RCA Engineer - WorldRadioHistory.Com · 5/6/1984 · RCA authors describe today's realities-the ongoing and successful efforts to automate the engineering workplace.-MRS Illustrations

RCA EngineerVol. 29 No. 3 May/June 1984

o.

Cover illustration by Mel Allen

.\\-2 O'

"1 % >

v.,\ N. ''. PO >

0 #:

Putting computer power at the engineers' fingertips.

Our cover depicts the possible shape of futuresystems that could allow engineers at differentsites to have system -wide access to tools on otherstations and on a superminicomputer. In this issue,RCA authors describe today's realities-theongoing and successful efforts to automate theengineering workplace.

-MRS

Illustrations on pages 7 and 12 by Mel Allen;page 13 by Diane Farkas; page 20 by Helen Mary;page 28 by Denise Miller; and page 61 JosephMcGarrity.

alma EngineerA technical journal published byRCA Technical Excellence Center13 Roszel RoadP.O. Box 432Princeton, NJ 08540TACNET: 226-3090 (609-734-3090)

Tom King

Mike Sweeny

Louise Carr

Frank Strobl

Betty Gutchigian

Phyllis Grimm

Jay Brandinger

John Christopher

Jim Feller

Mahlon Fisher

Tony Bianculli

Arch Luther

Howie Rosenthal

Joe Steoger

Bill Underwood

Bill Webster

Ed Burke

Walt Dennen

John Phillips

RCA Engineer Staff

Editor

Associate Editor

Art Editor

Contributing Editor

Composition

Secretary

Editorial Advisory Board

Staff Vice -President, Systems Engineering,RCA Laboratories

Vice -President, Technical Operations,RCA Americom

Division Vice -President, Engineering,Government Systems Division

Division Vice -President, Engineering,Video Component & Display Division

Manager, Engineering Information,Technical Excellence Center

Senior Staff Scientist, RCA Laboratories

Staff Vice -President, Engineering

Division Vice -President, Engineering,RCA Service Company

Director, TechnicalExcellence Center

Vice -President, Laboratories

Consulting Editors

Administrator, MarketingInformation and Communications,Government Systems Division

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

Manager, Business Developmentand Planning, RCA Service Company

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

C.A. Quinn

Bringing computer powerto the engineerThe strength and vitality of our engineering community sets thepace for RCA's future. Increasing pressures for shorter designcycles, better quality designs, and faster response to changingmarket needs, all focus on the effectiveness of the engineeringstaff. The growing complexity of the engineering task reachesbeyond the design cycle into vital areas of the manufacturingprocess. This demands better coordination and control.

Sheer numbers of engineers will not suffice in the solutions tothese types of problems. We must ensure that each engineer isoperating at maximum potential-increasingly, this meansengineers must have direct access to computer tools that improveindividual productivity. Personal computers, graphic work stations,electronic communications, retail software, and so on, offer newoptions that complement and extend the conventional mainframesolution. At the same time, the multitude of choices can lead tochaos if a broad plan is lacking.

RCA management is addressing these issues by investing heavilyin programs that bring computer power directly to the engineers.The productivity and performance gains that will ensure our futureare dependent on meeting this challenge.

Many exciting steps are underway across the Corporation. Thisissue of the RCA Engineer reports on some of the leadershipefforts.

Charles A. QuinnDivision Vice -President and General Manager,Video Component and Display Division

Iliffia EngineerVol. 291No. 3 May June1984

integrating 4 VCAD, an engineering productivity strategyindividual workstations J.C. Miller

10 Personal computers, an engineering productivity toolJ.C. Miller1W.T. Kelley

15 CADAM, an engineering productivity toolJ.C. Miller1P.J. Kunz

20 Emerging engineering design toolsE.M. Melendez

the buying 28 Choosing a CAD systemdecision C.A. Burton

35 The MSR local area networkW. Carey

services 39 An attached processor provides calculation speedand accuracyS.G. Handman

43 Adapting CMS for a varied user communityD. Zarodnansky

46 Analyzing design alternatives with the PRICE modelsA. DeMarco

general interest 54 CMOS technology moves towards IC leadershipH.G. Patterson 1J.L. Magos

61 A day in the life of an information -efficient scientistP. Schnitzler

departments Pen and Podium, 631 News and Highlights, 661 Obituary, 72Copyright © 1984 RCA CorporationAll rights reserved

in this issue ...automating the engineer's workplace

Miller: "The VCAD design philosophy has been to provide a locally -based distributed system in a standard environment with a logical andcomprehensive communication framework."

Miller/Kelley: "The PC was not viewed as some sort of super -designworkstation for the elite engineer."

Miller/Kunz: "We envisioned a need for a large number of terminals,fast response time, a tight focus on productivity, a common databaseacross all users, and a desire to access a wide set of software tools."

Melendez: "Both these subsystems will be linked together onto anetwork and will share an adaptive design database management sys-tem that will behave as the central data -monitoring and control facility."

Burton: "Outlined below are some of the major areas we investigatedwith each CAD company under consideration."

Carey: "Our hardest task was to decide which of a number of tech-nically excellent replies (to a request for proposal) was the best."

l'...NETWORKCONTROLNODE 100

Handman: "At RCA, the engineering community can take advantageof pipelined calculations in such diverse applications as circuit analysisand structural analysis."

Zarodnansky: "Experience has shown that if users can get imme-diate on -site assistance with a problem, they are more likely to continueto make use of the products provided on the system, and to ask forassistance when they need it."

DeMarco: "The potential use of the PRICE system throughout the lifeof a program was illustrated by the Programmable Signal Processorstudy."

39 ATTACHED

43 HELP!

Patterson/Magos: "CMOS use is growing in linear circuits such asoperational amplifiers and analog -to -digital converters, and it is fastbecoming a key technology for the emerging telecommunications anddata communications markets."

NMOS

82

HMOS

84YEAR

86 8 90

54

CMOS - 49

GaAs

in future issues...technical excellence,materials science applications,imaging technology,communications.

J.C. Miller

VCAD, an engineering productivity strategy

VCAD is an aggressive computer -based engineeringproductivity strategy. Video-Computer-Aided Design tools arealready dramatically changing the work style of engineers,technicians, secretaries, and managers at RCA VideoComponent and Display Division.

The Video Component and Display Di-vision (VCD) had made modest use ofcomputer technology in previous years; crit-ical design needs had been programmed,but most of the ancillary tasks were frag-mented across departmentally based sys-tems. Limited use was made of the Corpo-rate Computing Center due to its cost,features, and remoteness. Lacking a coher-ent plan, the Division had acquired a num-ber of small computers, each different andincompatible. There were few economiesof scale. Many tasks suitable for a comput-er remained in a manual mode. The dilem-ma was how to change to a modern sys-

Abstract: VCAD is an aggressivecomputer -based engineering productivitystrategy; the initials stand for the Division'sname as well as "Video-Computer-AidedDesign." This paper describes the VCADsystem from concept throughimplementation and covers the purpose,architecture, major subsystems, training,and early results. VCAD has been asuccessful program. Even at this earlystage the best testimonial is the strongdesire from the engineering managers formore VCAD tools. Although VCAD istechnically a computer system, from amanagerial perspective it is a productivitymultiplier. Companion papers in this issuecontain expanded treatment of two VCADsubsystems the CADAM graphics designsystem and the use of personal computers.

©1984 RCA CorporationFinal manuscript received March 22, 1984.Reprint RE -29-3-1

tern without destroying the existing base ofapplications.

Video-Computer-Aided Design(VCAD) began in early 1983 as a multi-year proposal to increase the productivityof the Engineering Department through acarefully designed and well -integrated com-puter system. By the end of 1983, Phase Iof the VCAD program was complete; abroadly conceived system had been installedand was in productive use.

ApproachThe VCAD effort began with a recogni-tion of the primary needs of the Engineer-ing Department. We sought increased pro-ductivity and concluded that a massiveinfusion of computer technology was need-ed, but we lacked the resources to writelarge amounts of code or to tackle majortechnological risks in the computer ap-proach. We recognized that our commu-nity required assistance in administrative(non -technical) computing, electronic shar-ing of information, coupling of discretesystems, increased local computer "horse-power," and graphical design aids. Signifi-cantly, VCD senior management was will-ing to back the VCAD project based uponintuitive productivity arguments as opposedto conventional cost/benefit analyses.

The VCAD design philosophy has beento provide a locally -based distributed sys-tem in a standard environment with a log-ical and comprehensive communicationframework (see sidebar, facing page). Fortu-nately, the evolution of the computer art

had resulted in technology that was notavailable in earlier years. An approachwas defined based upon several elements.

Use of the IBM 4341 computer withVM/CMS as the core of the system.This is a low- to mid -range machinewith very reliable hardware and a stableoperating system. The host can be upgrad-ed as the system load demands. TheIBM 4341 runs in an unattended mode24 hours a day, 7 days a week. All non -IBM computers and terminals link to thehost via the communication nework.

Construction of a local area network(LAN) to serve as the communicationbus joining the host and the many ter-minals and computers already in place.A LAN is a multi -port communicationswitch that allows one to logically andsimultaneously connect pairs of terminals;it uses translation tables and a networkcontroller to link devices with differentprotocols, transmission speeds, parity con-ventions, and so on. Within rather widelimits any two devices can be coupled.The LAN provides the technical meansto accommodate the variety of existingterminals and eliminates the need for anadditional investment.

Recognition of the personal computer asthe key to putting computer power directlyon the desk of the professional. The IBMpersonal computer (PC and XT) waschosen as our standard. Each PC wasattached to the LAN and can operate ineither a stand-alone or 3278 -terminalmode. The PC provides adequate dis-

4 RCA Engineer 29-3 May/June 1984

Video-Computer-Aided Design overview

The goal of VCAD is increased productivity. TheVCAD approach is to saturate the entire engineeringdepartment with common hardware and softwareunder a comprehensive integrated system design.

Central host computer. A common hub to the overallsystem, this is currently an IBM 4341 runningVM/CMS. It services CADAM, PROFS electronicmail, VSFORTRAN/VSPASCAL programming lan-guages, interactive and batch work. It also holdscommon databases and core software, and is easyto expand as requirements dictate. The IBM hostsupports various IBM, and non -IBM terminals via a3274 controller and local area network. The comput-er is highly stable and flexible; it runs unattended 24hours a day, non-stop.

Local area network. The LAN is an "electronic high-way" that links various terminals and computers in astandard framework. As a solution to our problem ofequipment from many vendors, the LAN handlesdevices with different protocols, baud rates, parityconventions, and so on. It supports the PC as a9600 -baud full -screen device, and can transfer filesbetween dissimilar terminals or computers. The LAN

is linked to the host via a Hydra protocol converter.

Personal computers. These off-load work to a local-ly -controlled environment. Powerful PC software letsusers solve their own problems where possible. PCssupport word processing, spreadsheets, planning,databases, analysis, and so on. They can act as327X -class terminals to the VCAD host whendesired. Personal computers are used by all levels ofthe organization as the basic computer terminal.

CADAM graphics design. CADAM offers a 0.3 -second response time to user requests, supports alarge number of users sharing a divisional database,and will eventually grow to 50 -plus CADAM termi-nals. CADAM's friendly design environment is basedupon well-known drafting methodology. Using thissystem, it's easy to train new users and get rapidproductivity gains. It generates high -quality outputdrawings on the Versatec plotter in 30 to 60seconds. CADAM comes in modules that extend thesystem to many application areas. The system drivesCAEDS and ANSYS to generate 3D designs andfinite -element analysis, and couples to a numericalcontrolled machining module.

tributed computing power on a local(and immediate) basis while retainingthe ability to link to the IBM 4341.Each PC has a common configurationand software complement (see compan-ion paper, by Miller and Kelley, fordetails).

