Our cover shows an electro-optic thin-filmwaveguide transmitting three different colorlaser beams. Waveguides of this type have beenused to make switches and modulators capableof placing information on light at microwavefrequencies, while requiring drive powers at leastan order of magnitude less than those requiredby previous methods. (See J. Hammer, p. 71.)

Photo credit: Tom Cook, RCA Laboratories,Princeton, N.J. Experimental setup: Clyde Neil

Mg] EngineerA technical journal published byRCA Research and EngineeringBldg 204-2Cherry Hill, N J 08101Tel PY-4254 1609-779-4254)Indexed annually in the Apr/May issue.

John PhillipsBill LautterJoan -oothillFrank Strobl

Pat Gibson

Joyce Davis

Jay Brandinger

Bill Br3ok

Joe Donahue

Hans ,.envy

Arch Luther

Howie Rosenthal

Carl Turner

Joe Vclpe

Bill Underwood

Bill Webster

Ed Burke

Walt I:lnnen

Charlie Foster

RCA Engineer Staff

Editor

Assistant EditorArt EditorContributing EditorCompositionEditorial Secretary

Editorial Advisory Board

Div. VP, Engineering,Consumer ElectronicsVP, Engineering,RCA AmericomDiv. VP, Engineering,Picture Tube DivisionManager, TechnicalInformation ProgramsChief Engineer, Engineering,Broadcast SystemsStaff VP, EngineeringDiv. VP, Solid StatePower Devices

Chief Engineer, Engineering,Missile and Surface RadarDirector, EngineeringProfessional ProgramsVP, Laboratories

Consulting Editors

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

To disseminate to RCA engineers technical information of professional value To pubrish in an appropriate manner important technical developments atRCA. anc the role of the engineer To serve as a medium of interchange of technical information betweenvarious groups at RCA To create a community of engineering interest within the company bystressing the interrelated nature of all technical contributions To help publicize engineering achievements in a manner that will promotethe interests and reputation of RCA in the engineering field To provide a convenient means by which the RCA engineer may review hisprofessional work before associates and engineering management To announce outstanding and unusual achievements of RCA engineers in amanner most likely to enhance their prestige and professional status

Electro-optics

Electro-optics is one of the twentieth century's most important andpervasive technologies.

The most obvious and well publicized electro-optic system is television,and the enhancement of this system has been a major RCA thrust for atleast three decades. Creative engineering efforts have extended ourelectro-optics capabilities far beyond the entertainment sphere to allowman to "see" inside the human body, at the bottom of oil wells, faroutside the solar system, and in almost total darkness.

But well beyond extending man's vision, electro-optics has openedimportant new vistas in the transmission, processing, and storage ofinformation by using the laser's ability to concentrate beams of coherentoptical energy. For example, optical fibers promise to replace both wiresand radio waves used for communications; optical memories are beingpushed toward packing densities approaching the wavelength of light;and integrated optical devices have been developed to access thesememories.

Finally, the exploitation of solar power in answer to our energy needswill define an even further expansion of electro-optics.

RCA has been, and continues to be, a leader in electro-opticstechnology and products, and the growth in applications seems certain.Thus, our broad technological base, coupled with sound businessjudgement, offers excellent opportunities for us in the future.

ERalph E. SimonDivision Vice -PresidentElectro-Optics and DevicesSolid State Div.Lancaster, Pa.

electro-opticsit all starts with sources, detectors, and media

LIGHT

PHOTOCATHODE

DYNODES

ANODE

PHOTOELECTRONSSIGNAL

36 photomultiplier tubes:single -photon detectionwith noise -free gain

fiber -optics passes its firstlarge-scale test

16 semiconductor lasers findcommunications applications

i8 design digital filters

Upcoming issues

Electro-optics is too big a subject for one issue,so our Apr/May issue will beabout E -O systems-image tracking, optical recording, the hand-heldlaser rangefinder, and CCTV surveillance, amongothers.

Microprocessors are the biggest thing to hitelectronics since the transistor. Our next issue(Feb/Mar) tells how they'll affect your particularline of work and how to get involved, besidesgiving a host of applications and information onRCA's COSMAC microprocessor.

2

R.E. Flory

O.F. Whitehead' D.E. Hutchison

H.E. Haynes 14

16

I. Gorog 18K.G. Hernqvistl R.W.Longsderff 19

H. Kressell H. LockwoodM. Ettenbergi I. Ladany 22

I. Gorogl P.V. Goedertierl J.D. KnoxI. Ladanyl J.P. Wittkel A.H. Firester 'n

E.D. Savoye 36R.G. Neuhauser 40G.A. Robinson 47

R.J. McIntyre' P.P. Webb 52

56

D.G. Herzog 58J.P. Wittke 60

D.A. de Wolf 64J.M. Hammer 71

L. Shapiro

R.E. Simonds' N.B. Mills J.F. Eagan 91R.E. Simonds 94

J. Lewin 98

105107109110

Copyright 1977 RCA CorporationAll rights reserved

ma Engineeron the job/off the jobElectronic astronomy

engineer and the corporationTechnical information: where to get it

Where the electro-optics action is at RCA

light sourcesLight -generating devices: an introduction

Inexpensive He-Ne laser tube construction

Light sources for fiber-optic communications

Applying injection lasers to information scanning

light detectorsLight -detecting devices: an introduction

The silicon -target vidiconThe silicon intensifier target tube" see ng in the dark

Avalanche photodiodes: no longer a laboratory curiosity

light transmission media and devicesOptical transmission media and devices: an introductionOptical -fiber communications linksOptical propagation through turbulent airIntegrated optics and optical waveguide modulators

special-short courseDigital electronics, Part III-Digital filters

general interest papersOperation of the RCA Frequency Bureau

How the Communications Act affects youGround -control system for Satcom satellites

departmentsPen and Podium

Patents

Dates and DeadlinesNews and Highlights

3

of 'ah@ ,cpWoff the jobElectronic astronomy

R.E. Flory

An engineer working in television has anatural conditioning in the grand -sounding concept of the "extension ofman's vision." A life-long interest inphotography and a familiarity with televi-sion equipment combined to lead me to aninterest in astro-photography. often withan electronic twist.

What are the characteristics of astro-photography that make it different fromSunday afternoon snapshot shooting orthe slide photography of your last vacationtrip? Generally. astro-photographic sub-jects are very dim and, or are apparentlyvery small. The apparent small site is, ofcourse, more a manifestation of the greatdistances involved. In my pursuit of theseeltisiNe images, I have used three basictechniques: television, image intensifica-tion. and conventional photography. Eachof these techniques seems to have a place inassisting me to get records of differentclasses of astral objects.

Conventional photography

Some of the most striking astro-photographs are those of relatively large

Star gazing and planet watching can be made morepleasurable by using photography and electronics to addnew dimensions to amateur astronomy.

astral objects of a sin that can be discernedOth the naked eye or with binoculars.I aking such photos involves the use of

conventional photographic optics and longexposure times to make a record of verydim objects. One such class of objects arediffuse galactic nebulae. They are oftenseveral degrees in extent but are so dim thatthey are seldom seen in populated areas, ifat all. In central New Jersey, it is rare toexperience a night so dark that you cannotsee your hand before your face. Yet,extreme darkness is a requirement for goodphotographs of these nebulae.

Fortunately, there is a large class of thesenebulae that emit only red ligfit becausethey are composed of glowing hydrogengas that emits only at certain characteristicwavelengths. It is possible to substantiallyincrease the contrast of these objects byusing a deep red filter over the camera lens.Because the nebulae are so dim,the film must be exposed forten minutes to an hourto get a good image.

The choice of film for such photographsbecomes very important. Conventionalphotographic film has a significant defectthat prevents satisfactory use for manyastro-photos. This is what is calledreciprocity failure. Simply stated, it is thatthe product of exposure time and intensityis not a constant over all light intensities.The result is that such films as TRI-X can-not record objects so dim that they requiremore than about five minutes exposure.Special films have been developed thatovercome this problem but at the expenseof more picture grain and poor keepingproperties. These films have also beenmade with more red sensitivity and highercontrast than usual; both characteristicsaid in recording these dim objects.

Bob Flory joined RCA Laboratories in 1959and has worked mainly on television andimaging and display systems. Heparticipated in the early VideoDisc andHolotape projects. He was involved in thedesign of the RCA Space Mountain exhibitin Walt Disney World and in the installationof an RCA television system on Kish Island.Iran. He is presently working on advancedvideo recording techniques, one of which isthe high -density video recording system fortv broadcast use. Mr. Flory received an RCALaboratories Achievement Award, and heshared the 1972 David Sarnoff Team Awardin Science. In 1976, he was appointed Fellowof The Technical Staff.

Contact him atCommunications Research LaboratoryRCA LaboratoriesPrinceton. N.J.Ext. 3320

Reprint RE -22-4-13Final manuscript received October 14. 1976

4

.:

.et. . : i 9.

. .4,.. 9 hir . ... .4... . kV ! '

..... ..... . ...; . .$4,,,. .

.. . it .1 - . . lifil. . .;; .... . 1 . . 7.. -.7 4/ 7% ' .... -._;....r":-..... ....,,.- *, .: r'. , . ." " tia .9. . , . 1. ' . . ..91 411". S. 141,40 .. 9... ,, . . 7,... . la .. .... . ..

. 4S .. . 3A.. .. , , -,.., a... -N.,'" :. i .:. i; ... .,,,:. ***.! oirib,.. 6.'. ,

t ' "'''. )."teb; ' St' x. - .* !lq1/4(..: 1 .:"..* A. .;. '411.; 1 '. 9 ./1114.4. : 91 i . ' .r it :.

'1".:*- .1'.',A . ' .' . *C 14 '.1:.1,1, .040:ip.-. .., .. A ..

:. .0; t . '.. ' ,;... 4 : ..

..11110......

..... ,

... .

.:

., .! .

:. i .' '1, . ' j

' 044

. P "1 ..... fili 41.. ' . '. . , ,ir... .fall Art co_ .4.: t ,jiit . ; - ' : -

.4*" :iik I'' - -*. a lir . .- IP %

. . " ' ' . 'C ' '.

Fig. 1"North American" nebula photographed with Kodak 130aE low -reciprocity -failure film using a Wratten 25 filter, a 300 -mm f, 2.5lens, and with an exposure time of 20 minutes. The picture is approximately four degrees wide.

Fig. I shows one such object,photographed in September 1976, when itwas straight overhead in the constellationof Cygnus. This is the so-called NorthAmerican nebula and its companion, thePelican. This picture was made withtwenty -minute red-light exposure. on 35-mm film, using a 300 -mm focal length lenshaving an/ number of 2.5. This is a ratherimpressive lens taken from a military sur-plus reconnaissance camera. However.135 -mm or 200 -mm camera lensesmounted on a 35 -mm camera can producegood results.

To achieve clear photographs with suchlong exposures, it is necessary to cancel outthe apparent movement of sky objects dueto the earth's rotation. This is done bymounting the camera on a platform thatrotates about an axis parallel to the earth'saxis once in twenty-four hours. The photo

on the first page of this article shows theequipment I use for nebula photography.Because it is impossible to achieve suf-ficient mechanical perfection in gearingand to eliminate errors due to at-mospheric refraction and mislocation ofthe camera's axis, it is necessary to monitorthe exposure through an auxiliarytelescope. This telescope must be of highmagnification, but quality is relativelyunimportant, as it is used only to keep a"guide" star centered on crosshairs in theeyepiece.

One recent technological improvementhere is the use of an LED to illuminate thecrosshairs. The star selected for guiding isoften rather dim, and it is a real obser-vational advantage to have the cross -hairimage appear in red and the star image inwhite. The correcting movements are madeby changing the rate at which the

clockwork rotates the platform axis. 1 his isusually done by means of a variable fre-quency oscillator driving a synchronousmotor.

Image intensification

Many sky objects are somewhat brighterthan the diffuse nebulae but are very muchsmaller. These include the globular clustersand planetary nebulae. The former objectsare clusters of stars so close together, andso far from us, that they appear to bepacked into a ball. The planetary nebulaeoften have the form of a disc or a ring andhave an appearance in the telescopesomewhat similar to the planets of oursolar system. All of these objects are manythousands of light years away from earth.A light year is defined as that distance thatlight travels in one year's time and is almost6 101- miles. At such great distances,these objects become very small.

5

I.- 10 arc -minutes -WI

,t

.. t

. tql

I

Fig. 2Globular cluster In Hercules taken in a one -second exposure on Kodak TRI-X film, usingan image intensifier. The telescope is a Celestron with a 2000 -mm focal length at f/10.Most stars fall inside a ten arc -minute diameter sphere.

Fig. 3Planet Jupiter photographed from a tvmonitor in 1/30 second on Kodak TRI-X film.The camera used contained a 2/3 -inchsilicon vidicon. The telescope is a Celestronwith a 4000 -mm focal length at f/20. Theplanet has a diameter of about 45 arc -seconds.

A well known example of a globular clusteris the one in the constellation of Hercules.Most of the stars in this cluster fall within asphere having an angular diameter of tenarc -minutes. Such an object is smallenough to require a true telescope toresolve. I use a 2000 -mm -focal -lengthCassegrain mirror system of 200 -mm aper-ture made by Celestron, Inc. Even thoughthe object is a cluster of stars, the greatdistance and the I number of 10 for thistelescope combine to make it a dim object.The manufacturer of my telescope haspublished a picture of this object whichrequired a ten-minute exposure on thesuper -fast, low -reciprocity -failure film. I

have used a three -stage RCA image in -

Fig. 4Planet Saturn taken from tv monitor withsame equipment and conditions as in Fig. 3.The planet diameter is 20 arc -seconds.

tensilier and a 1:1.112 relay system to imagethe intensifier output onto amateur TR I -X35 -mm film. With this equipment, I haveobtained an identical result with only atwo -second exposure.

Such short exposures are a great advantagewith small objects because it is extremelydifficult to make the precise position cor-rections needed for long exposures withlong focal lengths. Fig. 2 shows the Her-cules cluster as photographed with theimage intensifier. It "boggles the mind" tothink that after the light traveled 22.000years (1.32 X 10I7 miles), I only had toadmit light to my telescope for a fewseconds to see the object in some detail.

Television

For small objects of reasonable brightness,television has been a convenient viewingand recording aid. At the very highmagnifications needed to view the planetsor the moon, vibration of the telescope is aserious problem, either in the case of simpleobservation or photography. In fact, theoperation of the camera shutter is oftensufficient excitation to cause vibrationswhich may last an appreciable part of themany -second exposure required tophotograph the planets with any detail.

Objects as small as a few arc -seconds arehard to observe because of air turbulence-observers often sit for hours to get lucidglimpses when the air becomes calm for afew seconds. Television has been a usefultool for planetary observation because ofthe ability to operate the telescope com-pletely "hands-off." When used in conjunc-tion with a video tape recorder, it is

possible to accumulate an evening's obser-vation, and then, in the comfort of anarmchair, playback the tape to watch theclear views of the planet.

The surface features of the planets tend tohe of rather low contrast, and the resolu-tion roll -off of a telescope is a gradualfunction. It is possible to apply simplevideo processing, such as contrast enhance-ment or video peaking, to improve theimages. Figs. 3 and 4 show planetaryimages photographed from a nine -inchmonitor. Both have had contrast enhancedconsiderably by the simple expedient ofoperating the contrast control higher thanappropriate for a linear display. Scan linesare oriented vertically in these images toallow simple video peaking to enhance thelargely horizontal detail in both planets.

The telescope was operated at I./ 20 (4000 -mm focal length) by the use of a 35 -mmcamera "tel-extender" adapter. The vidiconis a 2 3 -inch silicon tube mounted in ahome-made camera. The camera was madein two pieces to allow a minimum weight tohe imposed on the telescope. A furthersensitivity gain was achieved by turningcamera beam current on only on everysixth tv field, allowing light integration forthe rest of the time. The raster vertical scanwas collapsed by a factor of two to get morescan lines in a given image sin on thevidicon. The monitor scan was similarlyreduced to achieve the proper aspect ratio.

At the same scale of magnification, the

6

moon is an especially interesting object. Ihave often found willing audiences fortelevision views of the moon and brighterplanets. With the moon, it is interesting to"travel" around by allowing the earth'srotation to scan the camera view over themoon's surface. Figs. 5, 6. and 7 aremonitor photographs made from a videotape recording of a quarter moon. Thecamera was again the 2/ 3 -inch siliconvidicon, but with normal 60 -field scan.

My most ambitious recent project was totelevise the eclipsing, or occulting, of a starby the planet Mars. On April 7, 1976, Marspassed in front of a star in the constellationof Gemini. We are accustomed to think ofplanets as fixed in the field of stars on agiven night. We are aware, however, thatthey move over a period of days. An objectas small as six arc -seconds, as Mars was atthis time, does not often pass very close to astar, so that we cannot see its motionsminute -by -minute.

On this occasion, however, the star ap-proached and passed behind Mars, allwithin half an hour. I planned to video tapethe entire sequence, and then make a time-lapse motion picture sequence that wouldcompress the event from about one halfhour to about one half minute. Mars smallangular diameter of six arc -seconds dic-tated a focal length of about 10,000 mmand an f number of 50. For such a large/number and such a dim object, a super-sensitive camera was necessary. I had in mypossession an RCA TCI030 silicon in-tensifier target camera which I was using ona research project. Rather than let it sit idleon the evening of April 7, I installed it onmy telescope, with the impressive resultsshown in Fig. 8. In this figure I havesuperimposed six negatives, taken from themonitor during video tape playback. Thesomewhat erratic shape of both objects isdue to air turbulence. Mars is not a roundobject in this view, being in a "gibbous"phase, not fully illuminated in earth view.The ability to see this event is impressive ifone considers that the objects are theequivalent of a basketball and a baseballviewed from about ten miles away. (Ofcourse, we are accustomed to viewing theprogress of baseballs and basketballs atmuch greater distances with the aid oftelevision.) The gain over photographicexposures is about 60:1. since one secondwould be a reasonable photographic ex-posure, as compared to 60 per second tvimages.

Fig. 5Short mountain range on the moonphotographed with equipment and con-ditions described in Fig. 6. Picture width is 5arc -minutes.

14.

'Nib

'1'

Fig. 7Moon's straight wall, which is about 60 mileslong and 800 feet high. It is probably due to a

the floor of old crater.Photographed with same equipment andconditions as in Fig. 6. Picture width 5 arc -minutes.

Summary

Astronomy is a character -building hobby.It teaches patience and equanimity in theface of disappointment. Some uniqueevents occur once in a lifetime. Other viewsare available only when the earth, sun, andmoon cooperate. All of them are availableonly when the weather and family plansallow. I have found that I have been able touse electronic imaging aids to enhance thelimited time and non -optimum locationavailable to me in pursuit of this rewardinghobby.

Bibliography

Astronomy magazine, published by AstroMedia Corp., 411 E. Mason St.,Milwuakee, Wisc., has been running aseries, "Photography in Astronomy," sinceApril 1976. The following, applying to thisarticle, is one in the series:

1>a%1,. J.: I ohm. W.. and Eaton. J.. "Rcd light sk)photo ,,,,,,,,,,, Vol. 4. No. K ENug 1976E

Fig. 6Moon photograph made from video tapeshows three large craters near the center ofthe moon's face. The largest of the threeattached craters, Ptolemaeus, is 115 milesacross. The picture was made with TRI-Xfilm in 1/30 second. Equipment used includea 2/3 -inch silicon vidicon camera and aCelestron telescope with a 4000 -mm focallength. at 1120. Picture width is about 5 arc -minutes.

Fig. 8Time lapse sequence shows occultation of astar by Mars. Six monitor photos, taken froma video tape playback, are superimposed toillustrate eclipse sequence. Exposures areabout two minutes apart. Equipment usedincluded a silicon intensifier target cameraand a Celestron telescope with a focallength of 10,000 mm and an number of 50.Photos were taken on TR I -X film at exposuretimes of 1/60 second.

7

O.F. WhiteheadD.E. Hutchison

Technological events that will shape theworld tomorrow are being announced inthe literature today. Because of the tremen-dous amount of work that gets reported,engineers and scientists are probablyfinding it difficult to keep up with theliterature. In RCA's libraries alone, thereare over 1300 different periodicals, and thisis just the tip of the iceberg. These samelibraries have thousands of books andreports as well as access to the documenta-tion of thousands of engineers andscientists in the U.S. and abroad.

Information is the basic commodity of theRCA Libraries. Such a collection,however, is of little value unless it is

organized to enable you to find informa-tion when you want it quickly. Read on,while we tell you how you can find theinformation you need-when you need it.

Answers to questions

The key to the contents of your library is, ofcourse, the card catalog, but if you can'tfind what you want, or you are not familiarwith the card catalog, don't hesitate to askyour librarian for help. Librarians are

Technical information:where to get it

The information explosion has produced such avast number of sources that RCA's libraries havehad to add to their traditional services to includeon-line computer searches and resource sharing.

accustomed to answering questions. Hereare a few of the more unusual ones we haveanswered at the Camden library during thepast year:

What is the length of the boundarybetween East and West Germany?

When were the first radio setsmarketed?

How do you spell the word from apopular movie-super something orother?

How do you apply for a Rhodesscholarship?

What is the name of the planegeometric figure that looks like a four-leaf clover?

What was the price of gold on FebruaryI, 1975?

Can you identify the source of "Sellingis the art of overcoming objections"?

Where can I get an illustration of aspacecraft passing through the VanAllen radiation belt?

Yes, your librarian can probably come upwith answers to all kinds of questions youmight need answered, but basically thepurpose of a company library is to providethe right information, in the right place, atthe right time. This requires a knowledge of

Reprint RE -22-4-19Final manuscript received October 1.1976

RCA's programs and products, carefulselection of resource materials, and speedywell -organized services to place thesedocuments in the hands of readers.

Information services

The activities of a library are so varied inthe fields of science and technology that thelibrarian must be capable of carrying onand supervising diversified operations.Where the size of the staff is minimal,which is the case in many of RCA'slibraries, the librarian executes manyoperations. Some of the directly user -oriented ones are:

Resources: Selection and purchase ofbooks, periodicals, conferenceproceedings, government reports, andother publications.

Services: Supervising readers' services,reference requests, and interlibrary loan.Executing literature searches. Organiz-ing non -book files, e.g., companyreports, contract reports, NASAliterature, federal and state documents,and engineering notebooks.

Publications: Compiling Tables of Con-tents of current periodicals and LibraryBulletins of recent acquisitions, in-cluding book reviews.

These main responsibilities are required inalmost every library. Where staff assistanceis available, providing these services is not aproblem and additional services may berendered.

8

New services

While the services listed above have beenavailable for many years, the nature of thecorporate library is changing, and RCAlibraries will need to change, too. Automa-tion is making information more accessi-ble, and with the increased volume ofinformation available, literature searchingwill, of necessity, need to be done withcomputers.

The library profession has recognized theimportance of library cooperation andresource sharing, and RCA's libraries arealso seeing this need. With the increasedcost of journals and conferenceproceedings, it is no longer possible forevery library to subscribe to every informa-tion source of possible interest to theirreaders. But, through cooperation withother RCA libraries, expensive ac-quisitions can be shared. In spite ofbudgetary restraints, our libraries will re-main committed to providing the sameservice, but by doing it with more borrow-ing and less purchasing. It was this needthat prompted the compilation ofPeriodicals in RCA Libraries (describedlater), and RCA Technical InformationSystems is planning to publish similarcompilations as needed.

The need for literature review beforebeginning research is vital and wellrecognized -within RCA, but also evidentwithin the company is the notion thatengineers and scientists who spend verymuch time in the library are probablywasting time or sleeping! This probablysprings from the work ethic of our Puritanheritage and the fact that reading is usuallydone for pleasure. Management apathy hassometimes placed a professional librarianin the role of "keeper of the books" ratherthan "information specialist." Some re-evaluation of the library's place in thecorporate structure is beginning to takeplace in industry, and corporate librariesare being called upon to meet the demandsmade by their users. High-technology com-panies must keep abreast of scientific andengineering developments but, oftentimes,the corporate library is relegated to aninsignificant place in the corporate struc-ture and its potential services areoverlooked. You can assist your librarianand increase the value of library services byexpressing your needs to your librarian andto your management.

If It's not in your I. brary, this compilation willtell you which of the other RCA libraries mayhave that periodical.

On-line search systems

The escalation of publications in the fieldsof physics, chemistry, and electronics hasmade machine-readable data bases anecessary ingredient in handling thisvolume of contemporary literatureeffectively. Approximately 200 such databases are presently available, a number ofwhich would adequately meet the needs ofRCA's engineers and scientists. Access tothese programs at RCA has just begun withour entry into two such search systems;librarians are training in their use now andthe systems should be on-line shortly.

The two systems presently available areSystem Development Corporation'sORIHT ///and Lockheed Missiles & SpaceInformation Systems' D/ALOG. Throughthese systems, you can search filesrepresenting millions of articles, reports,books, current and completed researchprojects, and other kinds of information.You may review the results of the search byhaving part or all of the citations printedon-line, or in the case of longbibliographies, you may have the resultsprinted on a high-speed printer and air-mailed to you on the same day. In mostcases the output includes an abstract as wellas a complete citation. Should you want acopy of the complete document, this wouldbe obtained through normal sources.

Some of the data bases that can be accessedthrough these systems include:

CHEMCONChemical Abstracts Condensates,prepared by Chemical Abstracts Service.

Olive Whitehead has administered the RCACamden library since 1969. Her professionalduties there include selecting new books,reports, and journals, in addition toreference work and literature searching.She is also highly active in the PhiladelphiaChapter of the Special Libraries Associa-tion.

Contact her at:Engineering LibraryGo,.ernment Communications Sys -emsCamden. N.J.Ext. PC -3438

Doris Hutchison joined CorporateEngineering in 1970 and is currentlyAdministrator of Technical InformationSystems. This group is concerned withpublishing RCA Technical Abstracts and isalso responsible for providing services tothe RCA Library Network, includingpublication of a Union Catalog ofPeriodicals in RCA Libraries.Contact her at:Technical Information SystemsCorporate EngineeringCherry Hil . N.J.Ext. PY-5412

Doris Hutchison (left) and Olive Whiteheadat a computer terminal used for on-line litera:Jre searching

9

A quick scan of Current Contents willpublications.

tell what articles may interest you in this week's

Covers biochemistry, organic chemistry,macromolecular chemistry, appliedchemistry and chemical engineering.

COMPENDEXPrepared by Engineering Index. Cor-responds to Engineering Index Monthly.Covers civil -environmental -geologicalengineering; mining -metals -petroleum -fuel engineering; mechanical -automotive -nuclear -aerospace engineer-ing; electrical -electronics -controlengineering; chemical -agricultural -foodengineering; and industrial engineering,management, mathematics, physics, andinstruments.

SSIEPrepared by the Smithsonian ScienceInformation Exchange. Covers on -goingand recently completed research in thelife, physical and social sciences-bothbasic and applied research projects.

INSPECPrepared by Institution of ElectricalEngineers. INS PEC data bases include:Physics Abstracts, Electrical and Elec-tronic Abstracts, and Computers andControl Abstracts.

SC/SEARCHProduced by the Institute for ScientificInformation. Over 850,000 citationsfrom 2,500 periodicals in the physicaland life sciences.

No library can subscribe to all the journals.

One of the most valuable resources in theRCA libraries are the periodicals receivedby paid subscription or by controlledcirculation.

At some locations, Current Contents, aweekly compilation of articles in variousjournals, provides an easy scanning

medium for recent issues. The Institute forScientific Information compiles the weeklyissues of Current Contents from advancecopies of title pages of periodicals suppliedby the publishers, or from the title pages ofjournals obtained by subscription. CurrentContents appears in several editions ofrelated subjects. The two editions closelyallied with RCA's interests areEngineering, Technology and AppliedSciences and Physical and ChemicalSciences, each covering approximately 700journals.

Several RCA libraries duplicate tables ofcontents from current issues of magazinesas they are received and compile a "Tablesof Contents" which is circulated to in-terested persons.

It is impossible for the library at each RCAlocation to subscribe to all the journals thatmight carry articles of interest. To assist inlocating periodicals quickly, RCATechnical Information Systems has

prepared Periodicals in RCA Libraries, aunion catalog of subscriptions andholdings of journals and selected serials ineach of the RCA libraries. If there is nosource within RCA for a journal, there areseveral tools that provide listings ofperiodical holdings in other organizations.(Consult your local library for titles.)

Interlibrary loan"Neither a borrower, nor a lender be."

Despite Shakespeare's admonition, nearlyevery RCA library must participate ininterlibrary loans to supplement itsresources. Cooperation between RCAlocations is excellent, borrowing andlending books, conference proceedings,reports, and copies of periodical articles asoften as possible. According to cir-cumstances, university libraries, statelibraries, area reference libraries, andspecial libraries may be contacted forliterature to supplement RCA's resources,and they may in turn call upon any RCAlibrary for assistance. There is a great dealof variation in regulations within librariescontrolling interlibrary loans, which willgovern how and where the interlibrary loanrequest is submitted. In placing a requestfor a loan, it is necessary to have accurateand complete references; otherwise therequest may not be honored.

Indexing sources provide a quick route to hundreds of thousands of articles. Specifics areon next page.

10

Book selection

The right book, in the right place, at the righttime.

[he librarian selects the "right book" byscanning publishers' announcements, bookreviews in periodicals, and the WeeklyRecord, which lists recently -publishedAmerican books. But most important arethe recommendations from RCA per-sonnel on which titles should be added.While building their resources to fill pre-sent and anticipated future needs, all RCAlibraries acquire numerous conferenceproceedings, reports from various govern-ment agencies and scientific organizations,and the periodicals most closely allied totheir interest. Announcements of new ac-quisitions are made through LibraryBulletins, displays of book jackets, and"new book" shelves.

Indexing sources

Have you ever tried to locate an article thatyou read 2 or 3 years ago in an IEEEpublication?

Or was it in Bell Systems TechnicalJournal? But now you need that articlequickly and don't have time to go back andlook at journals until you find the rightone. Here are some of the indexing sourcesthat you might consult:

IEEE Author and Subject Indexes toPublications.Recent annual issues cover all IEEEtechnical periodicals, technical itemsfrom IEEE magazines, conferencepapers, and IEEE standards.

Engineering Index Monthly and AnnualAbout 85,000 items, selected from 2000journals, are abstracted annually andclassified under broad subjects with ex-tensive cross-references.

Applied Science and Technology IndexA subject index to about 225 English -language periodicals in the fields ofaeronautics and space science, automa-tion, chemistry, earth sciences, electrici-ty and electronics, engineering.materials, mathematics, physics,telecommunications, and related sub-jects.

Mathematical Reviews.This monthly journal, which abstractsthe world literature in mathematicalresearch, is prepared by the AmericanMathematical Society.

44/4/0

116 -/to-s,/4/04:i_

Where to find what's newin government research reports.Biweekly publication has abstracts of allunclassified gcvernment reports hav ng unlimited distribution.

Physics Abstracts.An abstracting vehicle which covers thewhole field of physics, drawing fromworks in all countries and in alllanguages. Publishes 85,000 items peryear.

Electrical and Electronics Abstracts.A monthly abstracting publication,prepared by the Institution of ElectricalEngineers (British). Publishes 40,000items per year.

Computer and Control Abstracts.Includes literature from worldwidesources on computer science and controlengineering.

Science Citation IndexThis index, published by the Institute forScientific Information (ISI), lists alldocuments which are being referred to incurrent literature over a specific periodof time. It is arranged alphabetically bythe names of the first authors of thedocuments being cited. Below each ofthe cited items is a list of the newerarticles that cite them, also arrangedalphabetically by the names of theirauthors.Companion to the Citation Index is theSource Index, which is an alphabeticallyarranged author index of more than420,000 articles and other editorialmaterial indexed each year by ISI.

Chemical AbstractsUnder the auspices of the AmericanChemical Society, this publicationprovides excellent abstracts in English ofthe world's literature on chemistry andchemical engineering. In 1975, CAcontained over 300,000 abstracts from1000 primary journals and numeroussecondary ones.

External resourcesGovernment agencies, professionalassociations, and research organizationsare important sources.

Several important resources and servicesoutside the Company can be approachedthrough RCA libraries. Among those fre-quently called on by RCA librarians arevarious government agencies, professionalassociations, and research organizations.

The National Technical InformationService (NTIS), Springfield, Virginia, hasover 900,000 reports for sale, representingGovernment -sponsored research, develop-ment, and engineering reports and otheranalyses by Federal agencies, their contrac-tors, or grantees. Its customers are suppliedwith about 4 million documents andmicroforms annually. Many RCAlocations receive the NTIS biweeklypublication, Government Reports An-nouncements and Index, which abstractsnew research reports that are unclassifiedand have unlimited distribution. For rapidsearching, this publication carries indexesin the following categories: subject, cor-porate author, personal author, title, con-tract number, and accession reportnumber. To secure prompt service fromNTIS, most RCA locations maintain adeposit account there.

The services of the Defense Documenta-tion Center (DDC), Alexandria, Virginia,are available to RCA as contractor tovarious military agencies. However, accessto this data bank requires registration of anactive contract on the pertinent DoD form(DD 1540), commonly called "Field ofInterest Register." The technical reportscurrently in the DDC collection total morethan a million titles, of which 670,000 areunder computer control for quick retrieval.

II

MIT research is available through the Institute's Monthly List of Publications.

These include more than 65,000 titles in thefield of electronics and electrical engineer-ing and 90,000 in the subject of physics.Current classified reports and unclassifiedreports with distribution limitations areannounced in the Confidential DDCTechnical Abstract Bulletin.

At most RCA locations, the Library is thedesignated point -of -contact between theDefense Documentation Center and Com-pany personnel. Bibliographies supplied byDDC take three forms: ReportBibliography, prepared in response to aspecific request for references to technicalreports having a particular kind ofrelationship in subject matter; ScheduledBibliography, prepared for subject areasfor which requests are anticipated, andRapid Response Bibliographies, preparedfor user organizations having access toTELEX.

Registration with DDC also providesaccess to the services offered by theDefense -sponsored Information AnalysisCenters and the major DoD technicallibraries. There are eight DoD/ DSA Infor-mation Analysis Centers that collect, store,review, evaluate, synthesize, and dis-seminate authoritative scientific andtechnical information in a format useful toscientists, engineers, and technicians. Intheir literature surveys, these centers reviewtechnical reports from DoD, other Govern-ment agencies, industry, and academicinstitutions; open literature, includingforeign sources; and unpublished papersand similar sources.

Another source of Federal publications isthe Superintendent of Documents,Government Printing Office, Washington,

D.C., or the regional GPO bookstoreslocated in about 20 cities. Announcementsof a portion of the output of the Govern-ment Printing Office is made in theMonthly Catalog, United Stares Govern-ment Publications. These publications mayrange from Abacus to Zirconium in sub-ject, from elementary to postgraduate inscope, from pamphlets to multivolume setsin format. All purchases from theGovernment Printing Office require pay-ment in advance through Deposit Ac-counts, usually maintained by the RCAlibraries.

When the National Aeronautics and SpaceAdministration was established in 1958, itscharter stipulated that one of its functionsshould be "to provide for the widestpracticable and appropriate disseminationof information concerning NASA'sactivities and their results." Out of thisrequirement came an extensive publicationprogram of technical research reports,technical translations, contractor reports,bibliographies, and nontechnicaldescriptions of all aspects of space flightsfrom Mercury through Skylab. SeveralRCA libraries receive NASA publicationsregularly, and unclassified documents canbe purchased from the National TechnicalInformation Service.

To provide bibliographical control for thisextensive literature, NASA established anindexing and abstracting service to publishScientific and Technical AerospaceReports, frequently called "STAR." Thisabstracting journal, issued semimonthly,gives worldwide coverage to aerospace -report literature. By arrangement betweenNASA and the American Institute ofAeronautics and Astronautics, the publica-

tion International Aerospace Abstracts(IAA) abstracts books and scientific andtrade journals, to complement STAR'sreport coverage.

One of the most valuable non-governmental organizations whoseresources are available through RCAlibraries is the Industrial Liaison Office ofthe Massachusetts Institute of Technology.Through this office, reports coveringresearch at MIT may be obtained. Fre-quent symposia on important currentstudies are open to RCA personnel, andconsulting services with scientists on thestaff at MIT can be arranged. The In-dustrial Liaison Office issues a MonthlyList of Publications, which announces andabstracts current reports on research andeducational efforts in progress at MIT.

RCA's libraries

Meet your librarian-see what your librarycan do for you.

A recent survey of RCA's larger librariesrevealed that many are plagued by similarproblems-mainly, not encugh staff ormoney to serve the engineers at theirlocation to the extent that they would like.Depending on the requirements of theRCA division, the library staff size, andbudgetary constraints, library servicevaries from location to location. Somelibraries issue a Library Bulletin listing newacquisitions, while others display newitems in the library, requiring readers to goto the library to see what is new. Somelibraries circulate periodicals to those in-terested, while others require users toperuse them in the library, or will copypapers of interest. New acquisitions arebased on recommendations of a LibraryCommittee at some locations, and at othersthey are by request of the users, or at thediscretion of the librarian, or a combina-tion of all three.

If you have not been a library user in thepast, start now by visiting your local libraryand meeting your librarian. She will behappy to show you the library, tell youwhat services are available to you, and helpwith your information needs. If yourlibrary circulates a list of new acquisitions,get on the distribution list. If not, drop inoften enough to be an early borrower ofnew titles. Since RCA libraries do vary inscope, you will have to get acquainted withyours to find out just what services areavailable. Table I should give you a start inthis direction.

I2

Table IThe RCA library system has direct access to over 80,003 books and more than 300 periodicals.

Location Librarian Principal subjectsBooks & bound

periodicalsPeriodicals

receivedMicrofilm

readerMicoftche

reader

AS,Burlington

Veronica Hsu, Adm.,Library Resources

Electrical & ElectronicEngineering

5.200 150 No Yes

Computer TechnologyMathematicsPhysics

ManagementBusiness

GCS,Camden

Olive Whitehead, LibrarianVirginia Mattice,

Asst. Librarian

ElectronicsPhysics & OpticsTelecommunications

13,500 250 Yes Yes

Computer Technology

AE,Hightstown

Mary Pfann, Librarian Electronics andCommunications

3,660 143 No Yes

Computer ScienceMaterials EngineeringSpace ScienceSystems EngineeringMechanical EngineeringOpticsMathematicsSemiconductor Physics

CE,Indianapolis

Estella B. Perkins,Librarian

ElectronicsTV & Radio Engineering

2,370 143 Yes No

PhysicsMathematicsElectrical EngineeringMechanical Engineering

SSD/ PTD,Lancaster

Mary Kathryn Noll,Adm., Library Services

ElectronicsTelevision

6,650 140 Yes Yes

Physics & OpticsChemistryMaterials

M SR,Moorestown

Natalie J. Mamchur,Librarian

Radar & AssociatedElectronics

13,500 250 Yes Yes

MathematicsMilitary StrategyBusiness

Labs,Princeton

Wendy Chu. librarianHalina Kan.

Radio & TelevisionPhysics

30,000 275 Yes Yes

Associate Librarian ChemistryMathematicsElectronicsSolid State PhysicsComputer Science

RCA Ltd.,Ste. Annede Bellevue

Gladys Donaldson, Adm.,Company Library

Solid State PhysicsPlasma PhysicsCommunications

3.500 250 Yes Yes

Aeronautics & SpaceLaser PhysicsBusiness

SSD,Somerville

Esther Jankovics,Librarian

ElectronicsPhysics

5,740 150 Yes Yes

ChemistryMathematics

13

Ele

ctro

-opt

ics

at R

CA

SO

UR

CE

DE

TE

CT

OR

ST

RA

NS

MIS

SIO

N &

CO

NT

RO

L

Non

-im

agin

gIm

agin

gIm

agin

gN

on -

imag

ing

Geo

met

rical

Phy

sica

l

Lase

rS

igna

l1

Pow

erO

ptic

sO

ptic

s

Tec

hnol

ogy

cont

ribut

or*

1I

1M

odul

atio

nO

ther

Tec

hnol

ogy

user

4S

olid

LED

Sta

teG

asO

ptic

ally

Pum

ped

Cat

hode

Imag

eR

ay T

ube

Inte

nsifi

erC

amer

aS

olid

Tub

esS

tate

Sol

idP

hoto

Sta

teE

mitt

erP

hoto

volta

icC

ell

Opt

ical

Sys

tem

sH

olog

raph

yD

efle

ctio

nE

OE

ffect

s

AS

TR

CT

ELE

CT

RO

NIC

S-H

IGH

TS

TO

WN

. NJ

44

44

4.

..,,,.

.,, L

,E...

E.,

spa

Lecr

att p

ox.,

syS

tem

S1

0 m

ulte

spec

trai

Bite

r S

pace

born

e tv

sur

vert

ianr

e

AU

TO

MA

TE

D S

YS

TE

MS

-BU

RLI

NG

TO

N. M

A#

4*

44

*4

44

44.

44.

It4

4La

ser

Ran

gefin

ders

Lase

r ill

umin

ator

s R

econ

aiss

anti

syst

ems

Low

lignt

leye

l Iv

Opt

ical

cou

nter

mea

sure

,E

-0 a

utom

atic

test

scs

tern

st to

logr

aolv

e...-

1,E

E.0

poi

ntin

g an

d tr

acki

ng i

mag

e sc

anne

r,

GO

VE

RN

ME

NT

CO

MM

UN

ICA

TIO

NS

SY

ST

EM

S-C

AM

DE

N. N

J4

4 It

44

*4

4*

,ser

,,ag

e re

surd

my

L -

0 de

tect

., L

ignt

muc

lu,e

1,

,ste

ms

Rec

onna

issa

nce

imag

e da

ta s

yste

ms

CD

sca

nner

s

MIS

SIL

E I

SU

RF

AC

E R

AO

AR

-MO

OR

ES

TO

WS

L N

J4

44

.... i

,,,,,I

mg

and

trac

king

im

age

trac

king

Sys

tem

s

RC

A L

IMIT

ED

-ST

E A

NN

E D

E B

ELL

EV

UE

. OU

E#

* ,i

id s

tate

det

ecto

rs C

O-

lase

r,s

e' ,.

,mm

un.r

at.o

ns S

ySte

mS

BR

OA

DC

AS

T S

YS

TE

MS

-CA

MD

EN

. NJ

#4

4-4

44

4.,1

01 ly

Cam

eras

Cam

era

Oa:

..cD

imag

e p.

.11,

,,,,

.riser

0,S

C s

igna

l ter

neie

r

AD

VA

NC

ED

TE

CH

NO

LOG

Y L

AB

OR

AT

OR

IES

-C

AM

DE

N. N

J#

#4

44

44

44

*4

*iO

tiCai

',ra

ter',

cle

a.gn

Op,

cal E

np.,,

,.,,e

ni,cr

L a

ser

rada

rO

ptic

al in

spec

tion

SO

UO

ST

AT

E D

IV..

ELE

CT

RO

-OP

TIC

S I

OE

VIC

.ES

-LA

NC

AS

TE

R. P

A*

**

**

**

CC

D Im

age'

s C

amer

a IJ

oes

TV

cam

eras

CR

Ts

Sto

rage

tube

s Im

age

tube

s P

noto

muI

tiptie

rsP

noto

rnut

toph

ers

Sem

icon

duct

or la

irs L

ED

s

CO

NS

UM

ER

ELE

CT

RO

NIC

S-I

ND

IAN

AP

OLI

S. I

N#

TV

reG

e.ve

,s

PIC

TU

RE

TU

NE

DIV

.-LA

NC

AS

TE

R. P

A*

RC

A L

AB

OR

AT

OR

IES

-PR

INC

ET

ON

. NJ

**

**

**

**

0ax

arcn

in a

ll m

ajor

E -

O te

0,01

0g a

reas

Where the electro-optics action is at RCAH.E. Haynes Starting with "talkies" in the twenties, RCA's involvement with electro-optics

broadened to the point where it covers almost the entire corporation.

"Electro-optics" is a term that can mean rather differentthings to different people; however, for the purposes ofthis issue of the Engineer, its scope is limited somewhatarbitrarily to the categories of light sources, light detec-tors, and devices and media for transmitting and con-trolling light. (IR and UV are included under the heading of"light.") Such areas as systems applications, electronoptics, image processing, and the psychophysics of visionare, for practical reasons and with some regret, notincluded.

RCA came into existence nearly 60 years ago as acommunications company, and it continues to be a worldleader in this field. Having such a genesis, it might besurmised that its first venture into electro-optics wouldhave been brought about by the advent of television, butinterestingly this is not the case. What was probably RCA'sfirst significant plunge into electro-optics was its role inthe development of optical sound film recording, thetechnique that changed the movies into "talkies" almostovernight in the late twenties. An engineering group calledPhotophone Advanced Development, headed by the lateE.W. Kellogg, was a pioneer in this technology. (Inciden-tally, this group has evolved, through many stages, intothe Advanced Technology Laboratories in Camden.) Thegroup represented, in a sense, a microcosm of RCA'spresent electro-optics activities, since these engineersbrought forth innovations in all of the electro-opticscategories mentioned above. Light sources, lightmodulators, beam deflectors, unique optical -systemdesigns, photodetectors, photographic technology-allwere essential to produce a successful motion -picturesound -recording and reproduction system.

Televison, which obviously relies very heavily uponelectro-optics technology, launched a revolution of muchgreater magnitude in the forties; here RCA's role is farbetter known. And then in the fifties, television ex-perienced the full impact of color, calling for electro-optics technology of even greater complexity andsophistication. As Ralph Simon points out in the insidecover message for this issue, developments in televisioncontinue unabated.

In the more recent past, an explosive broadening hastaken place in the E -O field. Completely new phenomenaand devices, many of them resulting either directly orindirectly from military and space programs, have broughton this new expansion. Lasers, semiconductor lightsources and detectors, charge -coupled imaging arrays,new types of camera tubes, fiber optics, holography, liquidcrystals, optically active materials-the list is long and itcontinues to grow.

The lion's share of E -O technology applications areconcerned with some aspect of information handlir g-sensing, displaying, transmitting, recording, processing-

iilimmai

Harold E. Haynes has been Staff Technical Advisor in Govern-ment Engineering since 1972, with his major work emphasis on E-0 and related activities throughout the Government SystemsDivision. He has been with RCA since 1940, and has beenengaged it numerous advanced development programs in elec-tronics, recording systems, electro-optics, color reproductionsystems and the electro-optical aspects of a precision CRT -basedcomputerized photocomposition system.Contact him at Government Engineering, Government SystemsDiv., Moorestown, N.J., Ext. 3843

and so it is hardly surprising to see that electro-opticsapplications have found their way into a great many partsof an information -oriented company such as ours. Aglance at the accompanying diagram tells just howwidely the technology is being applied at RCA. The familytree at the top shows the wide range of electro-optics technology and devices that are available to drawon. Each of the organizational units listed is in some wayinvolved with some facet of electro-optics. Upward arrowsindicate those parts of RCA that are contributing tospecific areas of E -O technology, while downward arrowsindicate those RCA organization that are using this E -Otechnology to their benefit. Complicated as it may appear,the chart in still only an incomplete snapshot at one pointin time, and cannot be completely comprehensive. Theoverall picture it imparts-one of widespread and variedactivities and their interrelations-is the real message.

This issue contains articles relating to several-but farfrom all-of the topics shown in the diagram. It is dividedinto three sections on each of the E -O technology areas,each section beginning with a brief introductory sum-mary. Whether your attention is drawn to the "big picture"or to some of its details, or perhaps both, you should enjoythis guided tour through some of the complex andchanging world of electro-optics at RCA.

Reprint RE -22-4-11 Final manuscript received February 24. 197615

Light SourcesAn Electro-optic light source can be as basic assunlight, but many systems require the laser'scoherent light, narrow spectral linewidth, or highsource radiance.

Although there are a number of types of lasers, theoperating mechanism is essentially the same for all.First, electrons are excited to higher energy levels byan external means, such as an electric field or flashof lights. Since these higher levels are unstable, theelectrons spontaneously drop back to lower energylevels. In doing so, they emit photons of a

wavelength proportional to their energy drop. Thisordinary fluorescence provides incoherent light.

But, if these electrons can stimulate other electronswith light at a wavelength corresponding to one oftheir preferred energy drops, we have coherentradiation. This Light Amplification by StimulatedEmission of Radiation takes place in a resonantcavity called the Fabry-Perot resonator. There, theinitially radiated lightwaves reflect back and forth,stimulating more and more radiation. A small frac-tion of this light beam can be used as the laser'soutput by making one end of the resonant cavitysemireflective.

Laser types

Laser action can take place in many different media(it has even been reported in Jello), but solid, gas,and semiconductor lasers are among the mostcommon.

Solid lasers are more precisely "lasers based on ions insolid materials."

Solids, such as the ruby, were the first materials toshow laser activity. In a ruby laser a flashlampexcites the chromium impurity ions in the ruby'saluminum -oxide host. Light is produced as thechromium ions return to their base -level energy.The ruby crystal acts as the resonating cavity; itsends are coated with reflective surfaces, one ofwhich allows a small fraction of the light to passthrough as the output signal.

Gas lasers are low-cost,but require a high -voltage dc supply.

There are literally thousands of types of gas lasers,with the gas parameters determining the wavelengthof the light produced. Not all are practical, though,and He-Cd, He-Ne and CO2 are the dominant types.In gas lasers, the electrons are accelerated by a highvoltage, usually on the order of a few kilovolts, andtransfer energy to the gas molecules by collision.Mirrors at the ends of the gas tube form theresonating cavity for light amplification.

Hansch, T.W , Pernier. M.; and Schawlow, Ai.; "Laser action of dyes ingelatin," J. Quant. Elec.. Vol. OE -7 No. 1 (Jan 1971)

Size contrast: He-Cd gas laser and power supply (above); and GaAssolid-state injection laser, shown both unmounted (through the eye of aneedle) and in final -product form.

EXTERNALEXCITATIO'g

SPONTANEOUSDECAY

SPONTANEOUSANDSTIMULATEDEMISSION

E2-Eczhu

Simplified energy -level diagram for a solid (ruby) laser. A flashlampexcites the ruby s impurity chromium ions to the broad energy band at 3,from which they decay to their preferred level at 2, giving off heat as theydo. The ions then drop down to their base level at 1, emitting light at afrequency proportional to the energy change from 2 to 1. If the ionpopulation is aistributed among the levels properly, the spontantouslyemitted photons will then stimulate additional radiation.

16

LASER TUBETOTALLY

REFLECTINGMIRROR >BALLAST

0- 3000 VOLT DCPOWERSUPPLY

-o

NI % TRANSMITTING

MIRROR

Gas laser system. High -voltage dc supply excites the gas ions in the tube; light isproduced as the ions drop to lower energy levels. Hemispherical mirrors provide theresonant cavity for the stimulation of additional lightwaves.

Semiconductor lasers operate at low supply voltages and are highly efficient.

In certain semiconductors, such as GaAs, light emission takes place at thep -n junction when an external potential drives both electrons and holes intothe junction. The holes and electrons recombine very rapidly by transitionsfrom the conduction level to impurities near the valence band; eachtransition produces light at a frequency proportional to the energy drop.

So far, we have an LED emitting incoherent light. But. by optically polishingthe material in the direction perpendicular to the plane of the p -n junction,we create a resonant cavity. Light traveling back and forth between thepolished sides stimulates other transitions until all the light is oriented in thesame direction. This coherent light beam then can be coupled out oymaking one of the polished sides an imperfect reflector.

ELECTRONINJECTION

STIMULATION

IMPURITYLEVELS

VALENCEBAND

VEMISSION

P -TYPE N -TYPE

CONDUCTIONBAND

Solid-state laser energy -level diagram. In the forward -biased p -n junction, electronsinjected into the p -type region recombine with holes. Each recombination produceslightwaves, which then reflect back and forth between the optically polished sides ofthe device, stimulating more emissions as they do.

Laser fusion-maybeOne area of laser research outsideof RCA's communication-nformation activities, but one that

could greatly affect us all, is laserfusion. Such a system mayproduce "micro -explosions" ofclean thermonuclear power whenlaser energy is focused on a targetpellet This laser -produced ther-monuclear burn, which wasdemonstrated last summer, wasthe second of seven "milestones"that must be passed on the road toa commercial laser -fusion powerreactor. The breakeven point,where thermonuclear outputequals laser -energy input, is

scheduled for 1982. Despite thetremendous progress madetoward laser fusion in a relativelyshort time, its future remainscloudy-ERDA has said that itsresearch funds are highly con-tingent upon having the breakevenpoint in sight by 1982.

Projected specifications for a fu-sion power plant of the late 1990sinclude a peak power of up to 500TW (500 X 1012 W) and energy of500-1000 kJ. For comparison, themost powerful system presentlyavailable, the LivermoreLaboratories' Argus laser, iscapable of focusing over 2 TW.Their new 20 -beam Shiva system,scheduled for completior late thisyear, should be capable of deliver-ing 10 kJ in subnarosecondpulses. However, automaticpointing and focusing systemsmay help multiple -beam systemsbecome capable of the erergy andpower levels needed.

17

an introduction

Light -generating devicesI. Gorog

Electro-optics activities at RCA emphasizeinformation -handling applications, withoptical recording, reading, and com-munications all being actively pursued.Even though optical systems in all three ofthese areas had been successfully operatedbefore the discovery of lasers, cost andperformance considerations suggest that inthe future virtually all electro-opticalinformation -handling systems will use

laser sources.

The key laser characteristic used in mostapplications is their very high sourceradiance. (Another important lasercharacteristic is the narrow spectrallinewidth of the emitted radiation, which isused in many interferometric applications.But, although laser interferometric systemsare important measurement tools, theyrepresent a very small fraction of the totalpotential laser systems market.) A state-of-the-art 1-mW gas laser has an output -beamcross-sectional area of about 102 cm2, abeam divergence of IC radians, and thus acorresponding source radiance of about106 W/cm2/ steradian. The source radianceis important primarily because itdetermines the ultimate achievable spotirradiance (incident power per unit area).

For tv-rate information -recording, for ex-ample, the sensitivity of availablerecording materials implies recording spot -

irradiance requirements in the range of IW/cm2 to 106 W/cm2. The high powerdensity is required with the grainlessrecording media used for very -high -densityrecording applications; the low -powerlimit corresponds to low -resolution silverhalide films. The high -power end of thisrange of requirements cannot be achievedeven with the most powerful incoherentlamps, but low -power 1 -W output)lasers can easily be focused to provide theneeded power density. Even though thepower required to expose low -resolution

Reprint RE -22-4-8Final manuscript received May 10, 1976.

Optical information -handling systems need the laser's highsource radiance. Both gas and injection lasers are currentlybeing considered for these applications.

silver halide can be achieved with anincoherent source, overall system con-siderations make lasers the preferred alter-native even in this case.

The power required for informationreadouts is determined by signal-to-noiseconsiderations and is relatively insensitiveto the density of the recorded information.The spot power required for reading atypical fm -coded tv signal recording isapproximately 10-4W. Currently operatingrecording systems employ spot sizes in therange of 10-xcm2. In order to deliver l0 Winto a 10 -"-cm' spot, a typical state-of-the-art optical system requires a source with aradiance of approximately 104

/ steradian. A low -power laser caneasily supply this radiance. For com-parison, state-of-the-art light -emittingdiodes (LEDs) operate in the range of 10'-101 W / cm2, steradian, and tungsten bulbsprovide about 10 W/cm2/steradian.

The light source requirements of opticalfiber communications systems aredetermined by the required signal-to-nosieratio, the bandwidth, and transmissionmedium chosen. Typical optical fibers un-der development have a cross-sectionalarea of about 10' cm-, a light -acceptancecone angle of 10 l radian, and transmissionlosses in the vicinity of 10 dB/ km. Forshort -distance (< I km), narrowband(< 10 M Hz) communication applications,standard LEDs are suitable. Special-purpose LEDs have been developed forhigher -capacity intermediate -rangeapplications, but lasers must be used forlong -haul, wide -band applications.

