic phase distributionare assumed.' The fractional power dis-tribution of the complex pattern is

superimposed on this data so that thefractional power content of any angularportion of the feed pattern can bedetermined or the feed spillover losses canbe estimated. Data shown in Fig. 5 assumesa relative phase difference between modesof 0°, 20°, and 40°, coupled with a max-imum quadratic phase error of 40°, in-troduced by the horn flare angle. Thevariation of E- and H -plane patternbeamwidths and degradation of E -planesidelobe levels attributable to these effects

are evident in the figure. A rule of thumbfor high-performance multimode feeddesigns restricts the maximum -mode phaseerror to ±10° over the operational band-width, with a maximum quadratic phaseerror due to horn flaring of less than 50°.

Multimode feeddevelopments at MSR

The first of the high-performance mul-timode monopulse feeds for Cassegrainantennas developed at MSR was the ASIR(Apollo Ships Instrumentation Radar) feed.

This feed, shown in Fig. 6, was designed foroperation over the 5.4- to 5.9-GHz radar

3db

5db

10db

15db

20db

25db

30db

FRED -5.65 GI -1x IIH-PLANE)

MODE - TEI0 LSE12

FEED POL - VERTICAL

PLANE OF CUT - AZIMUTH

SUBREFLECTOR SUBTENDEDANGLE OF 270

36 48 60 712

9272 60 48 36 24 12 0 12 24

ANGLE IN DEGREES

OD

C -BAND MULTIMODE

MONOPULSE FEED

f - 5.4 GH: I E.PLANE H -PLANE

TE10 LSE12

ANGLE

C -BAND MULTIMODEMONOPULSE FEED

f 5 4 GH. AE AH,-,E-PLANE ERROR (LSEn)

,......H -PLANE ERROR (TE20

-1006

20DB

-3008

ANGLE

Fig. 7Measured sum and difference patterns of the ASIR multimode monopulse feed.

band, and was subsequently used in 16- and12 -ft -diameter Cassegrain reflector anten-na systems. The feed employed a circular -aperture hoin to obtain additional patternimprovements. Measured sum anddifference patterns obtained from the feed,shown in Fig. 7 demonstrate the low-sidelobe reference -pattern available withmultimode techniques. The antennaefficiency, determined by analysis of feedradiation pattern data and later verified bygain measurements on the antenna con-figuration, was nominally 60 to 65%.

A subsequent feed development at RCAMSR combined the advantages of mul-

3db

5db

10db

15db

20db

25db

30db

timode excitation and near -field hornradiation properties to achieve a furtherimprovement of Cassegrain antennaefficiency.9 The near -field multimode,monopulse feed employed the mode -generating network of the multimode feedin an electrically large pyramidal horn, andproduced reference patterns that were es-sentially free of sidelobes. Spilloverefficiencies obtained from this feed werethereby increased relative to the previousmultimode feed. Although only useful inlarge Cassegrain reflector antennaconfigurations,'° this marriage of tech-niques produced antenna efficienciesgreater than 70%. Fig. 8 is a photograph of

Fig. 8Near -field multlmode monopulse feed installed in a29 -foot diameter Cassegrain antenna. Operating fre-quency range was nominally 5.4 to 5.9 GHz.

the near -field, multimode monopulse feedin a 29 -ft. -diameter FPQ-6 Cassegrainreflector configuration. Measuredmonopulse feed pattern data is shown inFig. 9, illustrating the improved low-sidelobe reference -pattern characteristicsobtained.

Recent multimode, monopulse feeddevelopments at MSR have addressed in-creased bandwidth capabilities and reduc-ing the cross -polarization crosstalkproblems inherent in all monopulse feeds.

A K. -band multimode monopulse feed fora small Cassegrain antenna has beendeveloped that permits operation over a

Fig. 9Measured sum and difference patterns of the near -field feed. Note the virtual absence of sidelobes in the sum feed patterns,giving rise to additional efficiency enhancements.

FRED -5.65 GHz :.(E -PLANET

MODE - TE10 LSE12

FEED POL - HORIZONTAL

PLANE OF - AZIMUTH

SUEIREF LECTOR SUBTENDEDANGLE OF 27°

3db

5db

10db

15db

20db

25db

30db

60 48

FREC1 -5.65 GHz AH

MODE - TE20

FEED POL - VERTICALPLANE OF CUT - AZIMUTH

HD'

I is36 24 12 0 12 24 36 48 60 72

3db

5db

1001,

15db

20dh

25db

30db

72

FRED - 5.65 GHz

MODE LSE 11

FEED POL - HORIZONTAL

PLANE OF CUT AZIMUTH

60 48124 12 0 12 24 36 48

ANGLE IN DEGREES

so 7272 60 48 36 24 12 0 12

ANGLE IN DEGREES

24 36 48 60 72 72

ANGLE IN DEGREES

93

14.7- to 15.6-GHz frequency band.' Thisfeed uses a mode -generation techniquerequiring a r -radian relative phase shiftbetween TEio and LSE12 modes, therebyderiving the wide bandwidth capabilities.Fig. 10 is a photograph of this feed in a 26 -in. -diameter Cassegrain geometry.

The mode -generating technique developedfor the described feed above was incor-porated in a Ka -band (34.5 - 35.5 GHz) feeddesign because of its attractive low -

depolarization characteristic. This feed wascombined with other selective mode -tuningdevices to produce a multimode monopulsefeed system with significantly reduceddifference -channel, cross -plane depolariza-tion. Using this design has reduceddifference -channel depolarization bygreater than 10 dB.

Several unique array -feed concepts havebeen demonstrated and developed at MSRusing the low-sidelobe properties of thedual -mode illumination characteristic.

A millimeter -wave planar -array configura-tion, developed by MSR under contract to

Chuck Profera joined RCA MSR as amember of the Antenna/Microwave group in1959. With the exception of a two-yearperiod, during which he attended DrexelUniversity as a Sarnoff Fellow, he remaineda member of that activity until June 1977,when he assumed his present position asMSR's coordinator of IndependentResearch and Development.

Contact him at:Missile and Surface RadarMoorestown, N.J.Ext. PM -3210

.011°g-

Fig. 10A new wideband multimode monopulsefeed for a small Cassegrain antenna. Thenew mode -generating technique hasdoubled bandwidth capabilities.

NASA, employed a dual -mode E -planesectoral horn as a constrained powerdivider.5'6 This power divider, which isshown feeding a portion of the array in Fig.I1, was incorporated with a 50 -elementlinear array of H -plane sectoral hornelements. The result was a low -loss, low-sidelobe planar phased array with wide-angle, single -plane electronic scancapabilities. Power division was ac-complished by terminating the continuousaperture of the dual -mode, E -plane sec-toral horn with a linear E -plane array ofwaveguide elements. These elements fed, inturn, phase shifters and the H -plane sec-toral horn radiating elements of the array.

An experimental model of the completearray is shown under test conditions in Fig.12. The phased -array model was designedto operate over a 15.2- to 16.8-GHz fre-quency band with a ±60° E -plane scancapability. Principal -plane pattern dataobtained from this antenna for a 0° and 40°scanned beam is shown in Fig. 13. Thedual -mode technique produced broadsideE -plane array patterns at 16.0 GHz withmaximum sidelobes below -28 dB. H -

plane patterns of the array at this frequencywere characterized by a maximum sidelobeof -39 dB. Pattern characteristics werenominally equivalent over the full designbandwith.

Sidelobe deterioration with scan angle is anintrinsic property of planar arrays; hence,the 40° E -plane scan patterns exhibited asomewhat increased maximum sidelobelevel. Although not specifically developedfor a radar application, the addition of a

Fig. 11Dual -mode sectoral horn power divider for asingle -plane -scanned, low-sidelobe phasedarray. The low -loss "optical divider"properties of this horn make it attractive formillimeter -wave array applications.

monopulse tracking capability to this arraywas demonstrated by generating a

difference mode within the E -plane sectoralhorn power divider and measuring theresulting array difference patterns.

A second array application of dual -modepower division techniques was realized atMSR by using a dual -mode radialwaveguide power divider as a feed for acircular array antenna.' An X -band (7.9-8.5 GHz) model of a dual -mode, radial-waveguide, 16 -way power divider is shownin Fig. 14, and an experimental model of a64 -element circular array antenna employ-ing the power divider is shown in Fig. 15.Radiation patterns of this arraydemonstrated maximum sidelobes of -23dB.'2 A monopulse tracking capability isalso available for radar application of thecircular array by using an orthogonalradial waveguide mode of the alreadyavailable dual -mode set.

ConclusionsThe low-sidelobe properties of dual ormultimode illumination functionscharacterizing several waveguide typeshave been applied by MSR to a variety ofantenna applications. The most widelyused application of multimode technologyhas been for the design of high-performance monopulse feed systems forCassegrain antennas. The combination ofmultimode and near -field radiationproperties of horn feeds has resulted infurther feed performance improvements.Wideband applications of multimodetechniques have been demonstrated byminimizing the differential reference -mode

94

0 -E -PLANE 0° SCAN

--'1411111111111Fig. 12Low-sidelobe phased array with dual -modehorn power divider undergoing testing in anMSR anechoic chamber.

phase shift introduced by the feed system.These newer mode -generation techniqueshave led to monopulse feed designs withsignificantly reduced difference -channel,cross -plane depolarization.