Selection of a turnkey graphics designpackage CADAM (CADAM Inc.) asthe vehicle for converting the core of themanual design process into an integratedsystem. Centralizing all of the graphicsinformation into a common database elim-inates many duplicate manual steps andenhances productivity. A scaled -downvariant of CADAM, IBM's FAST -DRAFT, was selected for our remoteplants in Marion and Scranton (see com-panion paper, by Miller and Kunz, fordetails).

ConfigurationThe initial phase of VCAD hardware (Fig.1) consisted of a fully implemented LANcovering the entire Engineering Department,an IBM 4341 with four megabytes of mem-ory, three 800 -megabyte disks, a tape drive,a 3274 controller for the 3270 -class ter-minals (provision for 32), two 16 -portHydra (JDS Microprocessing) protocol con-verters, three Displaywriters and seventy -

LOCAL HOST

01

3274

CMX

CMX

CMX

-T

-T

T

1

PC

1 2 1 3 1

PC

415

PERIPHERALS

HYDRA

L A \ +- PC

REMOTE CORPORATECOMPUTER

3258

3255

1

ti.1771

3251WORKSTATION

Fig. 1. VCAD architecture. An overview of the communication arrangement, show-ing 3274 controller path, the Hydra link, the local area network, and the CADAMterminal arrangement. The local host serves as the gateway for communications.The LAN connects various terminals, PCs, and non -IBM computers as required.

Miller: VCAD, an engineering productivity strategy 5

three IBM PC/XTs-all linked to the LAN,six 3251 graphic design terminals and asso-ciated controllers, five printers, and a 36 -inch Versatec electrostatic plotter. Thereare now 168 devices (computers, PCs, ter-minals) attached to the LAN. The 3251sare 19 -inch video graphic design terminalsused exclusively with the CADAM designsoftware. We chose the NET/1 LAN hard-ware from Ungermann-Bass and linked itto the IBM host via the Hydra converter.IBM full -screen terminals (3178/3279)were supported by a 3274 driving Unger-mann-Bass's CMX multiplexors.

The communication arrangement is non -conventional. From the outset we werefaced with the need to support the existingterminals (from many vendors), and thenew IBM terminals. The LAN provided avehicle for linking the existing terminalsbut the question remained, "How couldwe best attach the LAN to the host?" Theusual solution would have been an IBM3705 front-end processor. The Hydra of-fered most of the necessary features at alower cost, and with a somewhat simplersystem interface. The Hydra is a protocolconverter that presents asynchronous screen -oriented devices to the host as 3278 ter-minals with cursor control. As a side bene-fit, the Hydra -supplied PC -3278 terminalemulator (ACOM, Computer Vectors) wasfound to be superior to the IBM 3101emulator. We have been very pleased withthe full -screen terminal mode of operationfor the PC. A suitable emulator was alsofound for the (existing) Hewlett-Packardterminals, which enables them to communi-cate with the host via the LAN.

It was desirable to avoid the star archi-tecture associated with the conventional3274 cabling for the IBM -class terminals.The CMX multiplexor operates in a daisy -chain mode that makes it possible to greatlyreduce the total cable length of the instal-lation. The communication arrangementoffers considerable flexibility in installingnew devices. On the other hand, the Hydradecision splits the functional responsibilitybetween two vendors in a critical area.

Both Hydra and IBM have been respon-sive in solving communication problemsand offering support. On balance, we havebeen pleased with the design choice.

Most of the engineering managers use apersonal computer as their terminal, andthe majority of the engineers have reason-able access to some kind of terminal. Ourplan is to have an appropriate terminalavailable for every user.

Software

It was our goal from the outset to staywith "plain vanilla" software as providedby the vendor and resist the temptation tomodify it. This made it possible to mini-mize the support effort required and im-prove the overall system stability. All newapplication software was purchased andapplied to the system without change. Manyof the programs developed in the pastalready ran on the same operating systemas the corporate computer at Cherry Hill,N.J., thus eliminating the need for a con-version-but this did require the installa-tion of twelve VM/CMS system patchesfrom Cherry Hill to retain compatibility.Overall, the "plain vanilla" and "buy, don'twrite" combination were good decisions.

For example, the host software com-plement includes VM/CMS, VSFOR-TRAN, VSPASCAL, PROFS, CADAM,CAEDS (SDRC Inc.), and ANSYS (Swan-son Analysis). By retaining total control ofthe project, it has been possible to fine-tune the operating system to achieve thedesired response times and performancefor the various user groups-this wouldhave been difficult to do under shared cir-cumstances. All engineering applicationspreviously run at Cherry Hill (except for avery large design program) have beenbrought back to the VCAD system. ThePCs run Lotus123, MultiMate (SoftWordSystems), and other purchased software.The PC 3278 emulator (ACOM) is a full -screen package obtained from ComputerVectors. It also supports file transfer be-tween host -PC and PC -PC.

The Lockheed -developed CADAM sys-tem was acquired as the basis for all graph-ic engineering design. This is a provenmodular product capable of running a largenumber of graphic design terminals on thesame system, all sharing a common data-base. CADAM offers extensive design fea-tures coupled with subsecond response to

operator inputs. The fast response time isthe key to productivity. CADAM is viewedas the most important application on thehost. CAEDS and ANSYS support finite -element modelling.

IBM's PROFS was adopted as the coreelectronic mail system, but is not the pri-mary tool for word processing. This deci-sion was approached with mixed emotionsdue to the rather negative observationsformed in the RCA Laboratories' experi-ment. The reasons for the relatively suc-cessful installation to date include: VCAD'sbroad base of installed PROFS users (youhave to span the people you normallycommunicate with), cautious start-up, care-ful training, follow-up support, but primar-ily, our nondependence upon the PROFSlimited word processing features.

Word processing is supported by a mixbetween standard PROFS features and Mul-tiMate on the PC. MultiMate relieves thepressure on DCF document preparation,the weakest part of PROFS. Except forthree Displaywriters, all Engineering secre-taries use MultiMate on their IBM per-sonalACOM emulator makes it possible to trans-fer files between MultiMate and PROFSin a straightforward fashion, making it auser decision as to where each task is

done.The PC not only supports useful appli-

cations locally, it also off -loads substantialwork that would otherwise have "cluttered"the host. In our environment, this reducesthe interaction between CADAM and otherhost -based programs while offering bettersupport to the user. The low cost andwide variety of PC software make it easierto pursue new applications.

Training

The VCAD program had high manage-ment visibility within the Division. Exten-sive training was provided for each facetof VCAD. Locally available talent wasmobilized to present a wide variety oftraining courses, lectures, demonstrations,visits, and so on. Much of the sucrfss ofVCAD may be attributed to the carefulplanning of the training effort.

A simple but effective "PC College"was established to indoctrinate users intothe use of their new tool and the standardsoftware. Participation in the PC Collegewas a prerequisite to receiving the comput-


er. Extensive and highly personalizedCADAM/CAEDS/ANSYS training pro-grams were developed ranging from intro-ductory level material to advanced topics.Training was provided for each softwarepackage and for all levels and skills. Train-ing is a demanding, expensive, but vitaltask. Failure to train leads to frustration,inefficiency, and ultimately to lack ofresults.

The VCAD training program used atop -down philosophy wherever possible.This simple management concept has prov-en itself to be a powerful one and hasremoved much of the mystery from thecomputer. Mr. C.A. Quinn, VCD Vice -President and General Manager, was thefirst CADAM student, one week after deliv-ery of the VCAD system; the DivisionController and MIS manager also receivedearly training.

Before an individual user could betrained to use a CADAM terminal, everymanager in the chain of command up tothe Divisional General Manager was re-quired to have CADAM training. Thistechnique required management commit-ment and paid substantial dividends in thefollowing months. The managers' intimateknowledge of CADAM has greatly facili-

tated their constructive use of the system.Similarly, in the PC area, the initial termi-nals were preferentially assigned to man-agers who took training as a condition forreceiving a PC. Their hands-on understand-ing made it easier to extend the process totheir subordinates. Reminiscent of the fa-mous Western Electric Hawthorne study,this active management involvement con-tributed to the success of the program.

Measurement

Since the major purpose of the VCADinvestment was to enhance engineering pro-ductivity, significant effort was devoted tomeasurement. Engineering productivity isa particularly elusive quarry. Prior to thedelivery of the system, detailed measure-ments were made of the number of hoursrequired to develop each type of engineer-ing drawing in the various departments.This data was used as a benchmark tomeasure the progress of the CADAM sys-tem. We have used results from GeneralMotors (who has reported a 3.81 produc-tivity improvement with CADAM) as ournominal goal. After only one month ofCADAM experience, all operators wereexceeding their manual rates and were

pleased with the CADAM system. Monthlyand daily data is being gathered and pub-lished to track our progress in detail. ACADAM user group was quickly estab-lished to reinforce good practices and pro-vide a feedback forum.

The impact of personal computers onproductivity had been sampled by a sur-vey of the users shortly after delivery ofthe machines and will be repeated after asix-month interval. PC user groups havealso been established, sponsored by theMIS Information Center.

Implementation

A conceptual plan for the VCAD projectwas submitted in April 1983, approvalswere obtained by midyear, and the equip-ment was installed and made fully opera-tional by December 7, 1983. All projectdates and goals were met. Corporate CISSStaff was helpful at all points and cooperat-ed in allowing a flexible approach to themyriad of hardware/software changes thatoccurred during execution of the plan.

The plan was implemented by two task -force teams: one interfacing with the vendor(IBM) and another (VCAD) with the usercommunity. Each task force met on abiweekly basis for two hours and repre-sented a collateral duty for its membersrather than a full-time assignment. TheVCAD task force consisted of a represen-tative from each user department and wasparticularly useful in hammering out thenecessary system -design compromises andcommunicating with the user community.The smaller IBM task force met with thevendor to deal with the technical details ofthe project. Excellent support was providedby the vendor. Two full-time professionalswere added to the engineering staff tosupport VCAD.

Expansion

At the outset of the VCAD program aPhase II expansion was envisioned for1984; in fact, this process is well under-way and will more than double the systemresources by midyear. Phase II will resultin additional PCs, 3178 and 3179 termi-nals, and triple the number of CADAMstations (using the new 5080 color termi-nals). Moreover, this phase will lead to anupgraded host, doubled disk storage, anA -size Versatec plotter, and new software


modules (CADVUE, TRANSGRAPH,PRANCE/CBDS, SQL, and so on).

The expansion will broaden theCADAM applications beyond the initialmechanical area into electrical designs (sche-matics, printed circuit board routing andplacement, simulations). PROFS usage willgo beyond routine electronic mail into mech-anizing a number of paper -based subsys-tems. We intend to merge EngineeringStandards into the CADAM database. Anew task force has been formed to beginthe process of extending CADAM into thenumerical controlled machine area. As withPhase I, the expansion will be coupledwith an active training program. The proj-ect is being planned using the HarvardProject Manager (Harvard Software Inc.)on an IBM XT. MIS has adopted theVCAD architecture as the basis for thenew divisional efforts.

AcknowledgmentsMy sincere thanks to M.B. Fisher andC.A. Quinn who authorized VCAD onfaith; J. Balling (CISS) who ran interfer-ence; T. Hart (MIS) who became an enthu-siast; P. Kunz (CADAM), W.T. Kelley(PCs), B. Mangolds (Communication), andK. Walker (System) for their key contri-butions; and to the remaining members of

James C. Miller is the Manager of Techni-cal Projects and Engineering Administra-tion at VCD in Lancaster, Pa. He earned aBSEE from Rensselaer Polytechnic Insti-tute and an MEng and PhD from Yale Uni-versity. He received the Honeywell Awardfrom Yale, three RCA LaboratoriesAchievement Awards, and shared theDavid Sarnoff Award for OutstandingTechnical Achievement. Dr. Miller holds 23U.S. patents and has published exten-sively. His current interest is in the applica-tion of computer technology to improvingorganizational productivity.Contact him at:Video Component and Display DivisionLancaster, Pa.TACNET: 227-6490

the VCAD Task Force-M. Renfro, H.Hillegass, R. Miller, R. Marsland, C.Lausman, and G. Gadbois-who repre-sented the user community in a profes-sional manner and whose cooperation wasessential.