RCA's laser development and productionactivities include work on He-Cd, He-Ne,and CO2 gas lasers, and on AlGaAs/ GaAsinjection lasers and LEDs. The applicationareas for the He-Cd product line are ininformation recording; low-cost He-Nelasers are the primary sources in high -density information -retrieval systems, andthe CO2 lasers developed by RCA Ltd. in

Canada are being considered fo line -of -sight terrestrial and outer -space com-munications. The GaAs lasers and LEDsare being developed primarily for fiber-optic communications; however, they arealso expected to find extensive applicationsin information -retrieval systems.

Istvan Gorog is Head of the Optical Elec-tronics Research Group at RCALaboratories. His main areas of interest havebeen quantum electronics and electro-optical systems. Dr. Gorog's researchactivities have included lasers and lasersystems, holography, pre -recorded -videorecording and playback techniques, dis-plays, and investigation of the psycho-phsyical aspects of electronic imaging. Hiscurrent activities include product andprocess development related to the RCAVideoDisc, manufacturing instrumentation,and electrochromics.

Contact him atOptical Electronics Research GroupRCA LaboratoriesPrinceton, N.J.Ext. 3202

18

Inexpensive He-Ne laser tube constructionK.G. HernqvistR.W. Longsderff

ova

..sommos.".

This redesigned laser tube has a potential price well belowthe current $100 level. Low-cost materials and automatedassembly should make this cost breakthrough possible.

The emerging large -volume laser marketswill affect laser production techniquesand manufacturing costs greatly.Currently, there is a great potential forusing lasers in information -handlingsystems, such as scanners for point -of -sale checkout systems, readers andmemories, character and pattern recogni-tion systems, and copiers. The He-Nelaser, today's most widely used gas laser,sells for approximately $100 per unit.This price must be lowered significantly ifthe lasers are to be produced and sold athigh -volume levels.

As a source of lower -power, high -qualityvisible laser radiation for theseapplications, the He-Ne laser is unsur-passed because of its simple constructionand method of operation. Using gas lasersalso simplifies the end product, since theyrequire no add-on optics or heat sinks.

Making cost predictions for the He-Nelaser is helped by the fact that it is

basically similar to a gas discharge diode,such as the neon glow tube that has beenmanufactured for some time. It differsmainly in its optics, which require highalignment accuracy and thus new produc-tion techniques. The materials cost of aHe-Ne laser can be predicted withreasonable accuracy, but the labor cost,which depends strongly on productionrate and labor-saving machinery, is lesscertain. In this paper we describe a He-Nelaser tube constructed with the least -expensive parts and labor-saving produc-tion techniques and then review the es-timated materials and labor costs forproduction quantities.

Reprint RE -22-4-2Final manuscript received April 13, 1976.

The most critical He-Ne laser productioncosts are in the high -precision areasassociated with the optical cavity. Theseareas include:

I) mirror grinding and coating;2) precision aperture for mode control;

3) seats for aligning the cavity accurate-ly.

The art of mirror -making is beyond thescope of this paper; suffice it to say that

Standard-HAND GROUNDMIRROR SEAT

PRECISIONBORE TUBING

Low -cos}

HEMISPHERICAL CAVITYLOW COST MIRRORS

CATHCDETERMINAL

ANODELEAD

techniques for grinding and coating formass production are well established inthe optical industry. Traditionally,functions 2) and 3) have been performedby glass parts that require more expensiveproduction methods than metal parts. Inthe He-Ne laser tube constructiondescribed in this paper,' these functionsare taken over by metal parts that can beinexpensively fabricated by drilling andpressing.

HARD GLASSSEALS

1

CATHODECONTACT

ALUMINUM FOILCATHODE

Be - CuSUPPORT

GETTER

LSOFT GLASSENVELOPE

CATHODE

LIMITINGAPERTURE

NONPRECISIONBORE

HANDGROUNDMIRROR

HARD GLASSSEALS

R F SEALED

EPDXY/ SEALED

ANODETERMINAL

Fig 1

Low-cost laser uses non -precision bore, aluminum -foil cathode, and rf-sealed metalendcaps on soft glass envelope. Metal endcaps and BeCu bore supper representsubstantial savings over the glass parts used in the standard laser.

19

He-Ne lasertube construction

At the very minimum a He-Ne laser tubeconsists of the following parts:

I) a tube envelope that also serves as aplatform for the cavity mirrors;2)a discharge bore and supports;3) cathode and anode electrodes;4) an aperture for optical mode selec-tion.

In the low-cost laser (Fig. I ), the envelopeconsists of a soft glass tube with metalendcaps, both inexpensive materials.Simple glass -to -metal sealing using eitherradio -frequency (rf) or fritting (low -temperature -melting glass) methods areapplicable. The metal endcaps aresupplied with spherical depressions asconventional seats for mirror -tuning andepoxy -sealing. One endcap has anoversized aperture, the other a precisionaperture for mode selection. These metalparts can be manufactured at relativelylow cost using standard high -volumemetal -forming techniques.

Usingmode selection makes the discharge -borediameter uncritical, as long as it is

somewhat oversized. This produces asavings over an expensive precisiondiameter. The bore needs to be supportedat two ends; one of the supports alsoisolates the two electrodes electrically.This is most easily accomplished byflaring out the bore at one end and lettingit become part of the end seal.

A thin -tongued beryllium -copper discsupports the other end. Another methodof supporting the bore uses a number ofmetal discs alternately connected to pre-vent electrical breakdown outside thebore. This method does not require anysealing to the bore, and has already beenused in He-Cd lasers.'

Since one of the metal endcaps can serveas the anode, the one tube part left todiscuss is the cathode. Since the cathodepower dissipation in a low -power He-Nelaser is small (less than I watt), a piece ofthin aluminum tubing or foil may serve asa cathode if the surface area is largeenough to support the emission current asnormal' cathode drop. An alternativemethod uses an aluminum -film coatingon the inside of the glass envelope as acathode; depositing this film may becomea part of the tube processing.

FRED SWEPTLASER

TOPINGMECHANISM

DETECTOR

Fig 2

Production alignment method transmits an external frequency -swept laser beam throughthe laser cavity. optimizing it for maximum transmission.

Tube assemblyand processing

The new laser tube shown in Fig. I wasassembled by stacking the parts on amandrel that aligns the endplates and thebore. The endplates were then rf-sealed, aprocess well -suited to mass production.

For large-scale production it is moreconvenient to align the laser cavity andepoxy the mirrors to the endplates if thetube is not operating; techniques fordoing this have been describedelsewhere.' This method treats the lasercavity as a Fabry-Perot resonator andoptimizes the transmission of an externalfrequency -swept laser beam through it(Fig. 2). It is easy to visualize adaptingthis method for large-scale production.

From this point on, processing in mass -production machinery may proceed aspreviously demonstrated' for small argonlasers.

Performance and life test

The major emphasis of this work wasproducing lasers at low cost, so the effectsof low cost on performance were ex-amined in detail.

Devices employing major components ofthe low-cost systems were constructedand life -tested. Both evaporated and foilaluminum were found to be as good orbetter than conventionally processedaluminum cathodes. Lasers using thesecathodes were life -tested to more than6500 hours with no performancedegradation.

Devices employing all the aspects of theproposed low-cost laser have been con-

structed, and their performance was evenbetter than conventional helium -neonlasers in many cases. Several of thesedevices were life -tested to more than 2500hours with no failures recorded. Usinglow-cost lasers also lowers overall systemcost because their lower operatingvoltage (roughly 60% of a conventionalHe-Ne laser's) and starting voltage in turnlower the power -supply cost.

See Table I for a summary of operatingcharacteristics.

Estimated materials andmanufacturing costs

I he volume of laser production to datehas permitted the use of hard -glass con-struction techniques. However, for low-cost construction, expensive hard -glassoperations cannot be tolerated. Withouthigh -temperature, hard -glass seals, thetube envelope becomes a simple piece ofnonprecision glass tubing cut to length.Such glass tubing can be purchased inlarge quantities for approximately 20% ofthe cost of hard glass. This glass alsolends itself to automated parts -making.

Using a limiting aperture eliminates theneed for precision, hand -shrunk plasma

Table ITypical operating characteristics of low-cost He-Ne laser.

Wavelength 632.8 nmPower output 1.5 mWReam diameter 0.6 mm

DC operating voltage(anode -to -cathode) 1200 V

DC operating current 4 mACylinder outline I" diameter by !IA" long

20

bore tubing. Again, standard low-costsoft -glass tubing cut to length providesbore materials at a cost substantially lessthan hard glass. Automatic rotary -sealing techniques can add the "funnel" atone end of the bore tubing for plasmaisolation. These concepts allow a com-plete bore assembly to be produced forless than 10% of the cost of a hard -glassprecision bore.

The cathode can be constructed by usingaluminum foil or evaporating aluminumdirectly onto the inner surface of the tubeenvelope. Both these methods eliminatethe expensive machining or formingcurrently used.

The mirror seats and end -closure caps areformed -metal parts fabricated withautomated stamping equipment, so thecost of these parts is almost that of thematerial. With proper design one part canperform the functions of many others-end-closure, mirror alignment seat, elec-trode contact, and limiting aperture. Theberyllium -copper spacer that supportsthe plasma tabulation at its non -flaredend is also fabricated by common small -parts forming techniques.

Mirrors are by far the most expensivematerial/ part used in the laser system.However, unique mirror constructiontechniques that are conducive to low-cost, high -volume production have beenreported." It is estimated that using thesetechniques could reduce mirror costs toless than 10% of their current value.

These parts represent the major com-ponents of a low-cost laser. Mis-cellaneous contacts, getters, epoxy, etc.complete the parts list. In summary,vendor price -quotations based onpurchasing materials for 100,000 unitsshow that the materials/ parts cost for thelaser described here amounts to about$1.15 per laser, exclusive of mirrors.

Assembly andprocessing costs

Using the techniques already described,the labor content needed to produce thelow-cost laser was estimated in order tocompare unit -cost estimates with currentlaser costs and the demands of high -volume markets.

The major components of the laser(plasma bore tubing, bore spacer, end -caps, cathode and tube envelope) all

come together in one basic operation.Using rf techniques along withautomated annealing furnace linespermits this operation to be completedwith less than 10 minutes of unskilledlabor. Proper fixturing and assemblylayout should hold down scrap levels toless than 5%.

The mirror -alignment techniquespreviously described consume an es-timated 12 minutes of labor time,although automated alignment equip-ment could reduce this amount further.

Exhausting and backfilling the laser

assembly usually involves time-consuming bake -out procedures and gas -filling and pumping cycles. Proper equip-ment design, including computer -programmed processing control, shouldhold labor for this operation to ap-proximately 5 minutes per laser.

The labor content of all the remainingmiscellaneous operations such as testaging, inspection, parts preparation, etc.should not exceed 15 minutes of direct

Karl Hernqvist joined RCA Laboratories in1952 and became a Fellow of theLaboratories in 1969. In 1956 he in-dependently conceived the thermionic con-verter and reduced it to practice. Dr Hern-qvist is presently doing research on gaslasers and high-pressure mercury lamps. Hehas received five RCA LaboratoriesAchievement Awards and the 1974 DavidSarnoff Award.Contact him at:Systems Research LaboratoryRCA LaboratoriesPrinceton,Ext. 2932

labor, resulting in a total of 50 minutes oflabor per laser.

Using these estimates for material andlabor, and applying nominal mark-upsand profit considerations, one can seethat it is possible to set an actual sellingprice well below current laser sellingprices.

Acknowledgment

The authors wish to express their thanksto W. Lynch for his encouragement andsupport during the progress of this work.

Referelces

I. /lei nos ist. K (i.; "Gas laser tube." U.S. Pal. 4904,9136.

2 liernysist. K.G.."Vented-bore He-Cd lasers." (OA RereiirVii .14 (19741 p. 401.

1. ('obine. J.17.: "Gaseous Conductors," Di.ner Publications.Neu 1 ork. 1957, p. 217.

4. liernysist. C.G. and Firester. A.H.: "Pre -alignment of gaslaser caiiiies." Rev. Ste. Inv. Vol. 46. (1975). p. 11)40.

5. Herniiiisi. "Adsances in gas laser technology ." RCAt.ti,eineer. Vol. 15. No. 5I Feb Mar 19701 p. 14.

6. Firester. A.H.. Heller. M.E.:and %Vitae. J. P.."1 lexpensiielaser mirrors... Omer. J. u/ Phi. Sit,. Vol. 41.1 1973i B. 1202.

Dick Longsderff came to RCA Lancaster in1959 and joined the Gas Laser Group in 1971as a development engineer, aiding in thedevelopment of He-Ne laser packaging forconstruction -alignment purposes. He wasalso responsible for developing the low-costlaser. This line of responsibility continueswith his present assignment-cost reduc-tion on high -production photomultipliers.

Ccntact him at:Electro-optics GroupSolid State DivisionLancaster. Pa.Ext. 2028

21

Light sources for fiber-optic communications

H. Kressell H. LockwoodM. Ettenberg I. Ladany

System requirements for optical com-munications are largely determined by theavailable components. Although through -the -atmosphere systems have someapplicability, they are generally restrictedto small installations where their inherentlimitations are tolerable. The more impor-tant optical -transmission media will beglass or plastic fibers, and thus, theproperties of the fiber (absorption, disper-sion, mechanical strength, etc.) are all-important. The other determining factorsin the system are the availability of com-patible sources, fibers, and detectors. For-tunately, this compatibility exists, and it isthe reason why optical communications issuch a rapidly expanding field. It is

generally true that commercially availableoptical fibers have low loss and smalldispersion (both modal and material) in thewavelength range of 0.8 pm to 1.1 pm.Therefore, in this article we discuss thecurrent status of semiconductor lightsources spanning this wavelength region,and since there are applications for bothcoherent and incoherent sources, we willtreat both lasers and LEDs (light -emittingdiodes).

Heterojunction structure

It is only since the advent of the firstpractical AlGaAs/ GaAs heterojunctionlaser structures" and the subsequenttechnological progress that the potential oflaser diodes for optical communicationshas been realized. Prior to that, GaAshomojunction lasers had such highthreshold and operating currents that theywere impractical for room=temperatureoperation. In some heterojunction devicesthat we describe, threshold currentdensities have been reduced two orders ofmagnitude, down to 500 A/ cm' from theearly homojunction values of about 50,000A/ cm'. This progress has made practical acompact high -radiance continuous -wave(cw) laser that can be directly modulated tohundreds of M Hz and has a power output

Research has shown that heterojunction lasers are ready forpractical systems applications. Heterojunction lasers andLEDs have now been operating continuously for over twoyears.

in the range of 10 mW. Since optical fiberswith losses under 20 dB/ km are nowcommercially available, multi -kilometer,high -data -rate transmission systems (evenwithout repeaters) are expected to becomecompetitive with coaxial systems.

here are also many applications for fiber-optic links over short distances and/or atlow data rates (e.g., local distributionsystems). The LED satisfies many of therequirements of such systems, and since it isnot a threshold device, its output power isless sensitive than the laser's to smallchanges in operating current or ambienttemperature. Much LED technology iscomparable to that of lasers, so the twodevices have tended to develop together.State-of-the-art 1.EDs now deliver severalmilliwatts of output power at modulationrates of 100 to 200 megahertz.

6 2

60

58

56

54

Choosing the alloy system

DIODE EMISSION40 i 3

GoSbIr.As

The most highly engineered and testedlasers and LEDs for optical com-municatons are those derived from theternary alloy system A1,Gai-As. Thinlayers (0.05 to I µm) are sequentially grownwith a high degree of perfection by liquidphase epitaxyb on substrates of GaAs. Byvarying the alloy composition of therecombination region, the emissionwavelength can be varied over a usefulrange of 0.8 to 0.9 pm. Emission, andindeed lasing, can be achieved at stillshorter wavelengths, but attenuation anddispersion in the fiber begin to becomeappreciable at shorter wavelengths. Thepurest (OH -free) fibers available haveattenuation and dispersion values thatdecrease to negligible values in the 1.0 to1.1 mm wavelength region, so there is a real

WAVELENGTH (

l0

GoAs

GoP

07 055

ALSb

00 0 20BANDGAP ENERGY (eV)

Fig. 1Extent of band gaps and lattice parameters covered by the quaternary alloysAlzGai-xAsySbi-y (light area) and InrGai-xAsyP1-y (dark area). The boundaries of the areasrepresent ternary, and the vertices binary, alloys. Two particularly useful alloys areindicated: the square locates Ino.aciGao.2oAso.35Po.65 the lattice parameter of which matchesInP substrates and would make diodes that emit at 1.1 micrometers; the triangle shows theposition of Alo./Gao.9As, which matches the GaAs lattice and emits at about 0.82micrometer.

22

interest in developing new alloy systemsand a heterojunction technology for emis-sion in this range also.

In the AI,Gal-,As system the latticeparameter varies by only 0.14% as x goesfrom iero to unity. lieterojunctions in thissystem consequently have negligible strain -induced defects and demonstrate long-termreliability. More typical of most ternaryalloys, the handgap and lattice parametervary significantly between the extremes ofthe binary alloys from which they arederived. In,Ga, ,As, In,Ga, ,P andGaAs,P, , are examples of such alloys. Asa result, heterojunction structures in thesematerials are inevitably strained to a highdegree unless compositional grading (mosteasily controlled in vapor -phase epitaxy) isused.

An alternatise approach to obtainingefficient device performance at I to 1.1 Almis to fabricate heterojunction structures inquaternary alloys, such as In,Gal ,As,P, ,

and Ga,All ,As,Sb, ,. In these alloys.which cover the desired emission range. thehandgap and lattice parameter can heindependently adjusted (Fig. 1 ). At somecost in simplicity, this extra degree offreedom permits the fabrication of strain -free heterojunction devices. As evidencedby a growing literature on the subject,progress toward useful - 1.1 µm lasers andLEDs is being made, although much workremains in ensuring their reliabilit . whichis handicapped by lattice defects. Whetherthe complexity of fabrication will he

justified by the moderate improvement inattenuation and dispersion will ultimatelyhe determined by cost and long-termreliability.

The problem of lattice -parameter match atheterojunctions can he discussed in termsof mismatch -dislocation networks. For ex-ample, for a lattice mismatch at aninterface, a dislocation will be generatedover approximately 100 atom planes. Sincethe dislocation core consists of non -radiative recombination centers, such ahigh dislocation density will depress thedevice's internal quantum efficiency.Furthermore, dislocations are not alwaysconfined to the lattice -mismatched inter-lace, hut can propagate through multilayerstructures. [he effect of dislocations onradiative efficiency is dramatically il-lustrated in Fig. 2. There, we compare atransmission electron photomicrographof a dislocation array in a mismatchedheterojunction structure and a

10 microns

Fig. 2Misfit -dislocation arrays in a compositionally graded Ini,Ga (-MP vapor -grown epitaxial layeron a GaP substrate. The left-hand photo is a transmission electron micrograph (Ref. 7). Atthe right is a cathodoluminescence scan, in which dislocations near the surface appear asnon -radiative regions. Photos are not to scale.

Michael Ettenoerg joined RCA Laboratoriesin 1969 and has been involved in studiesconcerning factors that affect the reliabilityof electrolum nescent diodes, improved li-quid phase techniques for the synthesis ofIII -V compounds, and theoretical studies ofthe thermodynamics of crystal synthesis.

Ivan Ladany has worked in GaP, GaAIP andGaAs luminescent -diode research sincecoming to RCA Laboratories in 1966.Recently he has worked on improved GaAsinfrared diodes, green -emitting GaP LEDsand III -V compound growth on insulatingsubstrates using liquid -phase epitaxy. Atpresent, he is devoting most of his time toinjection laser development, having madesignificant contributions to the develop-ment of long-lived room -temperature cwlasers.

Reprint RE -22-4-3This article originally appeared in similar form in Physics Princeto-. N.J.Today. May 1976 Ext. 2427

Harry Lcckwood was one of the early in-vestigators of injection luminescence in III -V compounds, in particular GaAs andGaAsP. This work preceded the injectionlaser development. Dr. Lockwood joinedRCA Laboratories in 1969 and is currentlyworking on material -synthesis problemsand luminescent devices, including injec-tion lasers.

Henry Kressel is presently Head ofSemiconductor Device Research at RCALaboratories. Dr. Kressel pioneered in thefield of (AlGa)As-GaAs heterojunctiondevices, in particular laser diodes, aid hasbeen actively engaged in the study ofdevices and luminescent processes invarious III -V compound materials.Contact him atSemiconductor Device ResearchRCA Laboratories

Authors Ettenberg. Ladany. Kressel. and Lockwood.

cathodoluminescence scan of a similarstructure. The dark lines and spots in thecathodoluminescence micrograph are

areas of low radiative efficiency, cor-responding to dislocations that lie paralleland perpendicular to the plane of thesurface viewed. Therefore, in designingheterojunction structures for LEDs andlasers, it is extremely important to chooseeither a totally lattice -matched system orremove dislocations from the active regionthrough compositional grading.

Continuous -operation lasers

\ 'though there are numerous potentialconfigurations for the cw laser diodes, thesymmetric double-heterojunction with astripe contact has been widely adopted asthe laser geometry most suitable for opticalcommunications. The schematic and actualcross section of a typical laser, Fig. 3, showsthe optical ca' it) (recombination region)defined by its higher bandgap dielectricwalls as well as the outer n- and p -typeGaAs regions to which ohmic contact ismade.

-the efficient operation of a laser dioderequires effective confinement of minoritycarriers and radiation to the optical cavity,which is also the recombination region ofthe device. Both functions are provided byheterojunctions. Radiation confinement isprovided by the dielectric discontinuity,while carrier confinement results from apotential barrier created by the bandgapdifference between the materials that formthe heterojunction. The average carrier -pair density injected into the recombina-tion region of a double heterojunctiondevice for a current density J is

AN a- Jr ed

where e is the electron charge, r is thecarrier lifetime and d is the width of therecombination region (i.e., the heterojunc-tion spacing). In a typical GaAs laser diode,

at lasing threshold is - 1.5 X 101s cmat room temperature." To minimize thethreshold current density, we restrict therecombination region's width by placingthe heterojunction forming the potentialbarrier for minority carriers at the diffusionlength from the injecting heterojunction. Itis, however, essential that the heterojunc-tion interfaces be relatively defect -free inorder to prevent excessive nonradiativerecombination 'of the injected carriers.

The nonradiative loss of carriers at an

interface is characterized by therecombination velocity S of that interface.Under the usual laser operating conditionswe can express the effective recombinationrate due to the presence of the interface as

I I 2S

r, r d

where 1/ r is the recombination rate in theabsence of an interface. The internal quan-tum efficiency is given by

T,.11 To

so that

( I + 2Sr., d)I

In typical cw laser diodes, d= 0.3µm and r= 10-9. Therefore, for an internal quantumefficiency of 50% (a reasonable lowerlimit), we would require S < 2 X l04 cm s.The single most important contribution toS is from nonradiative recombinationstates introduced at the heterojunction dueto the lattice -parameter mismatch betweenthe two materials. If this mismatch is keptbelow 0.1%, S will be under 2 X 104 cmi sand cw operation can be anticipated. Ex-perimental data concerning S in GaAs-ALGai ,As heterojunctions indicate that S

5 X 10' cm/ s in practical laser structures(where Ati; a S 0.07%), a value that isfully satisfactory for narrow recombina-tion region devices."

To ensure wave propagation along theplane of the junction, as well as lowthreshold current density and high efficien-cy, there must he a means of confining

5-01 - 0 3

STRIPE WIDTH - 13

stimulated radiation to the region of in-verted population (or its close proximity).Two heterojunctions, as indicated in Fig. 3,provide a controlled degree of radiationconfinement because of the higherrefractive index in the lower-bandgap-recombination region. The fraction of theradiation confined depends on the hetero-junction spacing d and the refractive indexsteps -In at the lasing wavelength. InAI,Gai-,As/ GaAs structures, An 0.612xat -4== 0.9 pm. In general, it is desirable toequalize the refractive index steps An ateach heterojunction to prevent the loss ofwaveguiding that can occur in thinasymmetrical waveguides. However, evenwith the symmetric double-heterojunctionlaser, wave confinement within the hetero-junction is gradually reduced as its spacingbecomes small.

Low -threshold operation

I he [ruction of the radiation confined tothe recombination region of the double-heterojunction laser affects the radiationpattern and threshold current density. Theradiation pattern (far -field intensity dis-tribution) is affected because it is

determined by the effective source size(near -field intensity distribution); thethreshold current density J,5 is determinedby the gain at threshold-only that portionof the optical flux within the recombina-tion region is amplified. Fig. 4 showscalculated and experimental values of Jsh asa function of d for various -in valuescorresponding to varying heterojunctionbarrier heights. The lowest J15 value" of

METAL1ZATION

GaAs SUBSTRATE-.

E; 1 8 eV

RECOMBINATION --4.REGION E 1 55 eV

E, - 183 eV-'OXIDEMETAL IZAT IONSOLDER

I--1100µm

Fig. 3The cross section of a typical laser in a schematic illustration (left, not to scale) and aNomarski photomicrograph of a sample that has been polished at a shallow angle toproduce high magnification in the transverse direction (right). The "terracing" effect evidenton the lower portion of the micrograph, a growth artifact, causes some interface roughness.

24

2500

E2000

4

)-/-(.7)

215000

zCr 100000_J0fn 500CC

0001 02 03 04 05 06

HE TEROJUNC T ION SPACING (p.m)

Fig. 4Threshold current density as a function ofthe heterojunction spacing for AI xGa /-xAsdouble-heterojunction lasers. The ex-perimental data points are for aluminum -concentration steps A x of 0.20 (triangles)and 0.65 (circles). The theoretical curves arefor the discontinuities in the refractive indexA n shown, where the relation A n=0.62 A xhas been assumed to apply.

475 A cm: is obtained with d= 0.1µm andAn = 0.4 (corresponding to anAl,,,,Gan,As GaAs, A10,,Cia,,,As struc-ture). It should be noted that the harrierheight also affects the high -temperatureperformance of lasers:'

The maximum desirable threshold densityfor cw operation is below 2000 A cm-. anddouble-heterojunction laser diodes have sofar routinely operated cw at roomtemperature with 0 to 12ci AlAs concentra-tion in the active recombination region.Because of the corresponding bandgapvariation, this produces an emissionwavelength range of 0.9 to 0.8µm. Room -temperature cw operation of lasers at awavelength of 1.0 to 1.1 µm has also beenobtained with Cia,Ali ,As,Sbi , diodes,"where Jo, = 2000 A, cm- was achieved, andfor InGaP InGaAs diodes."

Constructing thestripe -geometry laser

Laser diodes are prepared by cleaving twoparallel facets to form the Fabry-Perotcavity. The cw laser diode uses a stripegeometry to define the lateral dimension(Fig. 5) for several reasons:

The radiation is emitted from a smallregion. which simplifies coupling of theradiation into fibers with a lownumerical aperture.

The operating current can he

minimized because it is relatively simpleto form a small active area with photo-lithographic contacting procedures.

The thermal dissipation of the diodeis improved because the heat -generatingactive region is imbedded in an inactivesemiconductor medium.

The small active diode area makes itsimpler to obtain a reasonably defect -free area.

The active region is isolated from anopen surface along its two majordimensions, a factor believed to be im-portant for reliable long-term operation.

In the simplest stripe -contact structure, theactive area is defined by opening astripe in a deposited Si02 film. The surfaceof the diode is then metalized, with theohmic contact formed only in the open areaof the surface. Other methods that havebeen used to define the stripe contactinclude various implantation methods thatincrease the lateral resistance, and narrov,,mesas.i,

Diodes for cw operation are generallydesigned with the thin p -side mounted oncopper heat sinks to minimize the thermalresistance of the structure, using a softsolder such as indium to minimize strains inthe devices.

EPITAXIALLAYER

METAL LICSTRIPE CONTACT

Laser -diode operation

The lateral width of the emitting region canbe adjusted for a desired operating level byadjusting the stripe width. For example,100 mW of cw power is obtainable (fromone facet) for a strip 100 pm wide.However. for the typical power levelsneeded in optical communications (5-10mW), stripe widths of - 13 Aim are used.This dimension represents a suitable com-promise between low operating currentsand an appropriate power -emission level.A typical curve of power output as afunction of diode current is shown in Fig. 6.The junction temperature for such devicesis only a few degrees above the heat -sinktemperature. For example. a temperaturedifferential of 7 K is calculated with atypical power input of 0.5 W at a diodecurrent of 0.3 A and a thermal resistance of14 K W.

The electromagnetic modes of the laserdiode cavity are separable into two in-dependent sets with transverse electric (TE)and transverse magnetic (TM) polariza-tion. The mode numbers in, s, and q definethe number of sinusoidal half -wavevariations along the three axes of thecit% it transverse, lateral andlongitudinal, respectively.

The allowed longitudinal modes aredetermined from the average index ofrefraction and the dispersion seen by thepropagating wave. I he Fabry-Perot mode

10-50p. fn

CLEAVEDFACET

S 0,

CURRENTDISTRIBUTION

R EC OM B IN AT IONREGION

Fig. 5Typical cw neterojunction laser; devices are now commercially available.

25

45

40

35E

w 300

25

it 2090

PULSED(10-4 DUTY CYCLE)

100 200 300 400 500 600CURRENT (mA)

Fig. 6Optical power output from one facet of atypical AIGaAs cw laser as a function of thedrive current. Measured at roomtemperature in pulse and cw operation.

spacing is several angstrom units in typicallaser diodes. The lateral modes are depen-dent on the method used to define the twoedges of the diode. Generally. in stripe -geometry lasers only low -order modes areexcited; their mode spacings are 0.1 to 0.2A and they appear as satellites to eachlongitudinal mode. The transverse modesdepend on the dielectric variationperpendicular to the junction plane. In thedevices discussed here, only the fundamen-tal mode is excited. a condition achieved byrestricting the width of the waveguidingregion (i.e.. heterojunction spacing). tovalues well under I p.m. Therefore, the far -field radiation pattern consists of a singlelobe in the direction perpendicular to thejunction. ( Higher -order transverse modeswould give rise to "rabbit -ear" lobes, un-desirable for fiber coupling.)

For a laser operating in the fundamentaltransverse mode, the full angular beam -width at half power perpendicular to thejunction plane is a function of the near -fieldradiation distribution.The narrower theemitting region in the directionperpendicular to the junction plane, thelarger the heamwidth. In practical cw laserdiodes the heamwidth is 30 to 50°. Theheamwidth in the direction parallel to thejunction plane (lateral direction) is typical-ly 5 to 10° and varies only slightly with thediode topology and internal geometry. Atleast one-half of the power emitted fromone facet can be coupled into a multi -modestep -index fiber with NA =0.14 and a corediameter of 80 Aim.

While fundamental transverse modeoperation is easily achieved, most narrowstripe laser diodes operate with severallongitudinal modes, and therefore emitover a 10- to 30-A spectral width, but someunits can emit several milliwatts in thefundamental lateral and transverse modeand a single longitudinal mode. The line -

width is 0.15 A and the power emitted is 3mW (Fig. 7).

Methods of modulating the laser outputvary widely, depending upon the applica-tion. For fast -pulse modulation, the diodeis biased with a current near the thresholdcurrent and then pulsed to an appropriatelevel above threshold. If no bias is applied.there is a lasing delay (related to thespontaneous carrier lifetime) before thecarrier population becomes fully invertedand the device turns on.'" This delay isseveral nanoseconds long in a typicalsituation, but vanishes if the laser is biasedto threshold.

LED structure

The spectral bandwidth of the LED istypically I to 2 k T(300 to 600 A) at roomtemperature, hence I to 2 orders ofmagnitude broader than that of the typicallaser diode. Because of the spectral disper-sion in fibers. this may limit the bandwidthfor long-distance fiber communicationsusing LEDs. Furthermore, the couplingefficiency of LEDs into low -numerical -

aperture fibers is much lower than for laserdiodes. However, the LED has the advan-tage of simpler construction and less

temperature dependence of its emittedpower. For example. the spontaneous out-put from an LED may decrease by only afactor of 1.5 to 2 as the diode temperatureincreases from room temperature to 100°C(at constant current), while the output of a

0 05 A

laser diode would typically decrease bymore than a factor of 3.

LED topology is designed to minimizeinternal reabsorption of the radiation,allow high -current -density operation andmaximize the coupling efficiency intofibers. While the structures used areapplicable to all materials, most of thework on communications -type devicesreported so far has been on AlGai ,As.

Two basic diode configurations for opticalcommunications have been reported: sceemitters' and edge emitters.'"'" In thesurlace emitter, the recombination regionis placed close to a heat sink and a well isetched through the GaAs substrate toaccommodate a fiber. The emission fromsuch a diode is essentially isotropic. ;Theedge -emitting heterojunction structures,similar to the geometry of Fig. 5. use thepartial internal waveguiding of the spon-taneous radiation due to the hetero-junctions to obtain improved directionalityof the emitted power in the directionperpendicular to the junction plane. Thelateral width of the emitting region isadjusted for the fiber dimension, but istypically 50-100µm. Fig. 8 shows an edge -emitting

LED performance

Surface -emitting and edge -emitting struc-tures provide several milliwatts of poweroutput in the 0.8- to 0.9 -Aim spectral range.operated at drive currents of 100 to 200µA(2000 to 4000 A cm2). The coupling lossinto step -index fibers with a numericalaperture of 0.14 is about 17 to 20 dB forsurface emitters and 12 to 16 dB for edge -emitting diodes, compared to about 3 dBfor an injection laser., Since the couplingloss decreases as NA -, much more power

8070 80 71 8 072 8073 8074

WAVELENGTH ( 4 )

I4.- 3 mW

8075LETT8076 8077 8078

Fig. 7Spectral output from one facet of an AlGaAs cw laser. It emits 3 mW when operating in asingle longitudinal mode and in the fundamental transverse and lateral mode

26

INSULATION

AtrA, 101A. - "X.:1

!

CONTACT

LEDLENS

CONTACT

COPPE RHEAT SINK

EPDXIES

FIBER

0" 01" 02"

GROOVEDBLOCK

Fig. 8Edge -emitting LED designed for fiber-optic communications. Short fiber extending to rightcouples to the long fiber.

can. of course, be coupled into larger -numerical -aperture fibers. But with thesecoupling losses, Al. iGaoyAs double-heterojunction LEDs can provide about100 µW into an 0.14 -NA, 80 -pm -diameterfiber at drive currents of about 200 mAwith an applied voltage of 1.7 V.

1 urning to the modulation problem, therelation between the optical power outputof an LED (with constant peak current)and the modulation frequency. w. is givenby-

11w)

Po [I + (cora'

where r is the injected carrier lifetime in therecombination region, and P is the dcpower emission value. (However. parasiticcircuit elements can reduce the modulatedpower range below this value.)

It is evident that a high-speed diode re-quires the lowest value of r possible

without sacrificing internal quantumefficiency. Low values of r are obtained athigh doping levels; but in GaAs and relatedcompounds. a high density of nonradiativecenters is formed when the dopant concen-tration approaches the solubility limit atthe growth temperature. Of the devicesreported so far, germanium -doped double-heterojunction LEDs have exhibitedmodulation capability (at the 3 -dB point)to about 200 MHz." The use of Ge isadvantageous because it can he incor-porated into GaAs to concentrations ofabout 2 x 10" cm thus providinglifetimes on the order of one nanosecondwithout unduly reducing the internal quan-tum efficiency.

With regard to LEDs for 1.0- to I.1 -µmwavelengths, surface emission outputs of

about I mW have been achieved using(InGalAs structures:I Further progress isexpected. particularly with the use of thelattice -matched I n(GaAs)/ ( I nGa)P hetero-junct ion structures: -

Failure causes

In any practical optical communicationssystem. component reliability is of greatconcern. It has been a major research goalto identify and correct the myriad of failuremechanisms that seemed to plague earlyelectroluminescent devices.-' The failuremodes have since been identified as eitherfacet- or bulk -related; they can be ofgradual or catastrophic nature. Facetdegradation is specifically a laser problembecause the facets are the mirrors thatdefine the Fabry-Perot resonator. Bulkdegradation. on the other hand, can occurin both LEDs and lasers. Most failuremechanisms have been eliminated as theoverall technology (crystal growth. devicefabrication) has matured, hut active con-cern still exists for those remaining fewwhich may limit ultimate operating life.Insufficient data exist for mean time tofailure which, for telephone applications, isin the 100.000 -hr 'ange. A comprehensivemodel for laser facet failure does not exist.but considerable phenomenological datahave been accumulated.

Intense optical fields

Facet fail= due to intense optical fields isa well-known phenomenon in solid-statelasers; it occurs in all types of semiconduc-tor lasers under varying conditions whenthe optical power density in the recombina-tion region reaches the order of 106 W/ cm-.The appearance of the damaged laser facetssuggests local dissociation of the material,as well as "cracking" in some cases.

The critical damage level is also a functionof the pulse length, 1, decreasing as I overthe range of 20 to 2000 ns. It is notsurprising, therefore, that facet damage canoccur is room -temperature laser diodesoperating cw at their maximum emissionlevels (even with relatively low totalpower). Because of the nonuniform radia-tion distribution in the plane of the junc-tion in stripe -contact lasers, it is difficult toestablish precise linear power densitycriteria. However, the damage thresholdfor 100-ns pulses is about 10 times higherthan it is in cw operation of diodes selectedfrom the same group!'

In addition to the dependence on opticalflux density and pulse length, it has beenfound that ambient conditions, specificallymoisture, and surface flaws (scratches, dirtparticles) can lead to premature facetfailure. In order to remove theselimitations, lasers must be operated at aspecific maximum power level and thefacets must be passivated with protectivedielectric coatings for isolation from theirsurroundings. For cw lasers the operatingrange of linear power density is about I

mW pm of stripe width. Facet failure, atleast in its early stages, commonlymanifests itself as a decrease in differentialquantum efficiency without an increase inthreshold current density.

Bulk degradation

Bulk degradation, the other failure mode,is accompanied by an increase in thresholdcurrent density and a decrease indifferential quantum efficiency caused by areduction in internal quantum efficiencyand an increase in absorption coefficient.The reduction in output power may besmall, hut the device may turn off com-pletely if it is a laser operated nearthreshold. In either case, a small adjust-ment in current will restore the output to itsinitial value. With such a feedback system,a definition of operating life becomessomewhat arbitrary and system -dependent.

The available evidence suggests that thisgradual degradation process results froman increase in the concentration of non -radiative centers in the recombinationregion.-' These defects are initiated by thegrowth of flaws that are initially present inthe recombination region of the diode.Point defects may also diffuse from exter-nal flawed regions adjacent to the active

27

region. Detrimental flaws include dis-locations and impurity precipitates. Oneprominent effect in some degraded lasers isthe formation of "dark lines" in regionswhere the luminescence is graduallyextinguished.2" These regions have beenidentified as large dislocation networksthat, having started as smaller pre-existingdislocation networks, grow by the in-migration of vacancies or interstitials:27 Inaddition, more dispersed nonradiativecenters, such as native point defects, ap-parently contribute to the degradationprocess.

It has been suggested that a multi -phononemission process resulting from non -radiative electron -hole recombinationgives rise to an intense vibration of thecenter, which reduces its displacementenergy,'" (Whether any point defects areactually formed within the recombinationregion remains unclear.) Hence, if non -radiative electron -hole recombination oc-curs, say at the damaged surface of a diode(as in the case of the sawed -edge diodeexperiment described in Ref. 29), itaccelerates the motion of point defects intothe device's active region.

The stoichiometry of the material in orclose to the active region also can affect thelocal density of point defects;'" for exam-ple, the vacancy concentration under cer-tain conditions may be substantially abovethe equilibrium value. A modification ofthe initial stoichiometry may account forthe improved degradation resistance ofdiodes fabricated with Al11iGa09As, ratherthan GaAs, in the recombination region.Finally, regions where nonradiativerecombination occurs will tend to grow insize, leading to the strongly nonuniformdegradation process that is commonlyobserved.

Reliable devices available

Enormous progress has been made since1970 in eliminating many degradationmechanisms in lasers and LEDs by carefulattention to the liquid -phase epitaxialgrowth and device processing. Facetpassivation with dielectric coatings hasvirtually eliminated facet damage as animportant failure mode in cw lasers. Theultimate operating lifetime of state -of -theart devices remains undetermined, butaccumulated data exists on devices inoperation at constant current for well inexcess of 20,000 hr without a serious dropin output power. Two examples of life data

taken over many hours of operation onmore recently fabricated diodes show that

for a lot of (AlGa)As LEDs emittingabout I mW at 0.8µm, the emitted powerremained constant within 5% over 18,000hours of operation; and

for a lot of (AIGa)As cw lasers emittingbetween 5 and 10 mW from one facet at awavelength of 0.82 Aim, the maximumdeviation was less than 25% in 14,000hours.

These curves contrast sharply to those of afew years back, when the best operating lifeunder similar conditions was a few hun-dred hours.

Conclusions

Heterojunction structures of AIGaAsemitting in the 0.8- to 0.85 -pm spectralrange have progressed to the point wherepractical systems applications of lasers andLEDs are becoming possible. Extensiveresearch concerning properties of materialsand methods of synthesis, as well as

identifying the major factors affecting theoperating lifetime of these devices, hasmade commercial products possible.Research is now under way on devicesemitting in the 1.0- to 1.1-µm range, whichoffers some potential advantages inreduced fiber absorption and dispersion.These devices involve more complexmaterial problems than the (AlGa)Asdevices because of the need to use dis-similar alloys to match the lattice.

References

I. Kressel. H.: and Nelson. H.: "Close -confinement galliumarsenide pn-junction lasers with reduced optical loss at roomtemperature." RCA Review. Vol. 30. 11969) p. 106.

2. Hayashi, I.; Panish. M.B.; and Foy. P.W.; "A low -thresholdroom -temperature injection laser." IEEE J. QuantumElectron., Vol. 5, (1969) p. 211.

3. Alleros. Zh. I.; Andreev, V.M.; Portnoi, EL.: and I rukan.M.K.:"AlAs-GaAs heterojunction injection lasers with a lowroom -temperature threshold,- SM.. Phis. Semiconductors.

ol. 3 (19691 p. 1325. II rans.: Sor. Phis. Semiconductors.01.1 (197(1) p. 1107.)

5. K nowt. H.: and Flaurylo. 1-.Z.: "Fahry-Perot structure.As injection lasers with room -temperature threshold

current densities ol 2530A cm-." App/. Phys. Iris., Vol. 1711970) p. 169.

6. Fur recent papers see J. ('rrstal Growth. Vol. 24 11974). aspecial issue on liquid -phase epitariy. I he apparatus used inour laboratory is described by H.F. Lockwood and H. Kresselon pp. 97-105 ol that issue and in Lockwood. FEE.; andEttenberg. M.; "I hin-solution multiple -laser 1:pitass." J.Crystal Growth. Vol. 1511972) p. K2.

7 . Olsen. Ci.H.; "Interfacial lattice mismatch efforts in Ill -Vcompounds." J. (rival Growth. Vol. 31 (1975) p. 223.

X. Stern. 1-.A., "Gain -current relation Ior GaAs lasers with n -s pc and undoped attire layers." IEEE J. Quantum Electron..

Vol. 9 (1973) p. 290. For an introductory treatment tostimulated emission in semiconductors. see Variv. A.:Quantum Electronics 2nd ed.. John Wiley 1975, or Pankove.IL: Optical Processes of Semiconductors. Prentice -Hall.Engles. trod ('has. NJ 1971. For detailed discussion of devicerequirements for optical communications see Miller. S.F.:I ingse. I .: and Marcatili. E.A.J.; "Research toward optical -

fiber transmission systems, Part II: Devices and systemsconsiderations." Pro(. IEEE. Vol. 61 11973) pp. 1726-1751.

9. F t tenberg. M.; and Kressel. H.: "Interfacial recombination atAlCialAs Ga As heterojunction structures: J. Appl.

Apr 1976.

10. Analyses ol mode guiding in thin structures are discussed inHuller. J.K.: Kressel. H.: and Ladany. 1.; "Internal opticallosses in sery thin cw heterojunction laser diodes." IEEE J.Quantum Lieu .. Vol. I I (19751p. 402; Dumke, W.P.: "Theangular beam disergence in double-heterojunction lasers, withsery thin actise regions." /hid. p. 400: Selway, I'.R.; andGoodwin. A.R.: "the properties ot double heterostructurelasers with sers narrow actise regions." J. Phys. D.. Vol. 5119721 p. 904: and ('ases. H.C. Jr.: Parrish. M.B.: and Merv..1.1... "Beam disergence tit the emission from double -hew ))) ) ructurc injection lasers.- J. Appl.Phts, Vol. 4411973)p. 5470.

II. Ettenherg. M.: "Very -low -threshold double-heterojunction-ALCia, .As injection lasers," Appl. Phys. Isnt.. Vol. 27(1975) p. 652.

12. (ioodum. A.R.: Peters, J.R.: Pion. M.; Thompson. (i.H.B.;and %Vacua). J.F.A.: hreshold temperature characteristicsof double-heterostructrue ,ALAs lasers." J. Appl.Vol. 46 119751 p. 3126.

11. Nrahors. R.I . Pollak. M.A.: !lecke. DeWinter. J.C.:and Dixon. R.16 , "Continuous operation ol 1.0-gm-ua s elength GaAs, ,Sb, Al,Ga. ,As. ,Sb, double-heterostructure injection lasers at temperature." Appl.Phis Iris.. Vol. 28 119761 p. 19.

14. Nuese. C.J.. et al.. to he published.

15. Various aspects ol the technology are discussed in papers inthe special issue on semiconductor lasers. IEEE!. QuantumElectron.. Vol. I I (Jul 1975) pp. 351-562.

16. Konnerth. K.: and Lanra. C.: "Ilelay between current pulseand light emission of a gallium arsenide injection laser."Appl.Phys. Len.. Vol. 4119641 p. 1211: and Ripper. J.E.: "Measure -menu of spontaneous carrier Wet me I rom stimulated emissiondelays in semiconductor lasers: J. .4ppl. V01.43(1972)p. 1762.

17. Hurrus. C.A.; and Miller. WI.: "Small -area. double -hew rrr s i ructure aluminum -gallium arsenide deem, -luminescent diode sources for optical -fiber transmissionlines." Opt. Commun.. Vol. 41 11971) p. 307.

IS. Ettenberg. M.: Lockuood, H. F.: Wittke. J.: and Kressel. H.:"High radiance. high speed ALCia, ,As heterojunction diodesfor optical communications."lichniral Digest. 1973 Inter-national Electron Dolce. Meeting. Washington. p. 317.

19. Wittke. JP.; Ettenherg. M.: and Kressel. M.: -High radiancelor single -fiber optical links.- RCA Review, Vol. 37, No.

2 (Jun 1976).

2(1. Liu. V .S.: and Smith. 1).5.; "1 he frequency response of anamplitude modulated GaAs luminescence diode." Prix. IEEE.Vol. 63 119751 p. 542. J. Wittke independently derived thiscspression for heterojunction structures.

21. Nuese. C.J.: and Engstrom. R.E.; "Efficient 1.06 -gm calls -sum rum ,As electroluminescent diodes." IEEEIrani. Electron. Devices. Vol. 129119721 p. 1067.

22. S nese. CT.: and Olsen. "Room temperature heterojec-lion laser diodes of InCia,-As ,P with emissionwaselength between 0.9 and 1.15 jim:Appl. Phys. Lett_ Vol.26 119751 p. 525.

23. A comprehensive role% of the literature until 1973 waspresented in Kressel. H.: and Lockuirod. KV: "A resins ofgradual degradation phenomena in electroluminescentdiodes." J. de Phi (.3. Vol. 35 (19741 p. - 23.

24. Kressel. H.: and I adany. I.. "Reliability aspects and facetdamage in high -power emission from TAKialAs cw laserdiodes at room temperature.- RC.4 kern... Vol. 3611975)p.

25. IN:I °acts. Hakki. B.W.: Hartman. and D'Asara.1_A.: "Degradation ol cw GaAs double-heterostructurelasers." Prin. IEEE. Vol. 61 119731 p. 1042.

26. huh. R.: Nakashima. H.: Kishino. S.: and Nakada. O.:"Degradation sources in Cia As-AlCia As double-heti: rrr s r rut:lure lasers." IEEE J. Quantum Eke trim.. Vol. II(19751 p. 551.

27. Petroll. P.:and Hartman. R.I....-I)etect structure introducedduring operation ol heterojunction (iaAs lasers." Opt Phy,Lai. Vol. 21 11973) p. 469.

25. (iold, R.I ).; and Weisberg. 1..R.: "Permanent degradation ofGaAs tunnel diodes." Solid -State Lieu trim.. Vol. 7 (19641 p.511.

29. Ladany. 1.: and Kressel. H.; "Influence of device labricationparameters on gradual degradation of 1A1Cia)As cw laserdiodes." .4ppl. Phis. Ire.. Vol. 25 119741 p. 701).

31). Ettenberg. M.: and Kressel. H.: "Heterojunction diodes of1 AlfialAs-(iaA. with improved degradation resistance.".4rart Phi, I tit . 01. 2611975) p. 475. I he effect oil dopantson reliability acre studied by McMullin. P.(i.: Blum. J.: Shih.K.K.: Smith. A. V. .and Woolhouse. G.R.: "Effect ol dupingon degradation of GaAs-AEG:I; ,As injection lasers." Appl.Phys. let:.. Vol. 24 41974) p. 595.

28

Applying injection lasers to information scanning

I. Gorogl P.V. GoedertierJ.D. Knoxi I. LadanyJ.P. Wittkel A.H. Firester

One of the most important future areas oflaser application lies in high -density infor-mation storage and retrieval. High -densityoptical storage systems are apt to be bothread-only and read write types, and lasersare generally assumed to provide theoptimal light source for both applications.However, apart from holographicmemories, these systems do not use thetemporal coherence properties commonlyassociated with lasers, but use only thelaser's extremely high radiance. Gas lasers,such as the low -power He-Ne type, aregenerally used for optical -informationstorage systems, but long-lived, room -temperature cw injection lasers haverecently become serious contenders forthese applications.

Indeed, our experience with these small,efficient, readily modulated devices in-dicates that injection lasers may well

EPiTAXIALLAYER

METALLICSTRIPE CONTACT

Experimental investigations have shown that cw room -

temperature GaAlAs injection lasers compare favorably withHe-Ne gas lasers for use in information -retrieval systems.

replace gas lasers as the preferred source inhigh -density information -retrieval systemsand in flying -spot scanner applications.Possible stems that are currently receiv-ing considerable attention includefacsimile transmission terminals and com-puterized point -of -sale checkout systems.

Because the optical characteristics of injec-tion lasers differ rather markedly fromthose of a typical gas laser, the opticalsystems that must he used with them alsodiffer. In the next section we consider whatthese injection laser characteristics are andwhat sort of optics is required to produceuseful optical beams from them. Then wewill present experimental results showinghow these (suitably modified)characteristics can he applied to high -density read-out systems and to a facsimilescanner.

10 - 50/./. m

CLEAVEDFACET

Si 02

CURRENTDISTRIBUTION

RECOMBINATIONREGION

F1(3

Stripe -geometry injection laser has slit -shaped emitting region.15 to 20 times as wide as it is high

Characteristics ofinjection lasers

For the applications of present interest, aninjection laser that can be run (quasi-)continuously without cryogenic cooling isrequired. Fig. I is a schematic sketch of astate-of-the-art injection laser structure forsuch cw operation. Those used in this studyare planar -stripe. double-heterojunction(Al Cia)A.s lasers that have demonstratedgood reliability.' In these devices, theeffective laser width is defined by thenarrow stripe contact that confines thecurrent to a small part of the total junctionarea. In the lasers used for the presentexperiments, the stripe width is - 13 pm.and the laser length is - 500 pm, as definedby cleaved end (reflecting) surfaces.Current -spreading between the stripe andopposing contact leads to an emissionwidth several pm wider than the metallicstripe. Although the electron -holerecombination region is only 0.2-0.5 pmthick, the light is emitted from a region 0.7-1.0 mm high surrounding the junction planeat the laser facet. Thus, the emitting regionis slit -shaped, some 15-20 times as wide as itis high.

I he narrow height of the emitting region.on the order of a wavelength, forces thelaser to oscillate in a single transverse mode(in the direction perpendicular to the junc-tion plane). The beam thus emerges, in thisdirection, as a diffraction -limited beamwhose full width at half -power is about40''. Laterally, the active region is manywavelengths wide, and several lateralmodes can be sustained. Therefore, the far -field beam pattern in the plane of thejunction often shows some structure and

Reprint RE -22-4-4This article appeared in Applied Optics. June 1976. insimilar torn

29

RELATIVE INTENSITY

-80 -60 -40 -20 0 20 40 60BEAM ANGLE, DEGREES

80

Fig. 2Far -field pattern is not circular, and so requires additional opticsto produce a circular spot.

Arthur H. Firester has been involved withelectro-optics for some time-his doctoraldissertation was on the modulation of lightby optically pumped alkalai-metal vapors.His recent research has been in nonlinearoptical phenomena and their possibleapplication to image processing; he is alsoworking with coherent -light opticalproblems and holography.

Joseph D. Knox has worked extensively inthe design and development of opticalscanning systems. In addition to the onedescribed here, these systems have beenused for laser deflection displays, light valvedisplays and IC mask and wafer inspection.Dr. Knox's other activities include thedesign and fabrication of acousto-opticdeflectors, modulators, and cavity dumpersfor visible and infrared lasers.

represents the emission in several lateralmodes, in a beam pattern perhaps 10° inwidth. This behavior is illustrated in Fig. 2.

Because the laser source region is thuselongated, rather than circular in cross-section, simple imaging of the source onto atarget plane produces a similarly elongatedimage spot. While one can envision systemsin which such a spot shape is desirable,most applications require a round spot.Focusing the laser beam into such a spotrequires anamorphic optics. One simplesystem, to he described in more detail inconnection with the flying -spot documentscanner, consists of a microscope objectivelens, placed near the emitting laser facet.

Peter V. Goedertier was the co -discover ofthe He-Ne "cascade" laser and a pioneer inthe development of the cross -pumpedYAG:Nd:Cr laser. His other research pro-jects have ranged from ion physics and earlygaseous and solid-state optical masers toelectro-optics and the development ofvarious optical systems.

The biographies of Istvan Gorog, James Wittke, and Ivan Ladany appear with their other articles in this issue.

Authors Firester, Gorog, Goedertier, Wittke, Knox, and LadanyContact them at. Systems Research Laboratory, RCA Laboratories, Princeton, N.J. Ext. 3202.

30

40

35

30

25

20

15

10

0 100 200 300 400 500

DRIVE CURRENT (mA)

Fig. 3Past threshold, output power of injectionlaser is a linear function of drive current, somodulating drive current will modulate out-put power.

that images the facet onto the target. This isfollowed by a "telescope," made from twocylindrical lenses, that provides differentmagnifications in the plane of the cylinderlens axes and the perpendicular plane. Inthis way, the elongated source can beimaged into the symmetric, round spotusually required.

Another important laser characteristic isthe available optical power. The lasersoperate in an incoherent mode below thethreshold drive current of about 300 mA.Above threshold, the output power is alinear function of the drive current, asshown in Fig. 3. This permits the outputpower to be simply modulated bymodulating the drive current. Lasers

operate at 1.5-2.0V across the junctionat currents between threshold and perhaps100 mA above threshold, giving outputpowers of 10 mW or more. Thus, theoverall laser efficiency is in the 1-5% range.Typical gas lasers have efficiencies less than0.1% and require a high voltage to maintainthe discharge, so the advantage of using aninjection laser when feasible is obvious.

The noise characteristics of the laser outputare of great significance for anyinformation -handling application. Thesewere evaluated for both unmodulated andpure -tone -modulated lasers by detectingthe laser output with a fast photomultiplier

and scanning the photomultiplier outputwith an rf spectrum analyzer. In the 0-100MHz spectral region tested, the measurednoise was within a few dB of the valuepredicted from the shot noise formula:

(4,2) = 2e/GB.

Here (i,,2) is the average mean -squarecurrent fluctuation at the photodetectoroutput, e is the electronic charge, / is thephotocurrent at the detector output, G isthe detector gain, and B is the analyzerbandwidth. For example, when the laser

RF INPUT

DC BIF.SLEAC

output was 5 mW and the photodetectorwas arranged to produce 5-µA outputcurrent with a gain of 2.5 X 10,6 the

measured noise level into 50 fl in a 3 kHzbandwidth was -102 dBm.