The application of dual -mode, low-sidelobe illumination techniques has alsobeen extended to use with planar andcircular arrays. Most noteworthy perfor-mance has been obtained from a planar -array antenna using multimode sectoral-horn power -divider techniques. This arrayprovided wideband, low-sidelobe antenna-

pattern characteristics with a single -plane±60° electronic scan capability.

ReferencesI. Hannan, P.W.; "Microwave antennas derived from the

Cassegrain telescope," IEEE Trans. on Antennas andPropagation, Vol. AP -9 No. 2 (Mar 1961).

2. Potter, P.S.; "A new horn antenna with suppressed sidelobesand equal beam width," Microwave Journal, Vol. 6 (Jun1963).

3. Jensen, P.A.; "A low -noise multimode Cassegrainmonopulse with polarization diversity," NEREM Record,(Nov 1963).

4. Profera, C.E. and Yorinks, L.H.; "A high efficiency dualfrequency multimode monopulse antenna feed system."IEEE Trans., Vol. AES-2 (Nov 1966).

5. Profera, C.E.; "Evaluation of a low sidelobe phased arrayantenna," NASA CR-1471 (Nov 1969).

6. Profera, C.E.; "A low loss n -way optical power divider foraK.-band low sidelobe phased array," Microwave Journal,(Nov 1969).

7. Profera, C.E.; "X -band circular array with multimode radialwaveguide feed,"private correspondence.

8. Profera, C.E.; "Complex radiation patterns of dual modepyramidal horns," IEEE Trans. on Ant. and Prop. (May1977).

9. Profera, C.E. and Yorinks, L.H.; "An improved Cassegrainmonopulse feed system," RCA Review (Dec 1967).

10. Mikulich, P.; Delusic, R.; Profera, C.E.; and Yorinks, L.;"Ahigh gain Cassegrain monopulse antenna," 1968 IEEE P -GAP Symp. Digest.

I I. Profera, C.E.; "A J -band Cassegrain feed/antenna system forspace application," private correspondence.

12. Yorinks, L.; "X -band circular array antenna," private cor-respondence.

10 -

2o-

30 -

401 1 1 1

180° 150° 120° 90° 60° 30°

0-

10-

20 -

30-

401 I

0-

10 -

zo -

30-

E -P LANE 40° SCAN

AI

BROADSIDE 0° SCAN16.0 GHz

1 1 I I

30° 60° 90° 120° 150°

40° SCAN16.0 GHz

I I' I I

180° 150° 120° 90° 60° 30° 0 300 60° 90° 120° 150°

H -PLANE 0°SCAN

0° SCAN

16.0 GHz

900 60° 30° 30° 60°

Fig. 13E -(scan plane) and H -plane patterns of thelow-sidelobe phased array with dual -modesectoral horn power divider.

Fig. 14Low -loss dual -mode radial waveguidepower divider for feeding a circular cylin-drical array antenna.

0--PLANE 40° SCAN

400 SCAN

16.0 GHz

90° 60° 30°

H

I0°

610.

Fig. 15X -band model of a circular array antennawith the radial waveguide dual -mode powerdivider. This configuration is capable ofproducing a beam that can be scanned 360°in a single plane.

95

CR1 in@ j©Woff the jobElectronic speed control for model railroad realism

W.S. Pike Electronic speed control of miniature locomotives enablesthis hobbyist to simulate realistic operation of his trains.

The N -gauge Susquehanna Southern Railway existsprincipally for the purpose of hauling coal from the mines inthe mountains above the headwaters of the SusquehannaRiver to the port city of Susquehanna. Although coal trafficis the line's major source of revenue, passenger service hasrecently been expanded to cope with the increasing numberof tourists attracted to this picturesque region.

Why electronic speed control?In the initial stages of the railroad's existence, locomotivespeed was controlled by using a conventional variableseries resistance controller. This type of controller, as mostmodel railroaders know, is not very satisfactory, leading to"jack -rabbit" starts, unrealistically abrupt stops, anddifficulties in slow -speed running. One of the solutions tothe problem is to use unidirectional pulses of variable dutycycle rather than dc for traction power. However, this tendsto make the tiny permanent -magnet motors in the enginerun hot, so that many variable -duty -cycle controllers alsoincorporate provision for either automatic or manualswitchover to dc running. Our throttle is of the latter variety.

It is also desirable to incorporate an "inertia" circuit tosimulate the "coasting" of real trains. This may be done witha diode and capacitor, which delays the fall of track voltagewhen the throttle is closed. Often a "brake" control is added,comprising one or more resistors which can be switchedacross the "inertia" capacitor to increase its rate of dis-charge. We have found it prudent also to include a "panic"

button which short circuits the inertia capacitor so that aquick stop can be achieved in the event of imminentcatastrophic collision.

Throttle operationFig. 1 depicts the Susquehanna Southern Railway controlpanel on which the throttle circuit is mounted. The throttlecontrols are at the edges of the panel. Also on the panel isthe mimic diagram of the trackage. The duplicate throttlesystem permits each block of the layout to be powered fromeither throttle to permit simultaneous operation of twotrains.

The large knob labeled SEL and BRAKE controls thevarious operating modes of the throttle. In the extremecounter -clockwise position, labeled "pulse," pulses ofapproximately 12 V amplitude at a 100 -Hz repetition rate areapplied to the tracks. As the THROTTLE knob at the top isadvanced clockwise, the duty cycle of the pulses changessmoothly from 5% on and 95% off to 95% on and 5% off. Theaverage dc value of this waveform thus changes from about0.6 V to about 11.4 V. This range of control will permitextremely smooth starting of most N -gauge locomotives.They may also be made to inch along the track (if the trackand wheels are clean) at one or two scale miles per hour orrun up to any desired fraction of full speed.

No inertia is provided in the pulse mode of operation as it isadvantageous (though less challenging to the operator) for

96

Fig. 1Control panel of Susquehanna Southern Railway. Dual throttle control system permits simultaneous powering of two trains.Route selection and power switching of trains is accomplished by toggle and modified rotary switches.

some critical switching moves to have one throttle modewithout inertia.

Moving the SEL and BRAKE knob to the next positionchanges over to dc operation with inertia. Thus a train maybe started on pulse power and then changed over at anytime to dc running. In the dc position, most of the

Win Pike joined RCA Laboratories in 1946. The recipient of fiveRCA Laboratories Achievement Awards, he has worked on a varietyof projects including reading machines for the blind, standard andspecial-purpose television cameras and receivers, equipment forautomatic driving, and solid-state sensors. In addition to buildingmodel railroads for a hobby, he also works on pipe and electronicorgans.

Contact him at:Television Research LaboratoryRCA LaboratoriesPrinceton, N.J.Ext. 2638

Susquehanna Southern Railway's motive power will permita locomotive to coast about three -fourths of a scale milewhen unloaded and somewhat less when hauling a train.Braking action is available in five steps by advancing theSEL and BRAKE knob further clockwise. The PANIC buttonmay be seen at the bottom of the panel between a pair ofreversing switches.

Throttle circuit theoryAlthough there are many fine published circuits for ac-complishing all this, the engineering department of theSusquehanna Southern Railway decided to design its ownthrottle circuit. A block diagram of the result is shown in Fig.2. A COS/MOS CD4047 integrated circuit wired as anastable multivibrator clocks the entire system at about 100Hz. This is followed by a second CD4047 connected as aone-shot multivibrator. The THROTTLE knob adjusts thepulse width of the one shot within the limits previouslydescribed. In the "pulse" position of the mode selector

PANIC

BUTTON

THROTTLEBRAKE

-"PA- Y41'-TO

C04041ASTABLE

MU LTI VI BR ATOR

C04041ONE

SHOT

ACTIVE

LOW-PASS

FILTER

INERTIA

CIRCUIT

OC

PULSE

OUTPUT

CURRENT

AMPLIFIER

REVERSING

-.- SWITCH

AND

TRACK

100 hiIOR DC)

Fig. 2An advantage of the author's throttle circuit design is that the samethrottle potentiometer can be used for either dc or pulse operationwithout requiring complicated switching.

97

CI

0 033

r"RI

68k

ASTABLEMULTIVIBRATOR

ONE

SHOT

C2

100µF

D2

12VR2

68011

ACTIVE

LOW PASS FILTERINERTIACIRCUIT

OUTPUT

CURRENT AMPLIFIER

04 65 66

284126 40361 285298

CD

4047

FC3

0033T

CD

4047

R3

8 2k

PI

100k

THROTTLE

R4 R9

9 I k

R8

41k

"'GS047=

R9 DI

`91k 18914

C7

50,LF

RIO

62k

01

2N4124

J-LJLfL1.1-11-11-

Fig. 3Schematic diagram of throttle control circuit shows allcomponents needed except for power supply, whichrequires no special circuitry.

02284126

BRAKE

RESISTORS

control, the output of the one shot is applied without furtherprocessing via a Darlington current amplifier to the track. Inthe dc mode, an active low-pass filter with a turnoverfrequency of about 10 Hz (and a rate of cutoff of 18dB/octave) and the diode/capacitor inertia circuit with itsbrake resistor are interposed between the one-shot and thecurrent amplifier. The low-pass filter extracts the averagedc value of the 100 -Hz pulses and presents it to the outputamplifier, thus permitting the same throttle potentiometerto be used for both dc and pulse operation without anyswitching complications. A reverse -biased diode is shuntedacross the output, mainly to protect the throttle circuitryfrom damage if a train happens inadvertently to enter ablock already energized from the second throttle circuit andset for the opposite direction of travel (opposite polarity).