Resource guide:VCAD system hardwareand softwareThis set of articles on the VCDEngineering organization'sprogress toward a network ofautomated workstationsincludes numerous referencesto equipment and software. Tohelp you sort out the options,we have culled this brieflyannotated listing completewith company addresses fromour own site -specific expe-riences. If you have any ques-tions, why not call theauthors? Or write themanufacturers.

The asterisked items (*)denote PC hardware andsoftware that are considered"standard" for the VCD Engi-neering organization; individ-ual machines vary in soft-ware/hardware configuration.The double asterisk (**) indi-cates "under study." We donot necessarily purchase fromthe source indicated. All of theitems have been found to beuseful in VCD applications, butthis does not constitute ablanket endorsement. All soft-ware mentioned will work onour standard hardwareconfiguration.


Personal Computer HardwareComputer: IBM PC with Dual Floppy* or XT Sys-tem Unit*, with a memory of 256 Kb (minimum)*.

Adapter boards: Monochrome/Printer Adap-ter* [IBM]; Asynchronous* [IBM]; Color Gra-phics [IBM]; Printer Adapter [IBM].Alternates: STB Graphix Plus**. Monochromegraphics. [STB Systems Inc., 601 N. GlenvilleAvenue, Suite 125, Richardson, TX 75081].Hercules. Supports monochrome graphics.[Hercules Computer Technology, 160 BeechnutDrive, Hercules, CA 94547].

Display/monitor: Monochrome* [IBM]; Color[IBM].

Keyboard: IBM* [IBM].Alternate: Keytronic KB5151**. Separatenumeric and cursor -key clusters. [Key Tronic,P.O. Box 14687, Spokane, WA 99214].

Printer: Okidata ML93 with IBM Plug'n Play Kit*.Our standard PC printer [Okidata Corp., 111Gaither Drive, Mount Laurel, NJ 08054].Alternate: NEC 3550 with Dual Bin sheet feeder.Letter -quality printer for secretaries. [NEC HomeElectronics (U.S.A.), Inc., Personal Computer Di-vision, 1401 Estes Avenue, Elk Grove Village, IL60007].

Personal Computer Software

Operating system: DOS 2.0* [IBM] or DOS 2.1[IBM]

Database/file managers:

Pfs File/Report. Easy for the novice to use.Suitable for simple filing, reporting, and retrievalapplications. [Software Publishing Corp., 1901Landings Drive, Mountain View, CA 94043].

TIM IV. A relational base with good report -writ-ing features, and allows linked files. Rigid inputprotocol and limited number of fields/record.[Innovative Software, Inc., 9300 West 110thStreet, Suite 380, Overland Park, KS 66210].

Data Base Manager II-The Integrator. Conve-nient means to move information between dif-ferent application packages. Simple database,easy to use. [Alpha Software Corp., 30 B Street,Burlington, MA 01803].

R:Base4000. Powerful relational database sys-tem suitable for more complex applications.Good features. Not for the beginner. [Microrim,1750 112th Avenue, N.E., Bellevue, WA 98004].

ThinkTank-The Idea Processor. Unique sys-tem for handling outlines, text, concepts, "todo"lists, and so on. Billed as an idea organizer.

Very handy for managers. [Living Video Text,Inc., 1000 Elwell Court, Suite 232, Palo Alto, CA94303].

Plotting: Graphwriter**. A powerful, yet easy touse, package for presentation-quality technicalgraphs. [Graphic Communications, Inc., 200 FifthAvenue, Waltham, MA 02254].

Statistics: Statgraphics. A sophisticated and pow-erful statistical package; many features. [Statgra-phics Inc., P.O. Box 1558, Princeton, NJ 08540].

Project Management: Harvard Project Manager.An excellent CPM planning tool for the PC. Workswith monochrome graphics. Practical features.Easy to use. [Harvard Software Inc., 512 GreatRoad, Littleton, MA 01460].

Spreadsheet: Lotus123*. Justly called "king of themountain." Excellent spreadsheet, simple but use-ful database, fine graphics, very limited word pro-cessing. Outstanding features. [Lotus Develop-ment Corp., 161 First Street, Cambridge, MA02142).

Word Processing: MultiMate*. An excellent"Wang -clone" product with features usually foundonly in dedicated word processors. Well -receivedby secretaries and engineers. [MultiMate Interna-tional Inc., 52 Oakland Avenue North, East Hart-ford, CT 06108]. Volkswriter and VolkswriterDeluxe. Not used by VCD. Simple to learn anduse. Just what the name implies. [Lifetree Soft-ware, Inc., 411 Pacific Street, Suite 315, Monterey,CA 93940].

Communications

Local Area Network (LAN): Net/1. Links terminals,and so on, to common electronic data path. Cou-ples to host via Hydra. [Ungermann-Bass, 2560Mission College Boulevard, Santa Clara, CA95050].

Hydra: Channel -attached protocol converter.Permits PC to look like 327X -class device to host.Supports other devices as well. [JDS Micro -Pro-cessing, 25 Mitchell Boulevard, Suite 7, SanRafael, CA 94903].

Terminal Emulators: ACOM*. Used with Hydra.Full -screen 327X emulator with file -transfer ca-pability. [Computer Vectors, Inc., 24871 PylosWay, Mission Viejo, CA 92691].

CMX: Couples to 3274 terminal controller andallows daisy -chain cabling to various IBM de-vices. This saves cable length. [Ungermann-Bass,2560 Mission College Boulevard, Santa Clara, CA95050].


J.C. Miller W.T. Kelley

Personal computers,an engineering productivity tool

The personal computer is a powerful productivity multiplier. Asone manager said: "I never really wanted the PC, but now ifanyone tries to take it away, I'll break their arm."

When the first personal computer (PC)was brought into the Video Componentand Display Division (VCD) only a fewyears ago, it was a struggle; the authoriza-tion system was geared toward remotetime sharing and centralized computing.The PC was a somewhat heretical approachto computing, deemed suitable for hobbyapplications but not for "real" computing.The PC allowed users to do their owncomputing directly-in an uncontrolledfashion. Where would it all lead if ev-eryone could use a computer? Who woulddirect it? Wouldn't users just play games?Where is the economy of scale? Thesewere just a few of the objections.

In a short time, personal computers havegone from "persona non grata" to the"engineer's best friend." What caused thetransition and why has it occurred withsuch speed? The PC surge reflects the democ-

Abstract: The personal computer (PC) isa powerful tool for bringing computerpower directly to the user. The VideoComponent and Display Division hasmade extensive use of the IBM PC as partof an overall engineering productivitystrategy (see companion paper, by Miller,on VCAD). This paper describes theapproach taken, some of the thinkingbehind the hardware/software choices(risks and mistakes), system integration,training and results of an earlyproductivity measurement

01984 RCA CorporationFinal manuscript received March 22, 1984.Reprint RE -29-3-2

ratization of computing. Plummeting PCcosts and explosive varieties of PC soft-ware have made it irresistible. Daily ads inthe local papers and on TV have extolledthe virtues of the PC. Amidst the "hoopla"one can find a large kernel of truth: com-puters can be used by non -experts.

Well -thought-out software is the key toPC usage. Unlike historical computer users,the PC user views the computer merely asan optional tool to do his real job moreeffectively. If it is easy, well and good; ifnot, it won't be used. In fact, the market-place has generated an amazing range ofsoftware directly aimed at the nonprofes-sional user-these are tools to solve prob-lems. A dynamic cottage industry has beenspawned. No longer does the user requirethe time and skill to craft his own tools.Just as the do-it-yourself carpenter doestasks formerly done by the professional, sonow the PC user picks off -the -shelf soft-ware to do his work.

The Apple personal computer, with itscost, features, and software, was the triggerto this market avalanche. The now -famousVisiCalc (VisiCorp) spreadsheet was theprofessional's eye-opener-it provided a ge-neric solution to a wide class of importantproblems and attracted users exponentially.VisiCalc on the Apple was chic. IBM'slater entry into the market put the seal ofapproval on PCs and, through a strongmarketing strategy, rapidly rose to a dom-inant position. Writers of PC software adopt-ed the IBM PC with its open architectureas the machine of preference and gave itthe position of a de facto standard.

On the average, the professional spends

AL1111111111111MMIIMINNIIIIIIMIk

the bulk of his time communicating, doc-umenting, and planning-as opposed todesigning or testing-yet these "adminis-trative" tasks traditionally have the weak-est level of support (Fig. 1). In the VCADapproach, we targeted the PC usage directlyat the administrative functions (includingminor calculations) as opposed to the com-puter -intensive tasks.

COMMUNICATING40%

PRODUCTIVITYIMPROVEMENTTARGET(80% PIECE OF THE PIE)

Fig.1. Pie chart breaking out the typicalengineer's time profile, based upon datafrom Hewlett-Packard. Traditional com-puter support focuses on the test/de-sign segment. The bulk of the engineer'stime is spent upon tasks that can benefitfrom computer assistance, but this isnot widely provided. VCAD seeks to ser-vice the entire pie through a roundedset of tools.


ApproachThe VCAD project recognized the PC asa key part of its overall productivity strat-egy. It provided a means to distribute use-ful computing power directly to the engi-neering community, flexibility in selectingsoftware, and the potential to be used in adual mode-local or terminal. It was clearfrom the outset that the PC was the meansto reach a broad class of users and changetheir work habits. The PC was not viewedas some sort of super -design workstationfor the elite engineer.

The next decision was which vendorand what product? From a hardware per-spective, there were many suitable choicesand complex tradeoffs to consider. Butfrom the software side, the answer wasclear: we chose the IBM PC/XT (the XTis a PC with a Winchester disk) as thestandard. While remaining alert to the tur-moil in the operating system area, PC -DOS was adopted as the standard operat-ing system. We wanted to avoid the prob-lems that accompany multivendor installa-tions and resisted the temptation to selectseveral PC vendors, each offering somepurported advantage. In the VCD engineering environment, standardization offeredsignificant advantages over specializationand proliferation.

Standardization is a risky tactic in avolatile market environment; it is easy topick an Edsel. The VCAD design conceptdoes not depend upon the PC being anIBM one. So far, the choice has stood thetest of time although we are constantly onthe alert for other trends. There are overone hundred manufacturers of PCs andliterally thousands of software offerings.The risk of making a mistake is high. Wealso learned that there are many subtleerrors one can make in meshing PC soft-ware to specific hardware configurations.Because the field is relatively new andhighly fragmented, it is much more diffi-cult to make good decisions about PCsthan about the nominally more complexmainframe options.

Within this conceptual framework, astandard configuration was devised for thePC and the XT. The engineering standardPC/XT was a 256K machine with a mono-chrome monitor and Okidata 93 printer.A variant configuration was adopted forthe secretarial user (different printer). Forpurposes of investigation, some machineswere acquired with minor variations: colormonitors, Hercules Graphics cards, morememory, math co -processor, pen plotters,and so on, but these represented variationson a theme rather than radical changes.

Each PC/XT was also configured to beattached to the local area network (LAN)so that it could communicate with otherdevices and computers.

An intuitive judgment was made on theratio of PCs to XTs, balancing the follow-ing factors: cost, ease of use, number ofusers of the machine, and user sophistica-tion. The XT was felt to be suitable forthe more experienced user, the person whocould make intelligent use of the largerhard -disk capacity, or the person who hadan identified application requiring the disk.The XT user also tended to be someonewho would not be sharing the machinewith other users. Limited disk -security fea-tures made it too easy for someone toinadvertently ruin the disk's contents in ashared environment. The PC was a goodentry-level machine. We bought 73 ma-chines in 1983, with PCs to XTs in a 4:1ratio. All secretarial machines were PCs.