Finally, since a laser is a threshold device,to maintain constant output power, thedrive current above threshold must be heldstable. Since the threshold current of aninjection laser is temperature -dependent,'either the laser temperature must bestabilized or an optical feedback loop onthe drive current must be used to prevent

THERMOELECTICCOOLER

THERMISTORLEADS

Fig 4Developmental cw injection laser package uses thermistor thermoelectric cooler system tostabilize laser temperature

3

output drifts with changing ambient con-ditions. We have stabilized the lasertemperature using a thermistor sensor in acontrol loop that includes a small ther-moelectric cooler. This system can providea constant laser power output with aconstant injection current. Fig. 4 shows acomplete injection -laser package.

High -densityinformation retrieval

High -density optical storage systems canhe designed in a number of ways. Onemethod records the information on aplanar surface in the form of small pits ordepressions. The pit dimensions can be onthe order of I aim, comparable to thewavelength of the readout light used. Thespot power requirements are determined bythe desired readout rate, signal-to-noiseratio, type of detector used, and theefficiency of the optics. Fora spot power onthe order of 0.1 mW, one expects about IpW of detected signal power; with a state-of-the-art, shot -noise -limited avalanchedetector, this would permit digital informa-tion to he read out at 250 Mb/ sec, with biterror rates Lc. 10 III.

If the transverse mode structure of the( A 1 GalAs laser resembled an axiallysymmetric Gaussian spot, most of theoutput could he focused into a spot whosediameter would be approximately one-halfof a wavelength, or approximately 4200 A.

GaAsLASER

As discussed above, the injection laser'sthin -junction geometry makes the outputhighly asymmetric; it also can exhibit somestructure (multi -lateral mode). Since it wastherefore difficult to predict the laser'sactual focusing properties analytically, aseries of knife-edge scan experiments wereperformed to obtain quantitative data.

Fig. 5 shows the experimental arrange-ment. The injection laser output iscollected and shaped by an anamorphiclens system that consists of a low -powermicroscope objective and a two -cylindrical -lens telescope. the low -powerobjective acts as a collimator and thecylindrical lenses expand the beam in theplane containing the junction. A second,high -quality, high -power microscope ob-jective forms the finely focused spot in theplane of a moveable knife edge. A retro-reflector rigidly coupled to the movingknife edge forms the measuring arm of a

Michelson -type interferometer that uses a

He-Ne laser operating at 6328 A as thesource. The interferometer provides aprecise calibration for the knife-edge dis-placements. The light reflected by the knifeedge is collected by the spot -forming lensand projected through a beam splitter ontoa photodetector. The photodetector outputis thus the spatial step -function response ofthe beam; differentiating the knife-edgeresponse then gives the line -spread func-tion (spot profile) easily. The spotdiameters deduced from these

DIFFERENT IATORV2

PHOTODETECTOR

CYLINDRICALLENSES

.1/4F 94mmf=17mrr

MICROSCOPEOBJECTIVE

F-MICROSCOPEOBJECTIVE

ELECTRO-MECHANICALSCANNER

"KNIFE"EDGE

BEAM-SPLITTER

SYNC

PHOTODETECTOR

HeNe LASER

Fig 5

Experimental system determined laser's focusing properties by measuringmicrometer -sized spots.

Table ISpot size data for GaAIAs cw injection laser.Output power was 6.6 mW and focused spotpower 0.6 mW, giving overall efficiencyabout equal to the He-Ne system.

Lens magnification Spot sizenumerical aperture width at half intensity)

X 50 0.115

X 40 0.65

x 20 0.40

.64 Acm

1.03 Alm

1.7 µm

measurements are listed in Table I. In therange studied, the spot power was a cons-tant 0.6 mW; the optical insertion lossbetween the laser and the focused spot wasapproximately 10 dB. Because of the highinjection laser efficiency, the overallefficiency, defined as the power in thefocused spot divided by the laser outputpower, is about the same for an injectionlaser as it is for a He-Ne system.

Flying -spot scanner system

A second category of systems that canprofitably use injection -laser sources in-cludes flying -spot document scanners. Insuch systems, the object to be scanned isheld stationary, (at least in one dimension,)and the light beam is moved across theobject. The resolution requirement fordocument scanning is in the vicinity of 250Aim. This corresponds to a scanner resolu-tion of approximately 100 lines per inch,which is the highest resolution demandedby a human viewer from an I8 -inch viewingdistance.

Fig. 6 shows an experimental flying -spotscanner. The injection laser output firstpasses through a low -power microscopeobjective (10X) and then through a cylin-drical beam expander. Ideally, themicroscope objective is so chosen that itboth produces the required magnificationand has its entrance aperture filled by thediffraction -limited wide beam emitted inthe plane perpendicular to the junction.This plane is arbitrarily chosen to coincidewith the horizontal scan plane; its onlyactive imaging element is the low -powerobjective that magnifies the thin dimensiona few hundred times. In order to minimizeaberrations and facilitate focus adjust-ment, another spherical lens could be usedin conjunction with the microscope ob-jective. In the plane of the junction, which

was chosen to coincide with the plane of thevertical scan, the laser beam is imaged bythe combination of the microscope ob-jective and a cylindrical beam expander.One may understand the operation of thisarrangement by recalling that in the junc-tion plane the effective source size is

approximately 20 times larger than in theplane perpendicular to it and, therefore,the required magnification is approximate-ly 20 times smaller. Our design philosophyis based on the idea that we wish to operatethe scanner in both the vertical andhorizontal directions in a diffraction -limited mode because this results in bothminimum lens and mirror aperture sizesand a minimum deflector tilt angle.

The line -spread function obtained with theabove arrangement is shown in Fig. 7. Thisfigure was obtained by scanning the spotacross a narrow slit mounted in the scanplane. The spot profile shows reasonablecylindrical symmetry and has a full widthat half maximum of approximately 100lam. The optical efficiency of the system,defined as the spot power in the scan planedivided by the laser output, was againapproximately 10(;. A number of ex-perimental scanners were constructed, us-ing nominally similar lasers. The efficiencyfor all of these units was about the same;the spot size varied from 100 to 200 Aim.

Our flying -spot scanner' typically scanneda standard document (WA X 11 in.) from adistance of I meter. Its vertical resolutionwas limited by the number of scan lines to1000 lines per picture height. The horizon-tal response can he computed from the spotprofile data. With reference to an 8/ -inch -wide image, the horizontal modulationtransfer function, as calculated from thedata of Fig. 7, is shown in Fig. 8. Ourresults suggest that very high quality (fewhundred lines per inch) image transmitterscan be constructed using injection -laserscanners.

A recently developed application area forflying -spot laser scanners is reading codedproduct information in point -of -sale(POS) systems. The resolution require-ment in POS applications is about the sameas for document readers (about 200 Aim).hut the POS system has the unique re-quirements that the depth of focus shouldhe several inches and that adequate meansof ambient -light discrimination must beprovided. The depth of focus for a

diffraction -limited system is completelydetermined by the spot size and the

VERTICAL SCANGALVANOMETER

FOCUSINGOPTICS; HORIZONTAL SCAN

-.IEWHI+< GALVANOMETER..--:-(CW INJECTIONLASER

-PHOTO DETECTOR OBJECT

BEINGELECTRONICDRIVE a SYNC SIGNAL

PROCESSORSCANNED

SIGNALS

FACSIMILE SIGNAL OUT

I_

STORAGE BUFFER HIGH RESOLUTION[LASEllPRINTER SCOPE TORAGE CRT

Fig 6Scanner systems produced resolutions over 100 lines per inch, and implied that very -high -quality scans of a few hundred lines per inch could be produced.

RELATIVEINTENSITY

HORIZONTAL SCAN

-- VERTICAL SCAN

/ 0.6 \

-100 0 100

DISTANCE IN SCAN PLANE ,

Fig. 7Line -spread function shows reasonable cylindrical symmetry

20 40 60 80 100 120 140

SPATIAL FREQUENCY , LINE-PAIRS/cm

Fig 8Horizontal modulation transfer function

33

IN FOCUS 1 INCH 2 INCHESf H 9Spot profiles obtained by moving the spot out of the nominal scan plane showed that .the scanner had a satisfactory depth of focus.

,.% Jo, elength. Let us assume that the fielddistribution in the focal plane is ap-proximately Gaussian and denote the spotradius where intensity drops to I/ e- timesits peak by w. Then one can show' that theradius of the I/ e- intensity point willincrease to 2' 'w for an axial focus error of

where A is the light wavelength. For A =8200 A and = 100 z =- 15.3 cm. Fig. 9shows a set of spot profiles obtained byfocusing an approximately 200 -pm -radiusspot in various planes in the vicinity of thenominal scan plane. We found that anoperating range in excess of 4 inches isavailable without significantly increasingthe beam diameter.

Ambient -light discrimination

In principle, narrowband interferencefilters can effectively discriminate betweenthe spectrally -narrow laser output and thebroadband ambient background.However, the actual acceptable bandwidthof the required interference filter is oftendetermined, not by the laser linewidth, butby the scan angle. As the angle of incidencevaries, the peak transmission of an in-terference filter also varies. For example, ifa photodetector is located I meter from astandard -size document, the angle -

dependent detuning is approximately 100A from the center of the document to its topor bottom. This implies that, for thisgeometry, in order to achieve reasonableshading (i.e., an incidence -angle -independent detected signal) interferencefilters with a few hundred angstroms half -

power bandwidth would be required. Suchbroadband filters provide very poor dis-crimination against typical ambient lightlevels.

We have achieved very effective dis-crimination against unwanted ambientlight by using a modulated laser in conjunc-tion with synchronous detection of thereturn signal. Fig. 10 is a block diagram ofan experimental synchronous flying -spotscanner system designed to read standard -size documents at a nominal rate of 12s/ page, with 1000 scan lines resolution.The nominal bandwidth was 50 kHz (12ms/ line, 500 cycles/ line, 2 ms flybacktime). The injection laser was modulated at1.2 MHz, and the modulation signal wasalso used as the reference signal in thesynchronous detector. The output currentfrom the optical detector is fed to the

LASERI 2 MHz

DRIVER -SCANNER

OPTICSUNFILTERED

PHOTODETECTOR' I I 11" 111 OUTPUT

--{7 -

DETECTOR

12 MHz

FILTER

SYNCHRONOUS1 DETECTOR

THRESHOLD

DETECTOR

SIGNALOUT

Fig. 10Synchronous detector arrangementdiscriminates against ambient light.

TUNED -FILTEROUTPUT

SYNCHRONOUSDETECTOR

OUTPUT

THRESHOLDDETECTOR

OUTPUT

Light ambientFig 1 1

Oscilloscope traces for scan of black bar on white background.

Dark ambient

34

synchronous detector, which provides thevideo signal for images with gray scale. Fortwo -level images, additional discrimina-tion can be obtained by feeding the syn-chronous detector's output through athreshold detector.

The performance of the synchronousflying -spot scanner is indicated by theoscilloscope traces shown in Fig. II. Thetest pattern used for preparing these traceswas a wide black bar on a whitebackground. Note the difference in thesignal-to-noise ratios of the tuned filter andsynchronous detector outputs between thetraces taken in a darkened environmentand those taken in a normal ambient; it isprimarily caused by the added shot noisegenerated by the ambient light incident onthe photodetector.

For a given signal level at the detector(white object), the range of ambientsagainst which the synchronous detector isable to discriminate depends primarily ontwo factors: the object contrast and thedynamic range of the photodetector. If theobject contrast is very low, the noise addedby even very low levels of ambient light willcause the threshold detector to misfire.Objects of primary interest, however, tendto he of high contrast. For high -contrastobjects the dynamic range is limited in twoways: I) at sufficiently high ambient levelsthe photodetector saturates; and 2) theadded shot -noise level is so high that noreliable threshold signal level is available.In either case the threshold detector willmisfire.

[he most meaningful way to quantitativelydescribe the discriminator's dynamic rangeis to compare the flying -spot scanner'sphotodetector output current produced ina dark environment to one produced in anambient illumination high enough to causeobjectionable misfirings of the thresholddetector. With our standard black -bar testpattern we found that the photocurrentfrom the ambient illumination could ex-ceed approximately 600 times the currentdue to the detected laser radiation beforethe misfirings obliterated the signal.

Fig. 12 shows a sample image reproducedwith the aid of the injection -laser scannerdescribed above. When the original black -on -white two -level image was scanned, thethreshold -detected video signal was fed to astorage scope; the stored display is

photographically reproduced here. (Theprimary resolution limiting element in thisreproduction cycle is the storage scope.)

C,C II II I

Re.,

C?C2SX UNCIIII CHASSIS COMPONINT LOCATION 01,013t

o

0 .0 C134L1139-1115M -

PR.Cj

22 crop') atZPPI1C

2 ,

C105 _ - 17:

0RI

/p. m°

1, P

1914184

741

M130 I

CI27

1 "0 ..c:ierm r3 5

C136 --5110-'-

0

717132

471419 - - -

A126 -

71 129

I/102 - -R V101 -

A105 --8176

507102

0111175 -T101 - --VOGL/5R109

t)

C LiikC'32

107vio0

17544

IRMO

Rilss cloNka

c 439e137

011151

SG7037

C5-

ci!.!, 00,

pm; It

cia'D

ttO911/14

- metSC TO 2M12.

SG701

09,15

--- 71%3- Ryes

...... R113

V103CII4

- CI29RIM

. - - R190-- C1211

R10611

- 56101

0130w10?0136

- ROMAca101II1V101

Del". .4.3 C1 IA L. mho Litassor c-ris'...W Loraridm Goo*

A

_9

A

---C19

-11

T

Fig. 12Sample image made with injection -laser scanner system. Storage scope used limited theresolution of the reproduction.

Conclusions

Our experiments indicate that state-of-the-art, room -temperature cw injection lasersare suitable sources for opticalinformation -retrieval systems. As a conse-quence of the highly asymmetric angulardistribution of the laser output our ex-perimental spot -forming optics had ap-proximately 10% throughput efficiency.Nevertheless, because the injection -laser'sefficiency can be greater than 1%, theoverall optical -system efficiency comparesfavorably with systems using He-Ne lasersources.

Acknowledgments

The authors thank Henry Kressel fornumerous valuable discussions and for

suggesting a number of usefulmodifications to the original manuscript.Also, they acknowledge the assistance ofDavid Patterson and William Mitchell forthe circuits used in the laser temperaturestabiliser.

References

I I adany. I.: and Kressel. H.: "1 he influence of devicefabrication parameters on the gradual degradation ofIA 14ia11s co laser dunks." 4p/d. Plit'A. telt. Vol. 5(19741

P.2 Kressel. H.: Lockwood. H.1-.: I.adan). I.. and I tienherg. M.:"Heir:mum:non laser diodes for room temperature opera-tion:" Ort bog.. Vol. 13 (1974) p. 416.

1 Cohen. 14.W.: and I. g. I.: "Visual eapaeit an imagequalitr, descriptor lor dopla% es aluation.- Prin. SID Vol. 15

4 (iir9re7d4cI P 3.riier5.i'.V.: 6orog. I.: Knox. J.D.: Lauan). I.: andWittke IP.: "A lacsimile .}stem using room -temperatureInirchor laser scanning.' H( 1 Rerieu :V 4,1.15 (19741 p. 335.

S. lam. A.: Owns..... It I,, 111,1111 S. John Wile! and Sons. Inc..New Nod, 1967. p. 224.

35

Non

imag

ing

Imag

ing

LIG

HT

--1-

10 SIGNAL0

Pho

toco

nduc

tor,

in w

hich

ligh

tde

tect

ion

prod

uces

a tr

ansf

er o

f

min

ority

carr

iers

with

inso

lid

mat

eria

l, e.

g., t

his

p -n

junc

tion.

PH

OT

OC

AT

HO

DE

LIGHT

/DY

NO

DE

S

ICJ

PH

OT

OE

LEC

TR

ON

S

AN

C/D

E

SIG

NA

L

Pho

toem

itter

,in

whi

chlig

htde

tect

ion

prod

uces

the

emis

sion

of e

lect

rons

into

vac

uum

, as

inth

isph

otom

ultip

lier

tube

.S

ucce

ssiv

e se

cond

ary

emis

sion

1-al

ong

the

serie

sof

dyn

odes

Q)

prod

uces

am

plifi

catio

n.4-

14-

+

LIG

HT

SIG

NA

LO

UT

PU

T

ELE

CT

RO

NB

EA

M

PH

OT

OC

ON

DU

CT

IVE

TA

RG

ET

Vid

icon

use

s an

arr

ay o

f p -

n ju

nctio

ns a

s th

e ph

otoc

ondu

ctiv

e el

emen

t;sc

anni

ng b

eam

rea

ds o

ut th

e im

age

sequ

entia

lly.

LIG

HT

_p.

PH

OT

OC

AT

HO

DE

PH

OS

PH

OR

1-6.

LIG

HT

-. O

UT

PH

OT

OE

LEC

TR

ON

S

Imag

e tu

bes

conv

ert t

he o

ptic

al im

age

to a

n el

ectr

on -

emis

sion

pat

tern

by

mea

ns o

f the

ir ph

otoc

atho

des.

PH

OT

OC

AT

HO

DE

SIG

NA

LO

UT

PU

T

SIT

tube

com

bine

s a

phot

ocat

hode

det

ecto

r an

d S

i tar

get

to p

rodu

ce a

nea

rly id

eal d

evic

e.

Fig

. 1Li

ght d

etec

tors

can

be

divi

ded

into

two

clas

ses-

phot

ocon

duct

ors

and

phot

oem

itter

s-w

ith b

oth

imag

ing

and

noni

mag

ing

type

s w

ithin

eac

h cl

ass.

LIG

HT

PH

OT

OC

AT

HO

DE

PH

OT

O-

ELE

CT

RO

NS

GA

TE

ELE

CT

RO

DE

S

tt

Cha

rge

-cou

pled

dev

ice

uses

Si d

etec

tors

, sto

r-in

g th

e ch

arge

in p

oten

tial w

ells

und

er M

OS

gate

s. R

eado

ut is

via

a c

harg

e -t

rans

fer

proc

ess-

cloc

k vo

ltage

s ap

plie

d to

the

gate

s m

ove

the

char

ge fr

om w

ell t

o w

ell.

TA

RG

ET

ELE

CT

RO

NR

EA

M

FIR

ST

DY

NO

DE

OF

ELE

CT

RO

NM

ULT

IPLI

ER

ELE

CT

RO

NM

ULT

IPLI

ER

Imag

e or

thic

on tu

be u

ses

seco

ndar

y em

issi

on a

t a th

inta

rget

to e

nhan

ce th

e ph

otoe

lect

ron

imag

e.

LIG

HT

PH

OT

OC

AT

HO

OE

PH

OT

OE

LEC

TR

ON

S

TA

RG

ET FIE

LDM

ES

H

FIR

ST

DY

NO

DE

OF

ELE

CT

RO

NM

ULT

IPLI

ER

ELE

CT

RO

NM

ULT

IPLI

ER

LRE

FLE

CT

ED

RE

TU

RN

BE

AM

SC

AT

TE

RE

DR

ET

UR

N B

EA

MS

IGN

AL

OU

TP

UT

Imag

e is

ocon

is a

mod

ified

imag

e or

thoc

on th

at e

limin

ates

muc

h of

the

elec

tron

-sc

an -

beam

noi

se b

y us

ing

a re

turn

-be

am s

epar

ator

. With

the

sepa

rato

r, o

nly

thos

e el

ectr

ons

that

hav

e ac

tual

ly b

een

scat

tere

d af

ter

hitti

ng th

e ta

rget

get

ampl

ified

.

an introductior

Light -detecting devices:

E.D. Savoye Two different classes of light -detecting devices,photoconductors and photoemitters, are approaching tneideal detector from different directions.

The ideal light -detecting device mustsatisfy two basic criteria:

1) Each incoming photon must con-tribute to the detected signal, i.e, thedevice should have unity quantumefficiency (Q.E.).2) The detector or detector/amplifiercombination must have sufficientsensitivity to produce a useful outputsignal for a single -photon input.

The first of these criteria is important inorder to preserve the signal-to-noise ratio(SI N) of the photon signal; the second isrequired (in addition) for detecting andimaging low -light -level signals, which mayhave only a single photon event per resolu-tion element per integration time.

Practical light -detecting devices have ap-proached the ideal in two different con-figurations, which define the two broadclasses of photodetectors:

I) photoconductors, in which photo -excitation causes a transfer of minoritycarriers within a solid material, e.g.across a p -n junction; and2)photoemitters, in which photo -excitation produces the emission of elec-trons into vacuum.

In order to produce a broad range ofpractical devices in both classes, trade-offshave been made in a number of areas,including choice of materials, physicalconfiguration, and operating conditions.

The principal advantage of usingphotoconductors lies in their high quan-tum efficiency, provided by the fact thatinternal collection of photoexcited carrierscan be highly efficient when there is no

energy barrier against their collection.Their principal disadvantage is at low lightlevels, where the signals are too small to bedetected above the system noise.

In contrast, photoemissive detectors arerelatively limited in quantum efficiencybecause of the energy barrier against theemission of electrons at the surface of thedetector material. However, they have theadvantage that photoelectrons, onceemitted, can readily be accelerated invacuum to produce essentially noise -freegain by secondary emission or by impact -ionization in a solid material. Devicesbased on photoemitters are therebycapable of detecting single -photoelectronevents.

The principal practical devices that haveevolved based on these two classes ofphotodetectors are summarized in Fig. I,and discussed in the following pages.

Dick Savoye is presently Manager ofEngineering for Electro-optics Products atRCA Lancaster. Dr. Savoye began workingon photoemissive devices for RCA in 1966,and received the RCA Laboratories Out-standing Achievement Award in 1968 for hiswork on tunneling photoemitters. He hassupervised advanced development work onSi vidicons and CCDs, negative electronaffinity photoemitters and secondaryemitters, and the SIT tube.

Contact him at:Electro-optics Products EngineeringSolid State DivisionLancaster. Pa.Ext. 2020

Reprint RE -22-4-5Final manuscript received October 15 197E

37

"Tea -cup" photomultiplier is shorter and more efficient than the "venetian -blind- PMT it isexpected to replace. The three-inch tube shown is the forerunner of a family of tubesdesigned for scintillation counting in medical electronics equipment.

Nonimaging detectors

Nonimaging photoconductive detectors arebased upon the p -n junction.

The photoconductive detector, representedhere by a single p -n junction, also exists inmany other configurations, such as p-i-njunctions, avalanche detectors, simplephotoconductor materials with ap-propriate electrical contacts, Schottkybarrier detectors, etc. In general, however,incident light photoexcites carriers withinthe device to produce the output signal. Inthe case of the avalanche detector, internalgain greatly increases the low -light -levelcapability of the device.

Photoconductive devices of various typesare being developed and manufactured atRCA's Montreal plant and, with theproliferation of various electro-opticaldevices, have quadrupled their sales in thepast two years. Two typical end uses aremilitary optical -tracking equipment, usinga quadrant p-i-n detector, and a range-finder that uses both a YAG laser and anavalanche detector with a risetime of lessthan I ns. Other military applicationsinclude proximity fuses and weapon -fire -simulators. This last equipment uses a laserbeam as the "ammunition" and detectorslocated at critical positions on militarytargets as hit indicators.

A very promising consumer application liesin automatic ranging for photographiccameras. In these systems an ir-emittingdiode is the source and a p-i-n detectorsenses the time it takes for the reflectedradiation to return from the subject to thecamera. This information is used to controlthe camera focus. One may also anticipatea large demand for nonimaging detectorswith the expected explosive growth in thewhole field of optical communications.

Nonimaging photoemissive detectors arerepresented by the photomultiplier tube,which has very high noise -free gain.

The photomultiplier tube (PMT)represents this category. In this type oftube, electrons emitted from thephotocathode are accelerated to the firstdynode, where they land with several hun-dred eV energy, which is sufficient to giverise to several (4-7) secondary electronsemitted per incident primary. The secon-daries in turn are acclerated to the seconddynode, where further multiplication oc-curs. Using about 10 such stages can giveessentially noise -free gain on the order of10", sufficient for detecting a single photo-electron from the cathode.

PMTs are important in a broad range ofapplications, including astronomy, laserdetection, spectroscopy, and medical elec-tronics. RCA has been particularlysuccessful in dominating the latter field;our annual sales of 3 -inch PMTs in gammacameras exceed $3M. These gamma -rayimaging cameras are now to be found inalmost all major hospitals. A typicalcamera uses 37 PMTs, set up in a hex-agonal cluster arrangement, to sense thegamma -ray scintillations coming fromradioactive isotopes introduced into thepatient.

Imaging detectors

Imaging photoconductive detectors storethe optical image as a charge pattern.

This class is represented by the vidicon andthe charge -coupled device (CCD). In bothdevices, an optical image is detected andstored as a charge pattern within thephotoconductive element, which is anarray of p -n junctions for the vidicon. TheCCD uses Si as the detector, storingcharges under MOS gates. In each case the

Vidicon tube finds use in broadcast, in-dustrial, and surveillance applications,depending upon the quality level of the tube.

image is read out sequentially by a scanningprocess-an electron beam for the vidiconand the charge -transfer process for theCCD.

Today, the largest appliction of vidicons isin surveillance, both industrial andmilitary. Particular vidicon types are alsoused by the military in tv-guided missiles.Broadcast tv uses the PbO vidicon for"live" tv; film -pickup for tv reproductionuses the Sb2S3 vidicon.

An important application for the CCD isanticipated in tv journalism, where por-tability is very important. The world-widemarket in camera tubes is approximately$100M annually, of which RCA's share ispresently about 10%.

Imaging photoemissive detectors areexcellent for low -light -level operation.

In these devices the optical image is con-verted to an electron -emission pattern bythe photocathode. In the image tube, theemitted electrons are accelerated to strike aphosphor screen, giving rise to a light -emission pattern that corresponds tothe original optical image. Image tubes canbe made with fiber-optic input and outputfaceplates. This construction permits series

38

CCD camera's small sizevaluable in tv journalism.

and light weight

coupling of the image tubes with minimumresolution loss and electron gain figures of30-50, measured from photocathode tophotocathode. Three such stages canprovide enough gain so that a single photo-electron emitted from the inputphotocathode produces a visible output atthe final phosphor screen. This makesoperation at very low light levels possible.

Image tubes were developed primarily withgovernment sponsorship for use in militarynight operations. Initially, they used anauxilliary infrared illuminator-theoriginal "snooperscope" and"sniperscope." The ready detectability ofthe infrared source, however, led to thedevelopment of "passive" viewers, whichproduce images with the bare minimum ofnatural light available even on moonlesscloudy nights. Many police departmentshave taken advantage of the capability ofthese devices for nighttime surveillance.Today's uses, however, have expanded toinclude many scientific and commercialapplications. For example, magneticallyfocused image tubes, because of theirminimum of image distortion, have provedto be very important in augmenting thephotographic plate in astronomy. Since theimage tube has an effective quantumefficiency about 10 times that of thephotographic plate, its use is equivalent toincreasing the telescope's diameter by afactor of three. Also, combined with a tv-camera tube, the image tube can aid inreproducing the very dim fluorescent im-ages found in x-ray inspection systems.

The image orthicon was the quality pick-updevice during the early years of television.

(1 kg without lens) and ruggedness make it

In these tubes, the photoelectron image isenhanced by secondary emission at a thintarget and the charge pattern is scanned offby an electron beam. The return -beamsignal is amplified by further secondaryemission, but noise in the electron beamprevents the image orthicon fromproviding useful signals at very low lightlevels. A modified image orthicon, calledthe image isocon, eliminates much of theelectron -scan -beam noise by a novelreturn -beam separator that permits theamplification of only those return -beamelectrons that have actually been scatteredafter hitting the target. The image isoconthus approaches an ideal device, but it isstill limited by target secondary -emissionnoise and electron -beam separation noise.

The recently developed SIT (silicon -intensifier target) tube is very close to anideal device. The SIT tube combines thephotocathode detector with a Si-vidicontype of target. Photoelectrons areaccelerated to strike the Si target at severalkeV, giving rise to an impact -ionizationgain of several thousand within the target.The device thus functions at low lightlevels, and, when coupled to an image tubeto produce the I -SIT configuration,provides a distinguishable output signal fora single -photoelectron input.

In either the SIT or I -SIT configuration,the overall gain of the system can be variedelectronically by means of the acceleratingpotential, making both devices useful overa broad range of light levels. SIT tubes,because of this adaptability to differentlight levels, are becoming more and moreuseful in industrial surveillance. Such a

.a

SIT tubes operate at light levels rearly downto photoelectron noise limit

typical industrial application is in parkinglots, where it is then not necessary toprovide a high level of illumination.

Approaching the ideal

The potential of realizing the idealphotodetector, combining high quantumefficiency with detecting single -photonevents, is highly promising. This ideal isbeing approached from two majordirections:

I )CCD detectors can, in principle,operate at very low light levels. Withfurther development it is possible that wemay produce a cooled, buried -channel,distributed -gate on -chip amplifier devicethat would be capable of providing adistinguishable output for a single -

photon input.2) Photoemissive detectors can alreadydetect single -photoelectron events, andthe so-called "negative electron affinity"devices can provide revolutionary ad-vances in photoemissive quantumefficiency. In fact, laboratory ex-periments have demonstrated quantumefficiencies near those of photoconduc-tors.

With further development it is possible thatone or both of these methods may closelyapproach the performance of an idealphotodetector.

39

The silicon -target vidicon

R.G. Neuhauser

The concept of using a large-scale array ofdiodes as a mosaic photosensitive target fora camera tube was first proposed at BellLaboratories in the late 1960's. Rapidexpansions in silicon technology and large-scale integrated -circuit technology even-tually produced a vidicon tube with anarray of more than 600,000 individualdiodes on a silicon wafer. In this design, thesilicon generates charge carriers from thelight of an optical image focused on thewafer; the diodes store the carriers untilthey are scanned out as video information.

Robert Neuhauser has been employed intelevision camera tube development,manufacturing, and application work since1959, when he joined RCA in Lancaster. Heis a fellow of the Society of Motion Pictureand Television Engineers and received theSociety's 1964 David Sarnoff Award. He isalso the author of 19 technical articles ontelevision camera tubes.Contact him at:Application EngineeringElectro-Optics and DevicesSolid State DivisionLancaster, Pa.Ext. 2223

With 1800 diodes per inch on a silicon wafer, this vidicon isreplacing the standard vidicon in many applications and alsoperforming in completely new situations where previouscamera tubes were useless.

The tube's advantages

The silicon -diode vidicon is rugged and itssensitivity is excellent. The silicon detectorhas a quantum efficiency (charge carriersdetected per incident photon) of nearly 100percent in the visible portion of the spec-trum. This sensitivity extends in the in-frared to nearly 1200 nanometers, and alsowell into the ultraviolet. In addition, thetube is immune to any image burn -in fromintense light, even an unattenuated imageof the sun.

The most recent development with this typeof tube is a reduced -blooming target thatgreatly reduces the highlight -bloomingcharacteristic of the silicon -target tubes.(Blooming is a lateral spreading of thehighlight image that occurs when the signalamplitude is limited by either completedischarge of the target or by the use ofinsufficient beam current to handle thesignal that is being developed.) Continuingimprovements in manufacturing efficiencyand techniques have reduced the numberand prominence of blemishes resultingfrom faulty diodes and have also produceda lower -cost tube. The mechanical strengthof the tube has been increased by mountingthe target as a part of the faceplateassembly. All of these features expand theusefulness of the silicon -target tube inexisting applications and extend its use tonew areas.

The silicon target is now used in both theoriginal I -inch -diameter tubes and insmaller 2/3 -inch -diameter tubes. Withsome modification it can also be used insilicon -intensifier -type (SIT) tubes.'

The electronic mechanism

At first glance, the silicon -target vidiconlooks disarmingly like the more familiar

Reprint RE -22-4-7Final manuscript received June 17, 1976. charge -isolation mechanism in the silicon -

camera tube, the antimony trisulfidevidicon. The same bulb, basing, and elec-tron gun is used for both types; only thetarget is different. The target, of course, iswhere most of the action is. In any vidiconthe target has two distinct functions: I) toconvert a pattern of light and dark (theimage) into an electrical charge pattern,and 2) to accumulate and store this patternuntil the electron beam and scanningmechanism get around to reading out thesignal.

The conventional vidicon uses a continuous)hotoconductive coating as its target.

Fig. I illustrates the conventional photo-conductive vidicon target, a soot -like filmof antimony trisulfide evaporated over a

transparent conductive coating on the in-ner surface of the faceplate glass. The targetis a capacitor that has been charged byconnecting the scanned surface to groundpotential by means of a moving electronbeam. At each point of illumination, thelight produces electrical conduction, whichpartially discharges the capacitance in thatarea. When the scanning electron beamreaches this discharged area, it depositselectrons. This recharging current flowsthrough the target circuit and is used as theoutput signal. The thinness of thephotoconductive layer and its highresistance prevents any significant lateralleakage of the individual image charges.

The silicon target is an array of discretediodes.

The silicon target, Fig. 2, is a separatestructure, independent of the faceplate.The beam side of the silicon wafer containsan array of diodes that performs thestorage function; light detection takesplace throughout the n region. Silicon ishighly absorbent in the visible spectrum, somost of the incident light is absorbed anddoes generate charge carriers.

However, as illustrated in Fig. 3, the

40

target is different from that in the con-ventional target. Although silicon is highlyconductive, the diodes, as will be seen, arereverse -biased. They are thus like littleislands that store electrons left by the beamas it charges them negatively to thepotential of the cathode of the electron gun.These charges on the diodes are preventedfrom wandering about by the insulatingeffect of the reverse -biased junction.

The remainder of the target action of thesilicon -target vidicon is better understoodby looking at Fig. 4, which shows a smallsection of the target with its diode cellcharged; an energy -level plot takenthrough the section is at the top of thefigure. Somewhere in the bulk of thesilicon, an incoming photon of light yieldsits energy to the release of a pair of chargecarriers; the negative electron moves to theleft. The left side of the target has beendoped to facilitate the capture of electrons.The positive hole moves toward the right,"falls" up into the electron -rich p -region,and reduces the stored free -electron inven-tory there by one. Only when the scanningelectron beam recharges the diodes doescurrent flow through the Ercircuit (Fig. 3).This recharging current constitutes thevideo signal.

Silicon -target construction

Starting with a wafer of n -type silicon, andthen using techniques familiar to thetransistor and integrated -circuit manufac-turer, the wafer is oxidized, treated withphotoresist, exposed to a pattern of dots,and then etched, leaving openings in theoxide corresponding to the dots. A p -typedopant is then introduced through theseopenings into the underlying silicon; theresult is the diode array. By leaving theoxide on to cover the area between diodecenters, the scanning beam is preventedfrom landing outside of the diode and thusfinding a short-circuit path through thesilicon.

There are two methods of shielding theoxide from the scanning beam.

It is also important, however, to keep theoxide from being exposed to the beam, forif it is, it will accept and store electroncharge negatively to the point where it willrepel any further incoming beam; i.e., eventhe openings into the diode centers will bepinched -off from receiving any more elec-trons. To prevent this, a pattern of raisedconducting "beam -landing pads" have

LIGHT INPUT

GLASSFACEPLATE

I ET

SIGNAL PLATE - TRANSPARENTCONDUCT IVE COATING

SCANNING BEAM

PHOTOCONDUCTOR -EVAPORATED FILM

VIDEOOUTPUT

Fig. 1Conventional vidicon target consists of a photoconductive film evaporated onto the glassfaceplate. Target acts as a capacitor that is charged by connecting it to ground through thescanning beam. Lighted areas conduct, and so partially discharge the capacitance in thatarea, so the scanning beam will deposit electrons there. This recharging current is used asthe tube's output signal.

LIGHT INPUT

GLASSFACEPLATE

SCANNING BEAM

-T-

VIDEOOUTPUT

Fig. 2Silicon target is independent of faceplate. Beam side of the wafer consists of an array ofdiodes that are discharged by the presence of light. Although the structure (see Figs. 3 and4) is different from the vidicon, the beam's recharging current is similarly used as the videooutput.

PHOTOGENERATEDCHARGE CARRIERS

LIGHTINPUT

MONOCRYSTAL LINESILICON WAFER"N".TYPE(THICKNESS.10- 20mm)

= ET

o + 102 Ii Mn'

C 4 nF/c/n2

BEAMLi/RIDING PADS(METALLIC)

INSJLATION - SILICON OXIDE

DIODE CELLS FORMED BYINTRODUCING RTYPE DOPANT

(DIODE SPACING ft 14 ionl

Fig. 3Reverse -biased silicon diodes act as islands that store electrons left by the scanning beamas it charges them to the negative potential of the electron -gun cathode. Highly conductivebeam -landing pads shield oxide iayer from electron beam.

LIGHT ORELECTRON IMAGE

S V

Si SCHEMATIC DIAGRAM

ELECTRONSCANNINGBEAM

f-vsne- Rama RESISTANCE

TARGET CAPACITANCE

Fig. 4Each photon striking the target produces a pair of charge carriers in the bulk of the siliconlayer. The electron is attracted by the positively doped left-hand side of the target, while thepositive hc le moves into the p region and reduces the stored free -electron inventory there byone. The target capacitance stores the electron charge there until the scanning beamreplaces this electron and is read out as the video signal when it does so. 41

El Mil,WA ME!

MU Al IN

EWAN191.1311111111

Fig. 5Scanned side of the target in an SEMphotograph. Diodes are circles, beam -landing pads are squares, and the oxidelayer behind them is background. Diodesare on 0.014 -mm centers; patterned struc-ture must isolate diodes from each other, yetshield the oxide layer from the scanningbeam as completely as possible.

been added to cover all but a small portionof the exposed oxide (Fig. 3). All of this isin a pattern that repeats at approximately1800 to the inch!

Fig. 5 is a scanning -electron -microscopephotograph of the scanned side of a portionof the completed RCA silicon -diode target.The precision fabrication of these padsgives nearly complete shielding of thesilicon dioxide from the scanning beam, yetstill isolates each diode pad. An alternativemeans of protecting the oxide insulator,preventing it from charging, is shown inFig. 6. Some manufacturers use thismethod, in which the oxide and diodestructure is overlaid with a "somewhatconducting" layer called a "resistive sea,"instead of individual highly conductivebeam -landing pads. It is apparent that thisoverlayer interconnects the diodes, which,

DIODE CELLS FORMEDBY INTRODUCINGP -TYPE DOPANT

MONOCRYSTAL LINESILICON WAFER"N" -TYPE

up to now, had been isolated, and that thelayer must have better conductivity thanthe oxide or it will be useless.

The resistive -sea technique is advantageousfrom a manufacturing standpoint in that itis not a pattern structure, as are the landingpads, so it does not need to be registeredwith the diode pattern. Being un-patterned, the resistive sea does not con-tribute discrete defective areas (spotblemishes) of its own and even tends toblend or smooth out some of the less -profound spot blemishes in the underlyingpattern. However, using a resistive searesults in a loss of resolution, and withhigh -contrast images, the lateral leakagepath appears to promote prematureblooming.

Silicon -target characteristicsCapacitance stores the image charge.

The preceding sections have described thetarget's first function-converting a

light pattern into an electrical pattern. Thesecond function, pattern storage, is ac-complished in the silicon target by storingthe image charge with the capacitance ofthe diodes. The diode storage capacitanceis the combination of the capacitance to thesilicon through the depletion zones and thecapacitance of the diode pads to the siliconacross the silicon oxide insulator (Fig. 4).The storage capacitance is approximately4500 pF in the 1 -inch tube and 2200 pF inthe 2/3 -inch tubes. These values aresomewhat dependent on the target voltage;the depletion regions contract at lowervoltages, with a resultant increase incapacitance.

Dark current is typically about 10 nA.

Dark current, the signal developed in theabsence of light, has several components:

"RESISTIVE SEA"COATING

INSULATION -SILICON OXIDE

Fig. 6Resistive -sea construction is another method of isolating diodes and shielding the oxidelayer. A relatively easy -to -manufacture "somewhat conductive" coating performs the task,but lowers resolution because of diode interconnection.

the reverse -bias diode leakage, thermal -carrier generation in then material, surfaceinjection from the diode perimeter, andconductivity through the oxide insulator.Dark current is temperature -dependent,and doubles for approximately each 9°Cincrease in temperature. At 30°C, a typicalI -inch tube operates with a dark current ofbetween 7 and 12 nA.

Cosmetic characteristics vary from tube totube.

The problem of constructing a silicon waferincorporating 600,000 similarly function-ing diodes of relatively consistent andstable operating characteristics is no easytask. The level of blemishes has been and isbeing improved, and tubes are sold withvarying grades of freedom fromblemishes. Blemishes remain very stablewith tube operating life as long as the targetvoltage remains unchanged.

There are several typical varieties ofblemishes; they are caused by differentfactors. "White spots" appear under non -illuminated conditions and are also visiblewhen the target is illuminated. Two majorcauses of white spots are shorted diodesand "unscheduled" holes in the siliconoxide. "Black spots" appear only when thetarget is illuminated, and may be black orblack and white. Larger black spots oftenhave a white halo around a black core. Themajor causes of black spots are inoperativediodes, missing diodes, connected pads,and particles on the target. Larger, low -contrast defects, such as streaks andsmudges, can also be present. These may becaused by silicon -crystal defects, scratches,or contamination.

A low-level crosshatch pattern may also bediscernible under some conditions; thiseffect is only noticeable when the tube isoperated at very low signal currents andhigh amplifier gain. Shading or variationsin sensitivity may be observed under cer-tain conditions; this effect is onlynoticeable when longer wavelength light isused. At these wavelengths, the siliconabsorbs only part of the light, and thesensitivity variations represent thicknessvariations in the silicon wafer.

The tube has an optimum target voltage.

The silicon -target vidicon tube uses thesame gun design as a standard vidicon.Voltages are similar, except that thesilicon -target tube has limited wall -meshand focus -electrode voltage ratings and alow target voltage. The target voltage is

42

fixed and must not be changed. Evenmometary increases in target voltage dur-ing tube operation will increase the ap-parent size and contrast of spots and willaffect beam landing by charging the Si02insulator. The optimum target voltage forthe RCA silicon -target tube is 8V, for thefollowing reasons.

The upper limit of target voltage is

determined by excessive dark current andreverse -bias diode breakdown. The lowerlimit is determined by ineffective collectionof the generated charge carriers, high lag,the inability of the tube to handlereasonable levels of signal current, andlower resolution. Fig. 7 illustrates the darkcurrent and the charge -carrier collectionefficiency in terms of signal output, both asa function of the target voltage.

High dark current is generally undesirable,and should be held to a minimum. Ex-cessive diode leakage can appear as whitespots when the target voltage reaches 10 to12 V; hence, the voltage should not exceed8 to 10 V. The volt-ampere signal outputcurve of Fig. 7 shows that the carriercollection efficiency increases rapidly to amaximum at about 4 V. At this value, thevoltage generated at the depletion region issubstantially greater than the energy of thegenerated carriers, and most of them arecollected. No more "sensitivity" will beachieved at higher target voltages. Lag alsodecreases slightly at higher target voltagesas the larger depletion depth reduces theeffective capacitance.

If the target voltage is set too low, thecharge that can be stored in the targetcapacitance will be insufficient to producea high, noise -free signal current. At 8 V, thesignal current reaches a limit at about 1100nA; this voltage has been chosen toproduce an optimum tradeoff among thevarious factors to be considered.

Performance characteristics showadvantages and disadvantages.

The most obvious advantages of a silicon -target vidicon are its sensitivity and itsresistance to any image burn -in or damageresulting from overexposure. Directlyfocusing the tube on the sun for severalminutes, for example, will not permanentlydamage the target, and it is virtuallyimpossible to produce any type of retainedimage burn -in on this type of tube if it isoperated at the correct fixed target voltage.

The sensitivity approaches unity quantum

SIGNALOLTPUT -nA

250

200

150

100

50

I_ IOUTPUT

ILLUMINATION

ISIGNALCONSTANT

WITH / 50

40

0 30

JP20

0PQ10

00 4 8 12 16 20 24 28 32 36

TARGET VOLTAGE -V

DARKCURRENT -nA

Fig. 7Target voltage affects dark current and signal output. Optimum operating po nt is 8 V.

yield (one charge carrier detected per inci-dent photon) throughout the entire visiblespectrum and into the infrared, and ex-ceeds that of any other vidicon-typephotoconductor in total sensitivity.However, this wide spectral range ofsensitivity can be a liability as well as anasset. It is sometimes necessary to restrictthe light entering the tube to either thevisible or the infrared. This is becauselenses are not corrected for both the visibleand infrared spectrum, and the lens

coatings that are commonly used to reducethe internal lens reflections in the visibleportion of the spectrum are not effective inthe infrared. In fact, they begin to act asmirrors for infrared light. Internalreflections in the optical system then causelack of contrast and flare aroundhighlighted areas of the picture. To obtainmaximum resolution, it is desirable toexclude light of those wavelengths forwhich the lens is not corrected. However,the tube's infrared sensitivity can be used todetect hot objects or to view infrared -illuminated scenes where it is impossible orundesirable to use visible illumination.

Table ISensitivities of photoconductive cameratubes. Values are microamperes per lumenof 2856 K tungsten light.

Totalsensitivity

Infraredexcluded

Silicon target vidicon 4350 1100

PbO 450 450

PbO (extended red) 700 550

Vidicon 250* 225

*Medium sensitivityoperation

The high sensitivity might also appear to bedesirable in a color camera, where lack ofsensitivity is usually a major limitation.Table I compares the sensitivities of thevarious vidicon-type camera tubes inmicroamperes per lumen of 2856 Ktungsten light. An examination of thespectral sensitivity curves of Fig. 8 showsthat the silicon -target tube's sensitivity inthe blue portion of the spectrum is nearlyidentical to the sensitivity of the lead -oxidephotoconductor of the Vistacon andPlumbicon2 tubes used in most color-tvcameras.

Since the blue channel is usually thelimiting factor in either sensitivity (noise)or lag in color cameras, the silicon targetdoes not promise to increase the overallsensitivity of color television cameras byany significant amount. A major increasein sensitivity can be achieved, though, inthe red channel of a color camera, wherethe signal can be increased by a factor of 4above the lead oxide tube's or a factor of 2above the extended -red lead -oxide tube's.One version of the silicon -target tube with

400

350

300

250

°g X°150

100

50

0

Pb0

SI L CONVIDICON

EXTENDEDRED Pb0

1 14igh *IOC4

400 800 800 1000WAVELENGTH - nm

Fig. 8Spectral response of silicon -target tube isquite close to standard tubes' in blue region,but is such higher in red region.

43

very tight blemish specifications is beingused in the red channel of a well-knownthree -tube color camera, where its excellentred spectral response and high signal levelproduce noticeable improvements in noiselevel and red colorimetry.

"Blooming" is a problem being conquered.

The signal amplitude from a I -inch silicon -target tube is limited at about 1100 nA.This limiting occurs when the capacitanceis fully discharged, and is an asset, since itprevents the intense highlights fromdeveloping a very high signal spike thatcould overload the camera's video -amplifier system. This limiting action alsoproduces a second effect called "bloom-ing," which is a lateral spreading of thehighlight image. It occurs when the signalamplitude is limited by either completedischarge of the target or by the use ofinsufficient beam current to handle thesignal being developed on the target.

Blooming is caused by the relatively longlifetime and mobility of the hole carriersand the lack of trapping states in thesilicon. Since carriers are free to movelaterally to adjacent negatively chargeddiodes when the field of the diodes underthe illumination is reduced to zero, excesshole carriers do so when the highlightillumination completely discharges thetarget at such a highlighted point. Becauseof their excellent mobility and relativelylong lifetime, these carriers will not"disappear" or recombine with free elec-trons, but will be accounted for elsewhereon the target. Therefore, instead of produc-ing a high -amplitude signal at the point ofhighlight illumination, the carriers spreadlaterally and produce a large white blobthat is completely devoid of picture detail.

A new line of reduced -blooming targetsand tubes has been developed thatpromises to eventually replace the standardtubes. The comparison photos of Fig. 9show how the new tube reduces bloomingin a practial situation.

Light -sensitivity control is via a lens-iris/filter system.

The silicon -target tube, along with the lead -oxide tube, has a fixed sensitivity. Unlikethe Sb2S3 vidicon, its sensitivity cannot bereduced or varied by changing the targetvoltage or the voltage on any other elec-trode on the tube. Changes in illuminationlevel must be accommodated by changing

Fig. 9Improved picture of a reduced -blooming silicon -target tube (left) compared with picturefrom a conventional silicon -target tube (right).

the quantity of light impinging on the tubeby using a lens iris or appropriate light-attentuating filters. Either method presentssome operational problems in general-purpose surveillance -camera use. The tubeis sensitive enough to produce an accep-table picture with as little as 0.1 footcandleof illumination with an f/ 1.4 lens (fulldaylight illumination is 5 X 10' foot-candles).

If the silicon -target tubes are to be usefulover this entire 50,000:1 illuminationrange, the iris of a conventional lenssystem that can only handle an illumina-tion range of 250:1 must be supplementedwith filters capable of handling anadditional illumination range of 200:1. Toaccommodate this wide range of illumina-tion, a series of lenses that employ a gradedneutral -density dot in the center of the lenshas been developed by various manufac-turers. This further attenuates the lightlevel at high f numbers and providesautomatic operation with scene illumina-tion levels varying over a 10,000:1 range.

RESIDUALSIGNAL1%0F PEAK)

30

25

20

15

10

5

0

33

Lag is lower than for many other cameratubes.

Lag characteristics determine the ability ofa tube to capture motion. Carriers have avery long lifetime in silicon because thereare very few areas where they can betrapped, i.e., there are very few sites whererecombination of carriers can take place.However, the lifetime is much shorter thanthe television frame time. Since there isvery little charge -trapping taking place inthe target, the conduction mechanismstarts and stops very quickly as the lightchanges at any point on the target.Therefore any lag effects are causedprimarily by the storage capacitance of thetarget and the beam characteristics.

Referring to the equivalent circuit of Fig. 4,the beam resistance approaches infinity asthe charge voltage on the diode pad reaches0 V during the beam read-out period. Thishigh -RC series circuit produces a

noticeable delay in the signal read-out atlow signal levels. (The effect is negligible

!.. 4,,

%C.

,'oo00(

,L4

/A

.-..,,4

,S>

co4

sio,, ii ..'"1.L

''''...........

4,4

n_as00

50 66 83 100 116 133 150 1/2sec.

MILLISECONDS AFTER LIGHT REMOVAL

VIDICON

-1SiPb0

Fig. 10Lag is nearly undetectable in an Si tube operating with a 400-nA signal current. The tube'shigh sensitivity and finite dark current make this possible.

44

when large signals are being generated.)Fig. 10 shows the comparative lag

characteristics of the various types of tubes.The "early" part of the curve results in lossof "dynamic" resolution; the latter part ofthe curves illustrate the relative amount of"smear" of a moving object. Operating atthe higher signal currents obviouslydecreases the lag.

Since the silicon -target tube has muchgreater sensitivity than many other cameratubes, the lag for a given scene illuminationcan be much lower. In fact, it is nearlyundetectable when signal currents of 500nA are used. The finite dark current of thesilicon target also contributes to lower lagbecause it provides a bias or fixed dark -signal level that is well above zero -signallevel. Thus, it is easier for the beam torecharge the storage capacitance in darkareas of the picture.

The capacitance of the silicon target in a 1 -inch tube is approximately 4500 pF. (Bycomparison, the 30 -millimeter lead -oxidetube has a capacitance of 1500 pF, and theI -inch antimony trisulphide vidicon, 1200pF.) A small silicon -target tube will have areduced capacitance in direct proportion tothe reduction in scan area.

Resolution is presently lower than forstandard vidicons

The resolution in the I -inch silicon tube islimited partially by the number of diodesand partially by the electron -optics of thebeam -forming and deflecting systems. The1829 diodes per inch on the RCA targetcorresponds to 700 diodes per pictureheight in the I -inch tube, which limits theresolution of the tube to approximately 700

190 OM. SW7,, Mt60 MIL

i' 50 MUM3 wriq .immi1

2/3" L tomuno! 30 Si TARGE laNIE20 VIDICON GUN VOLTAGESW

10 VOLTAGES(500 V MAX.)

ION 1500 V MAX.)

EENIMILO GUN

1" VIDICON (Sb2 S3)HIGH GUN VOLTAGES1700.900 v)

1" VIDICON (Sb2 S3)LOW GUN VOLTAGES(300-500 VI

200 400 600

TV LINE NUMBER

Fig. 11Amplitude -response curve shows resolu-tion that is presently lower than for standardvidicons.

tv lines per picture height. Resolution,however, is less in the small -sized tube.

Resolution is best defined by a modulation -transfer -function curve or, as it is common-

ly called, an amplitude -response curve. Atypical response curve for a I -inch tube isshown in Fig. I I, along with the responsesfor an antimony trisulphide vidicon and a2/3 -inch silicon tube. The silicon -tubeamplitude response could be increasedsomewhat with higher focusing fields andelectron -gun voltages. (Note the improve-ment generally achieved in Sb2S3 vidiconsat higher voltages.) However, at the presentstate of the art, either X-rays or gas ionsgenerated in the tube at these highervoltages strike the silicon target and causethe dark current to rise rapidly and per-manently. At a wall mesh voltage of 700volts or more, the dark current will rise tounacceptable levels in a few hundred hours.At a maximum mesh electrode rating of500 volts, the predicted life of silicontargets is on the order of 50,000 hours.

Camera designconsiderationsThe 8-V operating target voltage places aspecial requirement on the deflecting andfocusing coil system. These coils should bedesigned so that the beam -landing errorwill be less than 0.5 V. This limit is

necessary to assure that the full targetvoltage will be applied uniformly over theentire target area. The dark current of asilicon target is reasonably low, in theneighborhood of 10 nA at 30°C; thiscurrent is only 3% of a typical 300-nAsignal and does change with temperature.Because, as described above, target -voltageadjustment cannot be used as a means toreduce dark current, the camera should bedesigned to keep the temperature of thetube from ranging too widely, a conditionthat could lead to unacceptably high darkcurrent.

The silicon -target tube is a linear device,and signal output up to the signal -limitation point is directly proportional tothe amount of light input. This conditionproduces an accurate signal that is an exactanalog of the light intensity at any point onthe image. Lead -oxide tubes have a similarlinear characteristic. However, the signalfrom a linear tube must be processed byblack stretching, or "gamma correction," ifproper tonal values are desired on thereproducing picture tube. This action, ofcourse, stretches the blacks and enhancesthe noticeability of any lag.

Quality levels and tube lifeIt is a continual challenge to manufacture asilicon tube that is completely free ofdefects. Since defects are caused primarilyby shorted, inoperative, or bridged diodes,available silicon tubes are classified ac-cording to the blemish defect level, andprice varies inversely as the number ofdefects. At present, few silicon -target tubescan be made that approach the qualityrequired for tv broadcast use in the colorchannel producing the bulk of the picturedetail and brightness signal.

The life of a silicon -target tube operatingwithin the maximum voltage ratings is

primarily determined by the life of theelectron gun, which is in the neighborhoodof 15,000 hours. Sensitivity, lag, and allother important performance character-istics remain constant through life, and noresidual -image signals accumulate withlife. Blemishes do not grow during life andonly occasionally will a diode shortproduce a white spot.

Typical uses

Silicon -target tubes are replacing standardvidicons in many areas.

The list of uses for the silicon -target vidicon;s growing constantly. One group ofapplications involves upgrading theperformance, especially sensitivity, of ex-isting general-purpose tv systems. Propertyprotection is a good example of an applica-tion in which existing camera installationsare being retrofitted with the silicon tube.The results of the retrofit are quitedramatic, especially where the lighting isproduced by a red -rich tungsten source.

The fact that the target is not destroyedwhen it is focused on the sun (or otherintense source) has been the basis for yetother retrofits. Of course, if the tubereceives too much light, the picture washesout completely. But, the tube has theimportant feature that, as soon as theoverlighting transient is removed, thesilicon vidicon is back in normal operation.

The similarity of the silicon vidicon's gam-ma characteristic to that of the Plumbicon2and Vistacon lead -oxide tubes has led toanother color -camera design improve-ment. The problem has been with the reddeficiency of lead oxide in color cameras.Silicon tubes, with red sensitivity to spare,are used in the red channels of thesecameras with the lead -oxide tubes remain-ing in the green and blue sockets.