Throttle circuit designThe complete throttle control circuit is shown in Fig. 3. TheCD4047 astable and monostable multivibrators require littlecomment. Potentiometer P1 is the throttle potentiometercontrolling the pulse width of the monostable. Transistors01, Q2, and Q3 comprise the active low-pass filter, basicallythe well-known Sallen and Key configuration with oneadded RC section. Diode D1 and capacitor C7 comprise the"inertia" circuit, R15 through R20 are the braking resistors,and 04, 05, and Q6 comprise the Darlington -connectedoutput current amplifier. Transistor Q6 is mounted on a heatsink. Diode D3 serves as the protective device.

Potentiometer P2 (visible in Fig. 1 just above the PANICbutton and labeled ADJ), provides a means of adjusting the

R18

20k

RI739k

03284126

R19 R20

10k 10k

RI6 SIB

82k DC

RI5

160k d S2

PANIC BUTTON

SIA

PULSE

R11;10k

I8V DCINPUT

OR DC TO

REVERSING

t. SWITCHES

RI2 n, AND

10k 1N3254" TRACKS

NOTE SELECTOR SWITCH, SI, SHOWN

IN PULSE POSITION

P2

110kSPEED MATCH

RI4 10k

match between locomotive speed when running on pulsepower and running on dc. It allows a limited adjustment ofthe pulse amplitude applied to the output amplifier. Inpractice, I have found that different locomotives sometimesrequire slightly different settings of this control to ensure asmooth mode transition from pulse to dc and vice versa.

If any reader chooses to duplicate this circuit, I shouldperhaps caution that depending on the tolerances of thetiming components in the multivibrators, R3 and C3 mayrequire tailoring to ensure that the one-shot varies smoothlyfrom minimum to maximum duty cycle without entering afrequency -halving region at full throttle or ceasing opera-tion at minimum throttle. Another possible modificationwould be to include "inertia" on opening the throttle. Thismay be done by adding resistance in series with D1.

I have not shown the details of the 18-V raw dc source whichenergizes the throttle circuit, as no special circuitry isrequired. Our supply comprises a transformer, bridgerectifier, and 5500-microfarad filter capacitor. Note,however, that the COS/MOS multivibrators are isolatedfrom this supply by a resistor, R2, Zener diode, D2, andcapacitor, C2.

The author will be glad to answer inquiries from anyoneinterested in duplicating this circuit, which has been insuccessful operation on the Susquehanna SouthernRailway for over two years.

Reprint RE -23-5-18Final manuscript received December 29, 1977.

98

Building control circuitry, collecting trains, and designing the tracklayout for an N -gauge system are just part of the fun for a modelrailroader. These views of the author's "Village of Gap" in theSusquehanna Valley show the realism in detail put into building thestructures and scenery.

Ed. Note: Although the subject of this article iselectronic speed control, model railroad buffs mostcertainly will want additional details on Win Pike'srailroad system. The following gives some constructionand operating particulars about his Lilliputiantransportation network. FJS

The Susquehanna Southern Railroad, an N gauge (9mm between running rails) model railroad is con-structed entirely on a standard 36 -inch wide, flushdoor. A sheet of Homasote is glued to the top surface ofthe door, and the raised portions of the railroad havebeen constructed by the "cookie cutter" method,raising and supporting the Homasote where necessarywith glued -in wooden supports.

Scenery is made up of plaster on wire screen, againsupported where necessary with wooden framing.Although this method of construction is not as flexibleas open benchwork, it is simple to make and, particular-ly with N gauge, results in quite a lot of readily portablerailroad in a small space.

The structures visible in the photographs are a mixtureof commercial plastic model kits and scratch -builtbuildings. The backdrop (as yet unfinished) is in theprocess of being painted by the author's son, Eric Pike.

The trackage contains 16 separate blocks and 15electrically operated track switches, including twodouble slip switches (to save space). All but two of

these switches are commercial units using con-ventional, double -coil mechanisms. The remainingtwo, which are more satisfactory, are driven by "hooklever" type, single -coil mechanisms "liberated" from adefunct pipe organ. Because of their size, thesemechanisms are concealed under the scenery.

Most of the wiring is simple; a double -throw, double -pole, center -off switch associated with each blockpermits powering it from either of the two powersupplies. However, in a few areas, notably the upperlevel wye (visible in the center of the control panel)route selection and power switching have been com-bined into a single rotary switch which has beenmodified to include a coaxial pushbutton. One selectsthe desired route by rotating the switch, then pushingthe button to align the track switches to the selectedroute. A similar system is used in the yard area of therailroad.

An extension running down the side wall of the room inwhich the railroad is housed is planned for a later date.This will connect on to the exsting trackage and willprobably require a second control panel.

Rolling stock comprises a mixture of European andAmerican models and includes 2 diesel locototives, 2steam locomotives, and 2 electric locomotives. Inoperation, most trains are kept fairly short because ofthe relatively sharp curves and the 4% gradient to theupper level.

99

Adam

Katagi

Feller

Aires

\\)/Sasaki

The 1978 David Sarnoff Awardsfor Outstanding Technical Achievement

1

RCA's highest technical honors have been announced for 1978.Each award consists of a gold medal and bronze replica, a framed

citation, and a cash prize.

Advanced Technology Laboratory, Camden, N.J.

For outstanding achievement in the use of computer -aided design techniquesfor large scale integrated circuits.

At a 1977 IEEE workshop on the use of computer -aided design approaches, it wasindicated that RCA's system for LSI design is the most advanced in the industry. Al Felleris the prime contributor to this advanced position. His work in the development ofcomputer -aided standard -cell LSI array design and technology resulted in thedevelopment of advanced signal -processing equipment for RCA communication andradar systems. These same design approaches have also allowed cost-effectivedevelopment of semiconductor products. He made specific contributions to: APAR, theautomatic placement and routing program for CMOS standard -cell arrays; the SUMC-DV computer for NASA, in CMOS and CMOS/SOS form; an automotive on -boardprocessor; the multiport two-dimensional placement and routing program; the ATMACmicroprocessor; and radiation -hardened CMOS/SOS standard -cell LSI arrays.

Clark Davis Gorski

Kenneth C. Adarnj Ramon H. AiresCharles A. Clark, Jr.lWilliam J. DavisJohn C. Gorski! Kazuo Katagil Atrir' cgsakiAvionics Systems, Van Nuys, Calif.

For outstanding technical achievement in the development of airborne colorweather radar indicators.

RCA's PriMUS-90 and -400 ColoRadar units were industry's first airborne weather radarsto use color indicators. The new designs also provide four times the resolution ofprevious digital -memory -refreshed indicators, and allowed the use of a standard colorcrt by converting the radar scan from polar coordinates to x -y (tv-style) scan. The designteam worked very quickly; RCA was thus able to announce color radar availabilitysignificantly ahead of the competition and even go to production ahead of schedule.Customer acceptance for the color -indicator radars has been above expectation, somuch so that production schedules had to be accelerated twice during the first ninemonths after announcement.

100

Consumer Electronics Division, Indianapolis, Ind.

For outstanding team achievement in the development of scan and powersupply systems for color television.

All of the color-tv sets now being manufactured by Consumer Electronics enjoy thebenefits of this team's design effort: better power efficiency; substantially lower cost; andimproved scan and high -voltage performance. These benefits come from combining thefunctions of horizontal deflection and power supply regulation. A weighted average ofpower consumption for the "XtendedLife" models is 92 W; the immediately precedingline had a weighted power average of about 135 W. The improved efficiency and reducedpower consumption are expected to contribute substantially to the reliability and life -cycle savings of the "XtendedLife" sets. Improved performance results from the excellentscan stability, which is independent of line and load variations. RCA's "XtendedLife" XL-100 and ColorTrak receivers, which use the new design, have been successful in terms ofsales and reliability.

Fernand F. Martini Samuel WaldsteinjJason H. WoodwardAutomated Systems, Burlington, Mass.

For excellence of team effort in the product development of a hand held laserrangefinder.

The AN/GVS-5 laser rangefinder is as much as ten times improved over previousrangefinder designs in terms of weight, power consumption, and cost, while meeting orexceeding the operational characteristics of its predecessors. For example, the previous25 -lb. units cost $30,000 apiece to produce, but the 5 -lb. AN/GVS-5 is producible forunder $4,000. Army users can now determine the range to targets instantly, and canmake up to seven hundred 10,000 -meter range measurements with one charge of theself-contained 8 -oz. battery. The success of the rangefinder design (present orders total$30 million) is based in large part on the rangefinder's design -to -unit -cost program,which combined the efforts of the design and production teams to produce the besttradeoff among size, weight, power consumption, availability of parts and materials, andcost.

Murray A, Proinsky'rPto H. Schade, Jr.Solid State Division, Somerville, N.J.

For outstanding technical achievement in the development of high perfor-mance BiMOS integrated circuits.