Although there is considerable enthusi-asm for the PC as an engineering tool, it isnot a cure-all. Many other kinds of PCsand workstations on the market offer morecomputing power, specialized analytic soft-ware, and so on. As the need for higher"horsepower" arises, these terminals canbe attached to the LAN without change tothe system. For example, experiments withthe PC as a laboratory instrumentationcomputer have shown it to be insufficientfor some of the more computation -inten-sive tasks. In such applications, a 16 -bitmachine with a faster processor is required.Also, we would characterize the standardconfiguration as having a weak graphicscapability; this can be improved by severalrelatively minor changes or by use of thehost.

The broad dilemma was whether towait for things to settle down and make asafer choice, or to proceed in the face ofcertain change. We have opted for the lat-ter in the belief that the overall systemarchitecture is sufficiently robust. Down-stream, we will acquire PC -type machineswith increasing capabilities.

Software

Adopting reasonable hardware standardsearly in the project made it possible toinvestigate software from a more systematicviewpoint. It was assumed that the pri-mary application for the PC was in simpleapplications with a high administrative fla-vor as opposed to computer -intensive orgraphics tasks. Early in the process, wordprocessing and spreadsheets were identi-fied as the basic software areas, in addition

Fig. 2. The PC connection. The personalcomputer is attached to the local areanetwork, which in turn is coupled to thehost via the Hydra protocol converter.ACOM software in the PC and hostcause the PC to be presented to theIBM 4341 as a full -screen 327X -classdevice with full -function key support.ACOM also supports two-way file trans-fer. When in the terminal mode, the PCis driven at 9600 baud. Upon request,the PC can be linked to other deviceson the LAN, or of course operate in astand-alone mode.

to ancillary areas such as database manag-ers, statistics, graphics, project planning,and so on. At all times we tried to balancethe tradeoff between doing tasks on theVCAD host versus the PC. This is indeeda delicate balance-on the one hand, put-ting too little on the PC emasculates theoriginal purpose, on the other, each PCtends to become a miniature computercenter.

Word processing (WP)

Perhaps no area has as many choices anddifferent user opinions as word processing.No one seems to agree on the ideal set offeatures. In fact there is no one best WP,but a surprising array of specialized offer-ings, each with strong and weak features.Volkswriter (Lifetree) was selected at thebeginning of the VCAD effort (based uponsimplicity and ease of learning). Beforepurchasing it we learned of MultiMate(Soft Word Systems Inc.) and made a deci-sion to change direction. MultiMate, a WPbased upon the popular Wang dedicated-

Miller/Kelley: Personal computers, an engineering productivity tool 11

WP system, has an extensive set of fea-tures, is well supported by the vendor, andapproximates the features of a dedicatedWP. From a psychological point of view,it was attractive because there were sev-eral Wang WPs already in use in the Di-vision; MIS was asked to concur in therecommendation and they did so. Engi-neering secretaries who had been trainedon the Wang, now prefer the PC withMultiMate.

Spreadsheet

Here the decision was easy-Lotus123(Lotus Development Corp.), perhaps thefinest software written. Lotus123 builds onthe excellent VisiCalc concept and addsextremely Lotus123 of-fers excellent spreadsheet features, usefulinteractive graphics, modest database capa-bilities, and very limited word processing.Standard output interfaces make it pos-sible to transfer information between Mul-tiMate and Lotus123 (as well as otherpackages).

Fig. 3. Secretary using the PC and Mul-tiMate for word processing. The generalreaction of the secretaries has beenhighly favorable. Some secretaries arealso using packages such as Lotus123for spreadsheet work. A turnkeyLotus123 template was created to fullyautomate the business expense reportfrom input to final form ready for signa-ture. Such techniques are improving ef-ficiency.

Communications

We made a mistake in this area, initiallyselecting the IBM 3101 emulator to linkthe PC to the LAN. After acquiring it, cer-tain deficiencies were found in the soft-ware that caused us to look for anothersolution. Fortunately, the ACOM emula-tor (Computer Vectors) was found; it sup-ports file transfer as well as PC -3278 emula-tion.

MultiMate, ACOM, and Lotus123 wereinstalled on virtually all of the PCs ac-quired. They became the basic softwarebuilding blocks. It has been a gratifyingprocess to watch users develop their ownapplications with these tools. Many tasksthat would never have reached the level ofapproval on an MIS Data Processing Re-quest were now being solved quickly andimaginatively, following only a few hoursof instruction, often by users with no com-puter background.

PROFS

Although the IBM Professional Office Sys-tem (PROFS) is not a PC application, eachPC owner is coupled to PROFS via theLAN. From a software point of view, it isa package directly available to each user.PROFS serves as the electronic mail linkbetween VCAD users. Using the ACOMemulator, files can be transferred betweenhost and PC; this allows a letter to be

composed in MultiMate, sent to anotheruser via PROFS, and so on. Of course thefull spectrum of PROFS features are avail-able to each PC user.

A number of packages (database man-agers, project planning, graphics plotting,statistics) have been reviewed. This is anongoing effort of the VCAD task force;new software is periodically acquired andstudied to determine its utility. Formal re-ports are generated to document the review.In general, these offerings are specializedand of interest to only a portion of the PCcommunity. Perhaps a dozen different pack-ages are in use within the VCAD com-munity. Authority to purchase PC hard-ware and software is focussed within engi-neering to minimize proliferation and encour-age synergy among users.

TrainingSeveral decisions were key to the PC instal-lation process. Every person assigned toPC had to agree to take training prior toreceiving the machine. Secondly, a "PCCollege" was established as a forum fortraining. The PC college was not a highlystructured school. Students were asked todevote adequate hours to learning DOS,typing skills, and Lotus 1 23 using pro-grammed tutors and selected problems. Thiscombined self -study with hands-on use ofa PC and short lectures. MultiMate was


Typical comments on the PC investment

66... Having had little previous exposure to comput-ers, I was slow getting started but I see more andmore applications as time goes on. I was not anx-ious to get the unit, but will now break the arm ofanyone who tries to take it away! Every phase of mywork has been improved by the use of my PC.Clearer memos, better project planning, and more in-depth analysis and documentation of experimentalresults ...

... The PC is an amazing and powerful tool. Theability to revise documents in MultiMate and havenew copy within minutes has been a tremendoustime saver ... The system provides an easy accessfile so that all of the information I need is at my fin-gertips without going through a lot of paper work ...

.. it is easier and less intimidating to write reportsusing the word processor than using pencil andpaper ... The most important application is large-scale data manipulation for factory test ... I look for-ward to retiring my typewriter, supplementing mycalculator for engineering calculations, and dispens-ing with hand tabulation of data and hand plotting ongraph paper ...... A returned Development Tube (DT) sample had apoor focus gun. By looking at a data comparison ofall DTs shipped, three other "bad" tubes were found

that were not previously known. The ability to findinteresting and important facts in large data fieldshas been too cumbersome to do in the past. This PCability will improve this situation ...

Lotus123 allows "after thoughts" to be quicklyand easily inserted and rapid comparisons madewith alternate input variables ...I am now able tokeep abreast of time schedules on all importantjobs ...The quality of work is improved particularlyon design analysis. In cases where formerly featureswere calculated at a limited number of locations,those same features can be analyzed at an unlimitednumber of locations....

. .. We're now able to set up project tracking systemfor each engineer and designer. The PC will alloweach individual to update their project as progress ismade Project outlining, planning and manpowerresources are now possible through worksheetanalysis by sorting against projects, group leaders,and so on ...... We're now able to combine 'Things -to-do' inputsfor MUSCLE, DT list, Marketing, Technical aid con-tracts, engineering meetings, Customer ContractReports, Test Requests, and so on. This was impos-sible to do manually and has never really beendone ...»

Miller/Kelley: Personal computers, an engineering productivity tool 13

taught to secretaries in a 4 -hour session;others learned it by themselves. Some ofthe training was devoted to housekeepingtasks such as care of floppy disks, how torun the printer, and so on. Although thetime required to train was insignificant,and it varied with the background of thestudent, it saved many mistakes in actualusage of the machines.

PC user groups have been established toprovide a means to share results, tips, prob-lems, and solutions. These user groups arenow sustained by the newly -formed MISInformation Center and service the Divisionrather than just engineering. PROFS train-ing was conducted on a more formal basisin March, using the standard IBM refer-ence material supplemented by an intro-duction to CMS. During this training, thestudents also were introduced to ACOM,the means to link to the host and transferfiles.

MeasurementIn February, 1984, a productivity studywas undertaken to measure the results ofthe PC investment. A detailed survey wastaken of each PC user querying them aboutthe utility of the machine (see sidebar).Some of the highlights of the study are:

Almost uniform enthusiasm, good re-sponse to the standard software, and asignificant increase in computer literacy.

73 machines installed. Productivity gainsof 25 percent per machine hour afteronly 1.8 months, and average machineuse of 4 hours per day, with 2.6 users/machine. Faster problem solving.Lotus123 productivity gain 2.6:1, andMultiMate productivity gain 1.5:1.

Better planning, communications, projecttracking. Creation of simple but usefuldatabases in all areas. A myriad of usefulanalyses reduced to routine.

ExpansionBased upon the favorable initial user reac-tions, we are in the process of defining thePhase II VCAD expansion in the PC area.The current plan is "to reward the rich,"that is, to add additional resources prefer-entially to the groups that have made thebest use of the first PCs. Our problem isslightly more complex the second timearound as there are additional PC offer-ings to consider (XT/370, XT/3270, . . .),

each of which has certain desirable fea-tures. In the laboratory measurement area,

we are reviewing the IBM 9000 wherehigher performance is a requirement.

AcknowledgmentSpecial thanks are due to the enthusiasticengineering community who adopted thenew approach while in the midst of majorproject responsibilities; M. Renfro and H.Hillegass who taught in the "PC College";B. Mangolds for communications support;and T. Hart who formed the InformationCenter.

James C. Miller is the Manager ofTechnical Projects and EngineeringAdministration at VCD in Lancaster, Pa.(see biography, page 8).

William Kelley joined the RCA/EC Micro-wave activity in 1950. He has held a varietyof Manufacturing, Engineering, ProgramManagement, Business Analysis, andAdministrative positions during his RCAcareer. He is currently focusing onoptimizing the use of personal computersand PROFS to help improve engineeringproductivity within VCD.Contact him at:Video Component and Display DivisionLancaster, Pa.TACNET: 227-6443


J.C. Miller P.J. Kunz

CADAM, an engineering productivity tool

We are using CADAM, a computer -graphics system, as the keyto upgrading our manual design methods into an integratedmodern approach. CADAM is already increasing productivityand design quality at VCD, and changing the way engineerswork.

The Video Component and Display Divi-sion had made modest use of graphicalcomputer -aided -design tools. Several yearsago, an Applicon design system had beenacquired and installed in one engineeringdepartment. Though a good decision atthe time, when viewed in hindsight, certainlimitations may be discerned: deficienciesin reliability, number of terminals, ease of

Abstract: CADAM (CADAM Inc.) isthe key part of a graphic design systemused in the VCAD program. It is anassemblage of powerful computerhardware and software tools. These tools,although internally complex, allowmanagers, engineers, designers, anddrafters to easily perform their normaldesign functions with greater speed andaccuracy. This paper documents theCADAM system from concept toimplementation. The rapid success ofCA DAM is seen by its early use in abroad range of design problems, its easyassimilation into the existing workenvironment, and the demand for moreterminals. Our motivation with CADAM issingle-minded increased engineeringproductivity. With 2 weeks of training and8 weeks of production usage, productivityhas already reached 2:1 versus manualmethods-our goal is 4:1 over the fullspan of user activity. The CADAM systemwill eventually be expanded to supportapproximately 50 terminals.