45

Fig. 12inspection application uses silicon vidicon's ir-sensitivity to "see" through silicon IC wafer.

The tubes are also finding uses in entirelynew areas.

The secondary category of uses includesnew, novel and/ or previously impracticalforms of television cameras. For example:

A silicon -target vidicon camera coupled toa microscope can "see" through materialsthat are opaque to the eye-if the materialstransmit infrared. One such material is thesilicon wafer itself used in making thecamera tube or in manufacturing transistorand integrated -circuit devices (Fig. 12).

The behavior of gallium -arsenide light -emitting diodes operating at their typical9000 A can be monitored in real time by theuse of the silicon -target vidicon.

Several companies are working with tv-style spectrometers that aid in pollutioncontrol by scanning the spectra producedin chemical phosphorescence or duringrapid oxidation (flame spectra), Fig. 13.The scanning rates are extremely fast whencompared with mechanical -scanning anddata -recording schemes, and the spec-trometer outputs are well suited to os-cilloscope display and computer process-ing.

Extensions to performanceand future prospects

Silicon -target tubes with very good ul-traviolet response can be made Theirresponse to ultraviolet light seems to belimited primarily by the optical transmis-sion characteristics of the tube's faceplateand secondarily by the surface

characteristics of the silicon target.(Strongly absorbed short -wavelength lightgenerates carriers in the silicon surfacelayers that are unavoidably n' type, andcarrier recombination takes place rathereasily in this region. This effect begins tolimit the effective sensitivity of the tube atthe shorter wavelengths.) Reasonably goodsensitivity out to wavelengths of 200 nmhas been measured on developmentalmodels.

The infrared sensitivity of the target can beincreased by increasing its thickness andhence the absorption of the more weaklyabsorbed longer wavelengths. This increasein thickness is accompanied by some loss inresolution, as the carriers diffuse laterallybefore falling under the influence of thefield generated by the depletion regionssurrounding the diode sites. Special tubeswith enhanced ultraviolet and infraredsensitivity are currently available on acustom basis, awaiting the development ofa viable market and a definition of generalperformance specifications that the tubeswould have to meet in new and uniquesituations.

The silicon -target vidicon is replacing theconventional vidicon in many applications,but more importantly, by virtue of its verywide spectral response and other uniquecharacteristics, it is performing in com-pletely new situations where previouscamera tubes were useless. Because thetarget relies on silicon technology and thefabrication techniques of large-scaleintegrated -circuit technology, its futuredevelopment will be advanced and aided bythe continuing progress in these alliedfields.

Acknowledgments

Much credit must be given to P.R. Rule,G.A. Robinson, and the late P.D. Huston,who assisted in preparing the lecture on thesilicon -target tube that is the basis of thispaper. The assistance of F. Wallace and R.Phillips in interpreting the performancecharacteristics of the target and the reducedblooming tube is gratefully acknowledged.

References

Fig. 13 2.

Rapid -scanning tv-style spectrometer usessilicon target tube as its detector. Thesecurves measure phosphorescence delay;other applications are with flame spectra. 4.

(Photo courtesy Tektronix.)

3.

Robinson, G.A.; -The silicon intensifier target (SIT) tube, -this issue.

Plumbicon is a registered trademark of N.V. Phillips.

Neuhauser, R.G.; and Miller, L.D.; -Beam landing error andsignal output uniformity of vidicons; J. SMITE, (Mar 19511)p. 149.

Rodgers, R.L., Ill; "Beam -scanned silicon targets for cameratubes,' ST -4936, RCA Solid State Division, Somerville, NJ.

46

The silicon intensifier target tube: seeing in the darkG.A. Robinson

PHOTOCATHODE

LENS

SCENE

FIBER OPTICFACEPLATE

The SIT tube's excellent low -light -level characteristics haveled to applications ranging from police surveillance tointernal inspection of jet engine parts.

SCANNING SECTION

FIELD MESH GUN FOCUSING ACCELERATING CONTROL(GRID No.41 GRID (GRID No.31 GRID (GRID No 21 GRID (GRID No.11

IMAGE FOCUSGRIDS

IMAGE SECTION

ANODE

HIGH VELOCITY PHOTOELECTRONBEAM IMAGED ON TARGET

ELECTRONGUN

SILICON TARGET

LOW VELOCITYSCANNING BEAM

CATHODE

Fig 1Light enters the SIT tube through a fiber-optic faceplate, which transfers the flat -scene image onto the curvedphotocathode. The light then travels through the focusing grids and strikes the target, wh ch is a matrix of over 1800 silicondiodes per inch. The image is typically stored there and read out by the scanning beam every 1/30 second.

George Robinson has contributed to thedesign of flying -spot kinescopes. 1.5 -inchmagnetic, 1 -inch hybrid and all -electrostatic vidicons since joining RCA in1955. His specialities include electron -gunwork, high -resolution vidicons, and cameratubes for military systems and unusualenvironments. In his present position he isinvolvec with applications at low light levels.

Contact him at:Applications EngineeringEiectro-Optics ProductsSolid State DivisionLancaster, Pa.Ext. 2073

The idea of using a silicon -diode -arraytarget in an intensifier tube (SIT) hasreceived wide acceptance in the last fewyears in cameras operating at low lightlevels. This paper discusses some of thecharacteristics of the SIT tube that make itthe leading low -light -level camera tube inuse today. Elsewhere is this issue. R.G.Neuhauser' discusses how the silicon targetis used in vidicon tubes.

Photocathode +silicon -diode arrayThe SIT tube (Fig. I) is a photodetectorusing a photocathode as the light-sensitivesurface and a silicon -diode -array target asthe surface upon which an image chargepattern can be stored. The SIT tube's gaincomes from the high number of hole -electron pairs generated in the target whenit is bombarded by high-energy electronsfrom the photocathode. In the silicon-vidicon target some fraction of the in-coming photons creates a hole -electronpair. but 'n the silicon target of the SIT tubeeach photoelectron can create many hole -electron pairs. A tube operating with 9000volts across the intensifier section will havean electron gain of about 1600.

The relationship of gain to intensifiervoltage is shown in Fig. 2. By introducingan energy -absorbing "buffer layer" in front

of the target, the useful gain characteristicis kept above 3000 volts, where the in-tensifier section performs best. The bufferlayer selectively absorbs photoelectrons. Ifall photoelectrons had equal energy. the

z3

INTENSIFIER VOLTAGE- KILOVOLTS

Fig. 2Electron gain (solid line) is an essentiallylineal function of the intensifier voltage. Anenergy -absorbing "buffer layer" keeps thegain above 3 kV, where the intensifier sec-tion performs best. Deviation from linearity(the dashed line) is caused by high-energyphotons penetrating the buffer layer.

Reprint RE -22-4-6Final manuscript received October 22. 1976

47

gain would be expected to continue itslinear decrease as voltage is reduced until acutoff voltage is reached. But as Fig. 2shows, some gain exists below the pro-jected cutoff voltage, suggesting that somehigher -energy electrons are getting throughthe buffer layer.

An SIT tube with a gain of 1600 and a

photocathode responsivity of 140 µA/ Imwill be approximately 50 times moresensitive to tungsten illumination than asilicon -target vidicon with a publishedsensitivity of 4350 µA / Im.

Photocathode

The photocathode surface is the basic S-20multi -alkali (Na-K-Cs-Sb), and is placedon the inside of a curved fiber-opticfaceplate. Since fiber-optic plates havepoor ultraviolet transmission, the tuberesponse is low to the shorter wavelengths.The 340-nanometer cutoff can be seen in

100

2

10

2

300 400 500 600 700

WAVELENGTH- NANOMETERS

900 900

Fig. 3Photocathode response is limited by the lowuv transmission of the fiber-optic faceplate.(This can be corrected by using a uv-sensitive scintillator, however.) Tube alsodoes not have the near-ir response of thevidicon.

the response curve of Fig. 3. Some usersovercome this deficiency by applying anultraviolet -sensitive scintillator to the out-side of the faceplate.

The response to the longer wavelengths islimited by tube processing of the S-20photocathode; the extended -red (ERMA)photocathode is not presently obtainable inthe SIT tube. Because the silicon target isbombarded by electrons rather thanphotons. the basic near-ir response of thetarget. which is shown in the vidicon, islost.

Image section

The image section of the SIT tube invertsthe photoelectron image and focuses itonto the silicon target. The electron opticsof this process requires a sphericalphotocathode surface, but a conventionallens system focuses the image of a sceneonto a flat plane. Therefore, it is convenientto have a fiber-optic faceplate transmit theflat -scene image onto the curvedphotocathode. The faceplate, which isthicker at its edge than at its center,introduces a fixed "shading" signal into aflat -light field because of the transmissiondifference across the faceplate. Thisshading accounts for a drop in signaloutput of approximately 17% at the cor-ners. Also, geometric distortion increasesby about 2% as a result of the transitionfrom a plane image to a spherical surface.This distortion is displayed as "pin-cushion" effect on the camera monitor.

Resolution

The matrix structure of the silicon targetlimits the resolution performance of theSIT tube. The existing target has a densityof over 1800 diodes per inch, or about 35line pairs per millimeter of resolution. In a

nominal I6 -mm optical image (I/ 2 X 3/8in.) this condition will result in a limitingresolution of about 700 tv lines per pictureheight. Higher resolution is obtained byusing a larger -sized target rather than atarget of higher diode density. A 27 -mmtarget with a limiting -resolution capabilityin excess of 1000 tv lines per picture heightis currently being used in 1.5-in.-bulb-dia.tubes.

The contrast transfer function (CTF), orsquare -wave amplitude responsecharacteristic, can be useful in determiningresolution performance. The completecurve shown in Fig. 4 can be obtained in the

laboratory by viewing high -contrastpatterns of parallel white and black bars.This curve shows the relationship of outputsignal to bar width. The signal reductionassociated with small images is significantin that the signal-to-noise ratio is directlyaffected and is important in determiningthe low -light limitation of operation. Thislimitation may be reached at light levelshigher than expected because actual scenesare not made up of parallel black and whitebars. The effect upon amplitude responsewhen looking at points rather than bars canhe approximated by squaring the CTFcharacteristic curve, Fig. 4.

Resolution also degrades as light level islowered because lower output signals affectthe signal-to-noise ratio. Fig. 5 shows therelationship of light level and limitingresolution; the curve was made using astatic scene consisting of black and whitebars. Two different contrast levels areshown: the 100% level is typical of alaboratory -type evaluation, while the 30%level corresponds more closely to typicaloutdoor scenes.

Lag

Resolution is only one of the importantcharacteristics in low -light -level operation.A second very important characteristic islag, which becomes worse as the signal leveldecreases. Lag is the residual signalmeasured in the dark and is expressed asthe percentage of the original signal presentafter three fields of scanning in the dark.

SIT tubes exhibit no photoconductive lag,but there is some capacitative lag resultingfrom the finite time it takes for the electronbeam to remove accumulated charge fromthe target. A target with high capacitancewill store relatively more charge with lessvoltage change than will a low -capacitancetarget (C = dQ/ d However, it will take alonger time for the beam to discharge thesignal because of the electron velocitydistribution within the beam. It is for thisreason that lag increases as light level (andcharge) decreases. Fig. 6 shows typical"third -field" lag for a 4804 tube as afunction of light level.

It is possible to improve lag by artificallyraising the "zero -signal" voltage so that thebeam electrons can discharge the targetmore effectively. This can be done either bysimulating an increased dark current (usingbias lighting) or by actually increasing thedark current (increasing target voltage orraising target temperature).

48

100>

O

a.

a

-

a

0o-

0,

W 00

Za

o 20

W

a

a

aU. 600CC

80

TYPICAL

0 (CIF?

200 400 600

TV LINES PER PICTURE HEIGHT

800

Fig 4

Response to black -and -white parallel -bar pattern depends onbar width. Curve is called the contrast tranfer function (CTF);squaring it gives the effect upon amplitude response when thetube sees points instead of bars.

aTUBEDARK

TYPECURRENT

, 4804:7 NANOAMPERES

6

1-za 4

ccWO.

4 2-I0JW_

I:

...s.

NNNN10

0IC aiI .-

6

4 \N.

!era6 6

FACEPLATE ILLUMINANCE LUMENS/ FOOT 2Fig. 6Lag is a measure of the residual signal present afterscanning the field in the dark. Curve shows lag for astandard 4804 tube after scanning three fields in thedark.

Dark current

Dark current manifests itself as a

background to the image. This backgroundis a result of thermally developed carriers inthe target and is usually not a problemexcept for a low-level grain or mottling thatis visible only at extremely low signal levels.It is normally desirable to keep the darkcurrent low so that the dark portion of thescene will have a voltage reference nearzero. There may be circumstances when thedark current will be increased to com-promise performance, such as to improvelag, as discussed previously. If this is doneby increasing target voltage above thenominal 8 V, then the maximum signal

700

600

S00

400

300

200

100

0

.1

/Al6 8 2 4 68

106 1064 6 80.4 2 4

2856 6 FACEPLATE ILLUMINANCE-LUMENS/FOOT,

810 3

Fig. 5Low light levels lower resolution -100% and 30%levels correspond to laboratory and outdoorenvironments.

10

s.aT. 6

4oZ4Z 4

ZaCC

0UY 2CC

40

1

TUBE TYPE: 4804TEMPERATURE 30°C

4.45"

2 610

TARGET VOLTAGE - VOLTS

Fig. 7Dark current is caused by thermally -developed carriers in thetarget, and is seen as a background to the image. Dark current is afunction of both target voltage and temperature.

capability of the tube will be increased aswell. Normally this will be of no concern tothe user, because the tube generally hassufficient signal -handling capability.However, whenever the target voltage isincreased. the maximum value should notexceed 15 V because of the increasedprominence of target defects. Fig. 7 showsthe typical relationship of dark current totarget voltage. This curve is for a 30°Coperating temperature; dark current in-creases with increased operatingtemperature.

Operation at low light levelsThe SIT tube is basically a low -light -leveldevice. Even though it has a variable -gain

feature, as shown in Fig. 2, around -the -clock operation should not be attemptedwithout auxiliary light -control capability.Rather than operate the tube at highillumination for long periods, it is better toincorporate the variable -gain feature of thetube into the camera design as a fast -actingautomatic light control (ALC). As anexample, the nominal operating voltagemight be chosen as about 4500 volts, wherethe gain is about 200 (3 lens stops awayfrom full gain). At this operating point thetube will have a good signal-to-noise ratio,the picture will be pleasing, and the tubewill have long life. The voltage (gain) can bevaried from this operating point to handlelight levels changing up or down by asmuch as a factor of 8 in less than a second.

49

Then, at a slower rate, lens -iris adjustmentand or filter insertion can bring the lightlevel back to the value that will allow 4500 -volt operation.

Tube life

When caution is exercised in using the SITtube, reduced thermionic cathode emissionwill end its life, as in other camera tubes. Acalculated mean -time -to -failure ( MTTF)in excess of 5500 hours (to 90 -percentconfidence) has been maintainedthroughout the last three years of produc-tion. Faceplate exposure must be con-trolled for long life. By followingrecommended camera design con-siderations, the operating time before theonset of damage will be 2000 hr. Protectionmust he provided against two types ofdamage mechanisms: target damage fromhigh-energy electron bombardment andphotocathode damage from ion bombard-ment.

Target damage

In addition to time and illumination, theenergy of photoelectrons that bombard thetarget determines the extent of targetdamage. Excessive exposure will cause apermanent increase of dark current at thepoint of impact on the target. If the damagebecomes severe, it will be evident in thedark portions of a scene or even in totaldarkness with the high voltage completelyremoved, Fig. 8. Damage can be related tosignal level on the target and can becontrolled by keeping the signal withinhounds.

Damage is most apt to occur where small,intense sources are present. However, in

1111111k

_

111C -

Fig. 8Severe target bum caused by excessiveexposure. Damage is evident here even withthe high voltage removed.

such an application it is usually necessaryto obtain information from the darkbackground surrounding the small source,so it is not practical to reduce gain andsignal level just to protect the target fromthe exposure of the small area. If prolongedoperation is necessary under these con-ditions, it is best to move the camera so thatthe small area does not remain in one spot.

Target damage is most likely to occur in anunattended camera. In an attended andcorrectly operating camera, the overex-posure of a small spot will usually bloom to

an unusable degree before a damage levelis reached. If the over -illuminated areabecomes large and takes up a significantportion of the picture area, the chance forphotocathode damage increases.

Photocathode damage

Ion damage to the photocathode results ina poorly defined dark spot in the center ofthe picture, Fig. 9. This dark spot is

actually an area of reduced photocathodesensitivity and cannot be seen in the dark orwith the high voltage removed. Thedamage results from the bombardment ofpositive ions, originating from collisionsbetween photoelectrons and residual gasmolecules, that are accelerated toward thenegative potential of the photocathode. Innormal operation the number of photoelec-trons is never high enough to generate adamaging number of ions; it is only whenthe photocathode current becomes ex-cessive that the number of ions reaches adamaging level.

Ion generation is a function of numbers ofphotoelectrons rather than energy. It is

possible to have such a small voltage (aslow as 100 volts) on the image section that,

Fig. 9Photocathode ion damage produces a poor-ly defined dark spot in the center of theimage.

although no picture is present, photoelec-trons are flowing as a result of lightexposure. Therefore, the only sure methodsof photocathode protection are to com-pletely remove all photocathode voltage orto limit the light level.

S/N at low light levels

The degradation of resolution at low lightlevels is closely related to signal-to-noiseratio. It is desirable to have a camerasystem that is so quiet and a tube sosensitive that performance will be limitedonly by the number of photons availablefrom the scene. The SIT tube comes closeto reaching this goal. Several factors areinvolved in evaluating the low -light -levellimit of operation: the detector quantumefficiency and its integration with thespectral distribution of available photons;the reflectivity and contrast of the scene;the lens aperture; the solid angle cor-responding to the picture element; and theintegration time. All these factors areimportant in establishing an S/ N at thetarget.

When the camera system processes thisinformation, it contributes its ownadditional noise, the significance of whichdepends upon the actual signal level com-ing out of the target. Fig. 10 shows a typical

N characteristic for a tube and camera.Note that at the higher light levels the S/ Ndoes not continue to increase along the"photocathode -limited" line. When fullsignal output is obtained from the tube, anyfurther light -level increase is accompaniedby a gain reduction brought about bydecreasing the intensifier voltage. As thislower voltage cuts into the photoelectronenergy distribution, the number of primaryelectrons entering the target through thebuffer layer will only be sufficient tomaintain signal level, thus flattening theS/ N characteristic.

Signal integration

The silicon target stores the charge imageuntil it is scanned off. For broadcastsystems and most closed-circuit systems inthe United States this integrating time isnominally I 30 s. If the photon flux inputcan be integrated for a longer time, moreinformation will be stored; the results willhave an improved signal-to-noise ratio buta loss of motion perception. Operation inan extended integrating mode will be

limited by dark -current build-up, which isusually proportional to integrating time.

50

a6

TUBE TYPEPHOTOCATHODEEQUIVALENT

4804

NOISE4 MHz

RESPONSIVITYBANDWIDTH

160OF

,,A/LM /f30dB

2

20 dB 10

6

4

PREAMP

-

5*oPREAMP NOISE EQUALSPHOTON NOISE

10 dB

2-

ode I

6

4

10 dB

2

ooh

Aee PREAMPLIFIER CONTRIBUTIONTO NOISE

1

-

10-T 106 10-5 104 10-3 ,I0-2FACEPLATE I LLUM NANCE - LUMENS / FOOT.

10 I

Fig. 10Typical S/N characteristic for tube and camera. At higher illumination levels, intensifiervoltage decreases to limit gain. This lowers the number of primary photoelectrons strikingthe target, so the S/N flattens out from the photocathode -limited line.

When the tube is operated in this mode,increased dark current resulting frombuild-up can usually be tolerated up toabout a one -second integrating time.Longer times will require target cooling ofabout 20°C for each order increase inintegrating time. In addition, for in-tegrating times in excess of ten minutes, ithas been found useful to turn off the gunheater until just before ready to read outthe information. Signals have been

successfully stored in this manner for up toeight hours before being read out.

BloomingBlooming is the spread of a highlight imageand is associated with most camera tubes.It occurs when portions of the pictures areoverloaded with excessive light; the

overloaded area then appears larger than itshould because excess charge on the targetspills over into adjacent areas. Low -light -level scenes are particularly susceptible toblooming because the general content of ascene may have detail with a rather narrowcontrast range exept for an occasionalbright light or flash.

Recently developed targets with a newreduced -blooming feature are now beingused with significantly improved results.The 4804H reduced -blooming tube, atextreme overload, can handle intensities 20times greater than those that can behandled by the conventional 4804 tube.

However, spot intensities greater than 1000times full signal introduce additional con-cerns for the user of the reduced -blooming

tube: either further distortion of the spotimage as a result of lens flare (internalreflections) or potential damage to thetarget from overexposure.

Alternate configurationsAlthough most camera applications in-volving simple surveillance can use a 16 -mm SI-I tube with no modificaton, thereare circumstances requiring additionalfeatures or characteristics. At extremelylow light levels, such as those found out-doors with no artificial illumination, it isnecessary to have the pickup device in-troduce as little noise as possible to theincoming image signal and to have enoughgain available so that the output signal-to-noise ratio is not degraded by amplifiernoise.

An image -intensifier tube can provideadditional gain to an SIT tube so thatoperation can be realized up to thephotoelectron noise limit. The fiber-opticoutput element of the image tube couplesdirectly to the SIT photocathode throughthe SIT tube's fiber-optic faceplate. Theadditional white -light gain of over 20:1 isthe ratio of the SIT tube photocathodecurrent to the photocathode current of theimage -intensifier.

For additional light collection, it is possibleto increase the size of the input aperturefrom 16 mm to 40 mm. In applicationswhere space is at a premium, an SIT with asmaller, 2/ 3 -inch gun section may beuseful. Tubes are also available for uniqueapplications that require either gate orzoom in the intensifier section. Where

improved resolution is required, a 27 -millimeter target is available that providesresolution in excess of 1000 tv lines perpicture height.

ApplicationsMany military, medical, and scientificapplications, in addition to surveillanceapplications, have developed for low -light -level television cameras employing the SITtube.

Airborne cameras used in gunfire controltake advantage of the tube's ability toprovide useful information under adverseconditions of low contrast and low lightlevel. Cameras on aircraft carriersoperating under similar adverse conditionsare used to help guide aircraft landings.Other shipboard cameras are used for nightmaneuvers and harbor piloting. Theseapplications encounter less rapid motion,but present a challenge to blooming controlbecause of bright lights in the harbor orrunning lights on other ships. Submarineperiscopes are also being outfitted withlow -light -level cameras.

Parking -lot surveillance applications inboth the private and public securitydomains are well known, as are the search -and -observe functions of the cameras inlaw -enforcement applications.

Various businesses have found uses forlow -light -level cameras, including observ-ing activity in film -processing plants andinspecting internal parts of jet enginesduring maintenance. Cameras go un-derwater to aid in oil drilling. Airlines uselow -light -level cameras with low-level X-ray systems for baggage inspection.Fishermen use airborne cameras to locatefish schools in the ocean by observingplankton fluorescence.

Scientists can use low -light -level camerasto observe the nocturnal habits of birds andanimals or to advance their knowledge inastronomy. Medical applications find low -light -level cameras used in low -light X-raysystems and eye fundus investigations.

References1 Acul.xuser. R er.. "the silicon target sidicons: new features

and expanded uses." this issue.

2 Engstrom. R.W.: and Robinson. G.A.."Choosc the tube forI o." Electra -aria of S.IlleHIS Design. Jun 1971.

Rodgers. R.1..111. "Beam scanned silicon targets lor cameratubes." IEEE Intercon. Mar 1973.

4 Mesner. M.: and Sensenig. W. "Final, report Nil sensormeasurement." RCA Astro-Electronies. unpublished.

51

Avalanche photodiodes:no longer a laboratory curiosity

R.J. McIntyre) P.P. Webb

Avalanche photodiodes (APDs), whichwere little more than laboratory curiositiesa few years ago, have begun to find theirway into many electro-optical systems. Inthis paper we wish to summarize some ofthese applications, and to point out someother applications in which APDs have notyet been exploited to their full capabilities.

New APD configurations, including large-area arrays anddetector/preamplifier modules, are adding numerousapplications to a fast-growing list that includes laserrangefinders and detectors for optical communications.

Basic properties ofavalanche photodiodes1 he properties of both p-i-n' and avalanchephotodiodes"' have been described fullyelsewhere. The more important propertiesof RCA's avalanche photodiodes are asfollows:

Reprint RE -22-4-151 Final manuscript received December 13. 1976.

Robert McIntyre has contributed to theunderstanding of avalanche multiplicationin semiconductor diodes, generatingtheories to explain the noise spectral densi-ty and gain distribution in avalanchephotodiodes, among others. The photosen-sor R&D program under his direction has ledto the establishment of a rapidly expandingbusiness in this area for RCA Ltd.Contact him atElectro-Optics DepartmentSolid State Div.RCA Ltd.Ste. Anne de Bellevue. Que.Ext. 340

Paul Webb has been involved in the develop-ment of Ge(Li) diodes for nuclear radiationdetection in addition to various types ofsilicon photodiodes. Most recently, heparticipated in developing a successfulprocess for fabricating large -area, fully -depleted "reach -through" avalanchediodes, both single -element and quadrant.Contact him at:Electro-Optics DepartmentSolid State Div.RCA Ltd.Ste. Anne de Bellevue. Que.Ext. 340

Quantum efficiency is high.

With proper design, avalanchephotodiodes can be made with quantumefficiencies greater than 70% from theultraviolet right through the visible toabout I pm. With selective coatings, quan-tum efficiencies of 80 to 95% can probablybe achieved at any desired wavelength inthe 0.5- to 0.9-tim range. At 1.06 pm theabsorption coefficient of silicon is not highenough to allow really high quantumefficiencies, but QEs in excess of 30% havebeen achieved with a wide (200 Am)depletion -layer device.

APDs are fast.

The response time of an avalanchephotodiode is limited either by the carriertransit time or, for very fast diodes, by thetime for multiplication. For devicesdesigned for use below 0.9 Aim, such as theC30884, the depletion layer is 55µm or lessin width, and the total carrier transit time isless than I ns. Devices designed for use at1.06 Aim are a little slower, with responsetimes of 3 ns for a 100-µm depletion -layerwidth (C30817, C30872, C30895), rangingup to 10 ns for devices with 200-µmdepletion width. For very fast diodes,(narrow depletion layers), the ultimatespeed limitation is the avalanche builduptime. This gives a frequency response of theform M(w) = Mo/( 1 + Mon2)'', where nis 4 to 5 X 10-11s. Thus, the device has again -bandwidth product limitation of 300to 400 G H7.

Their size is small.

Because of the extreme doping uniformityrequired (" 0.1%) to give a respectably

;1

Fig 1

Avalanche detector -preamplifier module desigred for laser rangefinder. Detector chip incenter has a sensitive area of 0.5 sq mn.

uniform detector, high -gain avalanchedetectors cannot be made very large withany acceptable yield. Most devices of I mmdiameter or less can be operated at gains upto about 200, with some capable of goingmuch higher. Larger devices are normallylimited to somewhat lower gains. Thelargest device made to date in any quantity,with an area of 5 X 5 mm, is designed tooperate at a gain of 40.

Capacitance depends on the junction areaand depletion layer width.

Small devices are in the 0.5- to 2-pF range.Larger devices are about I pF/ mm2.

APDs are quiet or noisy, depending on howyou look at it.

APDs are rather noisy amplifiers, havingexcess noise factors (the ratio of the output -noise power to the multiplied input -noisepower) on the order of 3 to 5 (depending ongain), as compared to 1.1 to 1.4 for aphotomultiplier (PMT). However, underbackground -limited conditions, which in-cludes most applications, the signal-to-noise ratio is proportional to 17/ F, where nis the quantum efficiency and F is theexcess noise factor. Because of the higherquantum efficiency of the APD, it is

competitive with the PMT throughout thevisible range and is better than the PMT inthe near infrared (0.8-1.1 µm). Darkcurrents are low enough (-10H A/ mm2before multiplication at room temperature)that the devices are background -limited atvery low light levels. Like photomulitpliers,the dark current can be reduced by coolingif necessary.

Applications ofavalanche photodiodes

The following partial list of applicationsshows where A PDs have been used to goodadvantage. New ones are arising every day.

One of the biggest current markets for APDsis for laser rangefinders using either pulsedor swept -frequency modulated cw lasers.

RCA's APDs have been particularly pop-ular for use at 1.06 ium because of theirrelatively high quantum efficiency(- 20%).Fig. I shows an APD detector module,with an integrated thick -film preamplifier,designed for this purpose. The module hasa responsivity of greater than 4 X V/ W,a bandwidth of over 15 MHz, and an NEP

(Noise Equivalent Power) in the dark ofless than 10-" W/ Hi'. This particularmodule is used in the AN / GVS-5 hand-held laser rangefinder developed byAutomated Systems in Burlington for U.S.Army ECOM.4

APDs are particularly appropriate detectorsfor optical communications, both for line -of -sight systems and with fiber optics.

Fig. 2 shows both the measured andcalculated pulse responses for a detectordesigned specifically for use with a 400Mb/ s PCM line -of -sight system at 1.06Am, which was being evaluated by NASAfor possible space use. Using this detectorat a gain of 300 to 400 straight into 50 fl(i.e., no low -noise preamplifier), bit -error -rates ( BERs) of less than 106 weremeasured with average optical signals ofabout 90 nW (i.e.,' 500 photoelectrons, or2400 photons/ pulse). This was con-siderably better than could be achievedwith specially developed cooledphotomultipliers that, when new, hadquantum efficiencies in the range of I to 2%at 1.06 µm. The use of a good low -noisepreamplifier should reduce the requiredsignal level by about a factor of two.

Detectors for use with fiber optics are nowbeing developed. Devices with sensitivediameters of 0.020 in. have been fabricatedwith NEPs of about 10"W/ Hz'' at 0.8 to0.9 Arn at a gain of about 200 (i.e., 120A/ W).s These devices have response timesunder I ns and show effective dark currentsequivalent to less than I thermallygenerated electron per µs, or one every 100bits in a 100 Mb/ s system. Thus, these

Fig. 2Response to 100 ps pulse at 1.06 micrometers for detector with 80 -micrometer depletionlayer width. Time scale = 0.5ns/div; gain = 470. Dashed curve shows theoretical pulse shapeon the basis of transit -time effects alone. Solid curve includes effects of avalanche build-uptime.

53

10-10200

.0 025 05

1 V

300NP NUMBER OF PRIMARY

400 500PHOTOELECTRONS

Fig. 3Very low bit -error rates have beencalculated for an optical communicationssystem at optimum gain. Here, a is the ratioof the 'off' to 'on' pulse. Broken lines: using apreamplifier with a noise equivalent chargeof 800 electrons; Solid lines: using apreamplifier with a noise equivalent chargeof 2000 electrons.

devices will be limited by the noise in thesignal itself, unless the on/ off ratio of thepulse signal is greater than about 4 X 104:1,which is unlikely.

Fig. 3 shows how the BER varies withon/off ratio and amplifier noise as afunction of signal level. The amplifiernoise -equivalent -charge values of 800 and2000 would correspond roughly to what isachievable today with 20 Mb/s and100 Mb/s systems respectively. Thus, aBER of 10-1' should be achievable withabout 425 photons per pulse (385photoelectrons per pulse), or an averageoptical power of about 5 nW in a 100 Mb/ ssystem. Since laser diodes are capable ofcoupling greater than I mW into a singlefiber, total fiber attenuations of = 55 dBcan be tolerated without affecting the errorrate appreciably. With the most recentfibers (<5 dB/ km) this can mean a repeaterspacing of over 10 km in a fiber -opticscommunications link.

Fig. 4 shows the design of a prototypedetector module for use with fiber bundles.This unit, which will accept a fiber bundleup to 0.045 inch in diameter (I mm2), has anNEP of about 1.5 X le' W/ Hz''' at 1.06/Am, or about 5 x 104 at 0.9µm. Such units

0090" DIA.

FOUR FIBER BUNDLES I DETECTOREACH 0043" DIA. PRE -AMP

550 V ± 5% -IP.

DETECTORPRE -AMP

DETECTORPRE -AMP

DETECTORPRE -AMP

TEMPSENSE

TEMP. COMPUNIT

7V!5%" TO ALL UNITS-I0V

BIASINGNETWORK

A

Fig. 5(a) Quadrant detector consists of fiber-optic lightpipe and four detector/preamp units.Additional circuitry provides bias adjustment and temperature compensation.

are not just for optical communications-one customer used 10 of them to make a 10 -element linear array approximately I mm X10 mm in size. Fig. 5 shows how four unitsare used with a quadrant light -pipe to givean avalanche quadrant detector with an0.090 -in. -diameter sensitive area.6Obviously, with a sufficient number ofdetector modules, any desired detectorgeometry is possible.

Large -area detectors can be built up fromsmaller modules.

Avalanche diodes up to 5 mm X 5 mm inarea have been fabricated. Fig. 6 shows thedesign of a detector module in which thesensitive area has increased from 5 mm x 5

mm to I cm X 1 cm through the use of atapered light -pipe. This module is designedso that the module dimensions are only afew thousandths of an inch greater thanthose of the light -pipe. Thus, any desirednumber of these modules can be mountedin an n X m array, with a dead spacebetween detectors of less than 0.010 inch.

As soft X-ray detectors, APDs operate atvery high count rates.

Avalanche photodiodes are excellentdetectors of soft X-rays in the I keV to 10-keV range.' Although the achievableresolutions are not particularly goodbecause of the difficulty of making uniformdetectors, resolutions of about 10 to 15%

FIBER OPTIC INPUT

Fig 4Developmental detector -preamplifier designed for use with fiber-optic bundles up to 0.045inch in diameter. The bundle, with its protective coating, is cemented into the collar at thebottom, and interfaces with a clad light -pipe that carries the signal to the APD chip andpreamplifier. Package is standard 5/8" X 5/8" flatpack.

54

(b) Assembled test unit with cover removed. (c) End view of quadrant fiber bundle

are possible in the I- to I0-keV range.These devices have the advantage over non -multiplying solid-state detectors of beingusable at much higher count rates (up to afew million/ sec), and thus could haveapplications in high-speed monitoring ofX-rays.

Cooling reduces dark current for noise -freelow -light -level operation.

For narrow bandwidths, the noise -equivalent power of an APD, at roomtemperature, is limited by its dark current.This can be reduced by cooling. Whensufficiently cooled, the dark currentbecomes insignificant and the NEP can getdown to the 1016 to Or" range, so the noise

LIGHTINPUT

in the signal usually determines the noiselimit. A gain of 20 to 25, for which theexcess noise factor is about 2.5, is sufficientto reach this limit.

APDs can count single photons at high ratesin a manner similar to a Geiger tube.

If cooled sufficiently, say to 77K, thethermal -generation rate of electrons in thedevice is reduced to a few per second.Under these conditions the APD can beused above the breakdown voltage like a

Geiger tube, with single -photon detectionefficiencies in excess of 50%. A quenchcircuit is required to "reset" the device afterbreakdown. If this is sufficiently fast, anAPD can be used for single -photon

AVALANCHE PHOTODIODEAND PR E AMP

TAPERED LIGHT PIPE

Fig. 6Large -area (1cm X 1cm) stackable APD module. Tapered light -pipe funnels light down to 5mm X 5 mm APD and preamplifier, indicated by broken lines.

counting at rates up to a few million persecond."

ConclusionsAPDs are gradually being exploited inmany electro-optic systems in which theyare the most sensitive detectors available.The next few years should see a rapidexpansion in both the number and varietiesof their applications.

AcknowledgmentsThe authors would like to acknowledge thecontributions of their co-workers-J. Conradi, R. Cardinal, J. Bignet and M.Teare- who assisted in the design, fabrica-tion. and measurement of the devicesdescribed here.

ReferencesI. Doyle. T.; and Sprigings. H.C.; "Silicon photosensors and

Theis applications." RCA Engineer. Vol. 19, No. 6 (Apr/ May1974) p. 60.

2. McIntyre. R.J.; Spriginp, H.C.; and Webb, P.P.; "Solid statedetectors for laser applications," RCA Engineer. Vol. IS. No. S(Feb/ Mar 1970) p. 32.

3. Webb. P.P.; McIntyre, R.J.; and Commit, J.; "Properties ofavalanche photodiodes," RCA Review, Vol. 35, No. 2 (Jun1974) Pp. 234-278.

4. Woodward, J. "Hand-held laser rangefinder," to be publishedin RCA Engineer Vol. 22-6 (Apr/ May 1977).

5. Conradi, J.; Webb, P.P.; and McIntyre, R.J.; "Silicon reach -through avalanche photodiodes for fiber-optic applications."Proc First European Conference on Optical Fibre Com-munications. London (Sep 16. 1975) p. 128.

6. McIntyre. R.J.; Webb, P.P.; Team, M.; and Bignet, J.; "Aquadkant avalanche detector for laser applications," Proc.Electro-Optics 1976 Conf., New York, (Sep 1976).

7. Webb, P.P.; and McIntyre. R.J.; "Large area reach -throughavalanche diodes for X-ray spectroscopy." IEEE Tram. Nucl.Sc. Vol. NS -23, No. I pp. 138-144.

8. McIntyre, R.J.; "On the avalanche initiation probability ofavalanche diodes above the breakdown voltages IEEE Trans.Eke. Dev, Vol. ED -20, No. 7, (Jul 1973) pp. 637-641.

55

The atmosphereDistorted laser bean, once coherent. after passing through 1500meters of the atmosphere. A still-coherer t beam would appear asa pure white circle.

Laser satellite tracker uses both the at -nosphere and free space asits -ransmission med a. The atmosphere isa low -loss medium, butcan cause signal distortions. This ma<ts t good for tracking andrangefinding, but Fcor for lightwave communication.

Correcting the atmosphere's distortion

The possibility of correcting the type of lightwave distortionsshown at the left with real-time adaptive optics seems to bedistinct. Recent developments with coherent laser beamscorrect the phase error of time -varying distortionsby using reflected returns from "glints" at the focusof the phase front to compensate for atmosphericeffects and then reshape the wavefront.

The Defense Advanced Research Projects Agencycurrently has several programs underway in thisarea, including work on a high -frequency(10 -kHz) deformable mirror and largeoptics with as many as 10° to 106control elements. The most far-reaching implications of adaptiveoptics are not necessarily defenserelated, however. Other possibilitiesbesides long-range lasercommunication are beam -focusingin laser -fusion systems andtight -tolerance integrated -circuit manufacturing viaaberration compensation.

fiberoptics

progress

E 100

100

MULTIMODESTEP -INDEX FIBERS

(7)crwa.(r)

SELFOC®.

NEAR\PARABOLIC .,

E 0.1PROFILE

1970 1975 1980

Dispersion, or pulse -spreading, hasdropped by a factor greater than 100.Caused by light of different wavelengthstravelling at different velocities in the fiber, itis not a problem in single -mode fibers.

(Source: Locale. Electro-optical Systems Design, Oct 1976)

1000-

100-

10-

I -

0.1-

1965 1970

Signal losses have droppedmagnitude in about 15 yparison, RG-58 coaxial cab230 dB/km at 10 MHz, 60 dB.and 15 dB/km at 1 GHz.

1970 1975 15.

Selling price of cablesignificantly. Corning Glastpilot -plant experience to$0.10 per meter when the100,000 km of communicalper year.

Pr Fiber optics

1975 1980

three orders ofars. For corn -e has a loss ofkm at 100 MHz,

m SINGLECABLE

80

has droppedWorks has usede an estimate ofnarket calls forons-grade fiber

How it has progressed......and what this means to communications

The dollar figures behind fiber-optic communicationsnow look very promising for metropolitan areas, where thecurrent underground distribution systems are verycrowded. Extending the present level of service by diggingfor additional conventional lines is expensive, but fiberoptics' great bandwidth can transmit many more com-munication channels than coaxial cable of the samediameter. The losses in coaxial systems also requirerepeaters, which in turn require manholes for servicing.Fiber-optic systems would cut down on both of theseexpensive items considerably.

The largest -scale test so far of an urban lightwavecommunications system has been a Bell Labs/WesternElectric setup in Atlanta that proved the technicalfeasibility of such a system. The experimental systemthere transmitted and received digitized voice, television,and data signals at 44.7 Mb/s, equivalent to 672 telephonechannels per fiber. Their results indicated that "highlyreliable transmission may be obtained over distances ofabout 7 km before regeneration is necessary. Sincetelephone central offices are often spaced closer than this,lightwave communications systems could connect manyof these offices without the need for intermediaterepeaters in manholes."

CoJr0sy Bell Laboratories

Practical lightguide cable connectors will be a r ecessity for fieldsplicing. This prototype aligns 144 fibers within 0.0001 inch.

What's a fiber optic?A fairly confusing word, we think. l-ave you evenseen one? Is it adjective or noun, fiber -optics or fiberoptic, or what? Optical fibers, lightguides, andlightwave communications are good replacementterms for fiber optics; they all get the point acrosswith less confusion.

Fiber optics has tinally reached the point where it is doingan awful lot more than pose for interesting pictures like theones below. Businesses now think fiber optics is lookinggood, and they're talking dollars, not aesthetics.

Courtesy Corning Glass

Transparent fibers will be jacketed in actual use, and the optimumtransmission wave engths are above the visible region. Thesefacts may prove disappointing to those who felt giber -optic phoneconversatiors would be much more spectacular than onesmoving through copper wires.

Courtesy Bell Laboratories

Over a kilometer long, the fiber on these reels glows with laserlight during a test. Several companies are now producingcommunicatons-grade (20 dB/km loss) fibers at these lengths.

57

an introduction

Optical transmission devices and mediaD.G. Herzog Optical information must travel through a medium between

source and detector. Our Understanding of media andpropagation remained constant for years after Einstein'sunifying theory until the invention of precision laser andradar systems forced a number of technologicaldevelopments to occur.

For any optical transmission and detectionto take place, three elements-a lightsource, a propagation medium, and adetector -are needed, as shown in Fig. I.An understanding and a definition of thelight source is necessary to predict andcontrol any optical transmission through amedium. Conversely, a detection criterionis also necessary because of the interplaybetween the transmission medium and thedetector. Light sources are treated inanother section of this issue (see Gorog, p.IX); detectors are also treated separately(see Savoye, p. 36). In this discussion, weare concerned with optical transmissiondevices and media.

Early concepts

Early man accepted that he "saw," but heconsidered vision to be perhaps a"mechanical transfer" from the object tothe viewer. The concept of light as radiatingwaves was far too difficult to understandbecause it deviated from man's sensoryexperience. The early Greek philosophersconsidered vision as something projectedfrom the eye that fell on the object,resulting in particles or small "repliculars"returned to the eye for analyzing, with lightsomehow playing a part. However, theGreeks also explained observations of theheavens by way of superficial analogieswith human and animal behavior-withvery little experiment. These philosophers

observed the stars with intellectualgratification, but felt that the laboratorysmacked of a workshop for gain.

From the theoryto experiment

Yet it was from the workshops-withcarefully planned experiments andobservations --that the understanding anduntangling of the puzzling nature of lightfinally came. Galileo Galilei (1564-1642)learned how to manage the optical -transmission character of light by produc-ing images that could be both magnifiedand demagnified. This provided thenecessary control so that other ex-perimenters could go further. Isaac New-ton (1642-1727) initiated experiments intothe nature of light and its effect as a wavephenomenon, leading to the concept thatlight had a sound -like method of propaga-tion. A.J. Fresnel (1788-1827) establishedthe initial mathematical basis from whichJames Clerk Maxwell (1831-1879) was ableto provide his propagation equations.* In

Sometimes labor pays ut1 in more than one area. Quoting froma letter In Maxwell. "I made out the equations in the countryWore I had arr suspicions of the nearness between the...%clock) of propagation of magnetic Meets and that of light soThat I think I hare reason to believe that the mngnetic andluminitcrous media are identical." (At that time. lummiterousmedia was also known as ether. some vet -unknown mediathrough which light waves were propagated like ripples caused bya pebble tossed in water.)

TRANSMISSIONLIGHT SOURCE MEDIUM

Fig 1

Necessary ingredients for an optical propagation system

1902, Max Planck demonstrated that lightenergy was inversely proportional towavelength.

During this same time frame, other ex-perimenters established that light had aparticle nature by measuring the minuteamounts of momentum associated withlight. Earlier observers had already notedthat pressure from the sun's light apparent-ly kept the tails of comets always pointingaway from the sun.

Thus, two camps were established- -onetheorized that light was a particle -transmission phenomenon; the otherthought that light was a wave -propagationphenomenon. Finally, in 1905, Albert Eins-tein completed the basis of our modernunderstanding of light transmission. Hisunifying theory held that the energy of lightis not distributed evenly over the wholewave front, as the classical pictureassumed, but is rather concentrated orlocalized in small discrete regions. Eachwavefront of light, therefore, could beimagined to be studded with photons.

In Einstein's E= mc1 equation, light travelsat the speed r (2.99 X 105 km/ s) accordingto Maxwell's equation. Light of energy E(per unit wavelength) thereby appears as aparticle of mass El c2

Thus, Einstein provided the concept oflight as we know it today-a small region ofthe electromagnetic spectrum, produced bythe transitions of electrons from states ofhigher energy to states of lower energy, andtravelling in basically a straight line fromsource to detector.

Until the last 25 years, little was ac-complished past this ideal propagation

DETECTOR

Reprint RE -22-4-9Final manuscript received November 9, 1976

Sti

model. Astronomers suffered with the at-mospheric distortion and did the best theycould. Technologists rolled up theirsleeves, though, and made a start duringWorld War I I when the need to understandradar forced some practical understandingof electromagnetic -wave propagationthrough the atmosphere.

With the discovery of the laser in 1961,scientists finally had a powerful, coherentlight source-the measurement tool anddriving force needed to extend propagationstudies. However, the laser met with somany initial limitations that practicalapplications were slow in coming.

It has happened many times in the past thatwhen the force of producing a productmeets a technical limitation, manytechnological developments occur. So itwas with many laser and laser -associatedelectro-optic systems over the past fifteenyears.

Recent developments

Fig. 2 summarizes the areas when un-derstanding, technology, components andproducts have expanded significantly overthis fifteen -year period.

We understand turbulent -mediapropagation better today.

Our understanding of propagation in freespace has remained constant since Einsteinestablished his criteria. However, our un-derstanding of transmission and propaga-tion through a turbulent medium, such asthe atmosphere or the sea, has advancedsignificantly, leading to practical techni-ques to counter and correct for the disrup-tion and noise generated by the medium.These approaches promise to improveastronomical resolution through the at-mosphere by an order of magnitude.De Wolf (p. 64) discusses the problem areasinvolved in establishing, and working with,a propagation model for random media.

Lens management has advancedsignificantly.

Optical lensing management has also ad-vanced significantly because of the need tooptically manage laser light, as well as theneed for night -vision, sensing, and

military -reconnaissance applications. Theneed was coupled through computertechnology to develop the significantrefinements necessary for the versatileelectro-optic applications of today andtomorrow.

FREE SPACE

TURBULENTMEDIUM

OPTICAL LENSINGMANAGEMENT

FIBER OPTICS

OPTICAL CONTROLMODULATION/DEFLECTION

Fig. 2Types of transmission media.

Fiber optics shows great promise.

Fiber-optic technology shows greatpromise of providing low-cost, wider -bandwidth, more -reliable com-munications, and so many corporationsand governments are investing heavily in itsdevelopment. For example, in the U.S.alone, more than $10 million will beinvested into developing fiber optics thisyear. Optical Spectra (Oct 1976) projectsthat, by the turn of the century, fiber opticswill represent more than $10 billion a yearin U.S. business alone.

The fiber optic rod-an optical analog tothe electrical wire guides an optical beamaround twists, corners, and over extremelylong distances with low loss. Recent ad-vances have improved transmissivity bymany orders of magnitude. An optical fiber0.01 inch in diameter (no thicker than ahair) has a bandwidth of 500 MHz and istherefore capable of carrying more com-munication information than the six -inch -diameter communication trunk lines thatjoin cities and continents. Optical fibers arediscussed in more depth by Wittke (p. 00).

We now have practical optical control oflight.

Optical control of light through modula-tion and deflection has developed, in some

cases, far beyond the capability envisionedeven by many far-sighted people a fewyears ago. Optical modulators that arealmost too small to be seen and consumelittle power are being introduced. In-tegrated optical devices-the opticalequivalent of integrated circuits-havebeen developed to modulate, steer, switch,trunk, and add and subtract optical beams.Hammer (p. 71) discusses some of theintegrated -optics work carried on at RCALaboratories.

Acousto-optics can provide low-cost beamsteering.

Solid-state deflection techniques that com-bine acoustical and optical interactionshow promise of providing the beam -steering management at low cost and smallsize, both of which have previously limitedsuccessful product development in manyelectro-optics applications.

The acousto-optics phenomenon modifiesthe optical transmission through a mediumvia an acoustic interaction with the

medium. This phenomenon imparts itsfrequency and amplitude characteristicsinto the optical beam, so it may truly becalled an "optical transmission modifier."

Don Herzog, as Manager of the ATL'sSystems and Applications Group, has ex-perience in a number of electro-optic andlaser -associated areas. He has worked on aninjection -laser intrusion alarm system, lasercommunication systems, optical trackingand ranging systems, and now recordingsystems.Contact him atSystems and Applications GroupAdvarced Technology LaboratoryCamden, N.J.Ext. PC 3171

59

Optical -fiber communications links

J.P. Wittke Before optimizing the design of a practical fiber link, it isnecessary to understand its signal-to-noise considerations,frequency response, and dispersion effects.

Optical -fiber communications systemscarry information in the form of intensity-modulated light along transparent opticalfiber waveguides. They are of growinginterest world-wide because of their verylow loss and potentially very highbandwidth. Fibers are commerciallyavailable now with losses under 20 d13/ km,and within a year this figure will probablydrop to under 10 dB/ km, or even lower.( For comparison, RG 59/ U coax has a lossof 66 d13/ km at 30 MHz.) Dispersioneffects in present fibers strongly limit theiruseful bandwidths, but with continueddevelopment of graded -index and single-mode fibers, bandwidths approaching agigahertz can be expected. These features,combined with small size, low weight, andfreedom from most pick-up and in-terference problems, make fiber -opticslook very attractive for future com-munications systems.

Cost, of course, will be a deciding factor inadopting such systems. A commercialcabled bundle of 6 independent fibers ispresently available for about $4 per ft, or 75cents per fiber -ft, which is about 5 times asexpensive as conventional coaxial cable ofequivalent bandwidth, such as the typeused in cable-tv installations. This is anintroductory price, however, that clearlywill drop significantly with volume produc-tion (and competition). Fiber optics canthus be expected to be cost -competitivewith more conventional transmissionmedia within a few years, and, ultimately,should prove to be significantly cheaper.

This paper discusses noise and distortion infiber-optic systems. Since these factorslimit the performance achievable, un-derstanding them makes performance es-timates possible for systems beingdesigned.

Reprint RE -22-4-10A lengthier version of this paper appeared in theProceedings of the 1975 S P I E Annual Meeting.

The basic function of a communicationslink is to transmit information from onepoint to another without introducing un-acceptable degradation in the transmittedsignal. The specification of what con-stitutes "unacceptable degradation"depends upon the way in which the infor-mation is sent. In an analog link, theinformation is contained in the detailedshape of the transmitted electromagneticwave. In a digital system, the signal consistsof a series of discrete symbols (waveforms)which must be distinguished from eachother on reception. Noise and distortiontherefore affect analog and digital systemssomewhat differently.

Noise and distortion

To see where the major sources of noiseand distortion in an optical link are,consider the system represented in Fig. I.The (electrical) input signal is first appliedto a signal shaper/ encoder. For analogtransmission, this element pre -distorts(pre -equalizes) the signal to compensatefor unavoidable distortions introducedlater in the system; for analog signals beingtransmitted digitally, the shaper/encoder

INPUTSIGNAL

(ELECTRICAL

SIGNALSHAPER/ENCODER

DETECTOR

performs the analog -to -digital conversionand generates the appropriate digital sym-bols; and for pure digital systems, it maydetect the incoming data stream andregenerate and retime the symbols ap-propriately for the optical driver. After theelectrical signal is thus suitably "shaped," itis applied to a source driver, whichmodulates the current flowing through theoptical source to produce the desiredoptical -signal output. This assumes acurrent -modulated optical source, such asa semiconductor incoherent light -emittingdiode (LED) or a (properly biased) injec-tion laser.

The light from the source is coupled intothe fiber, which transmits it to a (square-law) optical detector at the receiving end ofthe system. The electrical signal generatedin the detector is amplified and fed to asignal shaper/ decoder, which converts theraw electrical signal into the proper formfor use. In long links, repeaters may berequired. These would essentially act as thereceiver in the above system, except theoutput would be applied to another sourcedriver and optical source, whose outputlight would be coupled into a followingfiber section, etc.

SOURCEDRIVER

OPTIC AL FIBER

AMPLIFIER SIGNALSHAPER/DECODER

OPTICALSOURCE COUPLER

OUTPUT SIGNAL(ELECTRICAL)

Fig 1

Elements of a point-to-point fiber -opt c communications (ink. Most of the noise in thesystem is associated with the optical receiver.

60

Detector noise

We shall assume that the input signal andthe electronics in front of the optical sourceare noiseless. This is generally, (but notinvariably) a very good approximation,since the electrical signal at the source isusually at a high level compared to theassociated electrical -noise sources.Similarly, noise introduced by the sourceand spurious light coupled into the fiberare negligible at the detector in a well -designed system.

The basic remaining noise is associatedwith the optical receiver, which has twomain noise sources to be considered. One isassociated with the discreteness of theelectronic charge and the "quantization" ofthe light upon photodetection. This is

generally called "shot," or "quantum,"noise and arises from the essentiallystatistical nature of the process by whichthe incoming light generates electricalcarriers in the detector. The noise currentof this shot noise has a zero average value,but a mean square value of

(0= 2eIBG2 Fit (I)

Here e is the elemental electronic charge(1.60 X 1019 coulombs), / is the averagephotocurrent (in amperes) generated by theincident light signal, and B is the (informa-tion) bandwidth (in hertz) in which thenoise current is observed. The last twofactors depend on the nature of thephotodetector: G is the current gain withinthe detector, and F0 is the equivalent "noisefactor" associated with the gainmechanism. In a photoconductive detec-tor, such as a silicon PIN diode, there is nogain mechanism, and G = Fo = I. In anavalanche photodiode, on the other hand,G, adjustable via the bias voltage, canexceed 100, while the noise factor Fo isgenerally gain -dependent. For our pur-poses. the commonly used approximationof F0= (I': for avalanche photodiodes' willbe assumed valid.

Even if an avalanche detector (with gain) isused, the signal level leaving thephotodetector proper will be at a low level,and more amplification is required. Ther-mal (Johnson) noise fluctuations in theequivalent input resistor of the requiredamplifier, and amplifier noise, form thesecond significant noise source. This noisecan also be represented by a mean square"thermal" noise current, (ir):

(i7-2) = (4k TI Req)BF1 (2)

Here k is Boltzmann's constant (1.36 x10 24J / ° K), Tis the (absolute) temperature(in °K) of the equivalent detector loadresistance R, (in ohms), and F1 is the noisefactor of the amplifier.

The noise component of "dark," orleakage, currents that occur in the detectorin the absence of illumination must also beconsidered. However, with proper detectorchoice and operation, "dark" noisecurrents are smaller than those discussedabove, and may be neglected.