BiMOS integrated circuits, which combine bipolar and MOS transistors on the samechip, are a unique development of the Solid State Division. The circuits have theoutstanding performance characteristics of each type of transistor, and so have beenvery popular, with millions per year sold. The applications for BiMOS so far haveincluded operational amplifiers, comparators, smoke detectors, and television digitaltuning circuits. Murray Polinsky's process engineering work greatly simplified themethod of fabricating the MOS and bipolar devices on the same chip, and Otto Schadehad the prime responsibility for the circuit design.

,UZ

Martin

Woodward

Polinsky

McDonald

Waldslein

Schade

101

Gra. poisnzatmWen n

An instructor at a CorporateEngineering Education class onMicroprocessors gave copies ofthis reprint to each of his classmembers.

Automated Systems used thisreprint as part of a prop.

Broadcast Systems' marketinggroup took this reprint to theNational Association of Broad-casters convention.

Industrial Relations activitiesthroughout the corporation usethis reprint as an orientation forprospective employees.

These are a few examples of how our readers have usedreprints of RCA Engineer articles. Activities throughout thecorporation order about 50,000 copies of our reprints peryear, varying from a two -page, single sheet to a 200 -pageanthology with a four-color cover.

Could you use a reprint?102

Before you answer the question, maybe you can use a littlemore background.

When an author has an article published in the RCAEngineer, we send him 90 printed copies of his article so hecan show them to his boss, his mother, his colleagues, andmost importantly, answer inquiries from readers.

You'll find author -contact information with each article. Soif you simply want an extra copy, call the author. But if youneed 100 or more copies of any RCA Engineer article, or acombination of articles, read on.

How to get reprints of single articlesYou can order reprints of any article from any issue of theEngineer. The page negatives are on file, so gettingsomething reprinted costs you less than printing it for thefirst time. (It is never economical to print in quantities under100.) Pages are always printed in increments of four; so athree -page article will be a four -page reprint, a five -pagearticle will be an eight -page reprint, etc. When there's anextra page available we make a title page. It's also possibleto put such things as photographs or marketing contactinformation on these extra pages.*

You can expect delivery on a single -article reprint about twoweeks after you order it.

How to create your own reprint bookletYou may want to reprint several articles from several issuesabout a certain product, technology, or division. Thefinished product could be a 196 -page book with a four-colorcover (we can reprint any RCA Engineer cover) or a 16 -pagebooklet with a title printed on a self -cover. There are morepossibilities than I can tell you about here, and if you'reinterested, call the RCA Engineer office. We'll be happy towork with you.

Reprint booklets, of course, take longer. Turn -around timeis from four to eight weeks, depending upon the customfeatures you want to add.

We create reprints too!Occasionally we originate a reprint book (e.g., Lasers,Microprocessor Technology, Electro-Optics), take quantityorders from publications people throughout the corpora-tion, and advertise these reprints in TREND for single -copypurchases. Several of these books are still in stock at ouroffice and can be ordered individually.

'All policy on reprint content, availability and utilization is administered by the RCAEngineer editorial staff, since additional changes must be made to reprints before theycan be distributed.

How much do they cost?The table below gives you a pretty good estimate on the costof several different sizes and quantities of reprints. Ship-ping, and additional typesetting and photos, of course, costextra.

Pages Quantity

100 500 1000 5000

2 $ 25. $ 51. $ 62.

4 85. 103. 128.

8 99. 140. 171.

12 137. 184. 224.

16 $163. 223. 282.

32 435. 522. $1270.

48 599. 724. 1856.

64 723. 926. 2482.

80 954. 1140. 3050.

96 $1110. $1330. $3624.

2 -color cover (black plus another color) add to the above prices$129. $159. $280.

4 -color cover (from an RCA Engineer cover) add to the aboveprices

$828. $1080.

Could you use a reprint?Why should you want more copies of RCA Engineerarticles? Everyone's reasons are different; some are il-lustrated here. Call our office for more information. We'll beglad to help.

Joan Toothill

Joan Toothill, Art Editor of the RCA Engineer, also runs theEngineer's reprint business.Contact her at:RCA EngineerBldg. 204-2Cherry Hill, N.J.(609)-338-4256

103

Dates and Deadlines

Upcoming meetings

Ed. Note: Meetings are listed chronological-ly. Listed after the meeting title (in bold type)are the sponsor(s), the location, and theperson to contact for more information.

MAY 1-3, 1978 -Microwave Power TubeConf. (IEEE) Naval Postgraduate School,Monterey, CA Prog Info: Dr. George Caryo-takis, Varian Assoc., 611 Hansen Way,Palo Alto, CA 94308

MAY 6-11, 1978 -American Ceramic Soc.80th Annual Mtg. & Expo. (ACS) Cobo Hall,Detroit, MI Prog Info: Frank P. Reid, Exec.Director, The American Ceramic Society,Inc., 65 Ceramic Drive, Columbus, OH43214

MAY 10-12, 1978-Conf. on SoftwareEngrg. (IEEE, NBS) Hyatt Regency Hotel,Atlanta, GA Hayman,Conf. on Software Engrg., PO Box 639,Silver Spring, MD 20901

MAY 15-19, 1978 -Intl. IEEE/AP Symp. &USNC/URSI Mtg., (IEEE, AP, USNC/URSI)Adult Education Ctr., Univ. of MD, CollegePark, MD Prog Info: G. Hyde, Room 2146,Comsat Labs., Box 115, Clarksburg, MD20734

MAY 16-18, 1978-NAECON (NationalAerospace & Electronics Conf.) (IEEE)Dayton Convention Ctr., Dayton, OH ProgInfo: NAECON, 140 E. Mounument Ave.,Dayton, OH 45402

MAY 17-19, 1978 -Circuits & Systems Intl.Symp. (IEEE) Roosevelt Hotel, New York,NY Prog Info: H.E. Meadows, Dept of Elec.Engineering & Computer Science, Co-lumbia Univ., New York, NY 10027

MAY 18, 1978 -Trends & Applications:Distributed Processing (IEEE, NBS) Natl.Bureau of Standards Cony. Ctr., Gaithers-burg, MD Prog Info: Ms. Helen Wood, NBS,Bldg. 225 B 212, Washington, DC 20234

MAY 23-25, 1978-Electro/78 (IEEE) Bos-ton -Sheraton, Hynes Auditorium, Boston,MA Prog Infor: W.C. Weber, Jr., IEEEElectro, 31 Channing St., Newton MA02158

MAY 29 -JUN 1, 1978 -Intl. Quantum Elec-tric Conf. (10th) (IEEE) Atlanta, GA ProgInfo: Joseph Giordmaine, Bell Labs, 600Mountain Ave., Murray Hill, NJ 07974

JUN 4-7, 1978 -Intl. Conf. on Communica-tions (IEEE, et al) Sheraton Hotel, Toronto,Ont. Prog Info: F.J. Heath, Power SystemOperation Dept., Ontario Hydro ElectricPower System, 700 Univ. Ave., Toronto,Ont.

JUN 5-8, 1978-Natl. Computer Conf.(AFIPS, IEEE, ACM) Anaheim Cony. Ctr.,Disneyland Hotel Comp., Anaheim, CAProg Info: Stephen Miller, SRI Intl., ISEDivision, Menlo Park, CA 94025

JUN 5-8, 1978 -13th Photovoltaic Spec.Conf. (IEEE) Shoreham Americana Hotel,Washington, DC Prog Info: John Gold-smith, M/S 169/422, JPL, 4800 Oak GroveDr., Pasadena, CA 91103

JUN 12-13, 1978 -Microcomputer BasedInstrumentation (IEEE) Natl. Bureau ofStandards Conf. Facil., Gaithersburg, MDProg Info: Bradford Smith, A-130 Tech-nology, Natl. Bureau of Standards, Wash-ington, DC 20234

JUN 13-15. 1978 -Power Electronics Spe-cialist Conf. (IEEE) Syracuse, NY ProgInfo: F.B. Goldin, Mail Drop 30, GeneralElectric, Genessee Street, Auburn, NY13021

JUN 19-21, 1978 -Design Automation(IEEE) Caesar's Palace, Las Vegas, NV ProgInfo: S.A. Szygenda, Univ. of Texas at Aus-tin, Dept. of Elect. Engr., Austin, TX 78712

JUN 20-22, 1978 -Pulsed Power ModulatorConf. (IEEE) Statler Hilton, Buffalo, NYProg Info: Sol Schneider, U.S. Army,ERADCOM, Code DREL-TL-BG, Fort Mon-mouth, NJ 07703

JUN 21-23, 1978 -Machine Processing ofRemotely Sensed Data (IEEE) WestLafayette, IN Prog Info: D. Morrison, Pur-due Univ. LARS, 1220 Potter Drive, WestLafayette, IN 47906

JUN 26-28, 1978 -Device Research Conf.(IEEE) Univ. of Calif., Santa Barbara, CAProg Info: Dr. James McGroddy, IBM, T.J.Watson Research Ctr., Yorktown Heights,NY 10598

JUN 26-29, 1978-Conf. on Precision Elec-tromagnetic Measure (IEEE, NBS, URSI/USNC) Conf. Ctr., Ottawa, Ont. Prog Info:Dr. Andrew F. Dunn, Natl. Research Coun-cil, Montreal Road, Ottawa, Ont.