©1984 RCA CorporationFinal manuscript received March 22. 1984.Reprint RE -29-3-3

training, access, and response time made itinappropriate to extend the system to abroader base. The system was primarilydedicated to specific aspects of electricaldesign. Eventually it evolved into a closed -shop service in which dedicated operatorsdid a mix of design, plotter operation,trouble shooting, backup, and program-ming. The operators were physically andorganizationally separated from their parentdesign groups. As with any service center,work scheduling became an issue. Produc-tivity gains were asserted but difficult tomonitor routinely. The formal training pro-gram was weak, consisting of either do-it-yourself work or a short course at thevendor.

ApproachAs the VCAD program was planned, wesought a turnkey graphics design systemapproach that addressed these concerns andthat led to an integrated solution. We envi-sioned a need for a large number of ter-minals, fast system -response time, a tightfocus on productivity, a common databaseacross all users, and a desire to access awide set of software tools. A balance wassought between speed and advanced designabilities: it is all too easy to incorporate afeature that, while elegant, consumes exor-bitant system resource and slows all otherusers. Because of the need for many ter-minals, the incremental cost -per -terminalwas an important factor.

It was important to integrate the newapproach into the existing organizationalfabric; therefore, the new system had to beeasy to learn, and not require its users to

become computer specialists. The graphicssystem had to assist our communicationflow; properly designed and integrated viaa good database, it would reduce paperand eliminate duplicate manual steps. Ofcourse, it had to help shorten the overallprogram design cycle. Different users needto operate the system appropriately; letdesigners draw and engineers think-notoperate a CAD system. Reducing the num-ber of times a part or product is drawn,easy acc.ess to current designs, increasedquality level, and so on, were other con-cerns.

Overall, we concluded that a successfuland productive solution required a solidhost operating system, stable graphics soft-ware, and a formal training program. Thesystem was to be imbedded in the existingwork culture and organization, yet formedand guided by a standard overall outlook.

The Lockheed -developed Computer -Aided Design and Manufacturing(CADAM) software was selected as theheart of our graphic design approach. Thisis a comprehensive and stable set of pro-grams that drives a graphics database. Thesystem was augmented by acquiringCAEDS (SDRC Inc.) and ANSYS (Swan-son Analysis) to support finite -elementmodeling as an adjunct to CADAM. Theseproducts offer a broad base of engineeringtools to use in design from concept andanalysis to the final machined part. Oneman was assigned to implement the sys-tem and assumed software, system, andtraining responsibilities. This is the firstCADAM installation within RCA.

CADAM is a widely -used system ca -

RCA Engineer 29-3 May/June 1984 15

pable of supporting large numbers of ter-minals (one installation has 600+).CADAM provides design and manufac-turing advantages such as increased pro-ductivity, shorter design cycle, lower costs,better designs, greater accuracy, flexibilityto make changes, and the opportunity tohave more standardization of design. Gener-al Motors has reported 3.81:1 productivitygains via CADAM. Even higher ratios areachieved in specialized tasks. The CADAMsystem provides a high performance, broadfunction, design/drafting package togetherwith a number of aids for analysis. It thenmoves onto completing the job by provid-ing numerical control parts programmingabilities. The central database is the system"glue," enhancing communications and re-ducing manual drawing -retrieval efforts.

ConfigurationAll CADAM hardware is attached to anIBM 4341 (Model 9) computer with 4megabytes of memory and 2.4 billion bytesof disk storage. The current CADAM work-station consists of the IBM 3251 graphicsterminal, which uses a program function key-board, standard typewriter keyboard, andlight pen, to control the program and enterdata. Three of the 3251 workstations areattached to a single 3255 terminal con-troller stationed up to fifty feet away. The3255 is attached via coaxial cable to a3258 controller in the computer room; upto four 3255s can be attached to one3258. The Phase I installation supports six3251 terminals. The CADAM terminalsare placed directly in the user work arearather than in segregated computer spaces.In a very real sense this configuration is a"starter -kit" to prove the concept.

The designer sits at the terminal (Fig. 1)accessing work stored on the large disk ina special CADAM database. He can recallprevious work and make changes, startafresh, initiate an analysis, compare draw-ings, and so on. Each user can access hisown reserved portion of the database or,with system authorization, view and copy

designs stored in other work areas. All ofthe normal graphical commands are sup-ported. The terminal works in a menu -driven mode that makes it easier to learnand that always coaches the user as to thenext set of steps available. For practicalpurposes, the size of the database is lim-itless-all active work is kept on-line. Datais stored in a so-called "21/2D" CADAMdatabase. Every ten interactions the fullworking drawing is backed up by the sys-tem, preventing loss in the event of acrash. Each user designs with a 5K -sizeworking model to instill good practicesand avoid waste of resource. Files are pro-tected by a multilevel password schemeand are grouped into sets based upon organ-izational requirements.

The workhorse of the hardcopy facilityis a 36 -inch -wide Versatec plotter. Theplotter uses an electrostatic process to placeelements on the paper or film with a reso-lution of 200 dots/inch. Through a chan-nel -attached controller, it can produce typ-ical E -size plots in ninety seconds or less.Generally, a plot is completely finished bythe time the requester has walked to theplotter. This process saves time formerly

lost while searching for a drawing in themanual files. The Versatec Random Ele-ment controller and software are used toproduce plots with radii that look trulyround. Plot quality generally exceeds thatof manually drawn tracings.

The CADAM database structure isshown in a simplified form in Fig. 2. Inour implementation, we have opened thebase to most of the CADAM user com-munity to facilitate sharing of work. Elab-orate security provisions may be appendedas desired.

For tactical reasons an interim approachwas chosen for remote graphics in Phase Iof VCAD; IBM FASTDRAFT systemswere installed in Marion and Scranton.FASTDRAFT is an interactive, two-dimen-sional, isometric system for computer -aideddrafting. It uses the IBM 3251 terminalwith a light pen and typewriter keyboard.The program runs on a stand-alone 7361graphics processing unit. Output is pro-duced on an E -size drum plotter with upto eight colors. Unfortunately, FAST -DRAFT is not directly compatible withCADAM; this complicates efforts to linkLancaster with the manufacturing sites. In

Fig. 1. Designer working at CADAM terminal. The workstation consists of the IBM3251 graphic display, light pen, function keyboard, and conventional keyboard. Thesystem provides a 0.3 -second response to most requests. The 3251 operates in amode that shows the menu at the bottom of the screen and the required responseoptions for a selection at the top. The VCAD visor is part of the VCAD trainingpublicity effort.


the short run, software links betweenCADAM and FASTDRAFT via IGES (agraphics interchange facility) are being in-vestigated, but a longer -range solution maybe satellite CADAM stations.

Software

The turnkey CADAM software has theability to set various performance parame-ters and options that are installation depen-dent. The primary workhorse is the graph-ics design package CADAM (release 19.1).CADAM is an interactive graphics systemfor computer -aided design, developed byLockheed Corporation for over 15 years.CADAM is divided into modules and weare currently running the following:CAD/CAM, Data Management, Hard -copy, Statistical Data, Accounting, 3-DSurface, Geometry Interface, CADGRAM,CADEX, and 3-D Mesh Geometry.CADAM was installed without incidentand has proven to be very stable.

CADAM is a complex piece of soft-ware that can be tuned in a myriad ofways internally and with respect to theVM/CMS host operating system. This is acomplex process and must be done care-fully to properly balance the system re-sources. The users are not involved in tun-ing, database management, or backup-theyconcentrate on their work. In our view,system tuning in an engineering environ-ment is sufficiently subtle and dynamic asto require local skills. In the VCAD sys-tem, priorities have been arranged suchthat CADAM is the most important taskon the system. Each CADAM user nor-mally sees system response to a light pen"hit" within 0.3 seconds. This extremelyrapid response time-even with a largenumber of terminals and considerable back-ground work-is the key to productivity.A skilled operator can stack up commandseven at this fast rate. The machine doesnot normally limit the thinking process.

CAEDS and ANSYS are used in com-bination for finite -element modeling.CAEDS is an integrated program that ad-dresses the functions and applications ofmechanical product development. CAEDSemphasizes the use of analytical modelingand analysis of a design in the conceptualphase. It interfaces with the CADAM data-base and will extract 3-D data from amodel for mesh generation. Our configu-ration uses the modules FRAME and

DESIGN

&NECKING

APPROVAL

RELEASE

YES

MODEL

Fig. 2. CADAM database structure. A design is initiated and stored in the user'swork space. When complete, it is advanced through several levels for checkingand approval. The user work space is deliberately kept small to force passing workon for approval and release. Each level in the hierarchy blocks the subordinateusers from making changes. Final released drawings cannot be altered by any ofthe users; but they can be copied and used as the basis for starting a new design.While signed file copies are maintained for archival and legal purposes, the "real"design rests in the CADAM database. All drawings are maintained on-line and canbe accessed in a matter of seconds by any authorized CADAM user. This facilitatessharing and reduces redundant effort.

GRAPHICS and then interfaces to thenonlinear solver of ANSYS to analyze themodel. ANSYS is a large-scale, general-purpose program for the solution of engi-neering analyses. Figure 3 shows a typicalANSYS output generated on the Versatecplotter.

The combination of the two products,along with CADAM, provides a powerfultool to build analytic models interactivelyand solve them for any conditions that wepresently encounter.

TrainingTraining is an extremely important aspectof the entire operation. It is an investmentin future results yielding return as increasedoutput at a higher quality level. Extensivetraining is needed to fully use the optionsavailable-this not only applies to the every-day users but most importantly to the man-agers who will have control over the jobson the terminal.

We have followed a top -down trainingphilosophy in order to imbed CADAM

Miller/Kunz: CADAM, an engineering productivity tool 17

into the working culture as quickly as pos-sible. The first CADAM trainee was Mr.C.A. Quinn, the Division Vice -Presidentand General Manager. All managers in thecommand chain between him and the ulti-mate user were trained prior to trainingthe user. The benefits of this approachcannot be overestimated. Excellent stan-dardized training is available at the vendorbut we opted for a local approach. One ofthe authors (P.J.K.) took specialized train-ing at the vendor's site prior to installationof the equipment; he then served as thelocal trainer. This technique allows the

training to be conducted in a fashion thatbest matches the norms and work systemof the location with minimum disruptionto the normal job assignments, and hasproven to be extremely effective.

A broad range of courses are presented:CADAM BASIC, On -the -Job Training,BASIC II, BASIC NC, Advanced NC,FEM,. ANSYS, 3-D Surface. The basiccourse includes 40 hours of terminal timeand a second 40 hours of on-the-job train-ing doing productive work. Thirty-two peo-ple have been trained in 2.5 months, repre-senting over 2000 hours of class time. Allhave finished the courses and are at least1:1 with manual techniques at the end oftraining. Refresher courses are held to main-tain skill levels or review new system fea-tures. Demonstrations are given to acquaintuntrained persons with the CADAM poten-tial. Close follow-up is maintained witheach user to review performance and skilllevel. All graduates of training receive cer-

Fig. 3. Typical ANSYS plot. Finite -element models can be created directly throughANSYS procedures (the slowest method) or through the CAEDS/CADAM tech-niques starting with a CADAM drawing. VCAD offers good turnaround and excel-lent plotting through the Versatec graphic plotter. A drawing takes 45 seconds togenerate from the ANSYS output run.

tificates; buttons, hats, and so on, are issuedto gain attention.

MeasurementA major effort was invested to establish aneffective productivity measurement scheme.At a very early date, benchmark data andwork statements were received from themanagers of the groups that were targetedto receive CADAM. Benchmark data wasalso received from the Scranton and Mar-ion plants before using FASTDRAFT. Itwas not simple to obtain meaningful dataand a certain element of subjectivity wasinvolved. Each group is being monitoredagainst their individual measurement stan-dards.