With these simplifying approximations,the total mean square noise current (OW)can be written

(oft/2)=W>+ (iT2)

= 2eIBG2 Fo + 4KTBFII R, (3)

If an optical power P (transmitted downthe fiber) is incident on the photodetector,the generated signal photocurrent is givenby

I = (el he) XriP, (4)

where h is Planck's constant (6.624 X 10 44J -sec), cis the velocity of light (3 X 10' m/ s),

is the wavelength of the light (in meters),and n is the quantum efficiency of thedetector (for converting optical photonsinto mobile electrons). For a wavelength of830 nm, (where fiber losses are low andefficient (GaA I )As light sources areavailable), 77 0.85 for silicon and, fromEq. 4,

/(amps) 0.57 P(watts). (5)

For a sinusoidally modulated carrier with amodulation index m, the mean -squaresignal current (ii) is given by

(is) = 1/2 12Gm2 (6)

(Note that throughout the above develop-ment, both noise and signal currents havebeen referred to the input load resistor atthe amplifier following the photodetectorproper.) The system signal-to-noise ratiocan then be written

I/ 212 G2 M2(Si iV) (7)

2eIBG2 Fo + (4k TBF1 R)

Since both noise terms are proportional tothe system bandwidth B, this can berewritten in the form

m2/2(S/ N)x B= , (8)

4e I Fo + (8k TIC! I 1444-)

Eq. 8 is shown plotted in Fig. 2, for the(reasonable) parameter values m = 1

and F, = 4 (6 dB), again with the previouslyassumed detector sensitivity of (1 P)= 0.57A/ W. Two gain values are shown: G = 1corresponds to a PIN diode, and G= 100 toa "typical" avalanche detector. The curvesare drawn for three values of equivalentload resistor: 50, 4K, and 1M ohms. Thesecover the gamut of interest, from low loadresistance, where high (hundreds ofmagahertz) analog bandwidth areavailable without the need for equaliza-tion, to a very high load resistance, wheredigital signals can be detected with a goodsignal-to-noise ratio, but the pulse shapesare badly distorted and require equaliza-tion at high data rates.*

Fig. 2 shows that shot noise dominates athigher signal levels, while thermal noisedominates at low levels. Moreover, at low

I he advantages of using a high equivalent load resistor tomprove the signal-to-noise ratio are discussed in Ref. 2.

James P. Wittke has studied optical fibercommunications systems extensively, mak-ing measurements of the transmissionproperties of various fibers and on thefrequency response of light emitting diodes,and developing techniques for coupling thelight from semiconductor light sources intofibers. Currently, he is investigatingapplications of cw injection lasers foroptical information handling and workingon problems connected with laser process-ing of materials.

Contact him at:Systems Research LaboratoryRCA LaboratoriesPrinceton, N.J.Ext. 3261

2010

1018

m 10

a 14Z 10

x

a 10.z

0.o

0Z

10I0

10 '2 10 -gyp le 10-6 10 4 10 2 100

OPTICAL. POWER AT DETECTOR ( WATTS)

Fig. 2Typical S/N-bandwidth product available fordifferent detector gains and load resistors. Thermalnoise dominates at low levels; shot noise dominatesat higher levels.

received -light levels, the gain of anas alanche detector makes it superior to theI'I\ diode, while at high levels, the noiseassociated with the avalanche amplifica-tion process makes the l'IN better.

Source powerand fiber losses

A long-lived LED source can have anoptical output power of the order of a fewmilliwatts. How much of this can becoupled into a fiber transmission linedepends on LED geometry, fiber accep-tance angle (numerical aperture) andeffective fiber diameter. Typical valuesmight range from about 100 microwattsinto a small -aperture (NA = 0.1) fiber up toseveral milliwatts into a large -aperture ( NA= 0.6) fiber. An injection laser can readilycouple ten milliwatts into an NA = 0.15fiber.

How much of the power coupled into thefiber actually reaches the detector dependsprimarily upon fiber losses. With presentfibers. losses in the 800-900 nm low -losswavelength region range from 500 dB/ kmfor low-cost, high -aperture glass fiberbundles to under 20 dB: km for high -cost,low -aperture quartz fibers. Research fibershave shown losses of about 2 dB km.

Required S N ratio

The required S N ratio for the com-munications link is one of its most impor-tant specifications. For example, a S, Nratio of 43 dB for analog television signalsgi es good cable-tv quality, while a 25 dBratio would give unacceptable performancein most tv applications. On the other hind,when binary digital pulses are beingtransmitted, the bit error rate (BER, theprobability of error per transmitted bit)determines acceptable performance. An

N of 20 dB corresponds to a BER ofbetter than 10 " and, at SI N = 22 dB, theHER drops' to 10-'11. Roughly, then, goodanalog transmission corresponds to SI N=40 dB (noise current 1% of signal current),while S N = 25dB will provide essentiallyerror -free digital transmission.

Source and detectornonlinearities

Consider now a second basic difficulty thatmust be overcome in a communicationslink signal distortion. Distortion can beintroduced in any of a number of places inthe system of Fig. I. There are two types ofdistortion associated with the light sources:amplitude- and frequency -dependent non-linearities. In LEDs at small drive cur-rents, nonradiative process that saturateas the drive current is increasedcan introduce distortion unless the diode isbiased into, and operated in, a linearportion of its characteristic. At high drivelevels, diode heating can occur, with anattendant reduction in radiative efficiency.again leading to distortion.

I.0

0I

I he frequency response of an LED isdetermined by the recombination time ofthe injected carriers and by the diodecapacitance.' LEDs with response times asshort as 1.1 ns have been made, makingthem useful at frequencies of hundreds ofmegahertz. (See Fig. 3.) Their high-Irequency fall -off in output, however, willdistort signals with such high -frequencycomponents.

If an injection -laser light source is used, theassociated nonlinearities are somewhatdifferent. Since a laser is a threshold device.its output is (approximately) a linear func-tion of the drive current only above itsthreshold, and if the current is allowed todrop below that threshold, signal "clip-ping" can occur. Since stimulated, ratherthan spontaneous, emission occurs in alaser, the recombination time does notcontrol the response speed in the samesimple way as it does in an LED; injectionlasers have been modulated at rates of overI OW.' However, an internal responseresonance can lead to signal distortion.

The limited frequency response in detec-tors also leads to distortion. This frequencylimitation is generally associated with thetime. or the spread of times, required tocollect the photogenerated carriers. PINdetectors are available with response timeswell under 0.1 ns, and hence present no reallimitation to presently considered systems.Avalanche diodes, on the other hand,generally have somewhat slower responsetimes. on the order of 2-3 ns, and hence candistort signals with very -high -frequencycomponents.

10 100MODULATION FREQUENCY, MHz

1000

Fig. 3Frequency response for three developmental LEDs. Their high -frequency fall -off presentlylimits distortion -free use to hundreds of MHz: injection lasers, however, have beenmodulated at over 1 GHz.

62

Fiber -induced distortion

In many systems, the distortion introducedby dispersive effects in the fiber can be themost important. There are two major typesof distortion to be considered-materialdispersion and mulitmode distortion.

Material dispersion depends upon both thefiber material and the light source.

When the light from the source is dis-tributed over a band of wavelengths,material dispersion in the fiber will cause a"chromatic" spread of propagation timesthat will distort the shape of the receivedsignal. Low -loss fibers are often made of(nearly pure) quartz. while some of thehigher -loss, larger -aperture fibers are madeof borosilicate and other lower -meltingglasses. Gloget has shown that, expressedas a spread of propagation times, thematerial dispersion is about 4 ns/ km forLEDs, with their spectral widths of 30-50nm, and only about 0.1 ns/ km for thespectrally purer injection lasers. Thismaterial dispersion is lower for silica fibersthan for borosilicate and other higher -index glass fibers by about 40%. In general,with laser sources, material dispersion isnegligible; however, with LEDs, it cannotbe ignored at high modulation frequencies.

Multimode distortion is caused by changesin propagation velocities with changes inmode.

The ultimate in fiber bandwidth is achievedby making the fiber core so small that onlyone (low -loss) mode can propagate. Thisresults in core diameters of only a fewmicrometers, however, and it is extremelydifficult to couple significant amounts oflight into such a single -mode fiber in astable, reliable way. For this reason, mostpresent effort is being spent on multimodefibers. The propagation velocities ofoptical signals are different for the differentmodes of such fibers, leading to distortionof the optical signal. It is more serious forstep -index than graded -index fibers. Bothtypes of fiber can be modeled by the radialindex of refraction distribution

n(r) = [I - 2J (r/ a)"]" (9)

taking a - 00 for the step -index case.

The numerical aperture of a fiber is definedas the sine of the half -angle of its accep-tance cone. For a step -index fiber, this canbe expressed as

NA = (2 no AY' (10) unction, v, ill result in the least inters in hol

interference.

The solid angle of the light acceptancecone, and hence, to a first approximation,the amount of light that can be coupled intoa fiber, varies as (NA )2, and thus, for a step -index fiber at least, as A. If a preliminarylink design indicates that the system ispotter limited (inadequate SI N), oneshould consider fibers with a larger A,(larger NA). However, if dispersion limitslink performance, smaller -aperture fibersare advantageous, as we now show. Glogeand Marcatili" have shown that the spreadof propagation times for a step -index fiberis

(nL c) (II)

while for a parabolically -graded -indexfiber, it is only

(nL, c)(-1212). (12)

Since A, the (maximum) fractional indexchange across the fiber, is generally onlyabout 0.01. the advantages of graded -indexfibers are obvious.

A note of caution is due here. The abovediscussion assumes that the distortionvaries linearly with transmission -pathlength. For fibers greater than about I 2

km in length, it has been observed thatdispersive effects scale in a non-linear waywith fiber length, due to mode -mixing andother effects.'"

Eq. I I indicates that one has less disper-sion in fibers with lower A, i.e., lessindex difference between core andcladding. However, going to such a fiberhas two serious disadvantages: I) theguiding becomes weaker, and losses atbends and imperfections can becomeserious; and 2) it greatly reduces the accep-tance cone for light that can propagate inthe fiber, as mentioned above. In a digitalsystem, where distortion is tolerable until itbegins to cause decision errors at thereceiver/ regenerator, other criteria aregenerally employed. Intersymbol in-terference between succeeding symbols isof vital importance, and such concepts as"eye -diagrams"" become useful. The pulsedistortion will depend critically upon thedetailed frequency spectrum of the digitalsymbols transmitted (and on the phaserelations between them). This means thatthe designer must choose that set of sym-bols which, when convoluted with the full(amplitude and phase) modulation transfer

ConclusionsIn an optical fiber communications link,noise and distortion play vital roles. Mostnoise is associated with the photodetector.If a PIN detector is used, the most seriousnoise source is the thermal noise in thedetector load resistor, indicating that ashigh a load resistor as feasible should beused. With an avalanche detector, its gaingenerally makes the shot, or quantum,noise in the detection process dominate.Avalanche detectors are more sensitive, buttheir high bias -voltage requirements,higher cost and need for temperaturestabilization may make them less desirablefor links where the received light is strongenough to permit use of PIN detectors.

Using injection lasers makes greatly in-creased signal strengths possible. Highermodulation capabilities and lower dis-persions also result from using lasers.However, their threshold nature makesthem require careful temperature stabiliza-tion, with added circuit complexity, highpower requirements, and higher cost. Thechoice of fiber will depend upon therequired attenuation characteristics,numerical aperture, and dispersionproperties. In general, one expects proper-ly fabricated graded -index fibers to besuperior to comparable step -index ones formost purposes.

References

I. Anderson, L.K. and McMurty, B.J.;-High-speed photodetec-tors,- Proc. IEEE, Vol. 54 (1966) p. 1335.

2. Goell, J.E., "An optical repeater with high -impedance inputamplifier." Bell Sys:. Tech. J. Vol. 53. (1974) p. 629.

3. Ettenberg, M., Lockwood, H.F., Wittke, J.P. and Kressel, H.;"High radiance, high speed Al.Gai-,.As heterojunction diodesfor optical communications." Tech. Digest. Int. Elec. DevicesMeeting. (1974) p. 317.

4. Schwartz. M.; Information Transmission, Modulation, andNoise, McGraw-Hill Book Co., N.Y., 2nd Edition, 1959; Ch.5.

S. Lockwood. H.F., Wittke, and Ettenberg, M.; "LED forhigh data rate optical communiations," Opt. Comm. Vol. 16,(1976) p. 193.

6. Schicketanz, D.; "Modulation von plliumarsenid-laserdioden," Simens Forsch.-u. Entivickt-Ber, Vol. 2(1973)p. 218.

7. McCumber, D.E.; "Intensity fluctuations in the output of cwlaser oscillators-I," Phys. Rev., 141, 306 (1966).

8. Gloge, D.; "Dispersion in weakly guiding fibers," Appl. Opt.Vol. 1011971) p. 2442.

9. Gloge. D. and Miranda, E.A.J.,; "Multimode theory ofgraded -core fibers," Bell Sys:. Tech. J., Vol. 52(1973) p. 1563.

10. Chinnock, LL, Cohen, L.G., Holden, W.S..Standley, R.D.,and Keck, D.B.; "The length dependence of pulse spreading inCGW-Bell-10 optical fiber," Proc. IEEE Vol. 61 (1973) p.1499.

1 I. Lucky. R.W., Selz. J. and Weldon, E.J., Jr.; Principles ofData Communication. McGraw-Hill Book Co., N.Y., 1968,Ch. 4.

12. Oliver. B. M., Pierce, J. R. and Shannon, C.E.; -Philosophy ofPCM.- Proc. IRE Vol. 36 (1948) p. 1324.

63

Optical propagation through turbulent airD.A. de Wolf Turbulent air causes optical rays to deviate from

otherwise straight paths by irregular slight undulations,and it also changes their phase velocities irregularly. Thesefluctuations cause a host of unpleasant degradationsto communications signals.

If you have any prior notion about "at-mospheric turbulence," you probably havein mind some blurry stars and indistinctfeatures of the moon viewed throughtelescopes. That is not a bad starting point.Before 1945, no one appears to have givenmuch thought about the subject beyondthis, and astronomers usually shruggedtheir shoulders and toted their equipmentup to high, poorly accessible mountainpeaks where the effects of the atmosphere

David A. de Wolf has been at RCALaboratories since 1962, working on wavepropagation in the atmosphere, radiationfrom antennas, communications inplanetary atmospheres. and diffractionoptics, among other areas. Dr. de Wolf is oneof the early contributors to the un-derstanding of multiple -scattering effectson high -frequency waves propagatingthrough random media such as turbulentair.

Contact him at.Materials Research LaboratoryRCA LaboratoriesPrinceton, N.J.Ext. 3023

are much reduced.' This was a brute -forceapproach, highly effective at the time, andthis article would end here were it not forthe development of high-powered radarsafter World War II, and -around 1960-the advent of the laser, enabling stronglight beams to propagate through theatmosphere. It is not always practical toplace huge radars on mountaintops, andlaser beams are often required elsewhere.

I his created an interest in understandingthe influence of turbulent air on elec-tromagnetic waves propagating throughthe atmosphere, and lately this interest hasincreased because the prospect of doingsomething about it appears imminent. Thecapabilities of present-day computers andprocessors are such that novel techniquesare being contemplated: techniques thatallow one to gather, process, and makereal-time corrections to maneuverablemultifaceted lenses to compensate for thedistortions caused by atmospheric tur-bulence. But I am getting ahead of myself,

for I have yet to tell you what thesedistortions are.

The simplest description is often the best,and always the most attractive to the non-initiated! It seems helpful to use a radar -like picture because it covers the effects onboth radar and optical waves. One canconsider a point transmitter spewing out abeam in a certain sector of the sky at someangle as in Fig. I. You will find that thepower received at some location at distanceR, elevation angle 0, and azimuth 4, fromthis transmitter is the absolute square of acomplex electromagnetic field that consistsof three obvious factors in free space:

I) a I R dependence on the distance;

2) an angular gain factor telling you howmuch this field at 04 is less than it wouldbe on -axis at the same distance;

Reprint RE -22-4-12Final manuscript received May 27. 1976

RECEIVER

t\

el\\,r \1)."/.7 \

8 , ci)

TRANSMITTER

EDGE OF BEAM

Fig 1

In free -space propagation, power received is a function of radius, angle and phase

64

Fig. 2Intensity losses are caused by absorption and turbidity.

3)a phase factor essentially equal to thenumber of wavelengths in distance Rtimes 27r.

Laser -beam propagation in free space canbe more complicated in the presence offocusing optics, but for this survey it is

sufficient to group lasers under pointsources and to ignore the extra complexityof focusing. The above power is really anenergy flux and is measured' in W / in2 orsome other outmoded non -MKS system ofunits. At large distances, objects fillingonly a tiny arc of angle "see" a practically -plane constant -phase surface; for suchobjects one may consider the source to beinfinitely far away. On the other hand, thegain factor does vary with 0 and 47 atconstant distance R but-again-for dis-tant objects the change in this factor overthe small arc will be quite negligible. Theamplitude will appear, merely, to change asI / R with increasing distance-as it should,for energy must be conserved from onesphere to the next. By "amplitude," wereally mean the factors I and 2 thatmultiply the phase factor 3 above.

How the atmospherediffers from free space

In free space no more tactors are needed. Ina medium such as air, water, or glass,through which light can propagate, afourth factor is needed to help express thechanges caused by the medium. This fourthfactor can affect both amplitude and phase,and consequently it is complex. We name itu( R,04) and note that it is unity, of course,in free space. This survey focuses entirelyupon this new factor u(R,0,40).

Atmospheric gases absorb radiation.

Now for atmospheric effects: first of all,certain atmospheric gases-particularly

water vapor and carbon dioxide-are ableto absorb electromagnetic energy at opticalfrequencies from the beam and convert thisenergy to internal processes ultimatelyresulting in heat, i.e., convert light intoradiation at entirely different portions ofthe electromagnetic spectrum. Not only aredetectors at the receiver insensitive to suchradiation; more significantly, such radia-tion is reradiated more or less isotropicallyfrom each air molecule and thus essentiallylost. Because such absorption losses areusually proportional to the amount ofoptical radiation incident upon a sphericalslab and to the thickness of the slab, theselosses give rise to an exponential decreasein intensity 1(R) from that at thetransmitter (in any direction over andabove the 1 / R2 loss due to distance). If wewrite 1(R) = 1(0) exp(-2aR), then arepresents the loss per unit distance. Forsimplicity, I have assumed we're lookingalong the axis, and the medium is uniform.(See Fig. 2). This simple formula is easilyextended for other directions and non-uniform media (but editorial restraintprevents me from doing it here!) The majorpoint here is that all that u( R,04) consistsof is a simple attenuating factor,exp(-aR), for 0 = 0, = 0 in this case ofuniform absorption.

Turbidity can scatter radiation.

Atmospheric turbidity -to be carefully dis-tinguished from the concept ofturbulence-is so similar in concept toabsorption that confusion is easy. Turbidi-ty implies the presence of air particles in thebeam that do not absorb, but scatter,incident optical radiation in all directionsso that-again -scattered radiation for allpractical purposes does not reach thereceiver.' Like absorption, its effect is

expressed by a loss per unit length. The

Definitions

Cr

h

e

k

L

L.,

1

IF

n

an

R

T

U

r

0

0,,

0,A

P.

ar

43(k),

ct7

refractive -index structure constant

temperature -index structure constant

vapor pressure

frequency

altitudeintensity of radiation

wave number

pathlength

largest possible scalelength

smallest possible scalelength

Fresnel radius

refractive index

deviation of n from free -space value

pressure

spatial coordinate

distance from source

absolute temperature

float velocity

field factor used to characterizean electromagnetic field nottravelling in free space

absorption -loss coefficient

angular coordinate (elevation)angle of ray diffraction

angle of ray deflectionwavelength

radius vector

coherence length

scattering constant

turbulence spectrum

angular coordinate (azimuth)phase angle

auto- correlating an ordered series ofcorrelation observations with the same series

in altered order

study of these loss quantities is obviously ofparamount importance to tnose interestedin solar energy.

The atmosphere's refractive index is notconstant.

I he next item on the list is refraction. Inperfectly uniform air (i.e., in the absence ofdensity gradients, temperature, pressuredifferences, etc.), the wavelength is not justA. According to a famous optical theorem,the original incident free -space radiation isabsorbed by the medium and reradiated"instantaneouly" in the same direction at anew wavelength A l[n(r)] where n(r), therefractive index, varies very slowly overmany wavelengths. (Otherwise, the concept

65

of "medium" makes little sense.) Of coursen(r) is very close to unity for such sparsemedia as atmospheric gases.

Herein lies a new problem: the atmosphereis not uniform, but its density [andtherefore the deviation of n(r) from unity]varies with altitude because ofgravitational forces. The opticalpathlength [proportional to the refractiveindex n(r)] decreases somewhat withaltitude, and the locus of all locations withthe same optical pathlength as the distanceR along the central axis of radiation is nolonger a sphere. The situation of Fig. 3holds. Because rays are defined as lighttrajectories perpendicular to the those loci,the wavefronts, it follows that rays are nowdeflected towards regions of higher n(r),i.e., greater atmospheric density, i.e..downwards. Therefore, the apparent loca-tion of stars close to the horizon may beseveral microradians higher than their realdirectional location. Fortunately, thiseffect is not dependent on frequency (to agood approximation), and the densitygradient is slow enough that distortion ofobserved objects is negligible in clear -atmosphere conditions.

This finally leads us to the following:suppose now that we allow for localfluctuations in the refractive index n(r)around the mean profile, and suppose alsothat these fluctuations occur on some scalemuch less than the e -folding length of 8 km(or so) of the altitude profile of air density.If, on the other hand, these fluctuations arenot microscopic (e.g., at the molecularlevel), then a situation as sketched in Fig. 4may occur. Rays are deflected irregularlyfrom their otherwise straight paths andtheir optical pathlengths also deviate fromwhat they would have been in free space.

\

.1.14FREE V.-. --1 \\1__--

__I- \- \__.,..--_,L--_-::-1-- P- -4 ---1 \

1 \ \

IN AIR

Fig 3Rays are deflected toward denser airbecause the atmosphere's refractive indexchanges with altitude.

INTENSITYTURBULENCE

ABSORPTION LOSSES

SCATTERINGLOSSES

AIR

VACUUM

0//

Alp

BLOOMING

/ DEFOCUSING (BY TURBULENCE)

THERMAL DISTORTION

\\ WIND OR AIR MOTIONRELATIVE TO BEAM

//////////////////////////////////////////////////////////EARTH

Fig. 5A laser beam traveling through air faces a number of distorting effects.

This situation is a typical effect of clear -airturbulence, which produces just this type ofrefractive -index fluctuation.

Changes to the signal

All these ettects, and some others, aresummarized for laser beams in Fig. 5. Letus see what these irregular deflections andchanges in optical pathlength do to oursignal parameter u(r):

Phase changes are 'caused by refractive-

index variations.

Consider a narrow ray bundle passingthrough a bunch of fluctuations in n(r). By"narrow," I mean that each ray in thebundle goes through almost the same stringof fluctuations, e.g., as in Fig. 6. Theoptical pathlength, or phase, is 2.7r times the

Fig. 4Clear -air turbulence deflects rays from theirnormal straight path

number of wavelengths in the actualpathlength R. Because A cc n- I, it followsthat phase is proportional to refractiveindex. Indeed, phase in the turbulentmedium differs from that in free space,namely 27rR/ A, because R is replaced bynR if n is a constant, and by a sum of n,AR,

for each path segment of length O T, with itsown refractive index n,=--- I + on,. Fromthis sum you can see that the phaseincrement over and beyond the free -spacephase involves a sum of &OR,. This sumbehaves like a Gaussian random variablewith zero mean (6n, can be positive ornegative), and therefore the total phasefluctuates about the quiescent free -spacemean.

Because phase is a means of expressing theconcept of a wavefront, the coherence of awavefront is strongly related to coherence

Fig 6Phase fluctuations are caused by variationsin phase velocity when light travels throughirregular refractive -index changes.

66

(a) DIFFRACTION

( b) REFRACTION

Fig. 7Small -angle scattering redistributes energy unequally and soproduces amplitude changes.

of phase. ("Coherence" is just a measure ofhow closely one point of a wavefrontfollows motions of a nearby point.) And,because phase is linearly related torefractive index, it follows that phasecoherence is determined by refractive -index coherence, i.e., by the spatial ortemporal correlation of the refractive in-dices at two locations. Indeed, thisautocorrelation of n(r) is a basic propertywe must know in order to predict wavebehavior.

Amplitude changes can be caused by small -angle scattering.

Although none of the energy -lossmechanisms depicted in Fig. 2 hold here, itis possible that energy is redistributedunequally at distance R by two small -anglescattering mechanisms. One is by diffrac-tion (edge scattering) from more -or -lesscoherent fluctuations; the other is

cumulative refractive deflection in a

random -walk fashion from a string of suchregions. These are illustrated in Fig. 7;discussion will resume later.

Imaging andcommunications parameters

In a typical imaging application, light isgathered by a lens and focused to a smallspot at a distance f beyond thelens. The lens "sees" the electromagneticfield incident upon it, and in particular itsenses the extra factor u(r) over and abovethe normal free -space field (see discussion

it 0 S= Z

Fig. 8Turbulent medium deflects a very narrow light beam.

above). The mathematical property offocusing entails basically that the elec-tromagnetic field in the focal plane at fbeyond the spot is a Fourier transform ofthat part of the factor u(r) corresponding topoints on the lens [u(r) is set zeroelsewhere]. The focal -plane field is

therefore linearly related to the factor u(r),which contains the turbulent phase correc-tion (and also amplitude corrections).

Workers in optics often prefer to examinethe mutual transfer function (MTF) of thelens: this quantity is the spatial Fouriertransform of the intensity in the focalplane. To say it another way: analyze theintensity in the focal plane into sine -wavecomponents; the coefficient of each com-ponent forms a function of the wavelengthof that component that is just the above -mentioned MTF. As the MTF is a Fouriertransform of the square of a quantity that islinear in u(r), it follows mathematicallythat the MTF is linearly dependent uponthe autocorrelation of u(r). For long ex-posure times you can show that the lensMTF is simply a product of thisautocorrelation with the free -space lensMTF. The only point I wish to stress in thiscontext is that the deterioration of imagequality (blurring) is determined byparameters such as the wave -field spatialautocorrelation.

Another parameter is illustrated in Fig. 8: avery narrow beam twists and bends as itpasses through a turbulent medium. By"very narrow," I mean that all three rays

sketched in Fig. 8 "see" essentially the samerefractive index at each path point s. Thephase front is approximately plane andperpendicular to the middle rayeverywhere. By remembering that the

phase of rays I and 2 at s can be related tothat of ray 0 at s = z by simple geometry,one obtains

ill (z) + 1/2k pO,

i(z) - ihkpO, (1)

It follows that the ray -angle of deflection is

0, [1,112(z)- kP (2)

The above should also make clear that 0,,the cumulative angle change, is a fluc-tuating quantity. More importantly, thisangular deflection is given by the phasedifference 1#2(z) - Thus, certainstatistics, such as the variance of thisdifference, known as the phase -structureJunction, are important for determiningdirectional variations. Note that the meanis zero, so it is not very interesting.

In another application, narrow laser beamsare directed at objects to be illuminated.The important intensity factor at the il-luminated object is 1(r), where r is at z = L,and r= (p,L). The laser forms a spot in the z= L plane, and the intensity l(p,L)drops offsharply as p varies from the spot center.Problem: how do we describe the spotcenter? One methods is operational-just

67

weight the intensity with the radius vector pto obtain a formal spot radius po thatfluctuates in turbulent air because theintensity fluctuates both as a function of pand of time 1. Thus, statistics of p must becalculated, and in particular (po) and (po')are of interest. These two requireknowledge of the intensity and its spatialautocorrelation, two other importantparameters.

Ideally, we would wish to know all thestatistics of these quantities, and thatwould entail knowledge not only of intensi-ty and phase autocorrelations (which alsoimply mean square, of course), but also ofcrosscorrelations and a veritable Pandora'sbox of higher statistical moments. I wishedonly to point out here that a good deal ofbasic information is contained in these fewlowest -order statistics of phase and intensi-ty.

Amplitude andphase fluctuations

We have now reduced the problem ofdealing with imaging and communicationin a random medium to one of un-derstanding the fluctuations of phase tir andamplitude A (or intensity / cc A2) of thestandardized electromagnetic -field factoru(r). In particular, I have argued, withsome intentional oversimplification, thatmuch of what we need to know is given byphase and amplitude means and variances,and several spatial autocorrelations, inplanes normal to the line of sight. Havinghammered home this point, let me proceedto discuss the basics of phase andamplitude in random media:

Phase fluctuations can be very large overlong paths.

1 equated "phase" with "optical path -length," and indicated that the latter differsfrom a free -space optical pathlength byweighting each path segment by therefractive index. The phase increment(difference of optical pathlengths in ran-dom medium and in free space) has pathsegments weighted by 6n, the deviation of nfrom its free -space value of unity. It can beargued that this increment is a normalrandom variable with zero mean. It canalso be shown that its variance is given by

Variance of ils = k'LL0((0n)2)= LLos (3)

where k is the wavenumber, L thepathlength, L a representative scalelengthof the fluctuations (theory shows L. is the

largest possible scalelength), ((M)') is ameasure of the strength of the fluctuatingmechanism, and C' will be discussed short-ly. Observations indicate that 6n 10'and 1., I m, so that it follows that thevariance of tit can be very large at opticalfrequencies for kilometer -long paths or so.

Angular deflection depends upon the"refractive -index structure constant."

Given the above knowledge of Vs, and thefact that the angular deflection 0, is givenby Eq. 2, it can be inferred that 0, is also anormal random variable with zero mean.Only, in this case, the pathlength isweighted with 6n(1)- 6n(2), the differenceof the refractive index at points p/ 2 aboveand below the path segment. In order tocalculate further, a model for the statisticalbehavior of this difference is needed. Tur-bulence theory gives the fundamental "two-thirds" law:" when p lies between themicroscale / (smallest possible scalelength)and macroscale L., then

variance of [6n(1) -6n(2)] = C.2 p2/3 (4)

where C2 is known as the refractive -indexstructure constant.' It can then be shown,by delving into more detail (we shall not dothat here), that

variance of 0, = C.2 LI. (5)

Here, it is interesting to point out that, as pcannot exceed Lo in Eq. 4, it follows thatthe variance of the left side of Eq. 4 tends to(6n2), and the right to C.24,2/3. Substitutionfor (6n2) in Eq. 3 leads to the alternativeform for the variance of IP. In theatmosphere, /0 I mm and

121111-2' (queer unit, but that'swhat it is!); hence the rms angular deflec-tion for L = I km can be as much as 0.1mrad, according to Eq. 5. This may appearto be small, but it means that the rmstransverse shift of a ray, LO, 10 cm,which is not necessarily negligible. Thereare other consequences to this Gaussianrandom ray dancing with rms shift 10 cm

that will come into discussion; however, itis preferable to discuss intensity effects.

Eddies produce intensity scintillations.

Intensity scintillations, or amplitude fluc-tuations, are more difficult to clarify at anelementary level. In the scattering -of -lightregime under which turbulent deflectionfalls, the wavelength A is typically muchshorter than any of the refractive -indexfluctuation scalelengths /. In particular,< 1, which of course implies A < /. In Fig.

7a, the effect of one coherent fluctuation ofdiameter /-an eddy of size I,colloquially-is shown. It creates a

diffraction pattern at z = L because the"edges" bend light rays by an angle 040,--X/1 towards higher 6n.

There aren't edges in reality, but a 1/ edropin strength does define the diameter 1. Thesituation is a little different from classicalslit diffraction. The edge ray, diffracted byangle 04/) A / /, meets an undeflected rayat z = L that left z = 0 at a transverseseparation L0 (I)- A If this separationis much less than the Fresnel radius IF =(AL)", it follows that there is no in-terference effect. But LOa (I) < IF - AL/ I <(AL)"'- I > IF. Here is the first complica-tion. Only those eddies with / IF canpossibly give rise to interference, and hencediffraction effects, leading to scintillation.On the other hand, when / < IF, thediffraction pattern will be as sketched forthe top eddy in Fig. 7a: the diffractedenergy is spread over many Fresnel zones,but is diffused to such a weak strength thatit is of little importance.

Thus, it appears that out of the entire rangeof scale sizes / with / 1< L. (with /,, - 1mm and L. - I m in the lower at-mosphere), only those eddies with I ^ IFcontribute important amplitude fluc-tuations. It is easy to convince oneself withthese cited scalelength limits that /0 < IF<L. for km -length optical links; hencediffraction effects, superimposed on whatotherwise appears to be a geometrical -optics medium, must be accounted for.However, as only a small portion of thespectrum of eddy sizes is affected, it followsthat these diffraction effects remain veryweak.

However, as C'02 increases, cumulativescattering and interference effects increase,and intensity scintillations increase. Thiscan be clarified by returning to Eq. 5, whichis a special case of

variance of 0, (/) C''Lr' (6)

for cumulative refraction by eddies withsizes / and larger. Remembering that / IF

(Al.)''] for important diffraction effectsby one eddy, we set /- IF in Eq. 6 andcompare LO,(lF) to L04/F) to see ifcumulative effects outweigh single -eddydiffraction. That is, squaring the abovequantities, and rearranging the factors oneach side somewhat, we compare

C;k7 °LH6 to /1.2. or 6 Lu° to I.

68

NE

NC

TIME 8 SEPTEMBER, 1968

Fig 9Refractive -index structure constant is far from being a constant. "Turbulent -weather" dayproduces order -of -magnitude changes.

Thus, we note that the curiousdimensionless combination C.' Ic2 L'I'b-often referred to as or- or somethingsimilar -determines whether or not thescattering is strong. When T2 > I, it is

strong and effects will be very noticeable.The physical meaning, I think, of theseparameters, which you will encounter allthrough the literature, is thus clarified.

In summary, we note that several size andstrength parameters govern phase andamplitude fluctuation. Three lengths arecrucial:

/: the microscale of turbulence,It: the (intermediate) Fresnel scale

determing amplitude effects,L,,: the macroscale,

and three strength parameters,

C::the strength ofrefractive -index variations,

or2:the strength ofamplitude scintillation,

C2k2LL,,' `:the strength ofphase fluctuations.

Space precludes more exhaustive treat-ment but there are other importantparameters. For example, there is a

coherence length pc such that when I pi - p2I

exceeds Pr, the autocorrelation of theelectromagnetic field will be very small.This wave -field coherence length is givenby

(C,2 k2 !Far" (7)

It can he shown that IF is the

coherence length for weak amplitude(intensity scintillation), whereas when are>I (strong scintillation) the coherencelength reduces to p which is then less thanI. For weak scintillation, the elec-

tromagnetic is almost unchanged inamplitude: the phase SO fluctuates oververy many times 27r, and thus the fieldbehaves like a log -normally distributedrandom variable. Refraction is weak,hence is the important correlation length.For strong scintillation, p, decreases waybelow /t, and suddenly all kinds of rays thatformerly were in phase with undetectedones interfere with these neighboring onesdestructively. The field distribution thenchanges towards the expected Rayleighdistribution. The intermediate case, wherea 7- ^ I, gives all the headaches to workersin this areat the statistics are then almostunpredictable.

What causesthe fluctuationsWe have seen that communications -systemparameters are determined by the behaviorof an appropriately normalized elec-tromagnetic field, aside from normal free -space geometry and other parameters.Amplitude and phase of this u -function aredetermined in turbulent air by refractive -index fluctuations. It is time to explainwhat, in turn, causes these. For opticalfrequencies, a good approximation for n interms of pressure p (millibars), temperatureT (degrees Kelvin), and vapor pressure e(millibars). is

n = 1 + A (pi 7) [I + B(elp7)] (8)

where

A 77.5 x 10 " [I + 5.15 X 10-' A -2+ 1.07X 10 4 A 4]

with wavelength A in µm. B is a numericalconstant unimportant here. Often-butnot always! -humidity fluctuations or con-tributions can be ignored, and of course thehigher -order A contributions are unimpor-tant at visible or infrared frequencies.Because pressure fluctuations are veryshort-lived it follows that

Sn -Aporl = (I - n)dTl T (9)

and thus on is proportional to ST at anyaltitude (remember that p is a strongfunction of altitude). In practice, onemeasures not refractive -index, buttemperature fluctuations, and inparticular for turbulence one finds themean square of ST, or more precisely

([71n) - Tkr2)r) = Cr P" (10)

where In' - rA = p in the so-called inertialsubrange of turbulence (4, < p < L.). Ofcourse, this is just the fundamental "two-thirds" law mentioned in connection withEq. 4. You see from Eq. 9 that C,,2 is simplyproportional to Cr' by the factor A p I 7'2,and in practice that is how values of C2 areobtained. However, a caveat is highly inorder: there are often humidity fluc-tuations to be accounted for, and then therelationship between C.2 and Cr' is

different.

Now back to C2, the refractive -indexstructure constant that is not a constant.The behavior of C,2 (altitude, time) ishighly erratic. Fig. 9, taken from a

representative publication, indicates thatC,,- varies over several orders of magnitudeduring a typical "turbulent -weather" day.Fig. 10, taken from a now somewhatoutdated work, shows the behavior of Cwith altitude (unfortunately the unit of theC axis is different). More recent work byits author' indicates that C,,2 itself is a veryrandom variable. However, at ground level(and up to 150 m or so) a semi -empiricalmodel has been worked out,"' predictingC' cc h 2 to Cn2 «h' '; the former extremefor quiescent (night) conditions and thelatter for highly unstable (midday) con-ditions. These trends compare well with theinitial portions of dawn/dusk and sunny -day curves, respectively, of Fig. 10.

69

The estimates of macro -and microscales ofturbulence (L and /) are more com-plicated. Briefly. L,, is proportional toaltitude h, but as the strength factor C,,2dies out rapidly with increasing h, I mappears to be a reasonable estimate forhorizontal propagation just above the sur-face. But, for slant -path propagation, avariable L must be used. Fortunately,most system parameters can be expressedso that L,. does not enter into them: I.,. is inEq. 3, but not in most expressions in whichthe phase fluctuations usually enter intoresults! On the other hand, 1 is determinedby a heat cascade process: the largest eddiesbreak up into smaller ones, conservingtheir eddy -motion energy, until at somesmallest size / they prefer to give up theirenergy as heat (and the eddy dies out). ThusI., is a function of the Reynolds number.Fortunately, / enters into only somecalculations and then usually as somefractional power such as 1 i %. The readeris referred to the literature for judiciousguesses.

Something has been said about thestrength of on see Eqs. 9 and 10 -butthe fundamental parameter is its spectrum.That is to say, for homogeneous turbulenceone can write (6n) as an integral of thespectrum 43(k) for all possiblewavenumbers k (k = 2 70). The fundamen-tal theorem of turbulence similarity theorystates (and it is usually quoted as theKolmogorov or Obukhov-Kolmogorovspectrum) that the spectrum 47(k) is

proportional to C'k No form can hegiven for small wavenumbers k<27r/ 1,,, andsometimes a Gaussian roll -off is given atthe large k (small eddy) end, but is basicallymeaningless to do more than to absorbthese unknown sections into the definitionsof 10 and /. Filter factors usually enter intothe calculations as factors of 47,,(k) toprevent its behavior for both small or largek from seriously influencing the effect ofthe inertial subrange 27r/ L < k < 27r//.Another caveat: one often finds an "-II/ 3law for 47(k) cited in the literature. Thereason is that -in contrast to the above-athree-dimensional 414k) can be defined,which differs of course by a factor 47rk2 forisotropic turbulence.

Bandwidth

Just a word about time effects. Most of thebehavior with time is adequately describedby the frozen -flow hypothesis: things varyat one location in time because a "frozen"irregular spatial structure floats by. Fre-

SUNNY DAY

\ .".ci I

- DAM /DUSK \ I

10-9

10-10Ins

MINIMUM

( HE IGHT

1

10m1

100m

ALTITUDE

,N DISTURBEDI V LAYERS

II

I

4

,II

t1 `1t,

f.

CLEARNIGHT

(HEIGHT (2/3

1km 10km 100km

ABOVE LOCAL GROUND

Fig 10"Constant" varies with altitude as well as time. Model assumes that local ground is level, noton a mountaintop. Direct comparison with Fig. 9 is impossible because of the different units.

quently, scales are translated from / tofrequency! by the simple relationship f -U11, where U is the float velocity. Thus,Eq. 10 manifests itself at one detector but attwo different times as a frequency effectCif 1. In the lower atmosphere f issandwiched between U/ 1. and U/ 10, and sofor U^ 10 m/ s, between 10 Hz and 10/kHzor so, but at lower wind velocities, thisspectrum may shift.

Final commentsI n summary, the subject has beendeveloped actively in the last ten years.Pertinent literature is cited in Refs. 5 andI I. and even the casual peruser will noticethat most of it is in only four or fivejournals: J. Opt. Soc. Ant., App/. Optics,Radio Science, IEEE Trans. Antennas andPropagation, and the Russian journalRadiolizika--IVUZ (also available intranslation, as Radiophysics and QuantumElectrodynamics, but with several yearsbacklog). Tatarski's textbook (Ref. 6) isconsidered to be the standard primer, but isnow somewhat dated as it was written in1967.

The following recent development is in-teresting. With the advent of fast large -memory computers, it has become possibleto process atmospheric -turbulence infor-mation and correct for it in real time. Thisis especially valuable for astronomicalapplications, where the major error -causing effects are the undulations (and

overall retardations or advances) in thephase front. A telescope mirror is strucknot by a plane wave but by an undulatingwave. It is now possible to correct for thisby using deformable mirror elements,process information about the wave -tiltstriking each element, and then adjust eachelement in tilt by a servomechanism toobtain the sharpest image. The raisond'etre of this development is-of course-the dominance of phase shifts (over otheramplitude -modifying causes) caused byturbulent air layers.

ReferencesI. At an altitude of 3 km. the density of the atmosphere has

decayed to 60', of its sea -level value. and thus a good fractionof the atmosphere already lies under our astronomer.

2. 1 here is a photometric coterie that insists upon foisting acontusing terminology and notation upon the scientificcommunity see the RCA Tier tro-Opurs Hand/nut A. Sec-tion 2. tor a patient exegesis hut their recommendat ions areWien including here) ignored.

3. I am osersimplilying here. Actually. for very dense mediasuch as clouds. rescattered radiation is quite important. to theextent that reradiation in other than the original directioncan he comparabk in magnitude to the remainder of thebeam. On the other hand, this is not the case in the almost -clear atmosphere suitable lor optical propagation.

4. Fried. ITT.; J. Opt. Soc. Am.. Vol. 5611966) p. 1372.

5. 1 -ante. R.I..: PIO, IEEE. Vol. 63 119751 p. 1669.

6. I atarski. V'.1.; -El/era ol the Turbulent Atmosphere on HarePropagation." Itransl.) U.S. Dept. of Commerce IN Ilbw. 11 66-50464.

7. Not unlike many names coined by present-day researchers.this one is a misnomer: C.' is not constant at all, neither intime nor in space!

K. Lawrence. R.S., Ochs. OR.; and Clifford. S.F.; J. Opt. SOC.Am.. Vol 60 11970) p. 516.

9. Hulnagel, R.E., Digest of Proceedings. topical Meeting onOptical Propagation 1 hrough Turbulence. July 9-11. 1974.Boulder. Colo.. Paper WA I.

10. Wyngaard, J., 1/urni. Y.; and Collins, S.A; J. Opt. Sot-. Am..Vol. 61 11971) p. 1646.

II. Prokhoros. A.; Bunkin. F.; Gochelashvily. K.; and Shishov.V.; Proc. IEEE Vol. 63. 11975) p. 790.

70

Integrated opticsand optical waveguide modulators

J.M. Hammer Optical communications requires all the switching andmodulation functions of its electronic counterpart. Recentlydeveloped waveguide modulators are capable of operatingbelow 10V and 0.5 mW/MHz.

The term integrated optics usually appliesto a system performing a variety ofoperations on light guided in a single plane.Such a system should combinetwo or more active functions, such asmodulation, spatial switching, frequencyconversion, amplification, generation, anddetection. An ultimate integrated -opticssystem, then, would be a monolithic opticalwaveguide containing a variety of activeand passive subelements coupled togetherto achieve the desired functions.The driving signals are frequently thoughtof as being electrical, but may well beoptical. Similarly, the output could beeither an electrical signal or light. In thelatter case, the output light could be sent onto a transmission system, such as a fiber-optic waveguide, or viewed directly.

Comparing electronicsand integrated opticsAlthough the role of integrated optics is notyet fully established, the hope is that suchcircuits will not only replace electronics incertain applications, but will providefunctions not previously available usingelectronic methods. An example of such a"new function" would be the direct man -machine interface made possible by

reading out the light from an integratedoptical processor. Some of the hoped -foradvantages of integrated optics circuits are:ruggedness; insensitivity to temperatureextremes, high humidity, and high ambientradiation; resistance to electrical in-terference and cross -talk; operation speedslimited by light transit times rather than

Reprint RE -22-4-11Final manuscript received June 18, 1976.

The research in this paper was jointly supported by theOffice of Naval Research, Arlington, Va. AvionicsLaboratory. Wright -Patterson AFB, 0. and RCALaboratories, Princeton. N.J.

capacitive effects; bandwidth capabilitiescomparable to light frequencies; low drivepower requirements; and finally, relativeproduction simplicity. In addition,integrated -optics systems are compatiblewith fiber-optic communication and data -transmission systems, about which more issaid below.

To date there have been few examples ofexperimental integrated optic circuits hav-ing more than one active operation on asingle substrate. Since most work has beendevoted to single discrete devices, thisreview is mainly concerned with thin-film dielectric optical waveguides, whichcan he used for active operations such asswitching and modulation.

An encouragingly large number of discreteactive operations on light traveling in

planar optical guides has indeed beenreported. These include switching' (deflec-tion in the waveguide plane), modulation, -amplification,' and generation.' Theseactive devices do in fact show many of thehoped -for advantages. In addition, a largenumber of important passive operationshas been demonstrated. These include allthe refractive effects analogous to thoseperformed with conventional optics: dif-fraction, directional coupling betweenplanar guides, and coupling light into andout of optical waveguides.

Advantages ofoptical communication

The related fields of optical communica-tion and data transmission probably holdthe most important applications for in-tegrated optics in the near future. The ideaof sending information using optical -frequency carriers even predates lasers, butthe idea neared fruition when lasers

became available. This strong historicalinterest in light carrier systems is based onthe high frequency of optical waves, whichallows tremendous bandwidths to be con-templated. For example, as system usingonly 1% of the center frequency of 6000-Alight could transmit 5X10'2 Hz (50,000digital iv channels). The short wavelengthallows the use of small antennas, a fewmillimeters as compared to a few meters formicrowave systems. In addition, since lightquanta are fairly energetic, efficient detec-tors such as PIN or avalanche detectors canbe used. Because the atmospheric transmis-sion of light is severely limited by ab-sorptions and made hopeless by foulweather, it became necessary to considerclosed transmission systems or "pipes"enclosing more -or -less conventionallenses. More recently, the advent of bothmulti- and single -mode fiber-optictransmission lines gave additional reasonto study optical -frequency carrier systems.

Fiber-optic systems

The description of light propagating in atransparent fiber or dielectric film is similarto that for microwaves traveling in (or on)the dielectric waveguide already studied bymicrowave engineers.' Thus, a good un-derstanding of propagation of light in

fibers and thin film guides was available atthe outset of research in the area. Work onfiber -optics has progressed to the pointwhere multi -mode fibers with losses lessthan 5 dB/ km have been produced.6 Whilesingle -mode fibers with losses below 10dB/ km have also been demonstrated, theactual bandwidth of a fiber-optic system isrestricted by the dispersion of the fiber.This amounts to 101° s/ km for single -mode fibers and deteriorates to as much as10-7 s/ km for multi -mode fibers. The

71

single -mode dispersion, however, is lowenough not to be an important limitation ina practical system. For an ultimate system,the large bandwidth can be efficiently usedin an error- and noise -resistant codingscheme, such as pulse -code modulation.Such a system would consist of anintegrated -optic transmitting terminal thatencodes, multiplexes and amplifies thefinal light -pulse train and couples it to afiber-optic transmission line. The fiber-optic line would be equipped with suitablyspaced integrated -optic regenerativerepeaters and terminate in an integrated -optics receiving terminal.

While such an elaborate system still re-quires further development of its com-ponents, such as efficient couplers betweenthe fibers and the I / 0 devices, more modestcoherent systems have been realized. Forexample, over short distances, as in anaircraft data -transmission link, bundles ofsingle -mode fibers provide largeinformation -handling capabilities and arestill free of crosstalk, need little protection

Jacob Hammer has worked on low -noisemicrowave research, electron interactionswith atoms, and lasers since joining RCALaboratories in 1959. He has been studyingproblems in optical communication andwaveguides since 1969. Dr. Hammerreceived the RCA Laboratories OutstandingAchievement Awards in 1962, 1964, and1973.

Contact him at:Communications Research LaboratoryRCA LaboratoriesPrinceton, N.J.Ext. 3210

Author Hammer examining a waveguidemodulator

from the aircraft's environmental ex-tremes, avoid ground -loop and other elec-tromagnetic interference problems, and arelight in weight. Systems such as these couldbe fed by simple waveguide modulatorswith effective bandwidths greater than 10MHz per fiber. As another example, asimple cable-tv system can use similarmodulators and accommodate one tvchannel per fiber over distances on theorder of 5 miles!"

Components for opticalcorn mu n ications-switches,modulators, and couplersAt this point we will describe thedevelopments in switches, modulators, andfiber couplers that have been the mainfocus of the program at RCA Laboratories.The reason for this emphasis is that optical-waveguide modulators and switches arevital components of any sophisticatedoptical communication or data transmis-sion system.' Without electronically con-trolled optical switches, the switching in anoptical system must be performed eithermechanically or by detecting the light andthen switching the detected signal elec-trically and regenerating a new opticalsignal. The disadvantages of both theseapproaches are obvious.

In the absence of good modulators, infor-mation can only be impressed on opticalcarriers by modulating the light source.This requirement places constraints on thepossible sources that may be used, and,because of the very nonlinear pumpingcharacteristic of lasers, presents seriousproblems in obtaining analog modulationat high frequencies. Using a modulator,therefore, widens the choice of light sourcesand allows the chosen light source to beoptimized for such basic emissioncharacteristics as, for example, efficiencyand coherence.

The future of optical communications,therefore, is highly dependent on develop-ing efficient, high-speed, light modulatorsand switches. Without these components,optical systems will never be as flexible asconventional electronic methods.

The RCA Laboratories program hasproduced devices that meet these re-quirements and also operate at voltagesand powers consistent with those used inelectronic integrated circuits. The frequen-cy capabilities of these devices extend wellinto the microwave region. The waveguide

modulators are readily produced usingmass -production techniques similar tothose used to make semiconductor devices.

The Pockels effect

This paper covers only optical waveguideswitches and modulators based on the useof linear electro-optic (Pockels) effect inessentially insulating crystalline materials.The Pockels effect is shown only by crystalsthat lack a center of symmetry. Briefly, theeffect is a change in birefringence or achange in refractive index caused by, andlinearly proportional to, applied electricfield. Generally, refractive -index changeson the order of some parts in 104 can beobtained with high but finite electric fields.Clearly, to be useful, the index change musttake place in the crystal volume throughwhich the light flows.

Optical waveguides

Optical waveguides, which inherentlyrestrict light flow to regions with at leastone dimension roughly the size of anoptical wavelength, are ideally suited tomake optimal use of the Pockels effect.Their advantage is that a low voltage maybe applied across the narrow gap to give ahigh field. At the same time, the volumeoccupied by the optical wave is small, so thetotal stored energy in the applied field islow. As a result, waveguide electro-opticmodulators operate at far lower voltageand require far less power than bulkelectro-optic modulators.

The linear electro-optic effect in crystalshas an additional attractive feature in thatthe inherent frequency response is extreme-ly high, extending well about themillimeter -wave region. It is thus clear whythe interest in electro-optic waveguidemodulators is high.

In discussing the application of the Pockelseffect to optical waveguides, it is con-venient to distinguish two broad classes ofoperation. The first class makes use ofcrystal and waveguide orientations so thatthe applied electric field producesbirefringence. The guided wave thus un-dergoes a change in polarization that isproportional to the applied field. Thesecond class uses orientations andpolarizations in which the electric fieldproduces a simple refractive -index change.Here the guided wave undergoes phasechanges that are proportional to theapplied field. Both of these classes can he

72

used in conjunction with a variety ofelectrode configurations to construct phaseand amplitude modulators and switches.

Diffraction gratings +waveguides = optical switchingThis section is devoted to modulators andswitches based on electrode configurationsthat provide periodic index variations."This type of structure produces an elec-trically controlled Bragg diffractiongrating that spatially separates the dif-fracted and undiffracted beams in thewaveguide plane. A number of suchgratings may be cascaded on a single planarwaveguide and used to perform a complexof switching functions. Using this ap-proach, RCA Laboratories has developeddevices based on a new and simplewaveguide made by diffusing metallicniobium into LiTaO3 substrates.'° Withthem, the power and voltage required toswitch or modulate large fractions (over80%) of the input light at frequencies closeto the microwave range are low enough tobe compatible with the outputs of elec-tronic circuits.''

These waveguide films are formed byevaporating approximately 500 A ofmetallic Nb onto polished LiTaO3 sub-strates. These are placed in a weaklyoxidizing atmosphere for 6 to 10 hours atabout 1100°C so that the Nb diffuses intothe LiTaO3, forming a thin film ofLiN b,Ta (LNT). This processproduces maximum index differentialsbetween the film and substrate on the ordeiof 1%, which makes for a relatively strongoptical waveguide with losses on the orderof I dB/cm.

Potential forlight -on -light operationsWe later discuss how even more efficientswitches than the Bragg grating switch havebeen made using the waveguide properties.In addition to producing active waveguidedevices with these guides, we have madedetailed studies of the physical propertiesof the LNT films.'° The particular systemof diffusing niobium into LiTaO3 hasfurther interesting potential in that bothLiTaO3 and LiNbO3 can be doped withiron in such a way that they become light-sensitive. Since refractive -index changescan be thus induced in the waveguide by theaction of light, this property opens thepossibility for using these waveguides toperform light -on -light operations.

LIGHTINPUT

Li To 03

Ldp

Fig. 1Grating modulator on LNT waveguide. Guided light is diffracted through an angle 26e whena voltage is applied to the interdigital electrodes. S is 7.6 micrometers and L is 0.3 cm.

As shown in Fig. I, laser light is coupledinto the film with SrTiO3 prism couplers.The effective index for the guided light maybe calculated from the coupling angle andis found to be consistent with valuesexpected for this system. Single -modeoperation is obtained; this implies that thethickness at which the graded indexdifference between film and substrate fallsto 10% of its maximum value is between 0.7and 1.6 Am.I2

The waveguide modulator is produced byapplying a voltage to an interdigital elec-trode pattern deposited on the waveguidesurface as shown in Fig. I. Applying thevoltage to the electrodes results in a

periodic variation in the refractive index;this variation acts as a phase diffractiongrating. For L sufficiently large the gratingmay be considered a "thick" or Bragggrating. Light entering the grating at anangle 08 is diffracted through an angle 208in the waveguide plane, where

sin 0a = Ao: 4Sn8

and 'I, is the effective index for the guidedmode being considered. The fraction of theentering light diffracted is

// /0 = sin2A0/ 2

and to the first order in r', the phase shift is

= -r(n8)3(L1 Ao)E

E is the average in -plane field componentcaused by the voltage Vo, and r' is theeffective electro-optic coefficient. Thus,/// has the form sin2BVo.

Experimental wavegu deresults

Fig. 2 shows the percentage diffracted as afunction of voltage for three laser

wavelengths. The solid curves are plots ofsin'BV, normalized to the data at ///o =75%. The functional agreement is good. Weobserve no variation in these percentagesfrom dc up to pulses with rise times below 3ns, which is the limit of our measurementcapability. The measured capacitance is 20pF, which indicates capacitive power re-quirements below 0.2 mW/ MHz.

The pulse response is shown in Fig. 3. Theapparent 3-ns rise time of the light is limitedby the response of the photomultiplier.Since the Pockets effect is capable ofresponding to frequencies of about 10'4 Hz,the real limitation of this device is set bycircuit -match considerations.

100

90

80

70

5 60

e.-. 50

4

30

20

10

4976 5592A 63284/

6 8 10 12 14

Vo (VOLTS)

Fig 2Perceitage of light diffracted in waveguidemodu,ator as a function of voltage.

73

0.IV

I.-2ns

LIGHT

VOLTAGE

Fig. 3Pulse response of waveguide modulator.The apparent rise time is limited by theresponse of the photomultiplier.