JUN 27-29, 1978 -Intl. Microwave Symp.(IEEE, et al) Chateau Laurier, Ottawa,Ont. Prog Info: A.L. VanKoughnett, Com-munications Research Ctr., POB 11490,Station H, Ottawa, Ont. K2H 8S2

JUL 18-21, 1978 -Nuclear & Space Radia-tion Effects (IEEE) Univ. of New Mexico,Albuquerque, NM Prog Info: B.L. Gregory,Sandia Labs., Dept. 2140, Albuquerque,NM 87115

AUG 20-25, 1978-1 ntersociety Energy Con-version Engr. Conf. (IEEE) Town & Coun-try Hotel, San Diego, CA Prog Info: GeorgeP. Townsend, Hamilton Standard Div.,United Technologies Corp., WindsorLocks, CT 06096

AUG 22-25, 1978 -Intl. Conf. on ParallelProcessing (IEEE) Shanty Creek Lodge,Bellaire, MI Prog Info: Prof. T.Y. Feng,Dept. of Elect. & Comp. Engr., WayneState University, Detroit, MI 48202

SEP 5-8, 1978-COMPCON FALL (IEEE)Washington, DC Prog Info: COMPCONFALL, P.O. Box 639, Silver Spring, MD20901

SEP 6-8, 1978 -Intl. Optical ComputingConf. (IEEE) Imperial College, London,Eng. Prog Info: S. Horvitz, Box 274, Water-ford, CT 06385

Calls for papers

Ed. Note: Calls are listed chronologically bymeeting date. Listed after the meeting (inbold type) are the sponsor(s), the location,and deadline information for submittals.

JUN 13-15, 1978 -Automated Testing forElectronics Mfg. (ATE) Boston Park PlazaHotel, Boston, MA Deadline Info: ab. 200-500 wds. and short biog. to: Sheila Goggin,ATE Seminar/Exhibit Coordinator, Cir-cuits Mfg. Magazine, 1050 Common-wealth Ave., Boston, MA 02215

OCT 9-11, 1978 -Semiconductor LaserConf. (6th) (IEEE) Hyatt Regency Hotel,San Francisco, CA Deadline Info: 6/15/78to T.L. Paoli, Bell Laboratories, 600 Moun-tain Ave., Murray Hill, NJ 07974

OCT 11-12, 1978 -3rd Specialist Conf. onTech. of Electroluminescent Diodes (IEEE)Hyatt Regency Hotel, San Francisco, CADeadline Info: 6/15/78 to R.N. Bhargava,Phillips Lab., Briarcliff Manor, NY 10510

NOV 7-9, 1978 -PLANS '78 (Position Loca-tion & Navigation Symp.) (IEEE) San Die-go, CA Deadline Info: 5/15/78 to NelsonHarnois, Cubic, PO Box 80787, San Diego,CA 92138

DEC 4-6, 1978-Natl. TelecommunicationsConf. (IEEE) Hyatt Hotel, Birmingham,AL Deadline Info: 5/78 to H.T. Uthlaut, Jr.,South Central Bell, PO Box 771, Birming-ham, AL 35201

104

Pen and PodiumRecent RCA technical papers and presentations

To obtain copies of papers, check your library or contact the author or his divisional Technical Publications Adminis-trator (listed on back cover) for a reprint. For additional assistance in locating RCA technical literature, contact RCATechnical Communications, Bldg. 204-2, Cherry Hill, N.J., extension PY-4256.

Automated SystemsD.R. BartlettThe YAH-64 ATE support program-Auto-testcon '77, Hyannis, MA (11/2-4/77)

H.L. Fischer' R.E. HansonNew techniques for automated enginediagnostics-an update-Autotestcon '77,Hyannis, MA (11/3/77)

E. B. GaltonWelcoming address and introduction-Autotestcon '77 Hyannis, MA (11/2/77)

J.I. HerzlingerThe use of commercial equipment inmilitary mobile systems-Mobile Elec-tronics Systems Packaging Symp., Box -borough, MA (11/2/77)

Advanced TechnologyLaboratories

E. HuttoEmile Berliner, Eldridge Johnson and theVictor Talking Machine Company-J.Audio Engineering Soc., Centennial Issue,Vol. 25, No. 10/11 (Oct/Nov 77) pp. 666-73

R. Kenville1G.J. AmmoniC.W. RenoOptical video disc for high rate digital datarecording-Electro-Optics/Laser Conf. andExpo., Anaheim, CA (10/25/77)

W. ThomasIndustry views of government specifica-tions and standards-Defense Specifica-tions Management Course, Army Logis-tics Management Center, Ft. Lee, VA(12/8/77)

Globcom

M. E. Log iad isInternational field tests of digital fac-simile equipment-CCITT, Study GroupXIV, Geneva (11/14-18/77)

GovernmentCommunications Systems

M. Nguyeni R. PickholtzBounds for the queue in loop system -1977Intl. Mtg. Data Transmission, Liege,Belgium (11/14/77)

S. YankelewitzHigh density tape recording-Drexel Uni-versity, Philadelphia, PA (12/5/77)

LaboratoriesC.R. CarlsonThresholds for perceived image sharpness-Proc., SPSA Conf., Rochester, NY (10/24/77) pp. 76-80

R. DawsonlJ. PreisigJ. CarnesiJ. PridgenA CMOS buried -N -channel CCD compati-ble process for analog signal processingapplications-RCA Review, Vol. 28, No. 3(9/77) pp. 406-435

C.A. DeckertEtching of CVD Si3N4 in acidic fluoridemedia-Electrochemical Soc. Mtg., Phila.,PA (5/8-13/77)

R.E. EnstromiD.A. DoaneA finite element solution for stress anddeflection of a centrally loaded siliconwafer-The Electrochemical Soc., Seattle,WA (5/21-26/77)

K.G. HernqvistContinuous laser oscillation at 2703 ang-stroms in copper ion-IEEE J. QuantumElectronics, Vol. QE -13, No. 11 (11/17) p.929

K.G. HernqvistLong -life hollow cathode laser-IEEE J. ofQuantum Electronics, Vol. QE -14, No. 2(2/78) pp. 129-132

M.L. HitchmanAnalysis of equilibrium potentials of hy-drogen tungsten bronzes-J. of Electro-analytical Chemistry, Vol. 85 (1977) pp.135-44

P.J. ZanzucchilD.E. CarlsonOptical properties of discharge -produceda-Si-Electrochem. Soc. Meeting, At-lanta, GA (10/9-14/77)

P.J. ZanzucchilM.T. Duffy' R.C. AligOptical reflectance method for determin-ing the surface quality of sapphire (A1203)-J. Electrochemical Soc., Vol. 125, No. 2(1978) p. 299

P.J. ZanzucchilC.R. Wronskil D.E. CarlsonOptical and photoconductive properties ofdischarge -produced amorphous silicon-J. Appl. Phys.. Vol. 40, No. 12 (1977) p. 5227

Missile and Surface RadarR. J. Bannister' H. B. BoardmanM. D. BrazetORTS-A shipboard automatic test systemProc., Autotestcon '77, Hyannis, MA (11/2-4/77)

J.A. BauerLeadless carrier applications for avionicspackaging-Digital Avionics System Conf.,Los Angeles, CA (11/2-4/77)

J.A.BauerThe use of chip carriers for high packagingdensity, high reliability, high perform-ance products-Workshop on the Impactof LSI on Contact Systems (11/77)

M.W. Buckley, Jr.Project management-Education for Bus-iness and Industry, Ltd., London, England(11/1-2/77)

M.W. Buckley, Jr.Project management-Seminar, DrexelUniv., Phila., Pa. (11/9-11/77 and 12/7-9/77)

R. DiFelice[J. DrenikMechanical design and integration ofphased arrays naval sea systems com-mand-Proc., Mechanical Engineeringin Radar Symp., Washington, DC (11/8-10/77) pp.230-35

B. FellBasic radar concepts: an introduction toradar for optical engineers-Proc.BMDATC (12/77)

W.A. HarmeningStatic mass balancing with a torsionspring and four -bar linkage-Proc., Me-chanical Engineering in Radar Symp.,Washington, DC (11/8-10/77), pp. 169-72

J.W. HurleyIndustrial logistics management-Phila.Chapter of the Soc. of Logistics Engineers(10 -wk symp.) Philadelphia, PA (Oct -Dec77)

P.R. KalataOn system identification with and with-out certainty-J. of Cybernetics, Vol. 8,

Issue 1 (Jan -Mar 1978) pp. 31-50

R.J. KosichAEGIS ship combat system distributedcomputer system design-WINCON '78,Sheraton -Universal Hotel, CA (2/15/78)

E.J. NossenE R. StarnerOne-way doppler extractor-RCA Review,Vol. 38, No. 4 (12/77)

R.P. PerrylL.W. MartinsonRadar matched filtering-Chapter inRadar Technology, Artech (10/77) pp. 163-69

S.A. SteeleCharacteristics of managing real timesolftware development for military sys-tems-AIAA Computers in AerospaceConf., Los Angeles, CA (11/1/77)

L. WeinbergScheduling multifunction radar systems-Proc., IEEE Electronics and AerospaceSystems Cony. (12/77)

105

Patents

Astro-ElectronicsL. Muhlfelder1J.E. Keigled B. StewartMomentum biased active three -axis sat-ellite attitude control system -4071211