The CADAM system has been wellreceived and is meeting our expectations.Evidence of this is seen in the progressionof events. Prior to receiving the system,the managers did not believe 1:1 produc-tivity would be reached for eight to twelvemonths. After they were trained, their fig-ure was revised to six months, and cur-rently, after 8 weeks of production use, thesame managers will affirm that the opera-tors are at a level of 2:1 productivity orbetter. A further sign of acceptance ap-peared in February, when terminal utiliza-tion jumped to better than 90 percent dur-ing normal working hours. By March, wewere experiencing 90 percent usage over a12 -hour day plus some weekend time. Oneengineer, after only 3 weeks of productionexperience, solved a knotty design prob-lem with CADAM and avoided an esti-mated $100K loss that would otherwisehave occurred with a manual design. Thatsingle effort "paid" for the CPU.

Our goal is to move up the productivitycurve to a level of 4:1 or better. Manage-rial attention is focused on this factor. Overthe period ahead, the productivity moni-toring effort will be automated in order tohighlight performance and possible retrain-ing needs.

We have reached a point where newtraining has essentially stopped because eachperson cannot schedule adequate terminaltime. New designs and some rework arebeing done on the system, which is cur-rently being used primarily for mechanical,as opposed to electrical, design. Users haveembraced the new technique and voicedno desire to return to the older ways. It istoo early to determine the number of draw -


ing boards replaced by a terminal.The system has been available for use

around the clock since December 7, 1983.The host 4341 runs unattended and hasbeen solid (one unscheduled IPL since No-vember) since installation. This type ofperformance is crucial to a centralized sys-tem. During the first 45 days of operation(6 terminals, 10 -hour CADAM day), welogged 1688 production terminal hoursfrom an available 2700 hours (62 percentusage), and lost 21 terminal hours to anin -plant power failure.

Regular CADAM Users Group meet-ings have resulted in excellent feedback.All trained personnel have agreed that thesystem is stable, fast, and easy to learn.They are beginning to develop their owntechniques to improve their new-found skill.Several user areas have developed newways to perform their design problemsincluding applications where they had pre-viously been unable to get satisfactory an-swers. The Drafting Standards Committeehas started to integrate CADAM into thework standards.

Expansion

The enthusiastic reaction to CADAM andits auxiliary tools has justified an earlyexpansion of the system. By midyear, thehost computer will have been upgradedand a dozen additional CADAM termi-nals will be arriving. The present projec-tion calls for an ultimate need for morethan fifty CADAM terminals and will re-quire further expansion.

The new CADAM terminals will be ofthe IBM 5080 family, featuring color, ras-ter refresh, and tablet input. The 5080architecture has one controller per termi-nal, allowing them to be sited individuallyas desired. In future CADAM releases,more function will be downloaded to the5080 terminal, thus allowing more termi-nals per unit of CPU resource.

At the same time, we are broadeningour scope to consider the extension ofCADAM into numerical controlled machin-ing. A numerical control task force hasbeen established to do the preliminary plan-ning, and experiments are being conductedwith the existing CADAM numerical con-

trol modules to determine its viability. Linkshave been established between VCAD andthe existing NC equipment at Lancaster.We had acquired CADGRAM andCADEX, two electrical -design -orientedCADAM modules as a part of Phase I. Asthe new terminals arrive, these will allowus to extend the scope of CADAM intothe electrical design area. CADAM is alsobeing studied as a mechanism to put ourEngineering Standards on-line for the tech-nical community; it would be accessed byrelatively inexpensive 3179 color terminalsusing CADVUE. The list of potential im-provements is greater than our capacity toaddress them.

AcknowledgmentSincere thanks to K. Walker for his con-tributions in the system area.

James C. Miller is the Manager ofTechnical Projects and EngineeringAdministration at VCD in Lancaster, Pa.(see biography, page 8).

Peter Kunz received the BSME with aminor in Materials Science from theUniversity of Connecticut in 1979. Hejoined RCA in 1979 and his initial workcentered on machine design of productionequipment for the electron gun, includingcathode problems and the quick -heatbimetal cathode process. He received hisfirst patent in 1983 and has several patentspending. In 1983, Mr. Kunz was promotedto Member Technical Staff and joined theVCAD project. In his current capacity, he isresponsible for implementation andsupport of the CADAM computer -aideddesign system including training, databasemanagement, finite -element modeling,productivity measurement, and systemssupport.Contact him at:Video Component and Display DivisionLancaster, Pa.TACNET: 227-6492

Authors Miller (left) and Kunz.

Miller/Kunz: CADAM, an engineering productivity tool 19

E.M. Melendez

Emerging engineering design tools

Future workstations will allow flexibility, integration, ease of use,and prospects for creating proprietary sets of design tools.The tools will paradoxically and powerfully integrate broadindividual design freedom within highly organized andsynergistic group efforts.

During this decade and into the 1990s the design of embeddedmicrocomputer systems will require a new set of design tools.The design environment for the logic and system (hardware andsoftware) engineers will consist of highly intelligent design/de-velopment workstations containing highly integrated automatedsoftware tools on one hand and an automated software -supportenvironment residing on a superminicomputer on the other.Both these subsystems will be linked together onto a networkand will share an adaptive design database management systemthat will behave as the central data -monitoring and control facil-ity for the multi -engineer team. This design system will increaseproductivity, shorten the design cycle, and will allow manage-ment to better monitor and control the design cycle. Creativityand innovative use of these automated design tools by the designteam will ensure a competitive edge in the marketplace.

In pursuing research on what type of tools an engineer willneed and use in the future, an interesting cause -effect relationshipwas discovered. Today, most engineers are team players, relyingon each other to perform project -related tasks in taking a prod-uct from concept to completion. The size of most engineering -re-lated undertakings dictates that an engineer interface and workin cooperation with an integrated design team. However, in past

Abstract: This article addresses today's engineering designproblems and how they may be solved using technology togetherwith recent and future software applications. Aspects ofmanaging design projects in present and future environments willbe covered including system tools, IC design tools, PC -boarddesign tools and hardware/software integration tools. Thedistributed database environment is a key to achieving teamworkwithin a highly individualized working environment. Otheraspects of future systems will be artificial intelligence, rigorousproject management tools, superminicomputers, and powerfuluser intetfaces.

©1984 RCA CorporationFinal manuscript received March 29. 1984.Reprint RE -29-3-4

years engineering as a profession was thought of as an individualpursuit that demanded creativity and a working knowledge ofmost all of the engineering disciplines. If the recent and continu-ing evolution to low-cost memory becomes married to theemerging revolution on artificial intelligence, the engineer maymove from a team player environment back to an individual,less specialized way of working.

It is important for the reader to realize that the projectionsof what future engineering tools for the engineering professionwill be is an opinion of the author and not a stated policy ofRCA. In Burlington, Massachusetts, RCA Automated Systemshas been and continues to pursue many of the areas discussed inthis article. For example, Automated Systems has under devel-opment an artificial intelligence program for microprocessor -based test systems, a plant -wide local area network (LAN) fordevelopment activities that is married to today's most advancedengineering tools. Many of the projections made in this articlereflect the baseline Automated Systems has established in its

drive to provide engineers with the best available engineeringtools that will meet not only today's but tomorrow's technicalchallenges.

Managing design projectsUse of IC technologies such as gate arrays, programmable arraylogic, standard cells, custom VLSI, and the ever-increasing amountof software that needs to be developed on most design' projectshas placed new demands on the designer and the related envi-ronment. Today, an embedded microcomputer system requiresstrict packaging considerations and, as a result, PC -board designersdepend more than ever on the assistance of computers. In fact,microprocessor technology has outstripped the ability of engi-neers to program them without automated assistance.

Today, software maintenance can be as high as 70 percentof the total software costs. For example, a software maintenanceplan must include provisions not only for finding and fixing bugsbut also for enhancing system capability. These tasks are espe-cially difficult if design data and documentation is not well

Melendez: Emerging engineering design tools 21

Fig.1. Hardware/software integrated configuration allowsthe engineer's workstation to emulate two processors andperform timing analysis on a separate processor. The win-dow capabilities allow the user to view all pertinent data onone screen.

maintained. A design team must know the full impact of anychanges it proposes. As a result, it is apparent that the role

of managing today's and tomorrow's design projects will not beeasy.

Only a well conceived, tightly integrated design environ-ment can support the management, documentation and efficientuse of today's large databases. When a design team becomeslarge, successful system integration relies on project -managementtools that support partitioning of the many design tasks. But,effective system and design -task partitioning requires engineeringtools that can analyze circuit behavior, packaging, testing, andintegration attributes with respect to IC- and PC -board design.Also new tools will be needed that can assist with the design ofcomputer systems and software requirements.

Presently, most IC vendors have access to automated designtools, while PC -board tools are limited to only draftsmen andnot engineers. The main problem even with hardware and soft-ware integration tools is that they work as separate tools andthey only address a subset of the integration phase. As a result,there is an inherent underuse of narrowly focused expensive cap-ital equipment and longer design cycles. In the future, in order toalleviate this situation there will be a highly integrated designsystem that is network based. It will provide most of the designtools that will be needed for the entire design cycle. It will con-sist of computer -assisted engineering (CAE) workstations withcomputer -aided design (CAD) capabilities linked on a local areanetwork (LAN) to a superminicomputer. The design system willcontain a highly flexible architecture that will allow users to con-figure the network according to application. Each workstationwill possess multiple microprocessors working in parallel, localdatabase storage, a high -resolution color -graphic display withwindowing capabilities, and a highly interactive, easy -to -use, userinterface. The workstation will host a wide range of automated

design tools that will allow the user to configure it as an IC -de-sign station, as a PC -board design station, or as a hardware/ -software integration development system as depicted in Fig. 1.

The design system will behave as a virtually linked system.This will allow users at any station to access tools and designdata anywhere on the network without having to know its loca-tion. The superminicomputer will host the automated softwaresupport environment and the central project database (Fig. 2).

The workstations will typically include 4 megabytes of realmemory, 200 megabytes of disk memory, a plotter, and a printer(plotter and printer are optional; they can be shared). As per-sonal computers gain more processing power and obtain virtualmemory operations, they will be used as low-cost workstations.Presently, even though the applications are general, personalcomputers work with CAE design tools provided by second -and third -party vendors that allow them to currently assist thedesign engineer.

System tools

Most of the IC- and PC -board tools intended for the workstationwill be highly portable in that they can also be hosted on thesuperminicomputer. Yet for the most part they will operate in aworkstation environment. These tools are discussed according totheir application.

IC design tools

Although most IC automated design tools have existed for years,they have remained largely inaccessible to the circuit and sys-tems engineer. Conventional batch processed tools provide undulyslow response time and therefore can't significantly increaseengineering productivity. To increase engineering productivity,engineers must be able to interact with fast and capable analysistools, to rapidly assess new ideas and efficiently investigate theoperation of designs in progress. By providing local, easy -to -learn, graphics -intensive CAE systems that support the interactiv-ity needed, the designers can benefit from the capabilities of thecomputers they design.

The IC design tools will support most IC technologiesespecially for fully custom and semi -custom applications. Pres-ently the major sources for IC design tools are silicon vendorssince they are best able to keep up with changing technologiesand provide simulation tools that accurately predict device behav-ior. But this will change as more CAE vendors enter the picture,due to their nurturing of software expertise.

The IC design tools that will be available on a workstationare:

Functional libraries for different IC technologies (for example,standard cells, gate arrays)

VLSI mask design and verification routines

Logic and circuit simulators

Interactive waveform display routines for logic simulation results

Automatic and interactive function cell placement and routingsoftware

Schematic and logic capture packages

Cell compiler libraries and real chip libraries

Design analysis tools

Net -list extractors (for physical design) and net -list comparators(for logical design)


LOCAL

DATABASE

PLOTTER

PRINTER

SITE B

WORK

STATION

0

A

WORK

STATION

SMARTTERMINAL

L_

STPIR MINI

SMARTTERMINAL

GATEWAYTO REMOTE

SITES

SITE A

WORK

STATION

PLO TERPRINTER

LAN

PROJECT

DATABASE

177_SITE D

WORKSTATION

LOCAL BUS

LOCAL

DATABASEPRINTER/PLOTTER

Fig. 2. This system diagram illustrates the inherent flexibilityof this architecture. Engineers at different sites will have sys-tem -wide access to tools on other stations and on thesuperminicomputer.