The LiNbTai-.03 optical waveguidesdescribed here are relatively simple tomake, have excellent and controllablewaveguide properties, and can be orientedto make optimal use of the strong electro-optic effect of both LiNbO3 and LiTaO3.The high efficiency and low voltage andpower requirements of this modulatorrepresent an order -of -magnitude improve-ment over bulk devices and earlierwaveguide grating modulators.

Fig. 4 illustrates the waveguide modulator'ssize advantage over the bulk modulator.Here a commercial state-of-the-art bulkelectro-optic modulator is contrasted witha waveguide modulator hypotheticallypackaged to include input and outputgrating couplers and a protective case. Thewaveguide modulator operates at ap-proximately 10V and requires powersbelow 0.5 mW / M Hz; the bulk modulatorrequires at least ten times its power andvoltage.

Although the waveguide modulator couldbe packaged to act as a bench modulator, itis more likely to be used in conjunctionwith fiber-optic transmission lines. Propercoupling devices to connect the waveguidemodulator directly to fiber-opticwaveguides are under development atRCA Laboratories.

Overcoming couplingproblemsThere are many problems to be overcomein making direct couplers between fibersand planar electro-optic waveguidedevices. Direct coupling by end -firingtechniques is impractical because of bothsize and symmetry differences between the

fiber and the planar guide. In addition, thebeam spread of light from the end of amulti -mode fiber is too large to be acceptedby the modulator structures. Even with asingle -mode fiber, the beam spread from anend -fired coupler will be too large to acceptbecause of diffraction from the very smallcore of the fiber.

An alternative approach is to attempt tocouple the fiber to the film guide using theirevanescent fields. Ideally, simply bringingthe core of the fiber close to the film shouldallow some coupling between the twosystems. Here, the difficulty is the phase -velocity mismatch between light travelingon the fiber, which has a refractive index ofapproximately 1.5, and light on the film,with refractive indexes on the order of 2.2.

Coupling viaevanescent fieldsand a diffraction gratingNonetheless, our approach is indeed basedon the use of evanescent -field coupling. Weallow the evanescent fields of the twoguides to overlap in a planar region con-taining a phase diffraction grating, whichserves as the required phase -matching ele-ment (Fig. 5). Coupling then takes placeover a controllable interaction length. Theeffective coupling aperture, H, maytherefore be designed to reduce the dif-fractive beam spread in the film waveguideto values that the switch can accept. Withthis method, theoretical couplingefficiencies on the order of 90% have beenpredicted' between single -mode fibers andfilm guides using blazed backward wavegratings.

Multi -mode fibers may also be coupled tosingle -mode films. This converts the spreadin velocity among the fiber modes to an

Aim

A

angular spread in the film, and so reducescoupling efficiency. The spread, however,may be rather small (<0.2°) for the low -loss, low -optical -aperture fibers that aremost attractive for optical com-munications. Nevertheless, couplingefficiencies on the order of 50% may still berealized.

Since the coupler is located entirely on thesurface plane of the film waveguide, itsfabrication and packaging characteristicsare compatible with the requirements ex-pected for the mass production of planaroptical waveguide devices and systems.

To understand the operation of thecoupler, consider the core of an opticalfiber in intimate contact with a gratingformed in the surface of a planar (film)optical waveguide, as in Fig. 5.

If a mode with propagation constant IA -1 =27rn,/ Ito can propagate in the fiber and amode with propagation constant I#21 =2rn2/ Ao can propagate in the film, strongcoupling may take place if the fields of thewaves overlap in the grating region and ifthe grating vector Ike = 27r/d closes thevector triangle illustrated in Fig. 5. Thisoccurs when

dl X0 = (n22 + nc2 - 2n2riccosOCand ( I)

= arc sin (nal sinO/Xo).

Here, nr and n2 are the effective refractiveindexes in the fiber and film, respectively.The critical coupling length, L, for whichthere is maximum transfer between theguides depends on the gratingcharacteristics and conditions of fieldoverlap. For this geometry it is clear thatthe effective coupling aperture, H, depends

ELECTRO -OPTIC - WAVEGUFDE MODULATOR

5/14. X 1/4" x

CONVENTIONAL

MODULATOR

Fig. 4Waveguide modulator has a size advantage over the bulk modulator. The waveguidemodulator is pictured as it would appear if provided with grating couplers and packaged.

74

TOP VIEW

FIBERCLADDING

CORE

AtoPRISMCOUPLER

LNT ^2

LiTa03 ni

CROSS-SECTION

Fig. 5Coupling between fiber and film waveguide is done with a diffraction -grating system.

on L and the angle O. In addition, bychoosing L B, H becomes essentiallyindependent of B for 0 > 0.

Without the grating, a fiber mode wouldcouple by refraction to an unbound sub-strate wave.14 The grating may be viewed asscattering unbound waves into guideddirections, and the efficiency of this scatter-ing can be made high by using blazedgratings.° If a forward wave grating ischosen (0 < 90° ), light will be lost to highergrating orders, restricting the ideal efficien-cy to 50% or less. This restriction can beavoided with backward wave coupling (0>900).16

Working withmultimode fibersOur discussion has, to this point, beenbased on coupling a single -mode fiber or asingle mode of a multimode fiber to asingle -mode planar guide. The in -planepropagation direction of a single -modeplanar guide is unrestricted, so we mightexpect that many of the fiber modes couldbe coupled to the planar guide by assigninga different direction in the planar guide foreach fiber mode. We shall see that thegrating coupler can perform this function.

A multi -mode fiber, with core and claddingrefractive index of n, and n respectively,will sustain modes with a effectiverefractive index nw ranging from nem:, 'Trfor the lowest -order to ne n, for thehighest -order mode. If the coupling gratingis designed to couple the lowest -order(slowest) mode, higher modes may also

couple with varying efficiencies, dependingon the angular selectivity of the grating. Agrating coupler with a long critical coupl-ing length will have high angular selectivi-ty, and vice versa. Long critical couplinglengths are associated with a combinationof weak gratings and small field overlaps.

What happens is that as the angle d)changes by an amount Ack, the diffractionefficiency falls off as"

sin2(kgLA41/ 2)/(.10) - /0 (2)

(lc,1Lai4)12)2

and the coupling efficiency for the modesthat do not exactly satisfy the condition ofEq. 1 are reduced. Thus, the couplingprocess causes a beam spread, which con-verts the velocity spread in the multi -modefiber to an angular spread in the film. Inaddition, in both the multi -mode andsingle -mode cases, light coupled from fiberto film will have an angular spread (.10)L.caused by the finite size of the effectiveoptical aperture H. For a single mode thespread will be

(A0)L = 2n2k./ Lcosck (3)

In the multi -mode case this will be in-creased by adding (AO.. Clearly, a com-promise between coupling efficiency andangular spread will have to be made for anymulti -mode coupler intended for use withwaveguide devices having limited accep-tance angles.

We may summarize our discussion to thispoint as follows. In using a grating to

couple a multi -mode fiber to a single -modefilm waveguide, the spread in phase

velocities among the fiber modes is con-verted to an angular spread in the planarguide. In addition to the coupling spread,there is an additional spread due to thefinite aperture (Eq. 3).

If a single planar -guide mode withminimized angular spread, such as one thatmight be produced by coupling a TE00 laserbeam into the film guide with a prism, isgrating -coupled out into a multi -modefiber, the coupling will be unrestricted bythe angular characteristics. In this case wewould expect the coupling efficiency to bedetermined entirely by the characteristicsof the grating and field overlaps. Thisexpectation was used to experimentallycharacterize the developmental coupler bymeasuring the output coupling fromplanar -film waveguide to fiber when theinput to the film is a well -collimated laserbeam.

Coupling experimentsIn our experiment an unclad multi -mode(diameter 0.002 in.) plastic (PM MA) fiberis embossed onto a grating etched into thefilm-waveguide surface. The filmwaveguide is a single -mode (TE0) diffusedLNT waveguide with n2= 2.195 and b tt 1.0

um. The fiber index is close to 1.5 and,since there is no cladding, An = 0.5. Thegrating space d is 7715 A, which requires 0to be 13.8° and 4 to be 25.8°.

The grating coupling characteristic is firstdetermined by coupling 6328-A HeNe laserlight into the LNT film with a prism

20

18

16

14

12

10

/NOMINALCOUPLING

ANGLE(11.1

4 6 8 10 12

S (DEGREES I

'4

Fig. 6Push-pull stripe guide switch uses phasematching to couple the two guides.

75

coupler using a lens to produce a well -collimated (essentially parallel) guidedbeam that enters the grating couplingregion at the nominal coupling angle.When this is done, we find a couplingefficiency of 6% by measuring the ratio oflight emerging from the fiber to lightentering the prism and correcting for theprism input coupling, waveguide loss andfiber loss, which are all independentlymeasured. The angular dependence of thecoupling output can be verified by rotatingthe entry angle of the beam in the planarguide; this is plotted in Fig. 6. Obtainingthe expected angular dependence verifiesthe physical basis of the coupling.

v[he multi -mode fiber -to -film coupling isdetermined by focusing the 6328-A laserinto the free end of the plastic fiber, using a40 -power microscope objective. Thisoverfills the acceptance angle of the fiber,ensuring a broad excitation of the modes.The modes are additionally mixed by usingat least a meter of fiber bent in a ten -cmradius. The light intensity coupled out ofthe film waveguide at the prism coupler isthen measured and compared to the inputlight, correcting for the other losses, and wefind a fiber -to -planar coupling of 0.11%.This low number is readily understood interms of the theory outlined earlier, sincemost of the modes fall in a region where the(sins x) dependence is small.

These experiments demonstrate thecoupler principle and give promise that aswe learn to make improved gratings andincrease the quality of the physical contactbetween fiber and film, the theoreticalvalues of coupling will be obtained.

System expectationsWe thus briefly consider what might beexpected using a low -loss (<20 dB, km)multi -mode fiber of the type now com-mercially available. We assume thecladding is left in place to avoid the large-In of an unclad fiber. The cladding mightappear as indicated by the dotted region inFig. 5. For these fibers ..1n ,----- 0.006 and n, =1.457 at 6328 A.

Using a critical coupling length L = 100

Am. and requiring J9 = 0.2°, we predict/ / 0.46 for an ideal correctly -blazedgrating. This must be reduced by 50% if wechoose to work in the forward -wave region,giving a predicted 34% coupling efficiencyinto a high -index planar guide. If we hadchosen to work in the backward -waveregion and were willing to accept 0.5°spread, approximately 50% coupling

would be predicted. A single -mode fiber ina similar coupler would produce ap-proximately 90% coupling.

The grating -coupler approach can also beapplied to stripe or channel opticalwaveguides. In the next section we give abrief description of stripe guide switchesthat make use of the unique properties ofthe LNT waveguide system and promise togive even better performance than thegrating switch described earlier.

Stripe guide modulatorsand switchesWe now consider the device shownschematically in Fig. 7. The operation ofthis type of coupler -modulator has beendescribed in great detail in a number ofplaces." so we will only give the briefestoutline. If /31 and /32 are the propagationconstants in guides I and 2, under phase -match conditions $i = 132 and light will becoupled between the guides. If initially allthe light is in guide I after traveling adistance, L. equal to the critical couplinglength, all the light will be transferred toguide 2. If the interaction continues over along length. light will be coupled back andforth between the guides.

If #1* $2, the amount of light coupled willhe reduced and the coupling periodshortened. The coupling is described by"

/2//u= I-(/1/ /0)= [tc'i (K2 + (A/ 2))2]-

sinI(K2 + (A/ 2))1' .r] (4)

where L is the intensity initially enteringthe coupling region and, at x = 0, /2 = 0 and

= The coupling coefficient K is relatedto the critical coupling length through

K = 2/ wL. (5)

The phase mismatch is measured by -I,given by

J= l$,-$21 (6)

Voltage applied to the electrodes shown inFig. 7 changes .1 and for our geometryguides we have shown

ok, (27r/ X0) r' n'3 (V I a) (7)

where r' is the effective electro-opticcoefficient and n' is the effective refractiveindex as defined in Ref. 18. Thus, a voltagecan vary the amount of light coupledbetween the two guides.

The extinction ratio is defined by

=1/1-/21/(ii+h)=1/1-1200 (8)

where the values of hand /2 are taken afterthe switch and a lossless system is assumed.Extinction ratios close to I (100%) aredesired.

For the type of device illustrated in Fig. 7, 77will depend strongly on how closely theactual device length approaches the criticalcoupling length. The critical coupling

MN/ AV111111Li..11111111.11

1111W/MM 4,

iy

1,1" /NI/ AVI'MNAIII.

I I

1.-aligr-a+1

GUIDE

GUIDEb

SECTIONA - A'

Fig. 7Push-pull stripe guide switch uses p'iase matching to couple between the two guides

76

length is proportional to the reciprocal ofthe coupling constant, which, in turn, isextremely sensitive to the device

dimensions and propagationcharacteristics.

Conventional photolithography is used tofabricate the stripe guide couplers; Nb isdeposited on a LiTaO; substrate, coveredwith photoresist and exposed through amask. After development the Nb is etched,leaving the desired pattern.

Manufacturing thecoupled stripe waveguides

This relative simplicity also applies to howthe coupled stripe waveguides are formed.After guide formation, the switching elec-trodes are deposited using evaporatedchrome -gold and a second photolitho-graphic step. Fig. 8 is a photomicrographof this stripe guide showing an

entry horn on the upper guide. When thestripe guides are correctly formed, lightcoupled into the upper guide through theentry horn emerges from the lower guidethrough a second horn (not visible in thefigure). The dimensions are (referring toFig. 5) a= 5µm, g = 2 pm, and L= 2 mm.

electrode is 2 pm wide. Thus, itis obvious that careful photolithographictechniques must be used, although suchtechniques are well within present

manufacturing capabilities. The results ofour initial experiments with this type ofswitch" are shown in Fig. 9 -the voltagebehaves as expected theoretically.

Future versions of these devices arepredicted to require drive voltages on theorder of only 1 V and drive power less than10 AW/ MHz. These very low values holdpromise for system flexibility that wouldmake these types of switches extremely

Fig. 8Stripe guide directional coupler. Lightentering the upper guide through the entryhorn emerges from a similar exit horn (notshown in this photomicrograph) in the lowerguide.

20

I0

x

x

x

x

x2 4 6 8 10 12 14

APPLIED VOLTAGE (VOLTS)

Fig. 9Experimental variation of output of stripeguide coupling modulator as function ofapplied voltage. The small bump at 14 V issignificant because it confirms thetheoretically predicted behavior.

attractive in applying integrated optics totelephone switching, data processing, andrepeaters for optical -fiber communicationsystems.

Summary andfuture work

The work at RCA Laboratories on thin-film dielectric optical waveguides hasresulted in the production of opticalswitches and modulators operating withvoltages below 10V and powers below 0.5mW/ M H/. These values are more than anorder of magnitude lower than bulkmodulators and are compatible with theoutputs of conventional integrated -circuittechnology. While the chief use for thesedevices lies in the field of optical com-munications, applications to optical data-processing systems (such as the modulatorelement in optical video disc recording)may also be visualized.

An important missing link for using"integrated optics" elements is the couplerto optical fibers. Work at RCALaboratories on the development ofgrating couplers has progressed to thepoint that the basic correctness of thismethod has been demonstrated. More

work is required, however, to improve thecoupling efficiency to useful values (lossesless than 3 dB per coupler pair).

Optical waveguide modulators for video-disc -type applications require im-provements in the materials and couplingmethods to produce lower insertion losses,higher optical power handling capacity,and better beam quality in the blue-greenspectral region than is now available.

In the next few years, the optical waveguideprogram at RCA Laboratories will concen-trate on obtaining improved fiber/thin-film couplers and on improving the blue-green performance of the waveguide

modulators.

Ackrowledgments

I he research at RCA Laboratoriesreviewed in this paper is the result of thejoint efforts of W. Phillips, C.C. Neil, R.A.Bartolini, A. Miller, D.J. Channin, M.T.Duffy, and the author. Our efforts havebeen encouraged by the informed supportof B.F. Williams. Free use has been madeof much joint published material. Theauthor wishes to express his appreciationfor the long and fruitful association he hasenjoyed with these friends and colleagues.

References

I. Hammer. J. M. Channin, DJ.; Duffy, M.T.; and Neil. CC.;IEEE .1. Quantum Electron. Vol. II (1975) p. 138.

2. Reinhart. F.K.; and Miller, B.I.; App!. Phys. Letters Vol. 20(1972) p. 36.

3. Klein. M.B.; and Abrams, R.L.; IEEE!. Quantum Electron.Vol. 11 (1975) p. 609.

4. Zeidlei, G.; J. App!. Phys. Vol. 42 (1971) P. 884.

5. Ramo. S.; and Whinnery, J.R.; Fields and Waves in ModernRadio. 'And. ed., New York. John Wiley and Sons, 1953 pp.388-393.

6. See. forexample. Cohen, L.G.; Kaiser, P.; MacCliesney, J.B.;O'Connor. P.B.; and Presby, H.M.; App!. Phys. Letters Vol.26 (1975) p. 472.

7. Hammer. J.M.; Proc. Soc. Photo -Optical Inst. EngineersVol. 33 (1975) p. 60.

8. St. Ledger, J.F.; and Ash, E.A.; Electron. Letters, Vol. 4(1968) p. 99.

9. Hammer, J.M.; App!. Phys. Letters Vol. 18 (1971) p. 147

10. Phillips, W.; and Hamrnaer, J.M.; J. Electronics MaterialsVol. 24 (1974) p. 545.

12. Marcnse, D.; IEEE J. Quantum Electron. Vol. 9 (1973) p.1000.

13. Hammer, J.M.; Bartolini, R.A.; Miller. A.; and Neil. CC.;Appl. Phys. Letters Vol. 28 (1976) p. 192.

14. Peng, S.T.; and Tamir. T.; Opt. Commun. Vol. 11 (1974) p.405.

15. Dalgoutte, D.G.; Opt. Commun. Vol. 8 (1973) p. 124; andUlrich. R.; J. Opt. Soc. Am. Vol. 63 (1973) p. 1419.

16. Dakss, M.L.; Kuhn, L.; Heidrich, P.F.; and Scott. B.A.;App!. Phys. Letters, Vol. 16 (1970) p. 523.

17. Gordon, EL; Proc. IEEE Vol. 54 (1966) p. 1391.

18. Tamir. T.. ed.; Integrated Optics, New York. SpringerVerlag. 1975. see Hammer. J.M.; Ch. 4, pp. 185-189.

19. Phillips, W.; and Hammer, J.M.; Digest, Topical Meeting onIntegrated Optics, Salt Lake City, Jan 12-14, Paper Tu A5.

20. Wittke, J.; RCA Review Vol. 35, (1974) p. 198.

77

Short course in DIGITAL ELECTRONICS, Part 3 -Digital filtersL. Shapiro

In digital signal processing, we need a filterfor the same reasons that we need one inanalog signal processing, namely, toremove noise or to select or reject aparticular frequency or band of fre-quencies. Thus, a noisy analog signal x(t),as shown in Fig. I , may be passed through alowpass analog RC filter to obtain a"cleaned up" analog output signal y(r). Thedigital process is essentially the same. Wesample the noisy analog signal (Fig. 2a)and obtain noisy samples (Fig. 2b). Theproblem is to remove the overlay of noisefrom the digital samples and obtain acleaned up digital" signal (Fig. 2c).

Although the general objective-to removenoise from a noisy signal-is the same inboth analog and digital cases, the filteringproblem is quite different. Thus, the simpleanalog low-pass filter of Fig. I would bequite inappropriate and actually destroyour sample pulse, leaving only a remnantconsisting of a small piece of exponentiallyrising signal followed by a long exponentialtail -off. Evidently, we need a differentapproach to obtain our cleaned up digitalsignal.

If we compare our cleaned -up analog signal(Fig. Ic) with its noisy parent (Fig. la), wecome to the conclusion that its amplitude atany moment depends quite substantiallyupon the previous amplitude history of thenoisy signal. In addition, although notquite so obvious, the amplitude of thecleaned -up analog signal (at any moment)also depends quite substantially upon theprevious amplitude history of the cleaned -up signal. (Of course, all values of thecleaned -up analog signal naturally dependupon the action, or transfer function, of theanalog lowpass filter.)

`There is a bit of a semantic probkm here since our signal has not

yet been quantized and, hence. should be classified as discreterather than digital. However, to simplify this development wewill ignore this distinction and continue to refer to all pulse -typesignals as digital since. in practice, such signals would necessarily

be quantized at some point in the process before being acted uponby the digital filter.

Reprint RE -22-4-14Final manuscript received September 1, 1976.

This third article of the series introduces the concept of thedigital filter and lays the groundwork necessary to designinfinite -impulse -response filters.

In carrying these relationships over to therequirements of our digital filter, wenecessarily conclude that our digital filtermust have a memory (e.g., a storage and/ ordelay capability) as well as a computingcapability. This ability is necessary so thatthe digital filter will be able to adjust theamplitudes of the samples on an individualbasis while preserving the discrete nature ofthe information.

From a mathematical standpoint, in whichthe memory aspects of the digital filter aretaken into account, the relationshipbetween input and output pulse trains maybe expressed as a difference equation

y(n) + bly(n-1) + b2y(n-2)= aox(n) t alx(n- I) + a2x(n-2)

(I)Here, a, and b, are the weightingcoefficients and n is an arbitrarily selectedreference pulse along the time axis as

a) Noisy signal

0 Tb) Lowpass filter

y(t)

111.'t0

c) Cleaned -up signal

Fig. 1Filtering action needed to clean up a noisyanalog signal can be provided by a simpltlowpass RC circuit.

shown in Fig. 3, which compares the noisyinput x(nT) and the filtered output y(nT).[In accordance with previous articles ofthis series we will continue to use thenotation T for the (sampling) time intervalbetween adjacent pulses. For clarity,however, we have dropped the constant Tfrom the arguments.]

Digital computeras a digital filter

Actually, our digital filter must take theform of a digital computer. This computermay be either a general all-purpose com-puter, so that only software is needed, orelse a dedicated, or "hard -wired," com-puter designed to perform only a specificfiltering function.

This introduces a very interesting aspect ofour digital filter. A digital computer is

a

sIt)

T

s(nT s(n)

hI

On)

Fig. 2Digital filtering process starts with asampled noisy analog signal; thus, thesamples represent a mixture of informationand noise. The problem is then to removethe noise from the digital samples.

78

Sit)

(nT)s(n)

11110 T 2T 3Ty(nT).y(n)

11

s(n -2)

x(n -I)

sin)xl n 1)

al

ts0

"STRENGTH.OF 601. f Sltldt .1.0

(1-co

1 I t a) Usual representation of the delta functionST IOT

-es y(n-2)

0 T 2T 3T 51 si TI_ 1°T

Fig. 3Numbered samples along the time axiscorrespond to a difference equation thatrelates input x(n) to output y(n).

presently capable of doing very much morethan simply performing a routine filteringoperation. The available signalmanipulations can be made very complexand highly sophisticated. Hence, ourdigital filter can now perform processingon the input signal far beyond the

capabilities of any array of analog com-ponents. Thus, we have suddenlyprogressed from a simple RC filteringoperation into a new world in which almostanything is possible. We would like to referthe interested reader to the first article ofthis series for a brief account of what can bedone with our innocent -looking digitalfilter. For the present, however, we willproceed in a modest way and only gradual-ly extend our digital filter capabilities.

Our initial approach will deal withsituations where, at least from an analogstandpoint, a need is recognized that maybe met by a conventional analog filter. Thisneed is then expressed as a set of (analog)filter specifications from which bothanalog filter and desired digital filter arereadily obtained. Although this processmay sound artificial and of limited scope,such is not really the case, and it is

representative of a great deal of the digitalfilter design work being done at the presenttime. First, however, we should note thatwith this approach we are really dealingwith the design of a so-called "infinite -impulse -response" (IIR) digital filter, asdescribed below.

IIR vs FIR

The design approach for infinite -impulse -response filters is quite different than thatfor the finite -impulse type.

At this point, we want to use the impulsefunction, OM, which appeared in the first

b)

'0

b) For digital purposes, the impulse functionis represented as a narrow pulse of unityamplitude.

Fig. 4Strength of the Impulse function (deltafunction) is unity as normally used in digitalcalculations.

article of this series as a sampling operator(see Fig. 4a). We will now use this function!'as the input to a digital filter so that we mayobtain its impulse response. If this responsenever decays exactly to zero, no matter howlong a period of time elapses, the filter isclassified as an infinite -impulse -response(IIR) filter. If, however, the output of thedigital filter does fall exactly to zero after afinite period of time, we classify the filter asa finite -impulse -response (FIR) filter.

The design of IIR and FIR filters is usuallytreated separately since the design ap-proach is quite different for each: Asindicated above, in this article we will dealwith the IIR filter and illustrate the princi-pal methods of its design. In each case, wewill assume that we are dealing with a so-called "DTLTI" (discret-time linear -time -invariant) system so that the filterbehavior will not change during the time itis operating on our input pulse.

The z -transform can be quite advantageousin the design of digital filters.

The z -transform is a modified Laplacetransform that deals directly with pulsetrains or, perhaps more exactly, with timeseries (of numbers). We will now take noteof the z -transform approach to the extentthat it will be needed in our design ex-amples. Additional information about thez -transform will be found in Appendix A.

The z -transform of a discrete functionx(n7) may be defined as follows:

X(z)= E x(n)z" (2)

'Subsequent calculations are based on the assignment of unity to

the "strength" of this function. It is sometimes convenient torepresent the delta function as a narrow rectangular pulse ofunity amplitude as shown in Fig. 5b.

where z is related to the Laplace variable sby the relation z = e'r, and s has the usualvalue a + fi.o.

A very interesting aspect of Eq. 2 is that thetransform contains, within itself, the

original (untransformed) time series. Itfollows then, that whenever we are able torepresent the z -transform of a function inthe form of a series in negative powers of z,the associated time series will appearnaturally as the coefficients.' As shown inAppendix A, the application of the z -transform to our difference expression (Eq.II gives

Y(z) + biz 1Y (z) + b2z-2 Y(z)= aX(z) +aiz-IX(z)+ a2z-2 X(z)

(3)

We can now obtain our transfer functionH(z), which is simply the ratio of Y(z) toA(:), e.g.

Y(z) ao + at: + a2z-'H(z)=

X(z) I + biz -I + b2z-2(4)

Our digital transfer function H(z) is a keyquantity in digital signal processing and, ingeneral, can be used as follows:

I) We can apply H(z) to a given inputX(z) and obtain the output Y(z), e.g.,

Y(z) = H(z) X(z) (5)

The output time series, y(n), is thenobtained by taking the inverse z -

transform of Y(z), e.g.,

.r(n) = ZIY(z)]

2) Given a particular 11(z), we can imple-ment it in terms of hardware, or suitablesoftware (for a general-purpose digitalcomputer).

3) Given a desired prototype analogfiltering function, or set of filteringspecifications, we can develop the cor-responding H(z).

4) Given a particular 11(z), we can ex-amine it critically for its performanceparameters and stability.

Space limitations do not permit a practicaldiscussion on the physical realization ofdigital filters. Interested readers are

referred to texts by Stanley9 and by

'Another exceedingly interesting situation is that when X(z)represents a transfer function, the time ivies embodied in the :-transform represents the impulse response of the associated

system.

79

We can now find the transfer function.

Rabiner and Gold." A model for such arealization, however, is given in theproblem at the end of this article.

Design examples

We will design a digital filter that has thefrequency response of a simple low-passfilter.

The design approaches examined in thisarticle are based on the development of thedesired digital transfer function from aprototype analog filter or correspondingset of filter specifications. We generallyfollow the development in Stanley.9

The type of desired digital filter may be anyof the following three:

I) A digital filter having the samefrequency response as the prototypeanalog filter over the portion of thefrequency spectrum of interest. Thisobjective can, in general, be only ap-proximately achieved so that a frequencytolerance specification is necessary.2) A digital filter having the sameimpulse response as the prototypeanalog filter.

3) A digital filter having the same stepfunction response as the prototypeanalog filter.

The above three digital filters, althoughbased on the same prototype analog filter,will, in general, have substantially differentfrequency responses. Therefore, we shouldinvestigate each of these three responses forany given digital filter. In our designexamples, we will begin with a digital filter

Notation:These symbols are used throughout thepaper, but extensively in the design ex-amples. The lower case letters apply tosignals in the time (t, nT, or n) domain;upper-case letters refer to signals in thefrequency (s or z) domain.

x(t), X(s)y(t), Y(s)G(s)

x(n), X(z).17(n), Y(z)

11(z)

Analog input

Analog output

Analog transfer function= Y(s)/ X(s)

Digital inputDigital outputDigital transfer function

= Y(z)I X(z)

0-11ARA"--1-0x(t) C y(t)

0 1110 0I/RCH(s)

5+ I/RC

Fig. 5This low-pass analog filter can be thestarting point from which we deve.op theequivalent digital filter.

that is to have the frequency response of asimple low-pass analog filter.

Given the low-pass analog filter of Fig. 5,we want to develop a digital filter that has asimilar frequency response over a frequen-cy range extending from dc to cutoff.

What do we mean by a "similar" frequencyresponse on the part of a digital filter?

If a sinewave input signal x(t) is applied toour analog filter (as in Fig. 6a), the outputwill be a sinewave of the same frequencybut of different amplitude and phase asshown by y(t) in Fig. 6b. For the digitalfilter, the corresponding input and outputare shown in Figs. 6c and 6d, where oursample pulses possess amplitudes cor-responding to the envelope of theassociated analog signal. In each case, theargument of the digital -filter signal func-tion is given as nT indicating that thesesignals are defined only for multiples of thesampling interval T. As previously in-dicated, the symbol for the sampling inter-val, T, is often dropped for reasons ofclarity, and the argument is given simply asn. (See the Notation table.)

ANALOG

aft) It)

a) Analog input

0 T 2T

/ 1

c) Digital input d) Digital outputFig. 6Frequency response approach. Analogfiltering changes phase and amplitude butnot the frequency of a signal, our digitalfilter is designed to match the analogresponse.

DIGITAL

h) Analog oLtput

.(nT) y AT I

21 /T1

The procedure for obtaining the transferfunction of the desired digital filter (for afrequency response match) is straight-forward:

I) Obtain the transfer function of theprototype analog filter.2) Obtain and plot the frequencyresponse of the prototype analog filter.

3) Apply the bilinear transformation tothe prototype analog transfer function toobtain the corresponding digital transferfunction

4) Obtain and plot the frequencyresponse of the digital transfer functionand compare it with the frequencyresponse of the prototype analog filter.Comparison of the two plots will showthe spectral region of acceptable cor-respondence.

We organize our work in accordance withthe above sequence of steps:

I) The transfer function of our analogfilter of Fig. 5 is easily obtained.

1/sC 1

G(s) = (6)R+ IlsC I + sRC

2) The frequency response, A(f), is ob-tained by replacing the Laplace variables by its imaginary part feu and then takingthe absolute value of the result

A() = I GU to) = [I + (a)/ (1/01 1/2 (7)

where to, = 27f, = I / RC.

This frequency response is plotted onFig. 7 foil; = 400 Hz and on Fig. 8 for f;= 400 Hz and 357.13 Hz.

3) We now develop the transfer functionfor our digital filter. This is ac-complished by means of the bilineartransformations

s=2\ / l - 1. ' \(

\ I +

Application of this transformation toEq. 6 gives us the digital transfer func-tion which is conveniently placed in thefollowing form:

ao + Y(z)

(8)

H(z) = (9)I + X(z)

'In complex -variable theory, a bilinear transformation is linearin each of the variables separately (the other one being heldfixed). The transformation of Eq. 8 satisfies this condition for thevariables s and Geometrically speaking, a bilinear transforma-tion has the property that it transforms lines into lines and circlesinto circles [see, for example. "Complex Variables andApplications." R.V. Churchill. McGraw-Hill. 1960.]

80

where

au = al =T

T+ 2 RC

T-2RC=

T+ 2 RC

The difference equation connecting theinput and output of our digital filter isreadily obtained from Eq. 9 by workingback from the above transfer function,see Appendix A.

tin) + bly(n- I ) = aox(n) + aix(n -1 )(10)

4) The frequency response of the digitaltransfer function of Eq. (9) is obtainedby replacing the Laplace variable s by itsimaginary part jw, noting that z = e'T -er = cos wT + j sin wT. With someminor manipulation we obtain

H(e'r)ao + al cos wT - jai sin wT

I + bicos wT - jbi sin wT

We take the absolute amplitude of theabove expression to obtain its frequencyresponse'

A(/) = MilTA (II)

11

(Jo' + al' + 2aual cos cuT

I + bi2 + 2 b1 cos cuT

At this point, the spectral region that isvalid for our development can be evaluatednumerically. For the sake of our example,we establish the following set ofspecifications based on the prototype filterof Fig. 5.

I0

Cutoff frequency for analog filter, J. =400 Hz

Sampling frequency, f's = 2000 Hz

Folding frequency, 1/2 = 1000 Hz

Sampling interval, T = I /J, = 0.0005 s

Results are plotted on Fig. 7 for both theanalog and digital responses. The analogresponse tails off monotonically to zerowith increasing frequency. The digitalresponse, however, is periodic and repeatsat multiples of the sampling frequency,For a region in the neighborhood of dc,however, there is a fair agreement betweenthese two responses. This region is shownon an expanded scale on Fig. 8. We maynote the following:

I) Frequency responses converge to uni-ty at dc.

2)The digital filter exhibits a cutofffrequency of only 357.13 Hz, althoughthe cutoff frequency of the analog filterfrom which it was derived was 400 Hz,which may be a significant deviation.

3) For comparison, the frequencyresponse of a similar analog filter with acutoff frequency of 357.13 Hz (instead of400 Hz) is also given on Fig. 8. In thiscase, although frequency responsesmatch exactly at dc and cutoff, thematch at other points in the spectrummay not be as good.

The impulse responses of our filters areshown in Fig. 9; the step -functionresponses, in Fig. 10. For the analog filter,these responses were obtained in the usualway using Laplace transform techniques.The digital filter responses were obtainedby straightforward use of the differenceequation [Eq. 10], where the values of theconstants were computed as follows:

' I here is a problem in semantics hcrc since the quantity AO isusualh called the amplitude response to distinguish it from thephase response 01 the transfer function.

DIGITAL ANALOGRESPONSE RESPONSE

0 707

0 I 2 3 4

DIGITAL fc to is 2ts

.35713Hz ANALOG IC FREQUENCY (kHz)400 Hz

6

3fs

Fig. 7Comparison of frequency responses. Note that the steeper fall -off of the digital filterresponse has resulted in a lowering of its cutoff frequency from 400 Hz to 357.13 Hz.

LO

0.9

08

0.7

0.6

0.50

IH N Pir)I1DIGITAL I

IRO) I(ANALOG)tc.400 Hz

- - J.11 (ANALOG) fe 357.13 Hz\\ \ \

0.707

200 3010 0 400

FREQUENCY(Hz)

500 600

Fig. 8Expanded comparison of frequencyresponses of analog and digital filters.

1.0

1.9

1.8

07

06

05

0

03

02

0I

0

ANALOG FILTER RESPONSE

DIGITAL FILTERRESPONSE

T 0 0005s

T 2T 3T 4T ST 6T TT

Fig. 9Comparison of Impulse responses of analogand digital filters designed for a frequencyresponse match.

ANALOG FILTER RESPONSE-1 .1100,RE I-1

0

DIGITAL FILTERRESPONSET 0 0009,

Fig. 10Comparison of step -function responses ofanalog and digital filters designed for afrequency response match.

10

-21

1-4 -3 -1 0

1 I

-5 -4

I I

2 3 4 5

a) Digital impulse input

1.0

-3I

-2

b) Digital step -function input

1

n

Fig. 11Standard digital inputs. These types ofinputs were used to obtain the results shownin Figs. 9 and 10.

81

RC= I I co, = I/ [27r(400)]

Ta.= at -

and

T+ 2 RC

0.0005

0.0005 + 2/(8007r)

= 0.386

T- 2RC1)1 = = 0.228

T+ 2 RC

The impulse and step function digitalinputs were used as defined in Fig. 11.

Figs. 9 and 10 show substantial differencesbetween the prototype analog filterresponses for the impulse and step -function inputs as compared to the digitalfilter responses for the same inputs. Ap-parently, the bilinear transformation cangive only very approximate results whenwe seek to design a filter to obtain a givenimpulse or step function response. Wemust seek another method.

Impulse -invariance approachto digital filter design

In this section we explore the design of anII R digital filter to give the same impulseresponse as that of a given analog filter. Tomake the filtering problem more in-teresting, we have chosen a second -orderButterworth analog filter having a 3 -dBcutoff frequency of 50 Hz. The samplingrate j, is 500 Hz. In this case, the designprocedure is somewhat different from thatused to obtain the digital filter for afrequency response match. We proceed asfollows:

1)Obtain the transfer function for theprototype analog filter.2)Obtain the impulse response of theprototype analog filter. This is con-veniently done by applying the impulseresponse to the transfer function (usingthe Laplace transform) and then con-verting back to the time domain.3) Establish the desired digital responseby replacing the variable in the(continuous) analog response (i) by itsdigital equivalent (nT).4) Apply the z -transform to the desireddigital response to obtain this responsein the z domain.

5) Using Y(z) = H(z)X(z), obtain thedesired digital filter transfer function bysolving for H(z), e.g., 11(z)= Y{z)/ X(z).In the present case, the computation issimplified by the fact that the z -

transform of the input impulse, X(z), isunity.

1)Establish the transfer function of theprototype analog filter.

The basic form for the "amplitude -squared" function of the Butterworth max-imally flat filter isr

A'(/) -1+40/04.>"

I +(w00,)4

(12)

where k is the order of the filter and cur is theradial cutoff frequency = 100 ir rad/ s. Weimmediately obtain the amplitude frequen-cy response (plotted in Fig. 12). However,to obtain the impulse response we firstmust obtain the analog transfer function inthe Laplace "s domain" formulation. Webegin by normalizing our radial frequencyw to the cutoff frequency we.

A'(/) =+ 0,4

and then replace cu" by -s2.complex, we write"

(13)

Since s is

'See Stanley" for a good account of the amplitude -squaredfunction. This is the same amplitude function A(f) that we calledthe frequency response in the previous section of this article.

`Since s is a compkx number and the entire right-hand side of Eq.14 represents the square of a complex quantity, the left-hand side

necessarily becomes the product of this quantity and its complexconjugate. e.g.. Ms) and NI-s)

4

10

08

06

04

02

00

A2(f)

H(s)H(-s)= (14)1 + s4

We obtain the poles of Eq. 14 using theconventional method of obtaining thefourth root of a complex number!'

s4 = -I = exp [A180° + q360°)]s = seix4pU(341480° + q360°)/ 4]

(15)(for q = 0, I, 2, 3).

This operation leads to poles at the follow-ing four locations:

Si =117 IF

S2 = --47 47

1

St=

V7

s.= -j -42. 'VT

To obtain a stable transfer function, weassociate only the poles in the left half -plane with H(s) in Eq. I4.' This gives us

G(s)-(s - szXs - 53)

+ + I

(16)

We revert to our unnormalized form of thetransfer function with the simple transfor-mation

"See Churchill. R.V.; Compkx variahks and applications; 2ndEd. (McGraw-Hill Book Co.; New York; 1960) p. 14.

'An excellent account of these manipulations is given in chapter 6of Ref. 9.

Wt

Fig. 12Amplitude frequency response, A(0, of second -order Butterworth analog filter.

82

S S

w, 100 it

which leads to

- (17)

104 fr2

G(s) = (18)

s2 + loon- s + 104r2

2)Obtain the impulse response of the(analog) transfer function.

Since the Laplace transform of the inputfunction, X(s), is unity (see Table 1), wehave only to obtain the inverse transformof our transfer function of Eq. 18 to obtainour impulse response in the timedomain, e.g.,

Y(s) = G(s)X(s)= G(s)[I]= G (s)

We begin by completing the square in thedenominator

Y(s)

104w2

(s + 507 V2)2 + 500072

and then make use of the Laplace

transform pair:

1 e"' sin /31

(s + a)2 + 112 /3

and obtain our expression for the output,

AO,

y(t)= 100 WI:- exp (-50 Virri)sin (50 Irrrt) (19)

This impulse response for our prototypeanalog filter is shown in Fig. 13.

3) Establish the actual impulse response ofthe digital lowpass filter.

We replace the argument, 1, in Eq. 19 withnT.

y(nT) = 100 VTir exp (-50 VTirn T)sin (50 WnT) (20)

This desired response is also shown on Fig.13 (for pulses occurring at moments n7).

4) Transform the digital impulse response tothe z domain.

We do this by applying the z -transform toEq. 20, making use of the z -transform pair(see Table I):

140

120

100

iiso-

ANALOG RESPONSE

DES RED DOTAL RESPONSEr 0 002 SEC

18

12 14

Fig. 13Impulse response of second -order Butterworth filter. This analog filter response informa-tion will help us establish the actual impulse response of the digital lowpass filter.

z sin aTsin naTolli

z2 -2z cos aT + I

which evaluates to

Z [sin(222.14nT)] -0.4298 z

z2- 1.8058 z+ 1

in order to obtain a match of frequencyresponse (rather than impulse response)between digital and analog versions. Wewill not burden the reader with the rigorousjustification for this step. However, it maybe inferred from a study of Appendix A,

(21) Eqs. A9 and A10.

Our actual expression requiring transfor-mation to the z domain (Eq. 20), however,includes an exponential factor multiplyingthe sine function. Fortunately, there is 'a z -transform which takes care of this con-tingency

-ne *Tx(n)-4--85- z) (22)

Hence, we need only alter the argument, z,of Eqs. 20 and 21. Direct evaluation of ourmodified argument of Eq. 22 gives

z 1.5594z

With some manipulation, we may nowwrite

Z [exp(-222.14n T) sin(222.14n

0.2756 z-

1 - 1.158021 + 0.4113z-2(23)

We have so far ignored the coefficient100N/Tr in Eqs. 19 and 20. To check theImpulse response of our digital filteragainst the impulse of the prototype analogfilter (Fig. 13), it may be convenient tocarry this multiplying constant alongwithout change. However, it is more usualto multiply our z -transform transfer func-tion by the sampling interval Tat this stage

Our final expression for the digital filter,therefore, is obtained by multiplying theresult of Eq. 23 by 100 Niiir(0.002). Theresult is

0.2449Y(z) = (24)

I - 1.1580/I + 0.411322

Table IZ -Transform of various function pairs andoperation pairs used in this paper.

Function pairsx(n) 11:1 X(s) (Laplace

transform)

d(n)

Step function[1.0 or u(n)]

nT

sin noT

cos naT

1.0

z

z - I

Tz

(z- 1)2 s2

zsinaT

- 2z cos aT+ I

- zcosaT

z2 -2z cosaT+ 1

a

s2 +a2

s2 a2

Operations pairs:x(n)- X(z)e -""r - X(e"-z)

83

5) Obtain the desired digital -filter transferfunction.

It may be noted that, since the z -transformof the delta function is unity, the output ofthe system Y(z) is equal to the z -transformof the transfer function itself, e.g.

Y(z) = H(z) X(z) = 11(z)

We have therefore obtained our desireddigital filter transfer function in Eq. 24.

Our result is easily checked

The difference equation can bereconstructed from the transfer function

Y(z) 0.24491'H(z) -=X(z) 1- 1.5801' + 0.41I3z

We obtain

Y(z) - 1.1580/1 Y(z) + 0.4113/2 Y(z)= 0.24491 X(z)

Applying the inverse z -transform to theabove (see Appendix A, Eq. A16) weobtain the difference equation

y(n)- 1.1580y(n - 1) + 0.4113 y (n -2)= 0.2449 x(n - 1)

The impulse response of the digital filter isobtained by successive solutions of thedifference equation, beginning with thedigital -impulse input of Fig. 11, at n = 0.After dividing by T(= 0.002) we find that,within round -off error, the match is exact(see Table II).

Step -invariance method

The design of a digital filter having thesame step -function response as that of agiven analog filter follows closely themethod used above for the impulse -

invariance approach. The steps are

()Obtain the transfer function for theprototype analog filter.2) Obtain the step -function response ofthe prototype analog filter. This is con-veniently done utilizing Laplacetransforms.

3) Establish the desired digital responseby replacing the time variable in thecontinuous analog response (1), by itsdigital equivalent (n7).4) Apply the z -transform to the desireddigital response to obtain this responsein the z domain.5) Using the relationship Y(z) =H(z)X(z), obtain the desired digitaltransfer function by solving for H(z), e.g.H(z)= Y(z)I X(z), where X(z)= z,.(z- 1)(Appendix A).

Space does not permit detailing the varioussteps in the development of the step -function -invariant digital transfer functionfor the second -order Butterworth filterused in the impulse -invariant exampleabove. The result, however, is

H(Z)-0.14541' + 0.1078 12

1 - 1.15801' + 0.4112z-2

and the associated difference equation is

y(n)- 1.1580 y(n-1) + 0.41I2y(n -2)= 0.1454x(n - I) + 0.1078x(n - 2)

We can check the results.

The digital step -function response may becalculated directly by repeated use of theabove difference equation starting with thedigital step -function input of Fig. 12. Wefind that the correspondence between thetwo filters is exact.

It becomes interesting to compaze thefrequency responses of our various digitalfilters based on the same second -order

02 04 06 04 ID Vie 0 02 0.4 05 OS 1157°0

foFCLOING FREQUENCY I/214

Fig. 14Comparison of frequency responses ofdigital filters designed from the sameprototype analog (second -order Butter-worth) filter.

Butterworth analog filter when designedeither by the bilinear transformation for amatched frequency response, or by theabove methods for a matched impulseresponse or a matched step -functionresponse. These responses, on both a linearand logarithmic scale, are shown on Fig. 14where it is evident that there are widedifferences in each case, particularly in thecutoff region.

Concluding remarks

This article treats the design of digitalfilters based upon the action -or transferfunction -of a given analog filter.However, this approach (although widelyused) represents only a very small part ofthe tremendous flexibility and capability ofa computer -based digital filter. For exam-ple, it is a simple matter to take the(discrete) Fourier transform of a time seriesand obtain a frequency series. In eithercase, the information is in series form and,insofar as the filter (e.g., the digital com-puter) is concerned, it does not (and, in-deed, cannot) distinguish between the two.It blithely goes ahead and modifies thevarious terms in the series in accordancewith whatever program has been entered.Hence, we have the alternative of operatingin either one or the other of the time andfrequency domains.

Table IIImpulse -Invariance -designed digital filter. Comparison of analog and digital amplitudes of impulseresponse. The response match is exact within roundoff errors.

n= 0 I 2 3 4 5 6 7 8 9 10

Analog 0 122.46 141.82 113.87 73.55 38.34 14.16 0.63 -5.10 -6.16 -5.04Dig. X 500 0 122.45 141.8 113.8 73.5 38.3 14.1 0.6 -5.1 -6.15 -5.0

84

The possibilities here are breath -taking andwill be treated further in the next article. Ineffect -within certain limits -we will beable to write our own bill as to the nature ofthe filtering action we desire, quite in-dependently of what has traditionally beenconsidered possible for conventionalanalog filters.

A future article will deal with the FastFourier Transform (FFT) which, by virtueof its speed, has made possible many of thetotally unexpected signal -processing ap-proaches that have appeared in the digitaldomain.

ReferencesThe following list represents only a smallportion of the available literature in thisfield. However, it contains material which,in the opinion of the writer, would beparticularly suited for readers planning tobecome further acquaintedwith the designof digital filters.

Laplace transform: CEE courses offer agood introduction.

We highly recommend the CEE courseM52, Fourier Analysis and the Laplace and

Z -transforms, for an entrance into thisfield. Alternatively, CEE course M2,Engineering Mathematics II, treats theLaplace transform. Readers wishing tobegin with the general subject of Fourieranalysis would be advised to arrange forthe CEE course MI, EngineeringMathematics I, before registering for M2.

For an excellent introduction:

I. Skilling, H.H., Electrical engineering circuits. Second Edi-tion (John Wiley & Sons, Inc.; 1965).

For further work on a slightly higer level:

2. Holbrook, J.G.; Laplace transforms for electronic engineers,Second (Revised) Edition (Pergamon Press; 1966).

3. Cheng D.K., Analysis of linear systems, (Addison-Wesley;1959).

The z -transform -Appendix A summarizesthe subject.

4."Fourier analysis and the Laplace and z -transforms," CEEcourse M52.

For a first introduction:

5. Cheng, D.K.; Analysis of linear systems. (Addison-Wesley;1959).

6. Cadzow, J.A.. Discrete -time systems, an introduction withinterdisciplinary applications, (Prentice -Hal, Inc.; 1973).

Advanced treatment:

7.Doetsch, G.; Guide to the applications of the Laplace and z-transfors, (Van Nostrand Reinhold Co.; 1971).

8. Jury, ET.; Theory and application of the z -transform method,(Robert E. Krieger Publishing Co.; 1964).

Direct design of digital filters. Start withStanley.

9. Stanley, W.D.; Digital signal processing (Reston PublishingCo., Inc.; Prentice -Hall co.; 1975).

On a somewhat higher level:

10. Oppenheim, A.V.; and Schaefer, R.W.; Digital signalprocessing (Prentice -Hall, Inc.; 1975).

II. Rabiner, L.R.; and Gold, B.; Theory and application ofdigital signal processing, (Prentice -Hall, Inc.; 1975).

Series manipulations.

The following paperbacks represent a good accumulation ofmatenal on power series.

12. Knopp, K.; Infinite sequences and series, (DoverPublications, Inc.; 1956).

13. Jolky, L.B.W.; Summation of series, (Dover Publications,Inc.; 1961).

14. Moore, C.N.; Summabk series and convergence factors,(Dover Publications, Inc.; 1966).

The tutorial treatment of series in the standard mathematicaltexts seems to have fallen into disrepute. The following texts,however, do have chapters on the treatment and analysis ofseries:

IS. Boas, M.L.; Mathematical method., in the physical sciences,(John Wiley & Sons, Inc.; 1966).

16. Pipes. L.A.; Applied mathematics for engineers andphysicals. Second Edition (McGraw-Hill Book Co.; 1958).

17. Sok oloikoff, I.S.; and Sokolnikoff, E.S.; Higher mathematicsfor engineers and physicists, Second Edition (McGraw-HillBook Co., Inc.; 1941).

18. Reddck. H.W.; and Miller, F.H.; Advaaced mathematics,Third Edition (John Wiley & Sons, Inc.; :955).

Glossary -a collection of terms from parts 1, 2, and 3 of this series.

Aliasing: A phenomenon arising as a result of thesampling process in which high frequency com-ponents of the original analog signal (whetherinformation or noise) appear as lower fre-quencies in the sampled signal. Aliasing occurswhen the sampling rate is less than twice thehighest frequency existing in the original analogsignal.

Amplitude function: As used in filter design, theamplitude function represents the absolute value3f the frequency response. A filter is sometimesspecified by the expression for the square of thisfunction, e.g., A2(f).Analog/ digital converter: A circuit whichsamples an analog signal at specified periods oftime to produce a discrete signal which is thenquantized.

Analog signal: A signal that is continuous itboth time and amplitude.Bilinear transformation: A transformation incomplex variable theory which is linear in eachof the variables separately (the other one beingheld fixed). The usual variables in our work are sand z. Geometrically speaking, a bilineartransformation has the property that ittransforms lines into lines and circles into circles;see, for example, Complex Variables andApplications by R.V. Churchill (McGraw-Hill,1960).

Binary code: A language in which each symbol(or pulse) has only one of two possible meanings

(or levels). Each symbol (or pulse) thenrepresents one bit.Binary signal: A digital signal with only twoavailable amplitudes or levels, variously calledon/ off, one/ zero, or high/ low. A binary signalmay be "positive" in the sense that the "one" levelmay be a positive voltage, or it may be "negative"in the sense that the "one" level may be a negativevoltage. In either case the "zero" level is ground.A binary signal may also be "bipolar", in whichcase the "one" is usually a positive voltage whilethe "zero" is a negative voltage of the sameamplitude.Bipolar pulse: A two -level pulse with both levelsbeing equal in magnitude but opposite in polari-ty. The "one" is usually assigned to the positivelevel and the "zero" to the negative level.

Bit: A unit of information corresponding to theselection of one of two equally likely possiblealternatives. It appears when the logarithminherent in the definition of information (H),below, is taken to the base 2.

H = (TI r) loge nwhere T is message duration (seconds); r is thewidth of slot assigned to each pulse (seconds);and n is the number of information -bearinglevels in each pulse.Capacity: The capacity (C) of a system is themaximum number of bits it can process, ortransmit, per second, e.g., C = T where T istime required for the processing or transmission

and 11 is information content (in units of bits).Compender. A device consisting of a com-pressor at the transmitting end and an expanderat the receiving end which operate as nonlinearamplifiers to obtain a more advantageousamplitude -quantizing relationship for the reduction of noise. The process of companding isparticularly important with audio signals. Theprinciples have also been effectively used in theprocessing of picture information.

Continuous amplitude signal: A signal that isable to assume any amplitude value, usuallybetween certain prescribed limits.Continuous time signal: A signal that is definedfor all values of time, usually between certainprescribed limits.Dirac delta function: A function defined by thefollowing relationships:

0 , for t#0(50)=

arbitrarily large, for t = 0

b( t) di = 1.0

As a consequence,

f(t) Nt) dl =f(0)

The delta function is also called the impulsefunction.

85

Digital! analog converter: A circuit whichtransforms a digital signal into an analog signal,usually by some type of filtering action.Digital filter: A discrete system (as definedbelow) which usually operates to remove noiseor separate out a particular frequency band forseparate processing.Digital signal: A discrete signal in which theavailable amplitude values constitute a discreteseries, each member of which can be representedby a number having a finite number of digits.The terms digital and discrete are sometimesloosely used interchangeably.

Discrete signal: A signal defined only at aparticular set of time values. Between thesevalues the amplitude may be zero or have anamplitude of no additional information value.

Discrete system: A system which operates on aninput discrete signal x(n), to generate anotherdiscrete signal y(n) according to some well-defined transfer function.Encoding: The process of translating a messagefrom one language to another; e.g., we maytranslate a message of M possible meaningsconveyed by a signal pulse of M possible levels,into two pulses such that these M possiblemeanings may be represented by various com-binations of these two pulses. That is, M = m2.Here, m represents the required number of levelsin each of the two pulses involved in thetranslation. The value ofm is selected as requiredto establish the equality.

Finite -impulse -response filter: A filter in whichthe response to an impulse input reaches exactlyzero after a finite period of time.

Fourier methods: See the first article of thisseries including the AppendixFrequency -division multiplexing (FDM): Asystem in which a number of separate channelsare assigned to different frequency bands withinan overall channel bandwidth. This is con-veniently done by appropriate choice of carrierfrequencies. The individual channels are thenseparated out at the receiving end with filteringtechniques.

Frequency domain: A graphical way ofrepresenting signals in which the horizontal axisis calibrated in units of frequency. Alternatively

it may be applied to a mathematical representa-tion of signals in which the variable is in units offrequency.Holding circuit: A circuit used at the receivingterminal in time division multiplexing whichlengthens the individual pulses (after separationand routing to their respective channels) to thefull sampling time interval in order to aid in thedemodulation process.

Impulse function response (impulse response):The response of a system to an impulse input.Impulse -invariance: The condition in which twosystems have identical impulse responses. This issometimes applied to a pair of systems in whichone is continuous and the other discrete. Theresponse match is then taken only at the sampl-ing points of the discrete system.Infinite -impulse -response (11R) filter: A filter inwhich the response to an impulse input neverquite reaches zero after a finite period of time.

Impulse function: See delta function.Information content: The information contentof a message \(H) is defined in terms of theprobability of the occurrence of the event inquestion, e.g., H = log ( I / pi) where pk is theprobability for the occurrence of the particularevent. For the case where M events are equallylikely, the probability for the occurrence of anyone of these events is I Mand we obtain H= logM. If M has been encoded in the form of n", wecan write\ H = mlogn. For the commonlyoccurring case where the logarithm is taken tothe base 2, we obtain the information H in bits.Linear system: One in which the behavior of thesystem is not dependent upon the amplitude ofthe input signal, or upon the simultaneouspresence of other signals.Linear time -invariant system (LTI system): Asystem in which the transfer function does notchange during the processing of the pertinentinformation.Negative pulse: A two -level pulse in which the"one" level is a negative voltage and -the "zero"level is at ground potential.Normalized power. The power that would bedeveloped by the signal across a one -ohmresistor.