R.J. TreadwellBrushless phase locked servo drive -4072884

Automated Systems

R.F. CroceIG.T. BurtonHolographic high resolution contactprinter -4043653 (assigned to U.S. govern-ment)

C.S. WarrenData packets distribution loop -4071706

Avionics Systems

J.E. MillerMulti -target tracker -4072943

J.R. HallCircuit for inhibition of autogenetic falsealarms in a collision avoidance system -4076010

Broadcast SystemsL.J. BazinSignal comparator circuit -4069432

R.A. DischertIR.N. HurstIA.C. Luther, Jr.Signal processor using charge -coupled de-vices -4074307

E.P. HerrmannFour -quadrant multiplier -4071777

R.N. HurstArrangements for testing color televisionsystems -4069500

B.M. PradalNetwork for temperature compensationof an at cut quartz crystal oscillator -4072912

D.M. Schneider1L.J. BazinConstant pulse width sync regenerator -4064541

H.G. Seer, Jr.Apparatus for automatic color balancingof color television signals -4064529

Consumer ElectronicsD.J. CarlsonRemote control transmitter with an au-dible battery life indicator -4067000

M.N. NormanAutomatic beam current limiter -4067048

CommercialCommunicationsSystems Div. StaffL.P. OrchardFood product extrusion apparatus andmethod -4068008

GovernmentCommunications SystemsW.A. BorgeseApparatus for measuring a dimension ofan object -4063820 (assigned to U.S. gov-ernment)

M. LysobeyWide range frequency modulation ofnarrow loop bandwidth phase -locked os-cillators -4074209 (assigned to U.S. Gov-ernment)

L.P. Nahay20Hz ringdown solid state two -wire/four -wire converter -4037065 (assigned toU.S. government)

LaboratoriesC.H. AndersonFlat panel display with beam injectioncleanup -4069439

F. AschwandenSECAM identification circuit -4072983

D.E. CarlsonSemiconductor device having a body ofamorphous silicon -4064521

J.M. Cartwright, Jr.Current mirror amplifiers with program-mable current gains -4064506

A.G. DingwallMethod of making a semiconductordevice -4069577

A.G. Dingwall' B.D. RosenthalDirect -coupled cascade amplifier withautomatically adjusted quiescent outputsignal level -4068182

A.G. DingwalltB.D. RosenthalVoltage controlled oscillator having equal-ly controlled current source and currentsink -4072910

A.H. Firester1J.P. WalentineDefect plotting system -4069484

R.A. GeshnerjL.J. SciambiPhotomask dryer -4068390

W.G. GibsonDelay line network for processing a com-posite electrical signal -4074308

A.G. IpriSilicon resistive device for integratedcircuits -4072974

H.C. JohnsonMonotonically ranging fm-cw radar sig-nal processor -4072947

G. Kaganowiczkl.W. RobinsonVideo disc with a dielectric layer formedfrom styrene and nitrogen -4072985

P.J. KannamTransistor having improved junctionbreakdown protection integrated there-in -4071852

H. Kressell R.V. D'AielloIP.H. RobinsonPolycrystalline or amorphous semicon-ductor photovoltaic device having im-proved collection efficiency -4070206(RCA has title (waiver), NASA has license)

W.G. McGuffinAdaptive delta modulation system -4071825

J.I. PankoveApparatus and method for maskless ionimplantation -4074139

J.I. PankovelD.E. CarlsonElectroluminescent semiconductor de-vice having a body of amorphous silicon -4069492

H.L. Pinch B. Abeles1J.I. GittlemanMethod of making high resistance cer-met film -4071426

J.H. Reisner1W.H. MorewoodiG.H. RiddleElectron beam disc recorder -4074313

W. Rosnowskil R. DenningMethod of fabricating a semiconductordevice -4066485

J.M. Shawl K.H. ZainingerMethod of manufacturing aperturedaluminum oxide substrates -4069094 (as-signed to U.S. government)

Mobile CommunicationsR.M. KongelkaReleasable mounting system -4071217

106

Missile and Surface Radar

P.T. PattersonSystem and method for authenticatingan electronically transmitted document-4064389 (assigned to U.S. government)

Picture Tube Division

A.M. MorrellCorrugated shadow mask assembly for acathode ray tube -4072876

RCA Ltd., Canada

J.M. KeeltyWide acquisition range MSK demodu-lator input circuit -4072905

A. WaksbergAutomatic optical bias control for lightmodulators -4071751

A. WaksbergRetro-reflection communication system-4064434

Solid State Division

A.A. AhmedCurrent mirror amplifier -4068184

B. CrowleVoltage regulators of a type using a

common -base transistor amplifier in thecollector -to -base feedback of the regula-tor transistor -4074181

N.F. Gubitosel L. Gawelkol R.J. SatrianoMethod for cracking brittle material -4068788

R.C. HeunerIS.J. NiemiecM.B. GoldmanIG.I. MortonProtection circuitry for insulated -gatefield-effect transistor (IGFET) circuits -4066918

R.H. !sham, 2ndSawtooth voltage generator for constantamplitude sawtooth waveform fromvarying frequency control signal -4071776

J.R. JasinskilJ.B. PickardIC.E. DonerDouble tuned input circuit for televisiontransmitter amplifier -4070627

N. KucharewskiAmplifier circuit -4069431

A.J. LeidichAmplifier circuit -4064463

T.J. RobeInput stage for fast-slewing amplifier -4074205

Microprocessor technology, electro-optics, andautomotive electronics anthologies available

Three new comprehensive reprints that u-niquely display RCA's current skills in

microprocessors, electro-optics, andautomotive electronics are now available.These reprints consist of selected papersfrom four of the most popular recent issuesof the RCA Engineer. Only papers of currentvalue are included.

Microprocessor technology offers 118pages of general microprocessor informa-tion, discusses RCA's general-purposemicroprocessor-the COSMAC CDP1802-as well as advanced RCA designs,and describes more than a dozen specificapplications. This reprint contains 24 ar-ticles, presenting the work of more than 40RCA engineers.

Electro-optics covers electro-optics work atRCA and has specific sections on lightsources, light transmission devices andmedia, light detectors, electro-opticssystems, electro-optics for inspection anddetection, and television. This reprint con-tains 28 articles presenting the work of morethan 40 RCA engineers; it also containsseveral editorial features introducing themajor topics.

Automotive electronics presents thetechnical contributions of more than 30RCA engineers to research, design, andapplication of automotive devices, equip-ment, and systems.

In addition to these three reprints, a limitedsupply of past reprints is available. Allreprints have color covers. A complete listfollows:

Reprint Price Date

Microprocessor technology(118 pages)

Electro-op tics(145 pages)

Automotive electronics(112 pages)

COS/MOS technology(76 pages)

Solid state technology(135 pages)

Consumer electronics(60 pages)

Integrated circuits(88 pages)

$2.00 1977

$2.50 1977

$2.00 1977

$1.00 1973

$1.50 1971

$1.00 1970

$1.00 1967

How to Order ...Although single -copy orders will beaccepted, minimum orders of five arepreferred Send your name, location, andaccount number (if a personal order-sendcheck or money order, payable to RCATechnical Communication Programs) to:

ReprintsRCA Technical Communication ProgramsBldg. 204-2Cherry 141I, N.J. 08101Phone: PY-4256

107

Engineering News and HighlightsMore Fellows, more PEsApparently, one way to elicit readerresponse is to publish a list of people, andmiss some of them. Last issue we publishedtwo-a list of RCA engineers who hadattained the grade of Fellow of the IEEE anda list of Licensed Professional Engineers atRCA.

Thanks to all of you who wrote in tocomplete our lists, particularly to Dr.George Brown who called our attention toseven IEEE Fellows left off our original list.We will continue to update our lists as wereceive new information. Here are the up-dates we have thus far:

IEEE Fellows

Goldsmith, Alfred N. 1915VanDyck, Arthur F. 1925Shackelford, Ben E. 1938Guy, Raymond F. 1939Crosby, Murray G. 1943Seeley, Stuart 1943Burnap, Robert S. 1947Ingles, Harry C. 1948Bingley, F. James 1950Law, Russell R. 1953Firestone, William L. 1965Hittinger, William C. 1967Cohen, Robert M. 1967Dodds, Wellesley J. 1972Pritchard, Dalton H. 1976Clark, John F. 1976

Licensed professional engineers

Alascom

Henderson, W.Strid. G.Swanson, D.

AK-CE4488AK-EE4502AK-CE4507

Astro-Electronics

D'Amanda, A.W. NJ -18747Schnapf, A. NJ -9310

Avionics Systems

Aires, R.H. CA-CS2196

Broadcast Systems

Schacht, W.F. NJ -12594

Consumer Electronics Division

Chaney, H.E. IN -15932Olson, L.A. IN -12407

Government CommunicationsSystems

Bradshaw, J.L. NJ -24610

Laboratories

Gange. R.A. NJ -21860

Missile and Surface Radar

Groman, E.M.

Ringo, K.A.

PA -17495ENJ -11557OH -E-019719

Picture Tube Division

Ehrlich, M.A.Fanale, J.M.Ferguson, J.E.Follis, G.R.LeMay, B.B.Parker, W.A.

Pearlman, S.Schneider, R.K.Swander, T.E.Thall, E.S.