Testability analysis tools

On-line design and electrical rule checkers

Physical layout verification routines

Schematic and layout plotters

PLA generators and optimizers

Text editors and graphics editors for generating design specifica-tions

Format interchange (for putting output tape into proper dataformat for the intended silicon foundry) software.

Together these tools will support the system design from thewriting of specifications, through the logic design and simulation,to mask design and verification. The output will be a silicon -foundry input tape. The physical design tool will offer flexibilityfor full -custom as well as standard -cell and gate -array design.The logic design tools will support all IC design styles and maybe used as a design capture -and -analysis system for printed cir-cuit design.

WORK

STATION

WORKSTATION

SITE C

0 B

C

ASLOCAL

DATABASE

PRINTER

PLOTTER

J

PC -board design tools

Presently, IC -layout systems aren't the only CAE products mov-ing from drafting to system -design environments. PC -board designsystems will also be acquiring automated design capabilities. Asa result, we will see that these tools will also be available on anengineer's local workstation. They will play an increasinglyimportant role in front-end circuit -creation and analysis tasks.Future PC -board design tools will address automated data -entryneeds. Typically, PC -board designers enter schematics from pre -drawn pages and thus require a data -entry package that accom-modates package -oriented flat designs. However, engineers requiremore flexible schematic capture programs that support less re-strictive data -entry forms. Below is a list of tools that will besupported by the workstation for PC -board design.

Symbol creation Interconnection routing softwareroutine for multilayer board design

Schematic entry Schematic and componentroutine libraries


APSE-solution forsoftware development portability

The overall objective of the APSE is to offer cost-effective support to all functions a project teamengages in during the development, maintenanceand management of Ada application software,throughout the lifetime of the software project, par-ticularly in the embedded computer system field.

To promote portability in this software supportenvironment, the APSE will contain two lower levels:the kernel APSE (KAPSE), and the minimal APSE(MAPSE). This figure depicts the APSE structure.

Level 0-Hardware and host software asappropriate.

Level 1-KAPSE; provides database, communica-tion, and run-time support functions to enable theexecution of an Ada program (including a MAPSEtool) and which present a machine independent por-tability interface. That is, it behaves as a virtual sup-port environment for Ada programs, including toolswritten in Ada.

Level 2-MAPSE; provides a minimal set of tools,written in Ada and supported by the KAPSE, that areboth necessary and sufficient for the developmentand continuing support of Ada programs.

INTERFACE

SPECS

Level 3-APSE; are constructed by extensions of theMAPSE to provide fuller support of particular appli-cations or methodologies.

Gate allocationroutine

. Fine -line routing and variablegrid

Drawing duplication software

Automatic component -placement software

Logic capture and simulation software

Hardware/software integration tools

The major future requirement for hardware and software inte-gration tools is real-time interaction of all the various test sup-port units (such as analyzers, emulation, stimulation, and so on).Today, this is a serious problem since most development systemvendors don't offer complete integrated support. Future worksta-tions must allow test instrumentation to interact in a cohesive,real-time manner. The stations will support a variety of processors.

Today, in -circuit emulators are severely pressed to behaveexactly like the microprocessors in a target system when theROM code is tested. These instruments must faultlessly duplicatethe nanosecond propagation delay that the CPU exhibits withinits internal registers. Future emulators will allow users to emulatetwo or more target units at once. The emulator provides a win-dow into the inner operation of a microprocessor and simulatesits activity, giving the designer the feedback and control neces-sary for development work. A logic analyzer, alternatively, willevaluate logic timing and states on dozens of input channelssimultaneously, and then present results in an easily recognizableformat, determining whether the problems are hardware- orsoftware -related.

The station will contain a software -performance -analysispackage that will, in addition to symbolic debugging, supply anoverview of system operation, printing out the relative activityamong various software modules, including I/O routines, systeminterrupts, memory allocation instructions, or actual programexecution states. These activities will be visually presented astime distributions or histograms; as lists of average time spent oneach software task; as percentages of the program's total execu-tion time. The station will allow users to detect redundant DOloops, spot excessive time spent on processing interrupts, andpinpoint execution -mode amess times. Another tool that will beavailable is a real-time debugger intended for multitasking ofsoftware environments.

These tools intended for the hardware and software integra-tion phase will provide the workstation with a powerful stimu-lus -response interactive measurement system that can be appliedto the integration problem of multiprocessor or embedded sys-tems.

Distributed database environmentIn this design system, it is imperative that users have the abilityto create and format their own local database without violatingthe overall structure imposed by the central project data -man-agement programs (hosted on the superminicomputer). A localdatabase in the workstation will serve the individual designers ona temporary basis, holding working copies of portions to thecomplete base that is stored on the superminicomputer. All tools


on the network will have a common interface to the system-wide database structure via a standard object -oriented interface.Likewise, all tools will have a standard interface to the user (Fig.3). The database structure will support intertool communications.

The central project database will support not only theautomated support environment for the software developmenteffort but also the managing and controlling of all design activi-ties on the design system network. The database will consist ofraw data in relational tables, and object -mapping capabilitiesthat transform the data from a relational format into user objects(such as gates, wires, networks, pages, specifications, design doc-umentation, program source text, program documentation, testdata, and so on). The database manager, residing on the super -minicomputer, inserts and retrieves data from these tables. Whenrequested, the manager extracts data from the project databaseand transmits it to the local processor in the workstation thatmade the request. There the local data manager enters it on adisk. Application programs then obtain the data through thedata -access manager, a program responsible for translating be-tween user objects and the database formats. In this way, theuser can reference and alter a database portion in local memoryindefinitely, needing no interaction with the project database. Atthe end of a session, the local data manager sends the editedportion back to the project database, updating the relationaltables to reflect all the changes. The local database and its asso-ciated software thus act as a high-speed buffer between the userand the slow, centralized project database.

As a result, the individual designers benefit while perform-ing individual tasks; however, they are subtly required to ensurethat their designs' external characteristics conform to the overallstructure.

Both local and project databases will have characteristics ofa design database management system that elects how to struc-ture the design data within the database system. Although thedatabase system does not interpret the data it manages, a design -database management system "understands" how the structure ofthe data describes a project. The major components of thedesign -database management system include: the storage com-ponent (reliably storing design data on disk); the recovery sub-system (saving incremental changes and ensuring resiliency to"crashes" either at the workstations or at the superminicomput-er); the design librarian (supporting check-in/check-out of designparts from the database); the validation component (checkingthat design constraints remain in force after a change); the designtransaction component, using the recovery subsystem, the designlibrarian, and the validation component to control the creationof new versions of data objects and the browser (providing aninteractive front-end for creating and viewing the data structure).The database system combines design tools, project managementtools, and recalls pieces of design and validates correctness. Theproject management tools assist in planning the implementationeffort. The design database management system is responsible forstructuring the design and exploiting the structure to keep thetotal design consistent.

Artificial Intelligence (Al)Artificial Intelligence will be extended to databases in futuredesign systems. For example, a self -adaptive simulation databasewould learn from characterization of presimulated circuits. Asinvestigations of physical circuit implementations feed actual per-formance data back into the design system, the AI database

USER

INTERFACE -

LOCAL DATABASE

NETWORKINTERFACE

Fig. 3. Workstation -standardized interfaces will allow inter -tool communication and allow tools to be accessed eitherby the network or by the user interface.

manager would compare actual and predicted behavior. Then,the design system will automatically correct simulation modelsto minimize future discrepancies. Adaptive databases will alsoconfigure specific tools for specific designers. Presently, manydesign automation programs support a range of menu and setupoptions, requiring users to spend considerable time and effortconfiguring tools before they can do valuable work. Adaptivedatabases will also learn the user's preference and store setupparameters within specific design files.

Project management tools

The present problem facing the manager of large team effortsinvolves partitioning a design and assigning individual tasks,cross-checking and integrating separately developed pieces, con-trolling access to the project database and managing its content,enforcing design standards and methodologies amongst teammembers, and monitoring the overall progress of the design. It isthe responsibility of managers to establish clearly defined formsfor project documentation. At the outset of design projects, man-agers can delineate those issues that must be addressed in designdocuments; automated design tools can then monitor the com-pletion of documentation forms, preventing design storage untilall forms contain at least some text.

The recordkeeping will be made transparent to the designsystem users. Documentation can be added to drawings, compo-nent -library elements, and software modules as an attribute.Thus, recordkeeping becomes an integral part of the design data-base and users add new designs as they specify components ornew software modules depending on the application. By provid-ing on-line documentation aids, automated tools can subtly forcedesigners to record their activities and explain their design with-out interfering with the productivity and creativity that futuredesign automation tools promise.

Below is a list of some of the automated management toolsthat will be incorporated into the design system:

Automatic separation of information according to type, versionand variant

Controlled access of the information

A guaranteed audit trail for software for all information (whatchanges were made, by whom, why, when)

Document grouping to form libraries

Version and release control routines

Automated PERT, CPM, and GERT modeling routine

Report generation


Transparent recording of design and analysis transactions

Electronic mail utility

Capability of obtaining printout or plot of all relevant man-agement and design documentation

Aside from the mentioned tools, there will exist managementtools that will automate the tracking of product status and asso-ciated warning tools that will continuously compare activity sta-tus in real-time against present activity parameters, alerting theproject manager when activity levels (such as time to complete)exceed the parameters.

Superminicomputers and theautomated software support environmentThe superminicomputer will primarily be used for softwaredesign and development purposes via an automated supportenvironment (aside from hosting secondary storage, the projectcontrol database system, and project monitoring and controlprogram).

The Department of Defense's Ada® program is a commonprogramming language that can serve as the basis for a commonenvironment. The Ada program has adopted the concept of acommon automated support environment into which automatedtools may be conveniently installed. Through a community -wide interactive process, the STONEMAN requirements defini-tion for a system supporting work in the Ada language wasevolved, which defines an Ada programming support environ-ment (APSE) built upon common interfaces and data represen-tations for automated tools.

Due to the DoD's commitment for Ada, I believe theSTONEMAN requirements presently offer the baseline conceptsfor tommorrow's future host software development environment(see sidebar on APSE). The purpose of the APSE is to supportthe development and maintenance of Ada application softwarethrough its lifecycle. It will also promote portability of both userprograms and of the software tools within the APSE. The APSEadapts a host/target approach to software construction. That is,a program which will execute in an embedded target computerwill be developed on the superminicomputer (host) which offersextensive support facilities. The APSE will provide a well -coor-dinated set of useful tools, like the workstation tools, with uni-form tool interfaces and with communication through the centralproject database, which acts as the information source and prod-uct repository for all designers on the network.

The APSE contains certain characteristics of the designnetwork itself, that is it will be an open-ended system. This willfacilitate the development and integration of new tools and itwill permit improvements, updates, and replacement of tools.The APSE will support at a minimum:

Text manipulation

Requirement specification

Overall system design

Program design

Program verification

Project management

Cross development fortarget system

Documentation system

Project control system

Configuration control system

Measurement tools

Fault report system

Libraries

Object importers (to supportother languages, likePASCAL, FORTRAN)

User interface

The design system will contain an easy -to -learn, easy -to -use,powerful, consistent user interface with rapid and complete dataaccess capabilities. The user interface will support a wide rangeof design styles via the use of artificial intelligence. These interfa-ces will "learn" user's needs and habits. The design system couldthen track designers progress along learning curves and providecustom-tailored command -entry formats as individual users be-come accustomed to tool characteristics. Such intelligent inter-faces will also learn user's habits, recording and providing menuitems only for most frequently used and forgotten commands.Moreover, the design system would calculate a time -dependentregression factor for each user, automatically compensating forthe effects of time spent away from specific tools. The systemwill also keep track of what project was last worked on, andwhen, and will automatically return to the last -referenced designstep.