Appendix A: The z -transformWe continue to use the notation of our first article so that the sampledfunction is indicated by an asterisk. Thus, given a train of delta functions(or impulses), we represent the sampling process of our originalcontinuous analog time function, x(t), as

x(4=x(t) E (t-n7)

.4(0= E x(n7) - n7)

Eqs. Al and A2 represent the same information except that Eq. A2 is inmore convenient form for further manipulations. It might be argued thatx(nT) is exactly equal to x(t). This may be true, but we can also arguethat multiplication by the delta -function factor insures that we wipe outany signal occurring between integer multiples of the sampling interval T

Positive pulse: A two -level pulse in which the"one" level is a positive voltage and the "zero"level is at ground potential.Quantization: The process of constraining thevalues of a signal, whether continuous or dis-crete, to assume one or another of a discrete setof values. By quantizing a discrete signal weobtain a digital signal.Quantization noise: A type of noise inherent inthe quantizing process due to the difference inamplitude between the original analog signaland the quantized signal. This implies an uncer-tainty regarding the original information, whichis then construed as quantization noise.Single polarity pulse: A two -level pulse in whichone of the levels is at ground potential. Theground potential level is normally designated asthe "zero" level.Step -function response (step response): Theresponse of a system to a step -function input.Step -invariance: The condition in which twosystems have identical step -function responses.This is sometimes applied to a pair of systems inwhich one is continuous and the other discrete.The response match is then taken only at thesampling points of the discrete system.Time -division multiplexing (TDM): A system inwhich the sample pulses are very narrow ascompared to the sampling time. It is hencefeasible to insert a number of such pulses,corresponding to different informationchannels, into the time period between thesuccessive samples of any one channel. We henceobtain a sequence of pulses during each samplingperiod. The pulses are routed to their respectivechannels at the receiving end.Time domain: A graphical way of representingsignals in which the horizontal axis is calibratedin units of time; alternatively , it may be appliedto a mathematical representation of signals inwhich the variable is in units of time.Transfer function: The quantitatively expressedratio of the response at one point in a system toan input at another point in the system.Z -transform: A modification of the Fouriertransform for use with digital signals in whichthe Laplace variable s is replaced by z = eT. Themeaning and utilization of this transform isdeveloped in appendix A.

and makes for a more precisely defined pulse train.

We now take the Laplace transform of Eq. A2

X*(s) = E x(n7)e-'2.n=0

(A3)

Our result on the right-hand side is obtained by noting that our variable ist, and hence, only the delta function need be transformed. The transformof the undelayed delta function, 6(t), is unity. The second term in theargument of 5(t-nT) is nT and represents a delaying operation. InLaplace transform language, this delay is given by the exponential factorin Eq. A3.°

'An excellent source for review of the Laplace transform as well as a comprehensive table ofLaplace transform pairs is Laplace Transforms for Electronic Engineers by James G. Holbrook(Macmillan Co.; N.Y.; 1959).

86

We now make the substitution

z = esT (A4)

Eq. A4 defines the relationship between the Laplace variable 3 and the z -transform variable z. With this substition, Eq. A3 becomes

x(z)= E x(nT)z " (A5)

Eq. A5 defines the z -transform of the time series x(nT). We no longer needthe asterisk for X(z) since the variable z, by itself, indicates that we aredealing with a time series. Another extremely important property of therelationship in Eq. A5 is the fact that the original time series appears in thesummation as an inherent part of the z -transform. Hence, at any time thatwe can write the z -transform of a function in the form of Eq. A5, e.g., as asummation in negative powers of z, the associated time series automatical-ly appears in our result. A simple example may be helpful. For example,suppose that we are given the following z -transform

G(z) = Tz 1(z - 1)2 (A6)

We can immediately extract the associated time series by expanding thedenominator and performing a straightforward long division.

G(z)= Tz I (z2 - 2z + 1) = Tz + 2 Tz 2 + 3 Tz +

Our time series, then, is

T, 2 T, 3 T,

This implies that we can rewrite our expression for G(z) as

(A7)

(A8)

G(z) = E (nT) z " (A9)

A further important implication is that, in accordance with the definitionof the :-transform in Eq. A5, we can say that the z -transform of the seriescorresponding to nT is the right-hand side of Eq. A6 or

Z[n7)= Tzl (z - 1)2 (A10)

This brings up the question of z -transform pairs for use in our work. Thereare, unfortunately, only a few sources available (to the knowledge of thewriter) of z -transform tables. Probably the most complete has beenassembled by Jury."

Fortunately, use of the :-transform is quite analogous to that of theLaplace transform so that the same approach may be used in either case.We transform to the z representation as convenient during the course ofthe problem and then transform our final answer bac& to the time domainwhen we want to determine the appearance of our result along a time axis.

A few additional remarks about the z -transform may be in order. Forexample, since the z -transform represents the transform of a time series(which exists within the z -transform itself) it often becomes convenient tobe able to move easily from a series representation to a closed -formrepresentation of a function. Thus, readers planning to engage in suchcalculations may find it advantageous to become conversant with thestandard forms for series expansions, such as the binomial theorem andother forms of the power series.'

"Application of the 2.- Transform Method by E.I. Jury (Krieger Publishing Co.; Huntington. N.Y.;1964; reprinted with corrections 1973).

'See the references for source material on series expansions.

As another remark in the handling of the z -transform, due to the ratherincomplete nature of available z -transform pairs, the writer has found itquite advantageous to make extensive use of the reduction of z -transformfunctions to a number of simply transformable terms using the partial -fraction approach exactly as in the Laplace case.° One precaution shouldbe noted, however. Practically every z -transform appearing in the tablesof z -transform pairs contains z to at least the first power in the numerator.Therefore, in obtaining partial -fraction expansions of a function, themanipulation should be so arranged as to save at least one power of z forthe numerator in each term. There are various devices for accomplishingthis but space, unfortunately, does not permit us to expand upon thesetricks of the trade. A good deal of interesting and well written informationabout the z -transform may be found in Cadzow.`

The z -transform in difference equations

We have found that the operation of a digital filter may, ingeneral, be described by a difference equation of the type given inEq. I. Let us now apply the z -transform to the solution of thisequation. We may desire the solution in either of two ways, i.e.,with, or without, the capability of incorporating initial con-ditions. Let us consider, first, the case in which we wish toincorporate initial conditions into the solution. For this case wemake use of the so-called "left -shifting" z -transform

-Ifin + zmF(z)- E (All)

When we go back only two time intervals from the reference moment, n,we need only the following three versions of the above transform pair

fin) -410- F(z)fin+ 1).0 - zflO)

fin + 'F(z) - z2./(0) - zfil)(Al2)

It is now convenient to rewrite our difference equation as follows:

.(n+ 2) + b,y (n + I) + b2y(n) (A13)= aox(n + 2) + aix(n + I) + a2x(n)

We transform term by term and obtain the following:

[z2 Y(z) - z2 y(0) - zy(I)] + bi[z Y(z) - zy(0)]+ b2Y(z) = ao [z2 X(z) - z2 x(0) - zx(I)]+ al[zX(z) - zx(0)] + a2X(z)

We solve for Y(z) and write

Y(z) =aoz2 + alz + 111

X(z)z2 + biz + b,

AO) - a,,r(0)]+ ziY(1) + biy(0) - x(I) - aix(0))(A15)

z2 + biz + b2

The separation of terms in the solution for 1(z) is quite significant. Thus,the coefficient of X(z) is independent of initial conditions and constitutesthe transfer function connecting the output to the input (initial conditions

''See CEE course M52. Session 5. for the Heaviside Expansion Theorem in z -transform notation.

See Cheng for the Laplace version.

'Discrete Time Systems, James A. Cadzow (Prentice -Hall. 1973).

87

are not considered). The second term on the right represents thecontribution of the initial conditions and modifies the output accordingly.Notice the direct analogy to the Laplace formulation where, in a similarsituation, the transfer function appears separately from the initialconditions.

If we are interested in obtaining only the transfer function, things aresomewhat simpler and we may make use of the "right -shifting" z -transform

fin-m>411..e" F(z)Or

fin) 110.F (z)

nn- I ) Rs)fin -2) F(z) (A16)

We operate directly on our difference equation as it appears in Eq. I andobtain

Y(z) + b11' Y(z) + b2z-z Y(z)= aoX(z) + X(z) + a2z-2 X(z) (A17)

which gives us our transfer function with much less fuss

ao + + 0212Y(z) = X(:.)

I + bill + b212(A18)

The transfer functions as developed in Eqs. A 15 and A 18 are easilytransformed from one form to the other by multiplication of numeratorand denominator by the proper power of z. At this stage, however, it maybe of interest to take a quick look at the transfer function and draw a rapidconclusion regarding its behavior over the frequency spectrum. Thosereaders familiar with the Laplace transform will recall that the frequencyresponse of a function in the Laplace transform representation is obtainedby letting the Laplace variable s go directly to its imaginary part jw, e.g.,

s = a + fie (A19)

The resultant expression is then evaluated for various values of ru in thespectral regions of interest. We do exactly the same thing with our transferfunction in the z -transform representation. Now, however, s occurs in theexponent of Naperian e as follows:

s = a + jo.t.4tojw (A20)

We substitute the last expression on the right for z wherever it appears inour expression for the transfer function. We then examine the behavior ofthe transfer function, again, in the spectral regions of interest. Now,however, things are a bit different. Our basic quantity is an exponentialfunctioa with an imaginary exponent. The result is that it is periodic andrepeats every time the argument (e.g., the factor multiplying the imaginaryquantity)) increases by a multiple of 27r, or, in effect,

wT = 2ffp p = 0, 1, 2, (A21)

The above relationship tells us that the frequency response of our transferfunction repeats exactly for the various integer values of p. If we make useof the fact that our sampling interval, T, is the reciprocal of the samplingfrequency, f,, we easily arrive at the result that the frequency response ofour transfer function repeats exactly at integer multiples of the samplingfrequency, a result that we had previously discovered in the first article ofthis series.

Transfer function in the z -transform domain

As usual, we define the transfer function as the ratio of the output to theinput of system, in which initial conditions have been separated out and donot appear,

Y(z)H(z) -

X(z)(A22)

For our previous case, such as indicated in Eq. I, our transfer functionfollowed immediately from application of the z -transform to thedifference equation giving the results of Eqs. A15 and A18. We may nowwrite

ao + ad' + a2z-2H(z)

I + + b2z-2

In the general case, where input and output relations involve pulsespositioned over many points along the time axis, Eq. A23 may begeneralized,

Ea111- AR.)

H(z)

EbkikkM3

(A24)

88

A final problem to chew onThe mathematical tools explored and used in this articlehave value far beyond the field of digital signal processing.In fact, they may be resorted to whenever information isavailable in discrete form. The following example shouldillustrate this point.

The national income of this country is calculated quarterlyand consists of the sum of the follow ng:

1) Consumer expenditures (purchase of consumergoods).2) Induced private investments (leading directly or in-directly to the purchase of capital equipment for in-,:reasing production).3) Government expenditures.

We let y(k) = National Incomei(k) = Induced private investmentc(k) = Consumer expendituresu(k) = Government expenditures

Paul A. Samuelson, a noted economist,' postulated thefollowing in 1939:

1) Consumer expenditures in the k-th accounting quarterare proportional to the national income in the preceding[(k - 1) th], quarter, e.g., c(k) = a y(k - 1).2) Induced private investment in any accounting period isproportional to the increase in consumer expenditure ofthat period over the preceding period, e.g.,

1(k) = b [c(k) - c(k -1)]

3) Government expenditure is the same for all accountingperiods.

Problem

Show that we may express the national income for the k-thquarter as

y(k) u(k) + a (1 + b)y(k - 1) - aby(k-2)

Problem

Show that the national inccme of this country can bemodeled as shown below.

where z-' represents a delay operator of one -quarter of ayear (e.g., T = 3 months).

Although tnis model of our economy is admittedly drastical-ly oversimplified, there is yet enough substance it it to raisesome valid points. The interested reader, for example, mayexperiment with various values for the constants (assumingreasonable initial conditions). He may be surprised at thegovernment expenditure required to keep the economy onan even keel.

'Samuelson. P A.. "Interactions Between the Multiplier Analysis and the Principle ofAcceleration." Review of Economic Statisiics, 21 (19391. pp 75-78

Ed. note: The answers to these questions will be giver in the nextissue. If yoJ answer them yourself and don't want to wait, write toDr. L. Shapiro, Continuing Eng'neering Education, RCA Bldg. 204-2, Camden N.J. 08101 for the solutions.

Answers to review questions from Fart 2 of this series

1) The inherent properties of the human ear which favor thecompanding of voice signals are:

The relationship between the loudness sensation ex-perienced by the human ear and the actual physicalsound intensity in decibels (e.g., ogarithmically) is verynearly linear, see next page. [from Acoustical Engineer-ing, Harry F. Olson (Van Nostrand, 1957)]. Normal speech makes much more use of soft, ratherthan loud sounds. An approximate relationship based on

a linea - amplitude scale is shown on the next page [fromCommunication Systems Engineering Handbook,Donalc H. Hamsher, Editor, (McGraw-Hill, 1937) p. 10-12].

Hence, by use of companding we perforrr the dualfunctions of favorinc the weaker speech sounds to whichthe ear is so much more sensitive while at the same timefavoring the (same) low levels which occur much morefrequently in normal human speech.

89

Ico 1000FREQUENCY IN CYCLES PER SECOND

Loudness sensation of the human ear related to sound intensity.

2) Cross -talk occurs when transmission lines couple theirsignals into each other causing the spurious appearance ofvoice signals on lines that should not be transmitting suchsignals. By the process of companding we raise theamplitude level of the weaker noise signals (at the expenseof the reduced gain for the stronger signals) thus makingthese (originally) weaker voice signals relatively immune tothe risk of being swamped by spurious competing cross-talk signals. [For further discussion see CommunicationSystems and Techniques, Schwartz, Bennett, and Stein(McGraw-Hill, 1966)].

Every man should get married. If he marries a

good woman, he will be very happy. If hemarries a bad woman, he will become aphilosopher...

(This statement by Socrates was probably motivated by aconcern about the low marriage rate in Athens of the fifthcentury B.C.)

We will skip the obvious male chauvinistic implications ofthe above statement and consider only the two alternatives,e.g., the probability of the male member of the unionbecoming a very happy individual or else becoming aphilosopher. Almost the only objective data available is thedivorce rate versus the marriage rate. In recent years, withthe easing of the divorce laws, these two rates haveapproached each other, with the divorce rate, in the case ofa certain west coast state, even exceeding the marriage rate.We may then say, very approximately, that with an equalprobability for the occurrence of either of the two alter-natives, the informational aspects of the two alternatives isclose to maximum since we have a maximum uncertaintyregarding the outcome of a union between any twoindividuals of opposite sex.

510cc

0.8

() 060It 047/

10.20

0

NORMAL DISTRIBUTIONOF SPEECH LEVELS

INCREASING SIGNAL AMPLITUDE

Relationship of soft to loudsounds in normal humanspeech.

My candle burns at both ends;It will not last the night;But ah my foes, and, oh my friends-It gives a lovely light.

Edna St. Vincent Millay seems to be quite happy with herapproach to the consumption of candle wax. The writer,however, doubts that in general such excessive consump-tion would actually produce the "lovely light" that soentranced our poetess. If we accept this premise we wouldthen say that the above poem does indeed contain a greatdeal of information, since it states that, despite the im-probability for the production of a "lovely light" in the caseof excessive candle wax consumption (by the averageindividual) such "lovely light" did, in fact occur for Edna-orat least, so she wrote.

90

Operation of the RCA Frequency Bureau

R.E. SimondsN.B. MillsJ.F. Eagan

This RCA service agency assists divisions of the Corporationwith equipment approval, station licenses, frequency alloca-tion, and other matters involving contact with the FCC,governmental agencies, and international organizations.

During the forty years since its establishment, the basicfunctions of the RCA Frequency Bureau haveremained much the same, although its scope has in-creased substantially. Modes of radiocommunicationhave increased from simple telegraphy and telephony tocomplex satellite multi -channel digital pulse transmis-sion. The useful radio spectrum has expanded from 30MHz to above 30 GHz, and the number of users has grownmaterially. At the present time, the FCC has authorized theuse of over 5 million radio stations, so the administrativeand technical problems associated with maintainingorderly use of the radio spectrum have necessarilybecome more intricate. This is reflected in an increase involume and complexity of national and international rulesand technical standards governing the use of the radiospectrum.

Radio station licenses

Most RCA licenses are obtained from the Federal Com-munications Commission, but from time to time operatingauthorities from other Government agencies have beenacquired for work in connection with Government con-tracts. In addition, RCA maintains ship licenses issued byLiberia and Panama. The range of licenses processedthrough the Bureau includes such varied operations asSpace and Earth stations and associated terrestrialmicrowave facilities, Ship stations, Coast stations, fixedHF, VHF and microwave Point -to -Point, AM, FM and TVBroadcast stations, including their auxiliary Broadcaststations, Rural Radio and Domestic Public Radio, Aircraft,Business, Manufacturers', and Experimental Radio

stations.

Radio station licensing and operation must be in accor-dance with the FCC Rules and Regulations and theRegulations of international treaties to which the UnitedStates is a signatory. As a result, the Frequency Bureaumust concern itself with the details of both domestic andinternational rules and regulations that govern the varietyof services in which RCA operates stations or sellsequipment.

At the present time the Frequency Bureau is responsiblefor maintaining records on well over 1300 licenses issued

Reprint RE -22-4-16Final manuscript received July 19. 1976

Norman B. Mills hasbeen associated with theFrequency Bureau since1941, dealing with radiostation licenses as wellas use of frequenciesthroughout the radiospectrum. He is present-ly Manager of the NewYork office of theBureau and also serveson the Executive Com-mittee of the RadioTechnical Commissionfor Marine Services(RTCM) and is amember of the NationalIndustries AdvisoryCommittee (NIAC).Contact him at:RCA Frequency Bureau50 Broad St.New York, N.Y.Ext. 5003

John F. Eagan, Jr. hadcompleted thirty-fiveyears of service with theRCA Frecuency Bureauwhen he died this April.He had been Manager ofthe Cherry Hill Frequen-cy Bureau since 1953.He represented RCA onseveral committees, in-cluding the Inter-national E lec-trotechnical Commis-sion and the RadioTechnical Commissionfor Aeronautics.

Raymond E. Simonds'biography and photoappear with his otherarticle in this issue.

91

by the FCC and 490 ship licenses issued by foreigngovernments. License information must be reviewed andupdated from time to time to conform to various changesin the requirements and to accommodate advances incommunications technology that were not available whenthe original application was submitted. Asa result, there isa continuing need to file information with the FCC. Alllicenses have a specific renewal period that varies ac-cording to service, so the Bureau maintains computerizedrecords to insure timely renewal filing. Licenseapplications vary in degree of complexity from two orthree pages to more complex forms, including supportingdocuments. The Frequency Bureau also provides data forNational Broadcasting Company requests for operatingauthority for on -the -spot coverage of news stories andsports events.

Equipment approval

The greater demand for and increased use of the radiospectrum has made frequency management increasinglyimportant. Over the years, and especially recently, theCommission has therefore developed more stringenttechnical requirements for the radio frequency equipmentover which it has jurisdiction. To minimize the possibilityof interference to the various radio services it regulates,the Commission has adopted technical standards andevaluation procedures for equipments that emit rf energy.These procedures consist of type approval, type accep-tance and certification of equipment. The rules alsoprovide that equipment requiring any of these threeapprovals may not be sold or leased until the approval hasbeen granted.

Type approval calls for submitting one or more sampleunits of the equipment to the FCC Laboratory Divison inLaurel, Md. for examination and measurement. All unitssubsequently marketed by the grantee must be identical inall respects to the sample tested or include only changesauthorized by the Commission. This applies to suchequipment as Broadcast Modulation Monitors and Class I

Television Devices. Class I Television equipment mustoperate within a channel allocated for television broad-casting; its output signal is connected to a receivingdevice by either wires or coaxial cable in addition to otherrequirements. RCA "SelectaVision" and the various pingpong, tennis and hockey games using the televisionscreen as the play area are examples of this classification.

Type acceptance requires advance approval of licensedradio transmitting equipment by the Commission. Itdetermines that the equipment complies with the rf-interference standards based on tests conducted by themanufacturer or his agent. Again, all units subsequentlymarketed must be identical to the sample tested except forvariations authorized by the Commission. Type accep-tance is required for transmitters in the Land Mobile,

Marine, Aviation, Class A and D Citizens, and BroadcastRadio Services.

Under equipment certification, the Commissionacknowledges officially that the units tested by themanufacturer or his agent are designed to meet applicabletechnical standards. Certification is mandatory for anyradio receiver which is capable of tuning in the band from30 to 890 MHz, and therefore encompasses FM and TVBroadcast Receivers. (RCA has a tv broadcast receivercertification facility at Bloomington, Ind.) It also governsthe use of Wireless Microphones, Radio -controlled Gar-age Door Openers, Field Disturbance Sensors and similarequipment. At the present time, most of the non -Government rf equipments manufactured or marketed byRCA are subject to these approvals. The FrequencyBureau processes all RCA applications for equipmentapproval and the resulting Commission grants, assuringefficient handling on a cost-effective basis. The Bureaualso administers RCA plant certification, an additionalCommission procedure whereby all manufacturingfacilities using industrial heating or ultrasonic equipmentsare certified to meet prescribed rf radiation limits.

Committee work

Members of the RCA Frequency Bureau have assisted oracted as the RCA representative on a number of domesticand international technical committees concerned withradio regulations, frequency allocations and equipmentspecifications. The Bureau participates in InternationalWorld Administrative Radio Conferences and in FCCproceedings where there may be a joint interest by morethan one part of RCA. Since the Bureau has a primeresponsibility for World Administrative InternationalRadio Conference work of the International Telecom-munication Uniton (ITU), it coordinates the necessarypreparatory work. This work is particularly importantinasmuch as the agreements reached by internationalconferences become the law of the United States whenratified by Congress.

The Bureau also actively participates in the work of theInternational Radio Consultative Committee (CCIR), andpassively in the International Telegraph and TelephoneConsultative Committee (CCITT) and the InternationalElectrotechnical Commission (IEC). Usually thisparticipation is supplemented by assistance furnished byappropriate experts within the Corporation. RCA isrepresented on all government -industry committees con-cerned with the allocation of radio frequencies.

Members of the Bureau have served as technical con-sultants to such groups as the Joint Technical AdvisoryCouncil of the IEEE and EIA (JTAC), the ElectronicIndustry Association (EIA), the Frequency ManagementAdvisory Council of the President's Office of Telecom-munication Policy, and the National Association of Broad-casters (NAB) FCC De -regulation committee.

92

Technical liaison

As part of its technical liaison, the RCA Frequency Bureauis routinely called upon to render interpretations of thedomestic and international rules and regulations that areof a technical and operational nature. This may involve in-depth study of existing rules and case histories orconsultation with the staffs of the FCC or Office ofTelecommunications Policy.

Current rules and knowledge of changes being proposedare of importance to the RCA operating companies, aswell as manufacturing and sales personnel. The Bureausupplies appropriate parties with the sections of theserules and regulations of interest to them and furnishesproposed changes, amendments and associateddocuments on a continuous basis. This service is suppliedto over 600 people throughout RCA.

Many Frequency Bureau activities require engineeringstudies, in addition to frequency coordination with otherradio systems. Prior frequency coordination is required byFCC Rules and must be completed before filingapplications for earth stations in the space communica-tion services or terrestrial microwave stations. This in-volves notifying other system operators of the proposeduse of the desired frequencies. The other operators arethen required to respond to our notification within 30 days.A continuing analysis is also carried out to protect existingRCA systems. In order to provide this frequency coordina-tion service and rapid distribution of information, theFrequency Bureau performs many of its functions usingdata-processing techniques.

The records of the Frequency Bureau are a source ofinformation to many parts of RCA. The Frequency Bureaumaintains the most recent International Radio Con-sultative Committee (CCIR) and Consultative CommitteeInternational Telegraph and Telephone (CCITT)documents and provides copies of excerpts from them inresponse to a large number of requests from manycompanies and divisons of the Corporation. Through itsknowledge of the records, functions and organization ofthe FCC and other federal agencies, the Bureau is able toassist in advance coordination and preparation ofapplications and engineering and marketing studies.

Conclusions

For more than fourty years, the RCA Frequency Bureau

has contributed to RCA's manufacturing, service,and communications entities with its concepts of cen-tralized accumulation and dispersal of informatioi, cost-effective support in licensing and equipment approval,and Corporate representation before U.S. Governmentand international organizations. Effective participation inimportant growth areas in the future will be a cortinuinggoal of the Frequency Bureau.

U.S. Government and international agencies w,th whomthe Frequency Bureau dea s

International Nationalorganizations organizations

International FederalTelecommmunication CommunicationsUnion Commission

CCIR Office ofCCITT Telecommunications

PolicyInternationalElectrotechnical Department ofCommission Transportation

International U.S. Coast GuardMaritimeConsulatative Federal Aviation

Organization Administration

Other National Oceanographic

international and Atmospheric Adm.

Department of State

Other national

RCA entities assisted by the Frequency Bureau.

RCA GlobalCommunications

RCA AlaskaCommunications

RCA AmericanCommunications

National Broadcasting Co.

Picture TubeDivision

RCA Service Company

Distributor andSpecial Products Div.

David SarnoffResearch Center

Commercial CommunicationsSystems Division

GovernmentSystems Division

Avionics Systems

Missile and SurfaceRadar

Astro Electronics

Automated Systems

Consumer Electronics

Solid State Division

Other RCA

93

How the Communications Act affects you

R.E. Simonds If you design any equipment that transmits or receives rfenergy, FCC Rules and Regulations affect your work in someway. The FCC's power comes from the Communications Actof 1934 and its amendments.

Everyone recognizes that laws have an important role insociety. However, the laws governing use of the radiospectrum-and the engineer's role in writing, im-plementing, interpreting, and enforcing these theselaws --may not be widely recognized. The com-munications services and products of RCA are subject toFCC Rules and Regulations, which are based on theCommunications Act of 1934 and its subsequentamendments.

General powers of the FCC

Many of the FCC Rules and Regulations are of a technicalnature-about half of the general powers of the FCC listedin Section 303 of the Communications Act have engineer-ing aspects. Consider those relating to the assignment offrequencies to the several classes of stations used forbroadcasting and radiocommunication:

Sec 303

GENERAL POWERS OF THE 0011111BSION

Sac. 303. Except as otherwise provided in this Act, the Commis-sion from time to time, as public convenience, interest, or necessityrequires shall-

(c) Assign bands of frequencies to the various classes of stations,and assign frequencies for each individual station and determine thepower which each station shall use and the time during which it mayoperate;

(d) Determine the location of classes of stations or individualstations;

(f) Make such regulations not inconsistent with law as it maydeem necessary to prevent interference between stations and to carryout the provisions of this Act: Provided, however, that changes inthe frequencies, authorized power, or in the times of operation ofany station, shall not be made without the consent of the stationlicensee unless, after a public hearing, the Commission shall deter-mine that such changes will promote public convenience or interestor will serve public necessity, or the provisions of this Act will bemore fully complied with ;

(h) Have authority to establish areas or zones to be served by anystat ion:

The Radio Spectrum between 10 kHz and 275 GHz isdivided into bands of frequencies allocated to the variousradio services such as Broadcasting, Aeronautical, Landand Maritime Mobile, Radionavigation and Radio -

Reprint RE -22-4-17Final manuscript received May 19. 1976

location, Fixed (Point -to -Point) and Radio Services usingSpace techniques. Knowledge of propagationcharacteristics over the radio spectrum is necessary tomake adequate and suitable frequency assignments tostations that must meet the operational needs of thesediverse services. For example, in the United States,standard AM Broadcast Service frequencies are identifiedas Local, Regional or Clear -channel, and assignments aremade depending upon the size of the geographical area tobe served. In some instances directional antennas areemployed to obtain the most effective coverage or toreduce interference in certain undesired directions.Similarly. FM Broadcast stations are classified accordingto power and coverage area as Class A, B or C.

The frequency range allocated to Television Broadcastingextends from 54 MHz to 890 MHz. In order to compensatefor this wide difference in frequency and variation inpropagation characteristics between the lowest andhighest frequencies, stations operating on the lower vhfchannels are authorized a maximum effective radiatedpower of 100 kW at an antenna height of 500 ft, stations onchannels 7 through 13 have a maximum power of 316 kWat 500 ft, and in order to provide equal coverage, uhfstations employ a maximum power of 5 MW.

Television Broadcast Stations are assigned in accordancewith a P'an contained in the FCC Rules and Regulations,taking into account the Commission's obligation given inSection 307 (b):

Sec 307

(b)" In considering applications for licenses, and modificationsand renewals thereof, when and insofar as there is demand for thesame, the Commission shall make such distribution of licenses, fre-quencies, hours of operation, and of power among the several Statesand communities as to provide a fair, efficient, and equitable distribu-tion of radio service to each of the same.

The Plan also provides for assignment of stations in such away as to prevent interference by taking into account theso-called "taboos," whereby assignments made to uhftelevision stations consider the deficiencies of many hometv receivers in their ability to suppress local oscillatorradiation, images, and intermodulation products.

Similarly, additional frequencies have been madeavailable to Land Mobile Services in some of the larger

94

OMEGA SHIP -TO -SHORE SONAR DECCA LORAN C MARITIME TELEGRAPHY AM RADIO LORANA DISTRESS AIRBORNE AMATEUR INTERNATIONAL BROADCAST SHIP -TO -SHORE CrIZEN'SBAND RADIO ASTRONOMY VHF -TV FM BROADCAST AIRCRAFT AIR TRAFFIC CONTROL RECREATIONAL BOATS TELEMETRY EMERGENCY AIRCRAFT SURVIVAL SATELLITE CITIZEN'SBAND UHF -TV MOBILE AIR-TRAFFIC RADAR MICROWAVE OVENS SATELLITE UPLINK -DOWNLINK MICROWAVE RELAY AIRBORNE DOPPLER RADAR RADAR ALTIMETERS RADAR SATELLITE

cities using uhf-tv channels 14 through 20 on a sharedbasis in a manner designed to avoid interference to thetelevision broadcast service.

Many of the frequencies used for space communicationsare shared with terrestrial services. In order to make suchsharing effective without mutual interference, it is

necessary to coordinate the frequencies in planning thelocations of earth stations and terrestrial microwavestations. The coordination calculations are complex, buteach applicant for a station license must perform them.

It is obvious from the foregoing that in order to makeassignments to stations so that these broadcasting andcommunications services can be provided without mutualinterference, it is necessary to take into account thetechnical characteristics of the various parts of the radiospectrum together with terrain factors, areas to be servedand powers of transmitters. These objectives of the Actcan only be achieved by engineering solutions.

Raymond E. Simonds, Director of the RCA Frequency Bureau,has been associated with the Bureau since 1941. dealing with theallocation, assignment and use of frequencies throughout theradio spectrum. Mr. Simonds has been a member of the U.S.Delegation to a number of international radio conferences and iscurrently an industry advisor to the FCC Steering Committeepreparing for the 1979 ITU General World Administrative RadioConference.

Contact him at:RCA Frequency Bureau, Washington, D.C., Ext. 4236

Equipment standards

Sec 303

(e) Regulate the kind of apparatus to be used with respect to itsexternal effects and the purity and sharpness of the emissions fromeach station and from the a ppa rat JS therein;

This is the authority under which the FCC prescribestechnical specifications that must be met by transmittersused in the various radio services. The FCC Rules andRegulations governing each radio service containspecifica:ions of frequency stability that must be main-tained by transmitters, the level of suppression of spuriousemissions, and the modulation characteristics of thetransmitters. The FCC has a procedure of equipment type -acceptance that requires manufacturers to submitmeasurement data to the FCC on each model oftransmitter offered for sale, attesting that it meets therequirements of the rules. Transmitters meeting the rulesare subsequently accepted by the FCC for licensing.Type -acceptance recently became a newsworthy itemwhen the FCC did not appove a large number of citizen'sband traisceivers.

Sec 303

(g) Study new uses for radio, provide for experimental Wee offrequencies, and generally encourage the larger and more effectiveuse of radio in the public interest

By way of encouraging the larger and more effective use ofradio, the FCC issues experimental licenses when it is

necessary to radiate a signal in just about any part of theradio spectrum for any purpose found by the FCC to be inthe public interest. Perhaps the only proviso is that nointerference be caused to regular radio services.

Sec 303

(r) Make such rules and regulations and prescribe such restrictionsand cond.tions, not inconsistent with law, as may be necessary to carryout the provisions of this Act, or any international rad ,o or wirecommunications treaty or convention, or regulations annexed thereto,including any treaty or convention insofar as it relates to the use ofradio, to which the United States is or may hereafter become a party.

This section of the Act gives the FCC authority toincorporate provisions of iiternational treaties in its Rulesand Regulations. Most international treaty matters ofconcerr to the FCC are those developed within theInternational Telecommunication Union (ITU), a

specialized agency of the United Nations. Its purposes, as

95

INI ITTNA I IONAL TEL E COMMUNICA I IONS UNION id TU

INTERNATIONAL RADIO CONSULTATIVE COMMITTEE ICCIRI

STUDY GROUPS

- Spectrum Utilization and Monitoring2 Space Research and Radioastronorny Services3 Fired Service at Frequencies Below About 30 MHz4 - Fired Service Using Satellites5 Propagation In Non -Ionized Media6 - Ionospheric Propagation7 Standard Frequency And Time -Signal Services8 -- Mobile Services9 - Fixed Service Using Radio -Relay Systems

10 Broadcasting Service (Sound)11 Broadcasting Service (Television)

WORLD ADM NISTRATIVERADIO CONFERENCE

IWARC)

NAT( 36JAL IU SNAT F CAT, TN

FCC RUSE MAKING

U S PREPARATION

FCC NOTICES OF INQUIRY

FCC ADVISORYCOMMITTEES

INTERDEPAR MENT RADIOADVISORY COMMITTEE

ORAC-Federal Gov-ernment Requirements)

aat U S IMPLEMENTATION

Fig. 1

How international radio regulations are made. Shaded blocks indicate areas in which industry engineers normally participate.

contained in the ITU Convention, are quite similar to thosecontained in the Communications Act: To promotedevelopment of technical facilities with a view to im-proving the efficiency of telecommunications services, toeffect allocation of the radio frequency spectrum, and toregister radio frequency assignments in order to avoidinterference between radio stations of different countries.

Technical matters within the responsibility of the ITU aredealt with by two technical arms referred to as the CCIs.The CCITT is the consulative committee for telegraph andtelephone matters and the CCIR is for radiocommunica-tion matters.' Technical questions having internationalimplications are normally studied by the CCI's and theirrecommendations are then developed for internationaladoption.

Article 1 of both the ITU Telegraph Regulations andTelephone Regulations are indicative of the importance ofthe recommendations of the CCITT.

Article 12 of the Radio Regulations, entitled TechnicalCharacteristics of Equipment and Emissions, relies heavi-ly upon Recommendations developed by the CCIR forguidance in selecting transmitting and receiving equip-ment. Registration of frequencies by the InternationalFrequency Registration Board (IFRB) requires technicalevaluation as to the probability of producing interferencewith other systems. The criteria employed in evaluatinginterference levels are based upon Recommendations ofthe CCIR.

The development of CCIR Recommendations is an areathat is the province of the radiocommunications engineer.Where such Recommendations become a matter of inter-

The initials represent. In French. Comite Consultant International Telegraph etTelephone, and Comite Consultant International des Radiocommunications

Telegraph Regulations

Article I

Purpose of the Telegraph Regulations

I I.) I) The Telegraph Regulations lay down the general principles tohe observed m the international telegraph service.

(2) In implementing the principles of the Regulations. Administrations) should comply with the C.C.I.T.T. Recommendations, ineluding any Instructions forming part of those Recommendations, on anymatters not covered by the Regulations.

2 2. These Regulations shall apply regardless of the means oftransmission used, so far as the Radio Regulations and the AdditionalRadio Regulations do not provide otherwise.

I or recognized private operating agencylses1

Telephone Regulations

Article I

Purpose of the Telephone Regulations

I .1 I I The Telephone Regulations lay down the general principlesto be observed in the international telephone service.

12) In implementing the principles of the Regulations. Administrations.i should comply with the C.C.I.T.T. Recommendations, ineluding any Instructions forming pan of those Recommendations, on anymatters not emered the Regulations.

2 2. These Regulations shall apply regardless of the means oftransmission used. so far as the Radio Regulations and the AdditionalRadio Regulations do not provide otherwise.

I or recognized proate operating agenctilles1

96

national regulation, they are almost invariably incor-porated into the Rules and Regulations of the FCC. Theprocedure for developing international RadioRegulations, including the role of the CCIR, is shown inFig. 1.

Television receivers

Sc 303(a) Have authority to require that apparatus designed to receive

television pictures broadcast simultaneously with sound be capable ofadequately receiving all frequencies allocated by the Commission totelevision broadcasting when such apparatus is shipped in interstatecommerce, or is imported from any foreign country into the UnitedStates, for sale or resale to the public.

Generally speaking, FCC jurisdication is limited totransmitters; its jurisdiction over receivers is limited torestricting (radio) radiation from them. A 1962 amendmentto the Communications Act, however, authorized the FCCto require television receivers to be able to tune allchannels, both vhf and uhf. When the FCC did adopt rulesrequiring tv receivers to have comparable vhf and uhftuning capability, it produced a challenge for engineers todevise effective methods of achieving this comparability.

Inspections

Se( 103

(n) Have authority to inspect all radio .nstallations associatedwith stations required to be licensed by any Act or which are subjectto the provisions of any Act, treaty, or convention binding on theUnited States, to ascertain whether in construction, installation, andoperation they conform to the requirements of the rules and regula-tions of the Commission, the provisions of any Act, the terms of anytreaty or convention binding on the United States, and the conditionsof the license or other instrument of authorization under which theyare constructed, installed, or operated.

The FCC inspects radio installations, and in so doing itmay be necessary to perform technical measurements toinsure compliance with the terms of the station's license.Such measurements are normally carried out by anengineer. In the event that violations of the regulations areobserved or detected during such inspections, a notice ofviolation is served upon the licensee. It is thereforenecessary in such cases to take corrective measures tobring the station into compliance again and provideassurance that such corrective measures are effective.

Technical requirements for ships

Sec 355

(e) The main and reserve installations shall, when connected tothe main antenna, have a minimum normal range of two hundrednautical miles and one hundred nautical miles, respectively: that is,they must be capable of transmitting and receiving clearly perceptiblesignals from ship to ship by day and under normal conditions andcircumstances over the specified ranges.

In order to implement this provision of the Act, the FCChad to determine what constituted "clearly perceptiblesignals from ship to ship by day and under normal

conditions and circumstances over the specified ranges"on the international distress frequency of 500 kHz. Severalengineers sailed on ships and measured the signal-to-noise ratios under various atmospheric conditions.They took into account the average configuration of thevessels and the typical antennas in use and determinedthat a field intensity of 82.5 microvolts per meter wasrequired at 200 miles in order to produce a "clearlyperceptible signal." They calculated this to be theequivalent to 30 millivolts per meter at one nautical mile(over sea water) and determined that a transmitter having200 watts output into a typical ship antenna could producethe required field strength.

Administrative procedure act

The Administrative Procedure Act requires that govern-ment agencies give prior notice of proposed rule makingand provide opportunity for interested parties to filecomments. As has been pointed out, many FCC Rules andRegulations are of a technical nature, and althoughcomments are filed in a legal format, the substance ofcomments on technical matters is normally prepared byengineers. Fig. 2 shows the procedures that are usuallyfollowed to develop FCC Rules and Regulations

Conclusion

The foregoing examples demonstrate the important roleof the radiocommunication engineer in developing andimplementing national and internationa' radioregulations. The orderly administration of the radio spec-trum requires engineers having knowledge of the state ofthe art.

PUBLIC ONGRESS TREATIES

I

EXECUTIVE'BRANCH

JUDICIALBRANCH

FCCBUREAUOFFICE

NOTICEOF

INOUIRY

'COMMENTS I REPLIES

MEMORANDUMOPINION I ORDER

CONCLUDINGINQUIRY

CC

NOTICE OFPROPOSED

RULE MAKING

COMMENTS I REPLIES

IMPLEMENTATION

REPORT a ORDERAMENDING RULES

Fig. 2How FCC rules are made. Shaded blocks indicate areas in whichindustry engineers normally participate in technical rule -makingproceedings.

97

Ground -control system for Satcom satellitesJ. Lewin RCA American Communications two Satcom satellites, presently in U.S.

domestic communications service, are supported by a highly reliableground -control system to insure uninterrupted communicationscoverage. Accordingly, the Satcom ground system embodies a highdegree of automation, redundancy, and autonomy of operation.

The key mission requirements of the Sat-com ground segment are to control threesatellites in geostationary orbit; providecontinuous monitoring of health andstatus, and command and control; andkeep the satellites on station within ±0.1°in longitude and latitude. The groundsegment must also support transfer -orbitoperations for each spacecraft, with anexpected peak loading of two spacecraft ingeostationary orbits and one in a transferorbit. Fig. I represents the ground com-plex, set up for transfer -orbit operations.

The basic design considerations for theground segment were operational reliabili-ty and economy of operation,which dictated a solution relying on a highdegree of autonomy and automation.However, human decision is called for inthe system at critical junctures.

TRACK'S.ANTKIIIRIA

1.111.8.118111.1TRACK...

SK MVO

CO0'RAC OKI01

CARNARVONTTIC STATION

High -reliability system

This emphasis on high reliability, whichincluded the consideration of catastrophicground -station failures, led to the conceptof two redundant control stations. Eachstation is capable of performing the re-quired Tracking, Telemetry, and Controlfunctions (TT&C) in the orbital arcdesignated for domestic communications(99° to 132.5° W Longitude).

Locating these TT&C stations at VernonValley, N.J. and Moorpark, California.provides even wider coverage of the orbitalpath for the support of transfer -orbitoperations. The addition of leased Intelsatfacilities at Carnarvon, Australia, andFucino, Italy produced global coverage.(See Fig. 2). The Intelsat stations were

AMNA WESTGROUND STATION

rot.coamoime...was

TRACK..NNNNN

RACKIVIIR11..111.0CR

TRACK..MAYO

'k/ACC.. K.

1111..1.11..COORIAID.

AO.

1111.1111COMMA.

400...100.800

wA?.01.LAV

instrumented by RCA for compatibilitywith the Satcom spacecraft.

Operating economics require the SatelliteOperations Control Center (SOCC) andTT&C to be located together. The com-bined TT&Ci SOCC is then collocatedwith a major satellite -communicationsearth station. This set-up means that asingle operator can use a central controlconsole to monitor the performance, ascer-tain the orbital position, and commandeach of the three spacecraft.

Dual but autonomouscontrol

Normally, the Vernon ValleyTT&C/SOCC is the primary SatcomSystems Controller for all spacecraft, with

IFUCINO

TUC STATION

SATCONI EASTGROUND STATION

10.110

1

1.1.10.1.1, VAACKIMIO

TRA41.11.111

..v0

MAW.. 1 1 3COA.A.000 STly{

RANO. OM, v

SATCON OPIRATKIAISCONTROL COMER COMA' '71

COOPUT.

.111.11.0.1.1

RCM TOR .10

Ir RACKNNNNN

MAC..RARKIK

SATO= OPERATIONSCONTROL CENTER

.....VAC.

Fig. 1

Ground -station setup for a transfer -orbit operation. The two Intelsat stations are 'eased for such operations only; the two U.S.stations alone are sufficient for covering the arc needed for domestic communications.

,,i-,,, 6..

ittoA i.

1 '

,..0

7,----..."

s

l'''

'i,

-

S

- -.,

, -

........-

A -z

A'

..

illiiiii ..

Vil.

'II. ' :-

.0. lia.. 7

'N.

f4 - ' - -

4 .mil..

,

(_ VERNON

1,!211

,A.'''.

ak.-r

!MONO

.

(.... .'.

'-,-------Ar

'IL . '...i, - . .

MOOREPAR% 01111L.:W..4_ _ _''

')IIIIIIPI°'

i\ sr rn^'-' Iv 7 %.

i `----.',.2%...-,i._, 1St 1

l.EER

72. d

CAEINARV,

UUU

LI a----ON-

:,

1

....,'-r,

,

a A

.-f- r - - -'4,,,

.

. 1

i 1

i . . . PX ... / ir) cs"--.-....7-,...1\,....--1/z4ltr,..

.- .

1 1 1 1 : .

Fig. 2Four stations provide worldwide coverage for transfer-orbit operations when their antennas are at 10° elevation and the

satellites are at synchronous altitude. Numbers on chart are longitude expressed on a 360° basis.

Moorpark operating in a "hot" standbymode. At a moment's notice, however, theMoorpark site can assume the role ofSystems Controller should the need arise,as in case of equipment malfunction orscheduled maintenance downtime, for ex-ample. This concept yields additionaloperating flexibility by permitting sharingof control, where one site controls onespacecraft and the other site controls theremaining spacecraft. This kind of load -sharing was practiced during the launch ofSatcom F-2. In that case, Moorpark actedas controller for RCA Satcom F-1, ingeostationary orbit, while Vernon Valleyacted as Mission Controller for F-2, intransfer orbit. After the successful injectionof F-2 into its geosynchronous (drift) orbit,control of F -I reverted back to VernonValley.

The requirement of making each TT&Cstation autonomous has been achieved,except for orbit -related computations. Thehigh precision and speed required there,particularly during transfer -orbit

operations, requires the use of large-scalecomputers. Their relatively low utilization,coupled with the high cost of purchase, ledto the leasing of Univac 1108s from com-puter utilities. Again, reliability re-quirements, including the possibility ofcatastrophic failures, dictated the use oftwo physically separated computers, oneprime and one backup. During transfer -orbit operations the backup computer wasused as a "hot" standby: i.e., it was con-tinually updated to keep abreast of themission, as well as periodically exercised tocross-check with the primary computer.

State-of-the-art permitting, (i.e., the

availability of low-cost, high -precision,high-speed minicomputer) a self-containedorbit -computing capability is under con-sideration for the TT&C/ SOCC.

Computer complex

Two Hewlett-Packard 2100 mini-computers, designated "data" and "con-trol," are at the core of each TT&C/ SOCC.

These minicomputers and the two leasedUnivac 1108 computers, which are bothaccessible from each U.S. ground station,produce a system, and individualTT&C/ SOCCs, with a high degree ofautomation. An H P 2100 is also in use ateach of the two transfer -orbit stations atCarnarvon and Fucino.

The eight -computer complex is inter-connected via a communications networkthat enables computer -to -computertransfer of data and commands (Fig. 3).

Two dedicated, full -duplex, 4800 -baudlines interconnect the Moorpark and Ver-non Valley TT&C/ SOCCs: one line inter-connects the data computers, transferringspacecraft telemetry received by one com-puter to the other; the second interconnectsthe two control computers. Command -listtransfers and station coordination are doneusing this link. Finally, a 300 -baud line is

Reprint RE -22-4-18Final manuscript received September 27, 1978.

99

NTT SATSPACECRAFT

NT I LSAT1.1114

ST TiON

27:t

L J

NP 2100

DATA.

RI KONT 201A

ROGER

AUSTRAL IATRANSFERORB STA

RCA -W COASTTT&C"SOCC

WESTCOASTSOCE

L

RI r

120.SIETOEM

HP210n0T

.T2100CONTROL

ON100

RUDER

TEE CODIAL UP LINES

-1-111NTC AS E

OMMNETWORK

RI KON120.

MOM TA

EN KON

10.WOE M

P

-----+--1RAXON

03,TAME TA

UCC DALLAS, Tx

11081NIKON

UPWHAM

S TEL CODEAL UP LINES

AS fCOMM

NI TWO.

RIXON

Moo1

IN

Mt -IN 1

T201.MODEM

wP2100DATA

WOOER E

R IKON

CONTRONE1100

.TO

DATA

m2 COMPPIKON RixON

C 102E 103AMODEM MODEM

S

TECM

r- 71ITTESATSTCCWASH, DCH 4IX

0 fii

SANTA CLARA. CA

08 0]AON MIXON I

I' ELSA,EARTH

STA ION

RCACOAST

IT&CSOCC

EASTCOASTSOCC

1

Fig. 3Data and commands are transferred back and forth between eight computers in the system.Data computers deal with telemetry, ranging and attitude data; control computers deal withcommand sequences.

provided for data transfer to and from theUnivac 1108 computers.

TT&C/SOCC ground station

the TT&C/ SOCC has three basicfunctional areas: the rf receiving andtransmitting equipment, the signal -processing equipment, and the spacecraftoperations control center.

Both TT&C/ SOCC ground stations areidentical in capability and hardware.Within each station a high degree ofredundancy and "cross -strapping" isprovided, minimizing the effects of single-point failures. For example, should thecomputer -controlled automated com-manding fail, either one of two redundant"command and range tone generators"provides manual command generation (seeFig. 4).

Each ground station can command orrange with each of the three spacecraft, oneat a time: telemetry can be received andprocessed simultaneously from all threespacecraft.

Antennas

Each spacecraft in orbit has a dedicatedantenna, which performs two functionsduring commercial operations-communciations and command and con-trol.

All antennas have I3 -meter dishes and a

G/ Tof 32.4 dB/ °K. The rf feeds are of dualorthogonal polarization, capable of receiv-ing and transmitting simultaneously inboth polarizations. The feeds are designedfor 500 MHz of bandwidth, with 6 GHzuplink and 4 GHz downlink. The gain at 6and 4 G Hz is 56 dB and 53 dB, respectively.When driven by a 3 kW transmitter, theantenna system is capable of producing aneffective isotropic radiated power (EIRP)of 89 dBw, which is ample for ranging andcommanding during both transfer -orbitand on -station commercial operations.

Tracking system

To facilitate transfer -orbit operations, oneantenna at each TT&C/ SOCC is equipped

TT SC RECEIVING ANDTRANSMITTING EQUIP COMMUNICATION

SUBSYSTEM

ANTENNA AZ

TRACKINGANTENNA

NOTE -1

LOW NOISEPRE AMP

TO CIAO FEED(TRACKING ANT)NOTE -2

HA

SWITCH

H PA

NOTE -2

CONTROLCONSOLE

UP CONY

ANTENNA EL

1 MODULATOR

RANGING VIACOMMUNICATIONS

TRANSPONDER

4 UP CONN HMODULATOR

FROM PROBEI I

(TRACKING ANTI

'TO CMO FEEDI FIXED ANTI

FROM PROBE(FIXED ANT)

TEST TRANSLATOR

FIXEDANTENNA

NOTE

LOW NOISEPRE AMP

FROMADD TIONALANTENNAS

PATCH

DOWN CONYa LINE

DRIVERSNOTE 2

DOWN CONY6 LINE

DRIVERS

DOWN CONY6 LINE

DRIVERS

NOTE 2

NOTESI- LOCATED ON ANTENNAS2- LOCATED NEAR ANTENNAS

DOWN CONT.& LINE

DRIVERS

100

with auto -track capability. This antennahas an extended range (±270° in azimuthand 90° in elevation) and higher angularvelocity (3° / s) when compared with thestandard communications antenna. Fourtracking elements comprise the monopulsefeed system.

The pointing accuracy of the antenna is0.025° rms and the tracking errors in theauto -track mode are less than 0.008° rms.The shaft position readout for azimuth andelevation has a resolution of 0.01°; it istransmitted to the data computer forfurther processing, in conjuction withspacecraft ranging.

In addition to its auto -track mode, theantenna can also be controlled viacomputer -driven "programmed -track" andmanual -track modes. Both tracking modeswere successfully used at eachTT&C/ SOCC site during the transfer orbitof Satcom F -I and F-2 when attemping aninitial acquisiton (the first time the

spacecraft appears over the horizon). Pre-dictions for azimuth and elevationgenerated by the Univac I108s off-linesoftware are used to position the antenna.

When on -station, the auto -track antennacan track either of the two orthogonallypolarized spacecraft beacon frequencies:3700.5 MHz, horizontally polarized (east -west); and 4199.5 MHz, verticallypolarized (north -south), originating at thespacecraft communications antenna. Thecommanding frequency is 6423.5 M Hz andis horizontally polarized.

When in transfer orbit, the spacecraft'somni antenna is used, with both beaconscopolarized. The command signal is

orthogonal to that of the beacons. Thebeacon polarization is linear and parallel tothe spacecraft's spin -axis, which changes inthe course of the transfer mission as thespacecraft spin axis precesses. Consequent-ly, manual polarization adjustments haveto he made during this phase of tracking.

The tracking antenna has a polarizationadjustment over ±45° for both orthogonal-ly positioned rf feeds. Additionally, thepolarization of the uplink feeds can beadjusted relative to the downlink feeds by±10° in order to optimize polarizationalignment

Redundant low -noise receivers (LNRs),provided for each polarization, enhancethe reliability of the telemetry downlink.The primary pair is thermoelectricallycooled and has a maximum noise

temperature of 55 K. The standby LNRsare uncooked, with a noise temperature of62 K. Fig. 4 is a block diagram of the TT&Creceiving and transmitting system.

Commanding system

The command system is designed tominimize errors on several levels: coding,transmisson, and command generation.

Under routine operating conditions, corn -

Fig. 4Combined TT&C SOCC (iround station. Each of the two such Satcom stations is almost totally auto-.omous.

COUNTER 1

LEASEDFOR LAUNCH

PAPER TAPE H,PUNCH

CRT DISPLAY

SPECTRLFMANAL YiER NAG

RE YROARD rIAPER APEL.L READER JTAPE

HPOSSOB 32K CORECOMPUTER

MAC TAPE C csc

OSCILLOSCOPE SWEEP GEN I OSCILLATOR 1

TEST SE LECTOR PANEL 1 110 BUS AND CONTROLLERS,

CONTROL COMPUTER

---4.4- ,70,;(-...1 TO TRANSFER_ _ ORBIT STATION

r MODEMEAST NESTTO 1101

WW,/ RCVR

CMDSWITCHING

MATRIX

TIME CODE GEN

SPACECRAFT OPERATION& CONTROL CENTER

CRT DISPLAY

CRATSIMULATOR

PATCH

DOWN CONYA LINE

DRIVERS

RCVRINPUTPATCHPANEL

NPUTS

ANALOGTAPE

RECORDER

rTAPE CIEGAUSsE ft

IMP 40ANALOG

TAPE

RECORDER

TIME CODREADER

CUD A RANGETONE GEN

CMD 6 RANGETONE GEN

1

CONSOLE

GMT

KEYBOARD DISPLAY

AUX, OP.

MISSION?ME

CRT DISPLAY 32K CORE

KEYBOARD COMMUTER

4 DIAGNOSTIC USE ONLY

PHASEME TE R

MAC/ TAPE -2

DATA COMPUTER

SIC CUDRCVR

TE LEM 11.RANGING

RCVR

I TE LEM RANGING

RCVR

TELEMRANGING

RCVR

TE LETA &RANGING

RCVR

SWITCHINGMATRIX

SWITCH

GADVERIFICATION

PLL 3543 Mc71344 Ns

A 31612 HI

27,77 KEW

ATTTIMING

PROCESSOR

.3

K

3PK

L_PHONE .;

AUX OP

CRT 01JILAyKEYBOARD

CRT DISPLAY

KEYBOARO

SICSTATUSDISPLAY

S/CSTATUSDISPLAY

I

CRT DISPLAY

KEY110RD

S/CSTATUSDISPLAY

/AUL TIPROGRAMME R

$

CUDVERIFICATIONS/C

DISCRIMINATORIRO RME

DISCRIMINATORIRS KI4/

DISCRIMINATOR14 5 RH/

DISCRIMINATORI4 5 KIE/

STATION CONTROL PANEL

MCA iIHNATO145 KHz

SWITCHINGMATRIX

DISC

PAPER TAPEREADER

PAM SIGCONDITIONER

PAM SIG H..CONDITIONER

PAM SIGCONDITIONER

LINEPRINTER

MODEM

CHART RCORINPUT

PATCH PANEL

MEM

EAST/WEST DATA

--.1 CONDITIONERRmay,MODEM FROM TRANSFER ORBIT ST

M

LEASED FOR LAUNCH

L.ANALOG MODEM J RAW TE LEM DATA FROM E TR

: SIGNAL PROCESSING EQUIP.101

ma nd ing is performed using "canned" com-mand lists designed to execute the desiredfunctions. Generating and executing eachlist entails authorization. The control com-puter provides command -list generation,storage, and initiation. Commands aretime -tagged and, once authorized, will beexecuted in accordance with thepredetermined time sequence.