OH -E040294PA -007009EOH -E29169IN -06773OH -E034462PA -29874-EIN -12907PA -022917EOH -28908OH -19669Canadian 5131200CHEM-MET

Service Company

Williams, J.A., Jr. VI -325ENH -2464

Solid State Division

Cole, E.W. M.I.E.R.E.Davenport, F.A. M.I.E.R.E.Free, J.J. PA -012070EGershey, J.T. PA -23048-EMedway, N.W. M.I.E.E.Stevens, R. M.I.E.R.E.

Promotions

Astro-Electronics

L. Jones from Senior Engineer to Manager,(Spec) Engineering.

R. Horner from Senior Engineer to Manager,(Spec) Engineering.

Consumer Electronics

Ronald L. Hess from Head, Deflection andPower Supply Systems Research, RCALaboratories, to Manager, TV SystemsDevelopment, New Products Laboratory.

Arthur Kaiman from Member TechnicalStaff, RCA Laboratories, to Manager,Manufacturing Systems, New ProductsLaboratory.

Robert M. Rast from Head, TV SystemsTechnology Research, to Manager, NewProducts Development, New ProductsLaboratory.

Recent books by RCA authors

Dynamic Business Strategythe art of planning for success

Theodore A. SmithPublished by McGraw-Hill[$12.95]

Author Ted Smith was an Executive VicePresident of RCA for 13 years, holdingvarious positions as a general manager ofgroups of operating units, and serving asDirector of Corporate Planning. Subse-quently, as a consultant for The ConferenceBoard, he was Associate Director for studyof management trends. In this project,leaders in management sciences preparedreports which were published under thetitle: Challenge to Leadership: Managementin a Changing World.

The jacket of Dynamic Business Strategyoffers the following introduction:

"The myth that business strategy must beintuitive costs American industry millions ofdollars each year. Reflect on the numerousdivestitures, liquidations, and writeoffs thathave been taken by otherwise well -managed companies. Strategy is the for-mula for success . . . and it need not beintuitive."

"Here is a realistic guide that describes step-by-step methods for developing a winningbusiness strategy, for avoiding future pit-falls, and for amplifying current profitpotential."

10

Awards

Eight receive Technical Excellence Awards at Moorestown

William Beckett-for creativity, judgment,and team leadership in design, productionand integration of AN/SPY-1A InterfaceSimulator.

Glenn Everhart-for outstanding develop-ment effort on critical computer programmodules for the AN/TPQ-27 operation.

George Friedman-for creative andresourceful design effort leading to thedevelopment of a Requirements Traceabili-ty Program for a complex mass of re-quirements and functional threads on aclassified software development program.

Paul Horton -for personal proficiency andperseverance in developing systems andprocedures to assure total operating safetyof the AEGIS Combat System during in-stallation, checkout, and system testing atthe CSED site.

John Mehling-for special contributions indevelopment of hardware test supportsoftware for AN/TPQ-27 equipment sub-systems.

Herbert Olson-for technical leadership inthe formulation of circuitry and logic forimproved integration of the AEGIS GuidedMissile Launching System with overallsystem operation.

David Phillips-for outstanding perfor-mance in the definition of AEGIS Commandand Decision software performance re-quirements for advanced AN/SPY-1AECCM capability.

William Smith-for technical competenceand personal initiative demonstrated in thedesign and implementation of the real-timeportion of the AN/SPY-1A InterfaceSimulator.

Stiller and Wine are Labs Fellows

Wine Stiller

Dr. William M. Webster, Vice President, RCA Laboratories, recently appointed Thomas M.Stiller and Charles M. Wine Fellows of the Technical Staff, in recognition of theiroutstanding contributions. The designation of Fellow, which was established by RCALaboratories in 1959, is comparable to the same title used by universities and virtually alltechnical societies, and is given in recognition of a record of sustained technicalcontribution in the past and of anticipated continued technical contribution in the future.

Presently the Fellows of the Technical Staff at RCA Laboratories are:

Charles H. Anderson Karl G. Hernqvist Thomas M. StillerKern K.N. Chang Ralph W. Klopfenstein Chih Chun WangRoger L. Crane Simon Larach Paul K. WeimerAndrew G.F. Dingwall Jacques I. Pankove Richard WilliamsRobert E. Flory Dalton H. Pritchard Charles M. WineJames J. Gibson Allen H. Simon J. Guy WoodwardJoseph J. Hanak Henry S. Sommers, Jr.

Beckett

Friedman

Mehling

Phillips

Everhart

Horton

Olson

Smith

109

Joseph Corsowins annual technical excellence award

All the 1977 technical excellence award winners from Missile and SurfaceRadar gathered for this photo. Joe Corso, the annual award winner isholding his award-flanked by Max Lehrer (left), Div. VP and GeneralManager, MSR, and Joe Volpe, Chief Engineer.

Joseph T. Corso, Principal Member, Engineering Staff, at Missile andSurface Radar in Moorestown was recently selected as the winner of the1977 Annual Technical Excellence Award.

Joseph C. Volpe, Chief Engineer, MSR Engineering Department, inannouncing the award said:

"The choice of an annual award winner is never an easy one, and in 1977CETEC [Chief Engineer's Technical Excellence Committee] selected arecord 29 engineers for quarterly Technical Excellence awards. JoeCorso's selection as a winner of the 1977 Annual award, then reflects a levelof achievement that is truly extraordinary.

"Although Joe has been with MSR only since 1970, we've had the benefit ofhis missile guidance expertise for a considerably longer time-from theearly 60's when he served as a consultant on missile -related developmentsleading ultimately to the AEGIS Program. His work has been uniformlysuperior, as exemplified by ... his accomplishments on AEGIS during 1977.

Hap Easter receives two honors

Dr. William L. Firestone (left), Div. VP and GeneralManager, Avionics Systems, hands Hap Easter one of histwo awards in a ceremony in Dr. Firestone's office.

Finis C. ("Hap") Easter of RCA Avionics Systems in VanNuys, California, recently received two honors for hisprofessional and technical activities. He was elected tothe College of Fellows of the Institute for the Advance-ment of Engineering, "in recognition of his outstandingcontribution to the advancement of the engineeringprofession."

His second honor came from the San Fernando ValleyEngineers' Council, a group of 22 engineering societiesof the various disciplines. "Hap" was one of eightreceiving an Engineering Merit Award "for outstandingprofessional qualities and meritorious achievementwithin the field of engineering."

Hap, an engineer with RCA for 27 years, has five issuedpatents, three additional patents pending and severalactive disclosures. This year he is chairman of the SanFernando Section of IEEE with a membership of some1700.

Staff Announcements

RCA RecordsPaul Potashner, Group Vice President, ap-pointed Robert D. Summer President, RCARecords Division.

Consumer ElectronicsJ.B. Thomas, Manager, ManufacturingEngineering, appointed Douglas J.McGinnis Manager, Process Engineering.

President and Chief ExecutiveOfficerEdgar H. Griffiths, President and ChiefExecutive Officer, announced the followingappointments:

Eugene J. Beyer, Jr. is Senior Vice Presi-dent and General Counsel. In this capaci-ty, he is responsible for the RCA laworganization and the Secretary's office.

Steven S. Barone continues as Vice Presi-dent, Licensing, and reports to the Presi-dent and Chief Executive Officer.

John V. Regan continues as Vice Presi-dent, Patent Operations, and reports toWilliam C. Hittinger, Executive Vice Presi-dent, Research and Engineering.

Patent OperationsHarold Christoffersen, Director, Patents,Solid State and Electronic Systems, hasappointed Robert A. Hays, Resident PatentCounsel, Cherry Hill; and Birgit E. Morris,Sr. Managing Patent Attorney.

Edward J. Norton, Director, Patent Planningand Administration, has appointed SharonK. Stepno Manager, Patent OfficeWashington.

Eugene M. Whitacre, Director, Patents,Consumer Products and Broadcast Equip-ment, has appointed Paul J. Rasmussen, Sr.Managing Patent Attorney; and Joseph S.Tripoli, Managing Patent Attorney.

Commercial CommunicationsSystems DivisionAdron M. Miller, Manager, RCA PhotophoneSystems, has appointed Gordon E. Cordellas Manager, Product Management, forRCA's line of professional motion picturefilm records and projectors.

Automated SystemsFernand F. Martin has been appointedManager, Radiation Systems Engineering.

110

Advanced TechnologyLaboratoriesJames A. Colligan has been appointedManager, Marketing for AdvancedTechnology Laboratories.

RCA GlobcomEugene F. Murphy, President, RCA GlobalCommunications, Inc., has announced theelection of Dr. Thomas Mathai to the newlycreated position of Vice President, DataServices.

RCA LaboratoriesThomas 0. Stanley, Staff Vice PresidentResearch Programs, has announced thatthe organization formerly known asManagement Information Systems will berestructured as follows: Marvin Blecker isappointed Head, Systems AnalysisResearch; Emilie M. Lengel continues asManager, Automation and Computing Ser-vices; and Warren C. Sayre is appointedManager, Administrative Systems.

Solid State DivisionCarl R. Turner, Division Vice PresidentIntegrated Circuits, and Gerald B. Herzog,Staff Vice President, Technology Centers,announced that Joseph H. Scott, Director,Integrated Circuit Technology will assumeadditional responsibilities. He will provideall technical and strategic direction inSilicon and Sapphire Technology within theSolid State Division, RCA Laboratories andTechnology Centers. While continuing inhis present position under the Staff VicePresident, Technology Centers, for thisassignment he will report to Carl R. Turner,Division Vice President, Integrated Circuits,Solid State Division.