Spacecraft commanding entails the con-cept of load, then verify, then execute. Thereceived command is retransmitted via thespacecraft beacon to the TT&C/ SOCC,where it is compared to the original com-mand. Upon verification, an "execute"command is sent to the spacecraft. Inaddition to the space -to -ground link,verification is also provided via a ground-antenna feedback loop, facilitating rapidfault -isolation to the ground segment.

Finally, protection is provided in the com-mand format and code. A Hamming dis-tance of two exists between the spacecraftaddress and operation codes (op -codes), soa single -bit error at the spacecraft will notresult in acceptance of the command.

Beyond the preceding, commands whichcould adversely affect revenue -producingcapability or spacecraft life are classified as"hazardous" commands. These cannot betransmitted without activating the "hazar-dous command key," located on theoperations control console.

The format allows for 256 command op -codes, of which 210 have been im-plemented in the spacecraft. Commandtransmission is at 100 bits per second. Themodulation is PCM/ FSK/ FM, with ter-nary FSK; the execute tone is a differentfrequency from either the "mark" or"space," giving added protection againstfalse command execution.

Given the above described system, theprobability for false command reception bythe spacecraft is less than 1022.

TelemetryTelemetry is updated from each spacecraftevery 2 seconds; there are 128 channels perframe. One hundred and ten channelsrequire processing by the data computer inreal time. A total of 145 points of informa-tion (155 with both beacons) are relayedfrom the spacecraft either as status oranalog data.

Each Satcom spacecraft can transmit twotelemetry signals simultaneously, one on

each beacon. PAM/ FM/ PM is usedfor "housekeeping" telemetry;analog/ FM / PM is used with both attitudetelemetry during transfer orbit, and "dwell"telemetry where a given channel is ex-amined for closer evaluation of a givenparameter. A I4.5 -kHz subcarrier is fm -modulated by either PAM or analog data.The modulated subcarrier in turn phase-

modulates the beacon rf carrier.

After down -conversion (see Fig. 4),telemetry is patched to one of the fourtelemetry and ranging receivers, where it isdemodulated and switched to one of thefour 14.5 -kHz discriminators. Afterrecovery of the baseband signal, PAMtelemetry is switched to any one of fourPAM signal conditioners, where analog -to -digital (A/ D) conversion is made. Thedigitized data is received by the datacomputer for processing, logging on discand digital tape, and display on CRTs,stripcharts, or line -printer hard copy.

After it is received, attitude data is routedto the attitude and timing processor, whereit is quantized and transmitted to the datacomputer for display on stripcharts or forpre-processing into an "attitude data file,"which is transmitted to the off-line 1108 forsubsequent determination of thespacecraft's spin -axis orientation. Dwelldata is routed directly for display on one ofthe 18 available stripchart recorderchannels.

Telemetry is backed up on several levels,providing for greater system operatingreliability and flexibility. Two seven -trackanalog tape recorders can continually storereceived telemetry (including ranging andcommand verification data). Should thedata computer be unable to process in realtime because of a failure or scheduledmaintenance, the stored analog tape maybe played back to recover telemetry data.Analog tapes may be maintained for aperiod of several days as a backup to theprocessed data logged on the digital tape.The digital tapes are stored for severalmonths for use in intermediate term trendanalysis. Subsequently, "data compres-sion" is performed on these tapes. Com-pressed tapes are then stored for long-termtrend analyses and posterity.

RangingSlant range to the spacecraft is measuredby using the spacecraft command receiverand beacon transmitter in transponderfashion. Four coherently generated sine-

wave tones are transmitted one at a time viathe command uplink, then received via thebeacon downlink. The range is determinedby measuring the shift in phase anglebetween each transmitted and receivedtone.

The four tones, generated by the commandand range tone generator under data-computer control, are 35.4 Hz, 283.4 Hz,3968.1 Hz, and 27.777 kHz. One cycle ofthe low tone measures to within 4241 km;the next two higher tones resolveambiguities to within 530 and 37 km; andthe high -frequency tone resolves the rangeto within 5.409 km.

Using all four tones together, the resultantrange measurement is typically accurate towithin 25 meters (I sigma), when using one -station ranging, and 12 meters (I sigma)when using two -station ranging.

The ranging system employs FM/ FMmodulation on the uplink and FM/ PM onthe downlink.

Typically, ranging to each spacecraft isperformed subsequent to a maneuver. Datais collected hourly, alternating betweenVernon Valley and Moorpark, to achievegreater accuracy. Ranging tones areapplied for a period of approximately 15minutes, each time. Accumulated tracking -data files are transmitted to the off-line1108, via the control computer, for orbit-determination processing.

Spacecraft operationscontrol center (SOCC)

\ six -cabinet wraparound console is thelocal point for Satcom spacecraft systemoperations. Four CRT/ keyboards are theprime monitoring and controlling devices.Three of these keyboard terminals andtheir associated character -mode CRTscreens interface with the data computerand also request and display telemetrydata. All three CRTs can display telemetryfrom one spacecraft or each CRT candisplay telemetry from a differentspacecraft.

Three additional 23 -inch black -and -whitetv monitors suspended from the ceiling actas repeaters, enabling operations analyststo monitor spacecraft health and statuswithout crowding the console area.

A fourth character -mode CRT/ keyboardinterfaces with the control computer andacts as the executive -command input to the

102

system. Command generation, authoriza-tion, and spacecraft control are done here.

An "operator control panel" allows theconsole operator to monitor and controlthe TT&C equipment remotely: e.g., com-puter status monitoring, stripchartrecorder, and analog tape -recorder con-trol. The panel also contains the

hazardous -command lock and alarms.

Intersite voice and data communicationscontrol is performed at the controlconsole, where the operator can controldata transmission to the off-line 1108 or tothe other TT&C/ SOCC, for example. Amission -time clock and a GMT clockround out the console equipment. Bothclocks are driven from a time -code

generator that is synchronized with aW WV receiver.

In addition to the console and its equip-ment, a line printer, three six -channel strip -

chart recorders connected to the datacomputer. and a second printer, switchableto either the data or control computer, givethe control center area the required hard -copy capability.

Computer software

Development of new computer softwareIA as a major undertaking needed to accom-modate a new spacecraft.

The software system consists of two majorparts: real-time software, needed to acceptand operate on telemetry and providespacecraft command and control; and off-line software, associated with orbital com-putations, establishing orbital position,planning maneuvers to effect the desiredtransfer orbit and on -orbit station -keeping.

All the software except the orbitdetermination and prediction software arenew. the latter was based on the "space 360"program used by the U.S. Government.

The off-line software performs the keyfunction of orbit determination and predic-tion. spin -axis attitude determinations,spin -axis precession planning, apogee kick

Clockwise, from tcp:Control console at VernonValley during launci of Sat -corn F-1.TT&C;SOCC o3erationsroom at Vernon Valley.TT&C antenna, VernonValley.

I();

Table IReal-time software resident in data and control computers.

Key modules of the real-time software system

Data computer Control computer

lelcmetry calibration processingI elemetry loggingIelemetry data intersite tele-

communicationsVideo displayChart recorder displayAttitude data processingAntenna (AaitE 11 data collectionTT&C antenna controlRanging data processor

Command -list generationS. C Command -list executionStation Command -list executionEast coast/ West coast system

coordinationIntrasite communications1108 file processing

1108 execution control1108 file transmission

Jack Lewin is Principal Member ofAmericom's Space Systems Engineeringstaff. He has over 19 years of experience insystem design, analysis, and testing indigital communications, data processing,command and control, and computerprogramming and applications. He joinedRCA in 1959 and has worked on numerousprograms in missile- and spacecraft -relatedprojects in these fields. His most recentassignment was as project engineer on theSatcom spacecraft command, ranging, andtelemetry subsystem, the groundTT&C/SOCC subsystem, and missionoperations. His present assignment is asgroup leader on mission and systemssoftware.Contact him at:Space Systems EngineeringRCA American CommunicationsPiscataway. N.J.201-885-4189

motor firing planning, maintaining andupdating the thruster performanceparameters used in maneuver -planning,hydrazine fuel status, station acquisitionplanning, station -keeping planning, anddata -base management.

Table I lists the key software modulescomprising the "real-time" software systemresident in the data and control computers.

Performance testing

The common thread throughout the Sat-com ground system development was thenewness of its constituent parts. This meantthat meeting the launch date with thedesired degree of confidence became achallenging exercise in testing and rehear-sals.

To expedite testing, the orbit -determination and prediction software, forexample, were tested against spacecraftdata supplied by NASA and Telesat,Canada.

The ultimate way of testing a satellitetracking system, however, is with a satellitein orbit. This enables testing of not onlysoftware, but the antennas, the uplink, thedownlink, and data and control computers,as well as the personnel.

In order to have this testing done prior tothe launch of the Satcom spacecraft, RCAarranged to have NASA's ATS-I , ATS-5and Anik II exercise the ranging subsystemand the orbit -determination and predictionsoftware.

This accelerated system debugging im-measurably. Also, incorporating live rang-ing into the mission rehearsals lent a highdegree of reality to the exercises, and led tothe ultimate achievement of two successfulmissions.

Acknowledgments

A number of technical groups andorganizations each contributed in theirspecialized way to the success of the Sat-corn ground segment development andrealization.

At RCA American Communications, Inc.,the TT&C and Mission OperationsEngineering, Satcom Spacecraft Control,Earth Station Engineering, and the EarthStation Implementation activities eachcontributed significantly to the Satcommission success by insuring the timelyavailability of the required systems,facilities, and personnel, all of which met orexceeded performance requirements.

RCA Astro-Electronics deliveredhardware, computer software, and man-power that helped produce two flawlessSatcom launch and transfer -orbitoperations.

Felesat Canada's Computer SoftwareSystem and Mission Analysis activity con-tributed invaluable advice toward a well-functioning timely system.

NASA allowed RCA to use the ATS-1 andATS-5. which helped debug the SatcomTT&C, SOCC system, and contributed to ahighly reliable operating ground systemprior to the launch of the Satcom satellites.

References

I Napoli. J.; and Christopher. J.; "RCA Satcom system." RCAEngineer. Vol. 21 No. I (Jun Jul 19751 pp. 23-29.

2. Napoli..).; and Greenspan. J.; "RCA Swum. the nextgeneration domestic communications satellite system." (Sep16-19. 1975) Wescon Communication Satellite Systems.

3. Becken. T.D.. "Satellite communications." RCA Engineer.Vol. 22 No. I (Jun Jul 1976) pp. 39-41.

4. Brook. A.W.."RCA Satcom system." RCA Engineer. Vol. 22No. I (Jun Jul 19761 pp. 42-49.

5. height.. J.E., "RCA Satcom - maximum communicationscapacity per unit cost." RCA Engineer. V 01.22 No. ((Jun Jul1976) pp. 50-55,

6. Cuddihy J.; and Walsh. J.M.; "RCA Satcom earth stationMeddles." RCA Engineer. Vol. 22 No. I (Jun, Jul 19761pp. 58-63.

7. Christopher, J.. Greenspan. D.; and Plush. P.; "The launchand in -orbit test elements of the Satcom system." RCAEngmerr Vol. 22 No. 1 (Jun Jul 19761 pp. 64-70.

S. Bell. J.R., and Bell, CU; editors. Mini -Computer Sirftwore,North Holland Publishing Co.. 1976. See Smith. R.D., andMills. R.W.; "RCA Satcom programming technology."

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 Administrator(listed on back cover) for a reprint. For additional assistance in locating RCA technical literature, contact RCA TechnicalCommunications. Bldg. 204-2, Cherry Hill, N.J., extension PY-4156.

Astro Electronics

G.T. Tsengl K.J. PhillipsAttitude stability of a flexible dual spinspacecraft with active nutation dampingusing products of inertia, J. AstronomicalSc., Vol. 24, No. 3 (7-9/76)

Missile and Surface Radar

M.W. BuckleyProject management-planning, schedul-ing and control-AMA Seminar, Chicago, IL(10/11-14/76)

M.W. BuckleyProject/program management-Drexel Un-iversity, Phila., PA (11/8-10/76)

J.R. Foglebochi F.I. PalmerJ.O. Taylor' R.W. SudburySolid state transceiver time sidelobeperformance-Digest GOMAC, Lake BuenaVista, FL (11/9-11/76)

M.W. BuckleyProject/program management -PERT/CPM Workshop, The BehrendCollege, Erie, PA (Pennsylvania State U.Continuing Education) (9/30-10/2/76)

B.M. FellElectron mobility for a two-dimensionalsemiconductor at low temperatures-Condensed Matter Symp., SUNY, StonyBrook, L.I., NY (11/76)

J.C. VolpeRCA and radar-evolving technology in thereal world-Signal, Vol. 31, No. 2 (10/76) Pp.40-45

M.W. BuckleyManagement planning, scheduling and con-trol, AMA Meeting, Baltimore, MD (9/27-28/76)

M.W. BuckleyManagement by objectives-PhiladelphiaNaval Base Officers' Club, Phila., PA(9/15/76)

D.L. PruittThyristor switches for super power intermit-tent duty operation-Proc. Pulse PowerSystems Workshop, Naval Surface WeaponsCenter, White Oak, MD (9/21-23/76)

J.T. ThrestonEnhanced carrier operations through use ofAN/SPY-1 radar system-Proc. Amer. Soc.Naval Engineers, San Diego, CA, (10/7-8/76)

D. ShoreRPV: the future is here-National Defense,vol. 61, No. 339 (11/76)

Automated Systems

J.C. WillettThe soldering process and solder Jointinspection-Printed Circuit Technology QCSeminar, Burl ngton, MA (12/3/76)

M.L. Johnson C.F. MatthewsTransferring usable information to themaintenance man -11th Intl. LogisticsSymp., Valley Forge, PA (8/19/761

M.J. CantellaObservations of sky brightness spatialvariations-SPIE Technical Symp., SanDiego, CA (8/27/76).

M.J. CantellaHigh resolution camera tubes -1976Special Summer Program on Photoelec-tronic Imaging Devices, San Diego, CA(8/21/76)

C.A. AsbrandSurveillance with night vision equipment-Dept. of Criminal Justice Crime Lab, ClintonC.C., PlattsbLrgh, NY (9/76).

M.J. CantellaElectrical input-output storage tunes -1976Special Summer Program on Photoelec-tronics Imaging Devices, San Diego, CA(8/18/76)

O.T. CarverEQUATE-New technology ATE-DefenseExpo '76, Wiesbaden, Germany (10/6-8/76)

T.E. KupfrianCommercial ntegrated circuits In selectedmilitary applications-IEEE Boston SectionReliability Chapter, Burlington, MA(1/12/77)

E.W. Richter' V.D. HoladayDesigning microwave ATE to meet UUTrequirements-AUTOTESTCON. Arlington,TX (11/10-12/76)

Corporate Engineering

H. Kleir berg' F.M. OberlanderStandarization in a multidivision company-Standards Engineering, Vol. 27 No. 4 (8/75)pp. 62-64

Advanced TechnologyLaboratories

E.P. Herrmann' D.A. GandolfoA. Boomardi D.B. SteppsCCD programmable correlator-Proc. 3rdIntl. Conf. on Applications of ChargeCoupled Devices, Edinburgh, Scotland(9/28-30/76)

E.P. Herrmann' D.A. GandolfoA. Boornardi D.B. SteppsA high-speed, CCD-scanned pnotosensorfor gigabit recording applications-Proc.1976 Electro-Optical Sys. Design Conf./Intl.Laser Exposition, New York, NY (9/14-16/76)

GovernmentCommunciations Systems

O.E. BessetteRecent advances in magnetic recording ofdigital data-Intl. Telemetry Conf., LosAngeles, CA (9/28-30/76)

A.M. Burke' F.L. PutzrathNASA standard spacecraft tape recorder-int. Telemetry Conf., Los Angeles, CA (9/28-30/76)

0. BlackWater soluble residue fluxes-PracticalSoldering Technology, Anaheim, CA (9/21-23/76)

D. Harwell J. RothweilerAn LSI FFT signal detection and demodula-tion processor-Government MicrocircuitConf., FL (11/10/76)

Broadcast Systems

M. S. SiukolaNew multi -station top mounted fm

105

antenna -26th Annual IEEE BroadcastSymp., Washington, DC (9/23-24/76)

0. Ben-DovRadiation patterns of broadcast antennasfrom aperture illumination -26th AnnualIEEE Broadcast Symp., Washington, DC(9/23-24/76)

A.J. FandozziNoise measuring techniques for televisiontransmitters -26th Annual IEEE BroadcastSymp. Washington, DC (9/23-24/76)

Picture Tube Division

R. MarshallDesign criteria for platinum -rhodium alloysheath thermocouples for stable operationabove 1300C-Proc. Instrument Society ofAmerica, Houston, TX (10/11-14/76). Also inProc. ISA-'76 Annual Conf., Vol. 31, Part 3

LaboratoriesW.C. StewartOn differential phase contrast with an ex-tended illumination source-J. Optical Soc.America, Vol. 66, No. 8 (8/76) p. 813

J.E. Carnes' R.L. Rodgers IllRecent results on CCD imagers-Laser 75Opto-Electronics, p. 78

H. Kressell M. EttenbergLow -threshold double-heterojunctionAlGaAs/GaAs laser diodes: theory andexperiment-J. Appl. Phys., Vol. 47, No. 8(8/76) p. 3533

G.W. Hughes' R.J. PowellM.H. WoodsExide thickness dependece of high-energyelectron, VUV, and corona -Induced chargein MOS capacitors- Appl. Phys. Lett., Vol.29. No. 6 (9/76) pp. 377-79

G.L. SchnableApplications of electrochemistry to fabrica-tion of semiconductor devices-Fabricationof Semiconductor Devices, Vol. 123, (9/76)p. 310-C

L.C. UpadhyayulaOuasienhancement mode of operation oftransferred electron logic devices(TELDs)-Electronics Lett., Vol. 2, No. 1G(10/13/76)

C.R. Wronskil D.E. CarlsonR.E. Daniell A.R. TrianoElectrical properties of a -Si solar cells-Digest, 1976 IEEE Intl. Electron DevicesMtg., Washington. D.C. (12/6-8/76).

D.O. NorthTheory of model characters, field structure,and losses for semiconductor laser-IEEE J.Quantum Electronics, Vol. QE -12, No. 10(10/76) pp. 616-624.

A. Rosen' P.T. Hol J.B. KlatskinFabrication and thermal performance of a

novel trapatt diode structure-IEEE Trans.Electron Devices (2/77)

C.J. NueseDiode sources for 1.0 to 1.2 urn emission-Technical Digest, 1976 Intl. ElectronDevices Mtg., Washington, D.C. (12/6/76)pp. 125-128

R.J. HimicsI M. KaplanN.V. Desail E.S. PoliniakPoly (cyclopentene sulfones) as electronbeam resists-SPSE Business GraphicsAbstracts, Society of PhotographicScientists and Engineers Symp.,Washington, DC (11/9-13/76) pp. 82-86

D.R. Carter' K.G. HernqvistAdvances in helium -cadmium lasers-Proc.of Electro-Optical Systems Design Conf.,New York, NY (9/14-16-76)

E.S. KohnIR imaging with Schottky barrier CCDarrays-Solid State Circuits Committee ofthe IEEE Circuits and Systems Society(10/15/76)

M. Skolnini L.C. ThanhF. Levy' G. HarbekeHigh -field magneto -absorption in-vestigations of excitor states in Pb12-Int.Conf. on Layered Semiconductors andMetals. Bari, (9/6-10/76)

D. MeyerhoferNew technique of aligning liquid crystals onsurfaces-Appl. Phys. Lett., Vol. 9 No. 11(12/1/76) pp. 691-2

M.S. Abrahams' C.J. Buiocchil R.T. SmithC.W. Cullen' J.F. Corboy, Jr.' J. BlancEarly growth of silicon on sapphire-I:transmission electron microscopy-Electrochemical Soc. Fall '76 Mtg., LasVegas, NV (10/76)

J. Blanc) M.S. Abrahams' G.W. CullenJ.F. Corboy, Jr. I C.J. Buiocchil R.T. SmithEarly growth of silicon on sapphire-II:models-Electrochemical Soc. Fall '76 Mtg.,Las Vegas, NV (10/76)

P.H. Robinson' R.V. D'Aiellol H. KresselDichlorosilane: a silicon source for expitax-ial growth-Electrochemical Soc. Fall '76Mtg., Las Vegas, NV (10/76)

R.J. Himicsl D.L. RossThe examination of poly (5-hexene-2-cnesulfone) as a positive -working photoreslst-Technical Papers, Soc. Plastics EngineersConf., Ellenville, NY (10/13-15/76), pp. 26-33

R.J. HimicsI M. KaplanN.Y. Desail E.S. PoliniakThe synthesis, characterization, evaluation,and processing of selected sulfonecopolymers as electron beam resists-Technical Papers, Soc. Plastics EngineersConf., Ellenville, NY (10/13-15/76) pp. 269-283

P.T. Ho' A. Rosen' J. KlatskinPower combination of broadband rrappatt

amplifiers-Electronics Lett., Vol. 12, No. 1(1/8/76) p. 16

M.T. Galel K. KnopColor -encoded focused image holograms-Appl. Optics, Vol. 15, No. 9, (9/76) pp. 2189-2198

A.G.F. Dingwall' R.E. StrickerC2L: A new high speed, high density bulkCMOS technology-IEEE Intl. ElectronDevice Conf. Washington, DC (12/7/76) p. 3

J.I. PankoveNew materials for electroluminescentdevices -1976 Intl. Conf. Solid StateDevices, Tokyo (9/2/76)

R.A. GeshnerLSI photolithography-I.G.C. Conf. onMicrophotolithography, Ipswich, MA(9/13/76)

A.G. Ipri1J.C.SaraceCMOS/SOS process evaluationtechniques-GOMAC. Buena Vista, FL(11/9-11/76)

A. Bloom' R.A. BartoliniL.K. Hung' D.L. RossA non -polymeric host for recording volumephase holograms-Appl. Phys. Lett., Vol. 29,No. 8 (10/15/76) pp. 483-4

D. Vilkomersoni R. Mezrichl M.F. EtzoldAn improved sysem for visualizing andmeasuring ultrasonic wavefronts-VII Intl.Symp. on Acoustical Imaging andHolography

A. BloomI R.A. Bartolinil P.L.T. HungThe effect of host on volume phaseholographic recording properties-SPE 4thTechnical Conf. on Photopolymers,Ellenville, NY (10/13/76)

R.A. Bartolinil A. BloomReview of organic holographic recordingmedia-Optical Soc. Annual Mtg., Tuscon,AZ (10/76)

A. BloomI P.L.K. HangEffect of dye structure on order parameter ina nematic liquid crystalline host -6th Intl.Liquid Crystal Conf., Kent, OH (8/24/76)

C.R. Wronskil D.E. Carlson' R.E. DanielSchottky -barrier characteristics of metal -amorphous -silicon diodes-Appl. Phys.Lett., Vol. 28, No. 9 (11/1/76) pp. 602-605

R. Williams' R.S. CrandallI P.J. WojtowiczMelting of crystalline suspensions ofpolystyrene spheres-Physical ReviewLett., Vol. 37, No. 6 (8/9/76) pp. 348-351

D.A. deWolfWaves in random media: weak scatteringrevisited -1976 Intl. IEEE/APS Symp. &USNC/URSI Mtg., Amherst, MA (10/11-15/76)

G.L. SchnableReliability of MOS devices In plasticpackages-International Micro -electronicsConf. (6/9/76)

106

J.R. SandercockSimple stabilization scheme formaintenance of mirror alignment in a scann-ing Fabry-Perot interferometer-J. PhysicsE: Scientific Instrument, Vol. 9 (1976) p. 567

E. Sichell R. MillerThermal conductivity of highly orientedpryolytic boron nitride-Thermal Con-ductivity 14 (1976)

RCA RecordsA. Devarajani D. MishraComputerized technique for analyzing in-ventory control problems-American Inst.of Industrial Engineers Annual SystemsEngineering Conf., Boston, MA (12/3/76)

D. MishraComptuer-aided work sampling-Work-sampling workshop, U. of Wisc. Ext.,Milwaukee, WI (11/17/76)

PatentsAutomated Systems

S.C.HaddenIL.R. HullsP.J. Slaneyl E.M. Sutphin, Jr.Tachometer without physical connection tointernal combustion engine -3978719

GovernmentCommunications Systems

S.P. ClurmanReduction of hunting in synchronousmotor -3988653

E.J. Nosseni E.R. StarnerAccurate digital phase/frequency extractor-3984771

Advanced TechnologyLaboratories

B W SiryjObject orientation apparatus -3986604

K.C. Hudson' R.F. KenvilleOptical scanner with large depth of focus -3989348

B.W. SiryjLabeling apparatus -3984279

Solid State Division

A.A. AhmedEqualization of base current flow in twointerconnected transistor amplifiers -3987368

A.A. AhmedCurrent amplifier -3990017

D.R. CarleyMethod of making a semiconductordevice -3980507

A.A. AhmedAbsolute -value circuit -3989997

A.J. PikorMultiple mesa semiconductor structure -3988765

R. Denning' W.G. EinthovenThermally balanced PN junction -3988759

H.A. WittlingerSeries energized transistor amplifier -3986132

S.W. Kessler, Jr.Transcalent semiconductor device -3984861

C.F. Wheatley, Jr.Bridge -output amplifier with direct -coupleddifferential -mode feedback -3983502

A F ArnoldNickel -gold -cobalt contact for silicondevices -3982908

O.H. Schade, Jr.Radiation responsive voltage dividingcircuit -3982197

S. Berkman J.G. MartonGraphite susceptor structure for inductivelyheating semiconductor wafers -3980854

M. GlogoljaProtection circuit -3980930

H. Arnold'Transistor testing circuit -3979672

0 H Schade. JrDifferential amplifier circuit -3979689

M.I. PayneInformation storage circuit -3979735

M.I. Payne) B.S. DalaiTransistor switching circuit -3979611

A.A. AhmedCurrent level detector -3979606

H.W. JusticeEtchant for silicon nitride and borosilicateglasses and method -3979238

A.D. Check!' A.G FreyAutomatic assembly of semiconductordevices -3978579

Consumer Electronics

J. AvinsTransient suppression in television videosystems -3984865

D.H.WillisGating circuit for a video driver inc1uding aclamping circuit -3984864

B.W. Beyers, JrCharacter generator for television channelnumber display with edging provision -3984828

S.A Steckler) A.R. BalabanDeflection waveform correction signalgenerator -3984729

J B GeorgeHigh power remote control ultrasonictransmitter -3984705

R.L Shan ey. II J M YongueApparatus for accentuating amplitudetransistions-3983576

M.E. Miller) J.G. AmeryVelocity correction system with dampingmeans -3983318

J.C. SchoppTurntable speed control system -3983316

L.A. CochranSwitching arrangement for flesh tonecorrection and chrominance overload con-trol circuits -3982273

J.C. Peer' D.W. LuzPower supply for a television receiver -3980821

K R Woolling, JrCounter type remote control receiver in-cluding noise immunity system -3980956

J.B. GeorgeTuner bandswitching system for a elevisiontuning system -3980959

LaboratoriesA V TumaDigital remote control for electronic signalreceivers -3987414

107

C.R. Wronskil A.D. Cope) B. AbelesLow dark current photoconductive device-3987327

L S NapoliMethod of forming conductive coatings ofpredetermined thickness by vacuumdepositing conductive coating on a measur-ing body -3987214

E.C.Giairno, Jr.Method of increasing the image exposureand developing sensitivity of magneto -electric printing system -3986872

M.Ettenbergl H.KresselSolar cell device having twoheterojunctions-3990101

F.C. Duigonj S. LiuPlanar trapatt diode -3990099

C.J. Busanovichi R.M. MooreSelenium rectifier having hexagonalpolycrystalline selenium layer -3990095

C T WuDisplay device utilizing magnetic storage-3988738

H. Kressell V.L. DalaiSolar cell device having improvedefficiency -3988167

W.F. KosonockyIntroducing signal at low noise level tocharge -coupled circuit -3986198

P.K. WeimerCharge transfer memories -3986176

J.I. PankoveInsulating nitride compounds as electronemitters -3986065

H.L. Pinch) S.T. OpreskoVapor deposition of cermet layers -3985919

J.L. Vossen, Jr.1F.R. NymanG.F. NicholsAdherence of metal films to polymericmaterials -3984907

G.S. Kaplan' A.D. RitzieHomodyne communication system -3984835

J. Rosen) E.J. DenlingerTwo -inductor varactor tunable solid-statemicrowave oscillator -3984787

J. Gross] W.H. BarkowDisplay system utilizing beam stripecorrection -3984723

S LarachApparatus and method for analyzingbiological cells for malignancy -3984683

I GorogFlying spot scanner unaffected by ambientlight -3984629

N. FeldsteinMethod for etectrolessly depositing metals

using improved sensitizer composition-3982054

P.N. Yocoml J.P. DismukesLuminescent sulfides of monovalent andtrivalent cations -3981819

P.E. Hafer!Deflection circuit -3980927

H.R. Beelitz] D.R. PreslarElectrical circuit -3979607

S.A. LippChemical vapor deposition of luminescentfilms -3984587

F.Z. HawryloOhmic contact -3984261

P.M. RussoElectron image identifying system -3983571

H.E. SchadeVacuum electron device having directly -heated matrix -cathode -heater assembly-3983443

D.J. ChanninLiquid crystal display -3981559

P.M. Heyman' R.L. Quinn' I. GorogElectrochromic display device -3981560

Distributor and SpecialProducts Division

P.J J.D. CallaghanBroadband antenna system with the feedline conductors spaced on one side of asupport boom -3984841

Broadcast Systems

L.J.BazinHigh efficiency deflection circuit -3983452

A M. Goldschmidt1W.A. Dischert J.R. WestSystem for controlling tension of magnetictape -3982160

Missile and Surface Radar

J.L. ChristensenRF cross correlation radar -3981013(assigned to U.S. Gov't)

R.P. PerryData processor reorder shin registermemory -3988601

R.P. PerryDigital matched filtering using a steptransform process -3987285

J.R. FordApparatus for erecting a true vertical axis-3985033

J.O. HorsleyHigh-speed counter with reliable count ex-traction system -3982108

D. Olivieril R.J. SocciHeat spreader and low parasitic transistormounting -3982271

Picture Tube Division

M.H. Wardell, Jr.Cathode ray tube assembly fixture -3989233

B.K. SmithYoke mount assembly -3986156

T.A. SaulnierPhotograph method employing organiclight -scattering particles for producing aviewing -screen structure -3981729

D.D. VanOrmerShadow mask color picture tube havingnon -reflective material between elongatedphosphor areas and positive tolerance-3979630

L.J. DimattioElectron tube base -3979157

RCA Ltd. Canada

C.H. ToddMethod of air letting an evacuated cathoderay tube -3981554

G.J. Lol F.L. PapworthN.E. Terme-Sens' M.V. O'Donovan G. DziubMetal plated body composed of graphitefibre epoxy composite -3982215

Avionics Systems

J.E. MillerTracking gate servoed by relative range-3987441

SelectaVision Project

M.L. McNeeleyi H. ReesApparatus for producing injection moldedand centrally apertured disc records -3989436

F.R.Nymanl J.L. Vossen, Jr.D.G. Fisher' G.F. NicholsMetal coating for video disc -3982066

RCA Ltd. BelgiumH.J. DigneffeAlternating current control system -3990000

108

Dates and DeadlinesUpcoming 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.

JAN 10-14, 1977 - AIAA 13th Annual Mtg.and Technical Display, Wash. DC Prog Info:Martin Newman, AIAA, 1290 Sixth Ave., NewYork, NY 10019

JAN 18-20, 1977 - Reliability and Main-tainability Conf. (IEEE) Marriott, Phila., PAProg Info: J.M. Wiesen, Dept. 1220, SandiaLabs, Albuquerque, NM 87115

FEB 7-9, 1977 - Aerospace and ElectronicSystems Winter Convention (WINCON)(IEEE) Sheraton -Universal, NorthHollywood, CA Prog Info: H.S. Abrams,Litton Systems, Inc., G&CS Div., 5500Canoga Ave., Woodland Hills, CA 91364

FEB 16-18, 1977 - Intl. Solid State CircuitsConf. (IEEE) Sheraton, Phila., PA Prog Info:John H. Wuorinen, Bell Labs, Whippany, NJ07981

FEB 22-24, 1977 - Optical Fiber Transmis-sion Conf. (IEEE, OSA) WilliamsburgLodge, Williamsburg, VA Prog Info: R.D.Maurer, Sullivan Park, Corning GlassWorks, Corning, NY 14830

FEB 28 -MAR 3, 1977 - COMPCON Spring(IEEE) Jack Tar, San Fran., CA Prog Info:L.C. Hobbs, Hobbs Assoc., POB 686, Cor-ona Del Mar, CA 92625

MAR 16-18, 1977 - Vehicular TechnologyConf. (IEEE) Orlando Hyatt House, Orlan-do, FL Prog info: G.F. McClure, 1730 ShilohLane, Winter Park, FL 32789

MAR 17-19, 1977 - Simulation Symp.(IEEE) Tampa, FL Prog Info: Ira M. Kay, POB22573, Tampa, FL 33622

MAR 21-23, 1977 - Industrial Applicationsof Microprocessors (IEEE) Sheraton, Phila.,PA Prog Info: S.J. Vahaviolos, Eng.Research Ctr., Western Electric Corp., POB900, Princeton, NJ 08540

MAR 21-23, 1977 - American NationalMetric Council, Third Annual Conf. andExposition, McCormick Inn, Chicago, ILProg Info: ANMC, Conference Committee,1625 Massachusetts Ave., N.W., Suite 501,Wash., DC 20036

MAR 23-25, 1977 - Computer ArchitectureSymp. (IEEE, ACM) College Park, MD ProgInfo: Bruce Wald, Naval Research Lab., 4555Overlook Ave., Washington, DC 20390

MAR 18-21, 1977 - Semiconductor PowerConverter Intl. Conf. (IEEE) Walt DisneyContemporary, Orlando, FL Prog Info:

Eberhart Reimers, USAMERDC, AMXFB-EA, Electrical Equip. Div., Fort Belvoir, VA22060

APR 6-8, 1977 - Digital Processing ofSignals In Communication (IERE, IEEE)Loughborough, UK Prog Info: IERE, 8-9Beford Square, London WC1B 3RG,England

APR 18-20, 1977 - Fifth Conf. on Chemicaland Molecular Lasers, Stouffer's RiverfrontInn, St. Louis, MO Prog Info: Dr. W.Q.Jeffers, Helios Inc., POB 2190, Boulder, CO80302

APR 18-21, 1977 - Design EngineeringConf. and Show, New York, NY Prog Info:Technical Affairs Dept., ASME, UnitedEngineering Center, 345 East 47th Street,New York, NY 10017

APR 19-21. 1977 - Society for InformationDisplay Intl. Symp. and Exhibition,Sheraton -Boston Hotel, Boston, MA ProgInfo: Lewis Winner, 152 West 42nd St , NewYork, NY 10036

APR 21-22, 1977 - Picture Data Descriptionand Management (IEEE) U. of III., Chicago,IL Prog Info: K.S. Du. Purdue Univ., Schoolof EE, Lafayette, IN 47907

APR 23-28, 1977 - American Ceramic Soc.,Electronics Div., 179th Annual Mtg. & Ex-position, Conrad ILProg Info: Dr. Richard M. Rosenberg, E.I.duPont de Nemours & Co., Inc., PhotoProducts, Bldg. 428, Buffalo Ave., NiagaraFalls, NY 14302

APR 25-27, 1977 - Circuits & Systems Intl.Symp. (IEEE) Del Webb's Towne House,Phoenix, AZ Prog Info: W.G. Howard,Motorola Integrated Circuits Center, MailStop MR, POB 20906, Phoenix, AZ 85036

MAY 2-5, 1977 - Offshore TechnologyConf. (IEEE et al) Astrodome, Albert

Thomas Convention Ctr., Houston, TX ProgInfo: OTC, 6200 N. Central Expressway,Dallas, TX 75206

MAY 3-6, 1977 - EUROCON 77 (IEEE et al)Venice, Italy Prog Info: Alberto VandindButi, c/o AEi, Viale Monza 259, 20126 Milan,Italy

JUN 1-3, 1977 - Conf. on Laser Engineer-ing and Applications (IEEE/OSA) Wash.,DC Prog Info: Conf. Mgr.: Anne J.Morandiere. Courtesy Assoc., Suite 700,1629 K Street, N.W. Wash., DC 20006

JUN 20-22, 1977 -14th Design AutomationConf. (ACM/IEEE) International Hotel, NewOrleans, LA Prog Info: David W. Hightower,MS907, Texas Instruments, Inc., POB 5012,Dallas, TX 75222

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.

APR 26-28. 1977 - Microwave Power TubeConf., Naval Postgraduate School,Monterey, CA Deadline Info: (abst) 1/30,Lynwood Cosby, U.S. Naval ResearchLaboratory, 4555 Overlook Ave., S.W.,Wash., DC 20390

JUN 20-24, 1977 - Intl. ElectromagneticSymp. (IEEE, URSI) Stanford U., Palo Alto,CA Deadline Info: 3/1, Prof. K.K. Mei, Dept.of EE and CS U. of California, Berkely, CA94720

JUL 26-25. 1977 - Intl. Conf. on CrimeCountermeasure Science and Engineering,(IEEE et al) Oxford University Oxford,England Deadline Info: (Abst) 1/14, John S.Jackson, Proceedings Editor, Dept. of Elec-trical Engineering, U. of Kentucky, Lex-ington, KY 40506

Clip out and mail tb Editor, RCA Engineer, 204-2, Cherry Hill, N.J.

ma EngineerHave we your correct address?If not indicate the change below:

Name Code #

Street or Bldg

City and State or Plant

'Please indicate the code letter(s) that appear next to your nameon the envelope.

109

Engineering News and HighlightsAutomated Systems team of nine cited forlaser rangefinder work

Design to Unit Production Cost of the AN/GVS-5 Hand -Held Lase'Rangefinder brought this team a TE award. From left to right areAnthony Amato, Manager Products Engineering; team members FredPratt. Sam Waldstein, Jay Woodward, Walt Radcliffe: Dr. Harry Woll, DivVice President and General Manager. team members Bob Guyer, FerdMartin, Larry Blundell, Norm Roberts, and Jim Quinn; Gene Stockton.Chief Engineer. and Bill Hannan, Manager Radiation SystemsEngineering.

Bosselaers cited for Equateaccomplishmentsat Automated SystemsRobert Bosselaers received the TechnicalExcellence Individual Award for his ac-complishments in the advancement of theEquate RF Synthesizer design. RF Syn-thesizers. up to 18 GHz with varyingmodulation types. have been a key factor inmaintaining RCA's front running position inthe ATE product line.

Left to right Dr. Harry Wok, Div Vice President andGeneral Manager. Bob Bosselaers; and Gene Stockton.Chief Engineer

Behnen

Landry

Socci

H lal wins award at GovernmentCommunications Systems

Design and Development of TENLEY Dedicated LoopEncryption Device won a Technical Excellence Awardfor George Hilal. Left to right are Hilal: J.V. Fayer.Manager. N. VanDelft. Leader: and C.A. Schmidt.Manager

41116\Bronecke

_,_4111111kPschunder

Udicious

Six receive technicalexcellence awardsat MoorestownSteve Behnen-foi his outstandingachievements in developing the weaponselection function for the AEGIS ship Com-bat Direction System.

Pete Bronecke-for his work in transform-ing a standard automatic integrated -circuittester (DATRON 4400) into a special high-speed device for testing complex hybridcircuits.

Norm Landry-for his outstanding perfor-mance in the development of an extremelyshort waveguide horn radiating element forphased array radars.

Ralph Pschunder-for further ad-vancements in the DYNA series of structuralanalysis computer programs. His updatedDYNA-3 provides substantial cost reductionand greater ease of modeling of complexstructures. and further includes a shockanalysis subrouting incorporating specifiedNavy standards for shock analysis.

Bob Socci-for his contributions in theredesign of the AN/SPY-1 RF receiver in anew version that provides a 70° reduction inproduction cost with improved perfor-mance, accessibility, and reliability andmaintainability.

Ralph Udicious-for his extraordinaryachievement in the installation, checkout,and certification of the CMS -2Y compilecapability in the AN/TP(1)-27 program sup-port facility.

Licensed engineers

When you receive a professional license,send your name. PE number (and state inwhich registered). RCA division, location.and telephone number to: RCA Engineer,Bldg. 204-2, RCA, Cherry Hill. N.J. Newlistings (and corrections or changes toprevious listings) will be published in eachissue.

RCA Records DivisionAugust Skele, Indianapolis, Ind.; PE -16489Ind

Melvin K. Martin, Indianapolis,16685 Ind.

Robert E. Fuller, Indianapolis.16697 Ind.

Picture Tube Division

Ind.. PE -

Ind.: PE -

Paul Justus, Circleville. Ohio, PE -41191Ohio

Obituary

Frank J. Bingley, who retired from RCA in1971 after nine years as an engineer withAstro Electronics, died on November 16. Hewas 70. Mr. Bingley was a graduate of theUniversity of London in physics andmathematics. He came to this country 'n1931 to work for the Philco Corp. afterdeveloping some of the earliest scanningsystems for television for the Baird Televi-sion Co. of London.

As chief television engineer for Philco, hedesigned and helped construct one of theworld's first all -electric experimental tele-vision stations.

In 1962. Mr. Bingley joined RCA AstroElectronics where he helped in the develop-ment of a high resolution tv camera tube forthe Earth Resources Satellite and the RCAApollo color television camera used toprovide live telecasts from the moon.

He published numerous technical papers inhis field. and in 1956 was awarded theVladimir K. Zworykin prize of the Institute ofRadio Engineers as outstanding televisionengineer of the year.

Promotions

Missile and Surface Radar

C. Laible from Mbr. Engrg. Staff to Sr. Mbr.Engrg Staff (M Korsen. Pedestals/RF Trans )

R. Wasson from Mbr. Engrg. Staff toSr. MbrEngrg. Staff (M. Korsen. Electrical Integra-tion)

A. Dennison from Princ. Mbr. Engrg. Staff toLdr. Engrg. Sys. Projects (R. Howery, CompProg AN, SPY1A)

H. Millett from Mbr. Engrg. Staff to Sr. Mbr.Engrg. Staff (C. Martin. Install. Documenta-tion)

E. Britt from Sr Mbr. Engrg. Staff to Ldr. Sys.Engrg. (B. Irman, Ship Integration)

R. Creedon from Sr. Mbr. Engrg. Staff to Ldr.Syst Engrg. (B Inman. Ship Integration)

N. Ricciardi from Mbr. Engrg. Staff to Sr.Mbr. Engrg Staff (J Nessmith HybridRadar CAMEL)

C. Tipple from Sr. Mbr. Engrg. Staff to Ldr.Engrg. Sys. Projects (J. Nessmith.Man/Machine)

L. Clayton 'rpm Sr. Mbr. Engrg. Staff toPrinc. Mbr. Engrg. Staff (S. Steele, Comb.Software Systems)

D. Haimowitz 'rpm Mbr. Engrg. Staff toMbr. Engrg. Staff (S. Steele. Comp. Prog.lnteg Simu

R. Rader from Sr. Mbr. Engrg. Staff to Ld'.Engrg. Sys. Projects (S. Steele. Weap. Sys.Soft Devel.)

J. Schwaninger from Princ. Mbr Engrg.Staff to Ldr. Engrg. Sys. Projects (S. Steele.Dist Data Proc. Sys.)

C. White from Princ. Mbr. Engrg. Staff 10Ldr. Engrg. Sys. Projects (S. Steele. Weap.Sys. Soft Devel.)

C. Yeisley from Sr. Mbr. Engrg. Staff to Ldr.

Engrg Sys. Projects (S. Steele. Adv.Software Engrg )

H. Bharmal from Mbr. Engrg. Staff to Sr.Mbr. Engrg. Staff (S. Steele. Weep. Sys. SoftDevel.)

Astro Electronics

S. Rayner from Staff Systems Scientist toMgr.. Specialty Engineering (T.Auk stikalnis)

W. Lindorfer from Staff Systems Scientist toMgr . Specialty Engineering (H. Curtis)

J. Balcewicz from Engineer to Mgr.,Speciality Engineering (H. Curtis)

H. Curtis from Staff Systems Scientist toMgr Engineering (W Manger)

R. DeBastos from Staff Systems Scientist toMgr. Project (A Schnapf )

A. Garman from Mgr. Project to Mgr. Com-munications Satellite Programs (C. Con-stantino)

Consumer Electronics

L.A. Cochran from Sr. Mbr. EngineeringStaff to Mgr.. Signal Processing Engineer-ing (J P Bingham, Engineering Develop-ment)

Solid State DivisionJ. Sundburg from Ldr.. Technical Staff toMgr . Bipolar Assembly and Test, Manufac-turing

J. Arico from Mbr Technical Staff to Ldr .Technical Staff.

Staff announcements

RCA Laboratories

Kerns H. Powers, Director of the Com-municat ,Dns Research Laboratory. has an-nounced the organization as follows: GuyW. Beckley, Head, Image ProcessingResearch; Jon K. Clemens, Head, SignalSystems Research; James J. Gibson, FellowTechnical Staff, H. Nelson Crooks,Manager. High -Density Recording Project.William 0. Houghton, Head, Special Pro-jects Research. Eugene 0. Keizer, Head.Video Systems Research, Charles B.Oakley Head. Broadcast SystemsResearch, Robert E. Flory, Fellow.Technical Staff. J. Guy Woodward, Fellow.Technical Staff, and Daniel A. Walters,Head, Satellite Systems Research.

Paul Rappaport, Director of the Process andApplied Materials Research Laboratory, hasannounced the organization as follows:Glenn W. Cullen, Head. Materials Synthesis

Research; Chih C. Wang, Fellow, TechnicalStaff; Leonard P. Fox, Head, "SelectaVision"Processing Research; Bernard Hershenov,Staff Advisor; Richard E. Honig, Head,Materials Characterization Research; D.Alex Ross, Manager, Division Liaison;Daniel L. Ross, Head, Organic Materials andDevices Research; George L. Schnable,Head, Process Research; Charles W.Mueller, Fellow, Technical Staff; Brown F.Williams, Head, Optical Materials andDevices Research; and Richard Williams,Fellow, Technical Staff.

CommercialCommunicationsSystems Division

Andrew F. Inglis, Division Vice Presidentand General Manager of the CommercialCommunications Systems Division, has an-nounced the organization as follows:William L. Firestone, Division Vice Presidentand General Manager, Avionics Systems;Joseph P. Ulasewicz, Division Vice Presi-dent and General Manager, Mobile Com-municatons Systems; Jack F. Underwood,Division Vice President, Plans andPrograms; and Neil Vander Dussen, DivisionVice President and General Manager,Broadcast Systems.

Distributor and SpecialProducts Division

Paul B. Garver, Division Vice President andGeneral Manager of Distributor and SpecialProducts Division, has announced thefollowing appointments: Gene W.Duckworth, Division Vice President,Business Development and InternationalOperations; and Fred G. Wenger, DivisonVice President, Sales.

Solid State Division

Thomas T. Lewis, Director of Electro-OpticsOperations, has announced the organiza-tion as follows: Thaddeus J. Grabowski,Manager, Market Planning-Displays &Emitters; Clarence H. Groah, Manager,Operations Control; Leonard W. Grove,Manager, Manufacturing-Electro-optics;N. Richard Hangen, Manager, MarketDevelopment-Research & Development;Fred A. Helvy, Manager, ApplicationsEngineering-Electro-Optics; Edward F.McDonough, Manager, Market Planning-Photodetectors; Ronald G. Powers,Manager, Solid State Detectors-Canada;Carl L. Rintz, Manager, Market Planning-Imaging Devices; and Eugene D. Savoye,Manager, Engineering-Electro-Optics.

Harold R. Krall, Manager of Electro-OpticsEquipment Operations, has announced thatthe responsibility for the manufacturing,

marketing, and engineering of ChargeCoupled Device (CCD) Cameras will beassumed by Electro-Optics EquipmentOperations. He announced the organizationof Electro-Optics Equipment Operations asfollows: Carl F. Adams, Jr., Administrator,Operations Control; David E. Bowser,Leader, Engineering-Subassemblies &Equipment; Joseph T. Gote, Marager,Manufacturing-Electro-Optics Equipment;Victor C. Houk, Manager, MarketPlanning-Closed-Circuit Video Equip-ment; and Robert L. Rodgers, Manager,Engineering-Closed-Circuit Video Equip-ment.

Leonard W. Grove, Manager ofManufacturing-Electro-Optics, has ap-pointed William N. Henry as Manager,Manufacturing-Silicon Devices (CCD andSi targets).

Carl R. Turner, Division Vice President ofSolid State Power Devices, has announcedthe organization as follows: Dale M.Baugher, Director, Power Engineering;Melvin Bondy, Manager, Operations Plan-ning & Administration-Power; Ralph S.Hartz, Manager, Mountaintop Engineering;John E. Mainzer, Director, Power Manufac-turing Operations; and Donald Watson,Director, Product Marketing-Power.

Ralph S. Hartz, Manager of MountaintopEngineering, has announced the organiza-tion as follows: Donald E. Burke, Manager,Device Development Engineering; Robert J.

Equipment Engineering;and Louis V. Zampetti, Manager, TypeEngineering.

Dale M. Baugher, Director of PowerEngineering, has announced the organiza-tion as follows: Jacques M. Assour, Leader,Process Engineering; Dale M. Baugher,Acting Administrator, New Product Plan-ning; Leonard H. Gibbons, Manager,Applications Engineering-Power; andWallace D. Williams, Leader, Test Engineer-ing.

Norman C. Turner, Manager of SomervilleMOS Operations, has announced theorganization of Photomask Operations asfollows: David S. Jacobson, Manager,Photomask Engineering & Tooling; andEvan P. Zlock, Manager, PhotomaskProduction.

Picture Tube Division

D. Joseph Donahue, Division Vice Presidentof Engineering, has appointed James C.Miller, Manager, Technical Projects andEngineering Administration.

Matthew M. Bell, Manager of EquipmentEngineering, Circleville Glass Operations,has announced the following appointments:F.L. Armstrong, Manager Equipment DesignEngineering-Forming; B.B. LeMay,Manager, Equipment Design Engineering-Finishing; and R.W. Marshall, Manager, In-strumentation Engineering.

Tannenbaum is TPAfor GovernmentCommunications Systems

Dan Tannenbaum has been appointedTechnical Publications Administrator forGovernment Communications Systems. Inthis capacity, Mr. Tannenbaum is responsi-ble for reviewing and approving technicalpapers; for coordinating the technicalreporting program; and for promoting thepreparation of papers for the RCA Engineerand other internal and external journals.

Mr. Tannenbaum is currently Staff Engineerfor the Government CommunicationsSystems Division in Camden; his more re-cent responsibilities as Manager of DigitalEquipment Engineering have been in thedesign and development of digital equip-ment, tactical transmission processingequipment, and telephone switchingterminals. Mr. Tannenbaum has been withRCA for 27 years after receiving the BS inEngineering Physics from the University ofMichigan in 1949. He is a licensedprofessional engineer in New Jersey and isalso a senior member of the IEEE.

Mobile CommunicationsSystems

George J. Mitchell, Manager of MobileProduct Operations, has appointed Lee F.Crowley, Manager, CommunicationsProducts Engineering and Product Manage-ment.

Service Company

Raymond J. Sokolowski, Division Vice Prisi-dent, Consumer Services, has appointed J.William McGee Director, Technical Sup-port.

112

Editorial RepresentativesThe Editorial Representative in your group is the one you should contact inscheduling technical papers and announcements of you professional activities.

Commercial Communications Systems Division

Broadcast Systems

Mobile Communications Systems

W.S. SEPICI- Broadcast Systems Engineering, Camden, N.J.K. PRABA Broadcast Systems Antenna Equip. Eng., Gibbsboro, N.J.A.C. BILLIE Broadcast Engineering, Meadow Lands, Fa.

F.A. BARTON Advanced Development, Meadow Lands, Pa.

Avionics Systems C.S. METCHETTE Engineering, Van Nuys. Calif.J.McDONOUGH Equipment Engineering, Van Nuys, Calif.

Government Systems Division

Astro-Electronics

Automated Systems

Government Communications Systems

Government Engineering

Missile and Surface Radar

Research and Engineering

Corporate Engineering

Laboratories

Solid State Division

Consumer Electronics

SelectaVision Project

RCA Service Company

Distributor andSpecial Products Division

Picture Tube Division

RCA Communications

AlascomAmericomGlobcom

RCA Records

NBC

RCA Ltd

Patent Operations

Electronic Industrial Engineering

E.A. GOLDBERr Engineering, Princeton, N.J.

K.E. PALM Engineering, Burlington, Mass.A.J. SKAVICJS Engineering, Burlington, Mass.L.B. SMITH Engineering, Burlington, Mass.

D.A. TANNENBAUM Engineering, Camden, N.J.H.R. KETCHAM Engineering, Camden, N.J.

M.G. PIETZ Advanced Technology Laboratories, Camden, N.J.

D.R. HIGGS Engineering, Moorestown, N J.

H.K. JENNY Technical Information Programs, Cherry Hill, N.J.

C.W. SALL Research, Princeton, N.J.

J.E. SCHOEN Engineering Publications, Somerville, N.J.H.R. RONAN Power Devices, Mountaintop, Pa.S. SILVERSTEIN Power Transistors, Somerville, N.J.A.J. BIANCULLI Integrated Circuits and Special Devices, Somerville, N.J.J.D. YOUNG IC Manufacturing, Findlay, OhioR.W. ENGSTROM Electro-Optics and Devices, Lancaster, Pa.

C.W. HOYT Engineering, Indianapolis, Ind.R.J. BUTH Engineering, Indianapolis, Ind.P.E. CROOKSHANKS Television Engineering, Indianapolis, Ind.C.P. HILL Manufacturing Technology, Indianapolis, Ind.

F.R. HOLT SelectaVision VideoDisc Engineering, Indianapolis, Ind.

J.E. STEOGER Consumer Services Engineering, Cherry Hill, N.J.R. MacWILLIAMS Marketing Services, Government Services Division Cherry Hill, N.J.R.M. DOMBROSKY Technical Support, Cherry Hill, N.J.

C.C. REARICK Product Development Encineering, Deptford, N J

E.K. MADENFORD Engineering, Lancaster, Pa.N. MEENA Glass Operations, Circleville, OhioJ.I. NUBANI Television Picture Tube Operations, Scranton. Pa.C.W. BELL Engineering, Marion, Ind.

P.WEST RCA Alaska Communications, Inc., Anchoraze, AlaskaD.L. LUNDGREN RCA American Communications, Kingsbridge Campus, N.J.W.S. LEIS RCA Global Communications, Inc., New Yclic, N.Y.

J.F. WELLS Records Eng Indianapolis, Ind.

W.A. HOWARD Staff Eng., Technical Development, New York, N Y

H.A. LINKE Research & Eng.. Montreal, Canada

J.S. TRIPOL! Patent Plans and Services, Princeton, NUJ.

J. OVNICK Engineering, N. Hollywood, Calif.

'Technical Publications Administrators (asterisked above) areresponsible for review and approval of papers and presentations.

ROIl EngineerA TECHNICAL JOURNAL PUBLISHED BY CORPORATE TECHNICAL COMMUNICATIONS"BY AND FOR THE RCA ENGINEER"

Printed in U S AForm No RE -22-4