Picture Tube DivisionCharles W. Thierf elder, Division Vice Presi-dent Product Safety, Quality and Reliability,has announced the organization of ProductSafety, Quality and Reliability as follows:Sherman L. Babcock, Administrator,Technical Quality Programs; David C.Ballard, Manager, Product Safety; WellesleyJ. Dodds, Director, Quality and ReliabilityAssurance Operations Analysis; J. EdwardFagan, Administrator, Quality AssuranceCoordination; Frank J. Hinnekamp,Manager, Life Test, Reliability and Warran-ty; and J. Paul Sasso, Administrator,Customer Quality Acceptance.

Barr on IEEE committee

Ken Barr, Component Engineer, at the Con-sumer Electronics Division in Indianapoliswas recently appointed to a three-year termas RCA's representative to serve on theAdministrative Committee of the Broadcast,Cable TV and Consumer Electronics Groupof the IEEE. Currently, Ken is also thetreasurer of the Central Indiana IEEE.

Obituaries

Edward Bliss, an electrical engineer withRCA Global Communications, died April 1,1978.

Mr. Bliss designed the transmitter for thehistoric Relay Communications Satelliteand for the first Lunar Excursion Moduleused during the moon landing in 1969. Forhis accomplishments, he received the RCAEngineering Award in 1970. During histenure with RCA, he also pioneered theapplication of computer -aided designtechniques to the development of solid-state microwave devices and authoredseveral technical papers on that subject. Hewas a recognized authority on the design oftravelling wave tubes and contributed achapter on the subject to the RCA ElectronTube Design book.

He joined the RCA American Com-munications Division in 1974 to work on theRCA Satcom Satellite Program. He was partof the launch team that supported twosuccessful launches and conducted the on -order performance checkout.

Nils E. Lindenblad, one of RCA's mostprolific inventcrs, died on February 18. Heretired from RCA Laboratories in 1960.

A pioneer in transoceanic radio com-munications in the 1920's and 1930's, hejoined the RCA Corporation in 1920 at thecompany's transmitting station at RockyPoint, L.I. He transferred to RCALaboratories it Princeton in 1950.

Mr. Lindenblad was credited with more than300 patents. He was widely known for hisdevelopment of the rhombic and slot anten-nas, the travelling wave tube, and the basicelements of thermoelectric coolingsystems. He helped to develop the firstwideband television antenna placed on topof New York's Empire State Building in 1938.

A Fellow of the IEEE and the RCALaboratories, Mr. Lindenblad was awardedthe 1958 David Sarnoff OutstandingAchievement Award "for his invention andpioneering development of many importantelectronic devices and for his research onthermoelectric cooling apparatus."

Finite Element Symposium Program Held

An RCA Finite Element Symposium was held on March 13 and 14 at RCALaboratories in Princeton. About 60 engineers and managers assembled to listento formal presentations, discuss mutual problems, and, for those with no directexperience, to gain an understanding of this powerful new tool.

Ron Enstrom, RCA Laboratories, was chairman of the program which includedspeakers from Astro Electronics, Missile and Surface Radar, Solid State Division,Picture Tube Division, and RCA Laboratories as well as a guest speaker fromMassachusetts Institute of Technology.

Phillips receivesGoldsmith award

John Phillips, Editor of the RCA Engineerrecently received the 1977 Alfred N.

Goldsmith Award of the IEEE Group onProfessional Communication.

Established by the Professional Com-munication Group in 1974, the Alfred N.Goldsmith Award is given in recognition ofservice within the Group's organization toimprove the quality of engineering corn -m un ication.

Lauffer promoted to AssociateEditor, RCA Engineer

Bill Lauffer, who has worked as AssistantEditor of the RCA Engineer since May 1976,was recently promoted to Associate Editor.

Bill has made important contributions to thequality of the RCA Engineer and has con-tributed to, or initiated, several innovativefeatures.

Authors and inventors honored at Moorestown

Fifty-one individuals werehonored on February 23 at Mis-sile and Surface Radar'sAuthors' Reception held inMoorestown, N.J. This recep-tion, hosted by Joe Volpe, ChiefEngineer, was the eleventh in aseries to honor people who havepresented or published papersor received patents.

In congratulating the MSRauthors and inventors, Mr. Volpecalled attention to the effortsinvolved in professionalauthorship.

Willis is new Ed Repat Consumer Electronics

Don Willis has been appointed EditorialRepresentative for Consumer Electronics inIndianapolis. Don is a senior engineer andhas worked at CE for 15 years on signalcircuiting, video tape recorders, and deflec-tion circuits. He has written several articlesand holds 17 patents.

As Editorial Representative, Don will assistCE authors with papers for the RCAEngineer, and will keep the editors informedof new developments as well as professionalactivities, awards, publications, andpromotions in their areas.

112

Editorial RepresentativesContact your Editorial Representative, at the extensions listed here, to scheduletechnical papers and announce your professional activities.

Commercial CommunicationsSystems DivisionBroadcast SystemsBILL SEPICH* Camden, N.J. Ext. PC -2156KRISHNA PRABA Gibbsboro, N.J. Ext. PC -3605ANDREW BILLIE Meadow Lands, Pa. Ext. 6231

Mobile Communications SystemsFRED BARTON* Meadow Lands, Pa. Ext. 6428

Avionics SystemsSTEWART METCHETTE* Van Nuys, Cal. Ext. 3806JOHN McDONOUGH Van Nuys, Cal. Ext. 3353

Electronic Industrial EngineeringJOHN OVNICK* N. Hollywood, Cal. Ext. 241

Government Systems DivisionAstro-ElectronicsED GOLDBERG* Hightstown, N.J. Ext. 2544

Automated SystemsKEN PALM* Burlington, Mass. Ext. 3797AL SKAVICUS Burlington, Mass. Ext. 2582LARRY SMITH Burlington, Mass. Ext. 2010

Government Communications SystemsDAN TANNENBAUM* Camden, N.J. Ext. PC -5410HARRY KETCHAM Camden, N.J. Ext. PC -3913

Government EngineeringMERLE PIETZ* Camden, N.J. Ext. PC -4155

Missile and Surface RadarDON HIGGS* Moorestown, N.J. Ext. PM -2836JACK FRIEDMAN Moorestown, N.J. Ext. PM -2112

Solid State DivisionJOHN SCHOEN* Somerville, N.J. Ext. 6467

Power DevicesHAROLD RONAN Mountaintop, Pa. Ext.827SY SILVERSTEIN Somerville, N.J. Ext. 6168

Integrated CircuitsFRED FOERSTER Somerville, N.J. Ext. 7452JOHN YOUNG Findlay, Ohio Ext. 307

Electro-Optics and DevicesRALPH ENGSTROM Lancaster, Pa. Ext. 2503

Consumer ElectronicsCLYDE HOYT*Indianapolis, Ind. Ext. VH-2462FRANCIS HOLT Indianapolis, Ind. Ext. VH-4393PAUL CROOKSHANKS Indianapolis, Ind. Ext. VH-2849DON WILLIS Indianapolis, Ind. Ext. VH-2829

SelectaVision ProjectROBERT MOORE Indianapolis, Ind. Ext. VR-3313

RCA Service CompanyJOE STEOGER* Cherry Hill, N.J. Ext. PY-5547RAY MacWILLIAMS Cherry Hill, N.J. Ext. PY-5986DICK DOMBROSKY Cherry Hill, N.J. Ext. PY-4414

Distributor andSpecial Products DivisionCHARLES REARICK* Deptford, N.J. Ext. PT -513

Picture Tube DivisionED MADENFORD* Lancaster, Pa. Ext. 3657NICK MEENA Circleville, Ohio Ext. 228JACK NUBANI Scranton, Pa. Ext. 499J.R. REECE Marion, Ind. Ext. 566

Alascom

PETE WEST* Anchorage, Alaska Ext. 7657

AmericomMURRAY ROSENTHAL* Kingsbridge Campus, N.J. Ext. 4363

GlobcomWALT LEIS* New York, N.Y. Ext. 5029

RCA RecordsJOSEPH WELLS* Indianapolis, Ind. Ext. VT -5507

NBC

BILL HOWARD' New York, N.Y. Ext. 4385

Patent OperationsJOSEPH TRIPOLI Princeton, N.J. Ext. 2491

Research and EngineeringCorporate EngineeringHANS JENNY* Cherry Hill, N.J. Ext. PY-4251

LaboratoriesMAUCIE MILLER Princeton, N.J.Ext. 2321LESLIE ADAMS Somerville, N.J. Ext. 7357

*Technical Publications Administrator, responsible forreview and approval of papers and presentations.

44 it: T.

4

RC/1 EngineerA technical jourria, 3::ubiihed by Corporate Technical Communications"by and foY the7RCA Err tneer"

Printed in USA Form Np. RE -23-5

CLIFFORD WINNER, JR.

Documents

radar (raidar). n. the presence and location of an …...dem. tot ,eterrntnnv.$15,,et by n. return 1,0711u 1144.1 MU, frelthe, J(eurtifyr) Our cover combines radar's past and present