A graphics screen and keyboard will be the primary userinterface, with a touch -screen mode as an option for some opera-tions. Multiple windows may be opened on a single screen,moved about, and enlarged or shrunk as needed. The systemworkstations will contain context -oriented help functions thatwill provide guidance and explain errors.

Benefits of reconfigurable architectureOne of the major characteristics of future engineering design sys-tems will be inherent flexibility due to a standardized reconfigur-able modular architecture. Major benefits can be obtained: tech-nologically independent tools, technology insertion, and the abilityto configure the design system according to the design applica-tion and phase.

With the present effort of standardizing microcomputerbuses, the hardware architecture of most workstations will allowthem to add on new tools and capabilities via the use of plug-inmodules in an expansion chassis. Also due to the present accep-tance of a de -facto operating system standard, UNIX, devicedrivers that will be needed in order to add new tools to a work-station would be readily available. An operating system standardwill promote design tool portability.

Aside from the standards mentioned above, most impor-tantly there will exist design tool standards that define the meansfor transmitting data from one tool to another. Such standardswill allow tools to be highly integrated and free designers toinvestigate alternative tool sources.

Presently these types of design tool standards are not avail-able but, through cooperation between workstation vendors,instrumentation vendors and silicon foundries, standardization atthe design description level will materialize. With these types ofstandards, a workstation can be easily configured to behaveeither as an IC -design station, PC -board design station, or ahardware/software development system.

The tools supported by the APSE, will be portable to otherhost computers due to the KAPSE. The distributed processingafforded by an LAN will also provide a measure of obsoles-cence. Presently, LANs are gaining wider industry acceptanceand through the efforts of the IEEE -802 committee and theInternational Standard Organization, LANs will be the standardinformation bus of future automated design systems. Newerworkstations or superminicomputers can be added to replaceolder ones as required with minimum hardware and softwareimpact. The other benefit of a LAN is that, with the use of


Enrique Melendez joined RCA in 1977 as an engineering co-opstudent while attending Northeastern University. He earned aBSEE degree (magna cum laude) from Northeastern University in1981. Enrique enrolled in RCA's Corporate Graduate Study pro-gram and received his MSEM from Northeastern University in1982. He will be awarded his MSCS from Northeastern Universityin June 1984. Since joining RCA, he has worked on microproces-sor hardware, software designs, and IR&D projects in the ATEfield. Presently, he is working on LAN communications softwarefor the next generation of ATE.Contact him at:RCA Automated SystemsBurlington, Mass.TACNET: 326-5075

gateways, the design system can be tied to other automateddesign systems that may exist at remote sites.

ConclusionAs new automated design tools become available, it is importantthat design engineers study their respective applications to seewhat tools they may require. Likewise, after examining thecomplete design cycle requirements, design managers shoulddecide on tools that will not only fulfill the design and develop-ment requirements but also the project management require-ments. In this context, the future design tools must be easy touse, configurable, and must have standardized (network, user,database) interfaces. By incorporating this flexibility, the designteam will be able to explore and create a class of proprietarytools. This is an important factor if, as predicted, CAE toolssaturate the market. Unless a company has developed its ownproprietary tools, they will be forced to compete with the sametools that its competitors have. As a result, the creativity andinnovative use of these tools by design teams will play in thefuture an even more important role.

References

I. L. Badagliacca, "New Tools Build Microsystems," Computer Design (January1984).

2. J. Marshall, "Integrating the Design Effort," Mini -Micro Systems (April 1982).

3. M. Marsh, "Designing PC Boards in the '80s with CAD/CAM," Design News

(February 6, 1984).

4. J. Borrel, "Engineering Workstations Meet Demand for Individual DesignNeeds," Digital Design (December 1983).

5. W.J. Kitchen, Jr. and L.D. Sikes, "VHSIC/VLSI Capabilities for DoD?" Defense

Electronics (February 1984).

6. D. Babcock, J. Williams, and D. Saliba, "Work Center Integrates ADA Equip-ment Needs," Military Electronics/Countermeasures (January 1983).

7. R.H. Katz, "Managing the Chip Design Database," Computer (December 1983).

8. H.O. Lubbes, "The Project Management Task Area," Computer (November1983).

9. Department of Defense, "Requirements for Ada Programming Support Environ-ments, 'Stoneman- (February 1980).


C.A. Burton

Choosing a CAD system

RCA Government Communications Systems realized that beforethey could use a CAD system effectively they had to make theright buying decision. Here's a look at that process.

Combine the wide variety and increasedcomplexity of work being done at RCAwith the intense competition in the elec-tronics industry. The result is that RCAmust use every advantage available to stayin the mainstream. One of the more impor-tant recent developments in the area ofnew technical design tools has been thecomputer graphics design terminal. To staycompetitive, RCA must implement the useof computer -aided design (CAD) terminalsin the areas of engineering design, analysis,and drafting. But the first question to beanswered is: "Which system to buy?"

RCA Government Communications Sys-tems (GCS) recently went through an exten-sive benchmarking procedure for the selec-tion of a mechanical design turnkey CADsystem. Outlined below are some of themajor areas we investigated with each CAD

Abstract: RCA GovernmentCommunications Systems, in an effort tostrengthen and upgrade their computer -aided design capabilities for mechanicalengineering, recently completed abenchmarking project for the evaluation offive vendors' systems. The techniques usedare described here in general termsbecause they could apply to otherengineering equipment evaluations.Sidebars throughout the article givespecific details on the actual workevaluation and equipment at GCS, leadingto the selection of Auto-Trol CADequipment

©1984 RCA CorporationFinal manuscript received April 10,1984.Reprint RE -29-3-5

company under consideration. This infor-mation should help increase awareness ofthe important issues during the selection ofa computer graphics design system. Al-though these techniques were used in theselection of a CAD system, these sameinvestigation techniques could also be ap-plied to other tradeoff studies in advanceof a major equipment purchase.

BeginningThe first task undertaken was to outline indetail the long-term and short-term require-ments of the system to be acquired. Thevariety of turnkey systems available todayis astounding, but few are suitable for everyapplication. Each product will usually showa specialization or concentration in one orseveral areas of expertise (that is, PCB


GCS's mechanical CAD choice

A benchmark of mechanical CAD systems was per-formed from September 1982 through February1983. Originally, two vendors were to be investi-gated, but before the benchmarking procedure gotinto full swing, the view was expanded to includefive additional companies. The company chosen tosupply mechanical CAD systems to GCS was theAuto-Trol Technology Corporation.

The first Auto-Trol equipment for GCS was deliv-ered in mid -September of 1983. RCA's system con-sists of three Advanced Graphics Workstations(AGW), a magtape unit, a TI line printer, and anHP7580B plotter. Three engineers and four drafts-men/designers have been trained to date on thesystem. Training has included two weeks of basicoperator's training, three days of finite -elementmodelling, and three days of Fortran interface use.

Currently, two terminals are operated on a two -shift basis for drafting production work. Productionwork on the terminals was started in November of1983. The third terminal is for engineering analysis,development, and system management.

A variety of projects have been handled includingtwo-dimensional detail drawings, three-dimensionalmodelling, flex cable development, finite -elementanalysis, mass properties analysis, and clearanceanalysis. Further development work is planned infinite -element modelling and analysis, three-dimen-sional modelling, and mass properties calculations.In addition, factory automation and MIS interfacesare planned for development. The emphasis will beon integrating all areas of design, analysis, and pro-duction through the use of the Auto-Trolworkstations.

design, IC design, two-dimensional draft-ing, technical illustration, and so on).

In addition, different companies are atdifferent levels of development in theseareas of expertise. For example, an exten-sion of mechanical CAD is the develop-ment of numerical control interfaces. Somecompanies have expanded into this area,while others have limited their commit-ment to strictly drafting and documenta-tion functions. If factory automation inter-faces are unimportant to you, or if youintend to develop them in-house, this willaffect your selection of CAD vendors tobe considered.

Other applications that should be con-sidered before initiating an investigationare finite -element modelling packages, flat -pattern development, solids modelling, pat-tern nesting, and so on. All possible appli-cations that may be required of the system,short-term and long-term, need to be ad-dressed prior to focusing on several vendorsfor the investigation.

Some systems will be good at acceptinguser -developed automated routines that cancustomize the system. Others may havehardware that allows the use of manythird -party -developed software packages toenhance the system. Therefore, the amountof in-house development to be committedto the CAD area is an issue that must bedecided in order to focus the benchmarkinvestigation properly.

If there is more than one type of appli-cation for CAD in an area, there are twooptions for filling those needs with CADterminals. One is to choose a system thatcan satisfy both needs. By doing this, train-ing needs are simplified, a working knowl-edge of the system can be shared, and

direct interfaces for transmitting informa-tion are readily available. The result ofsuch a decision is that one vendor willprobably not be the best for both applica-tions. Therefore, capabilities in one pack-age may be sacrificed in an effort to gainthese advantages in system commonality.In such situations, more commitment tosystem customizing, whereby automatedroutines are developed specifically for in-house use, may be required.

The other school of thought seeks tochoose the best system for each applica-tion and, if necessary, develop interfacelinks in-house or use translation languagessuch as IGES, an all-purpose graphics lan-guage, to allow the two systems to inter-face. GCS has chosen to purchase the sys-tem we believe is the best for a particularapplication and to handle interface opera-tions separately. GCS decided that con-tending with the added complexity of train-ing needs was worthwhile, in order tohave the best package in an applicationarea. The decision became whether or notto devote effort to developing interfacecapabilities or to customizing the system.

The benchmarkOne of the most important aspects of thedata collecting/decision-making process isthe execution of a typical application byall competing vendors. The best test of asystem is a significant exercise as close aspossible to the customer's actual work.

This benchmark project was a require-ment for considering a company as a pos-sible supplier of CAD terminals to GCS.Although a lot of companies do not wantto spend the time and effort a true bench-

mark effort takes, the response to GCS'sinvitations to compete in the benchmarkcompetition was very good. As soon asone company agreed to do the benchmarkproject, others joined suit and made thecommitment to show their capabilities byexecuting a controlled, timed benchmarkproject. The CAD industry is very com-petitive and with a big name like RCA,the implications are enough to really inten-sify the competition.

In choosing a benchmark project, it wasimportant to consider the type of work,the time involved for the vendors to com-plete it, and the availability of data forcomparison to manual effort. Much of themechanical work at GCS involves the de-sign of sheet -metal housings and drawersfor electronic equipment. Therefore, thistype of design was chosen for the bench-mark (Fig. I). Care was taken to includeas many typical aspects of the work doneat GCS as possible. The mounting of elec-trical components is a common occurrencein designs, and therefore was included inthe requested benchmark design. How eachvendor handles the creation and implemen-tation of standard components in a designwas watched carefully for its contributionto better efficiency through the use of CAD.

The benchmark project used by GCS totest mechanical design systems was onethat had been done previously by a designeron a drafting board and, therefore, thetimes required to complete the design onthe drafting board were known. Thesetimes, along with the CAD vendors' esti-mates of productivity enhancement ratios,allowed an estimate of the time that wasrequired by each vendor to complete thedesign on their company's system.

Burton: Choosing a CAD system 29

(a)

Fig. 1. The benchmark project selected byGCS before buying a mechanical CAD sys-tem was a folded sheet -metal chassis withswitches, circuit breakers, connectors, andterminal boards mounted on the outside.Stenciling was required on the front panel.The design went through three phases: (a)First, a sketch of the design was preparedas a guide to the vendors. (b) The CADvendors then created a three-dimensionalmodel of the assembly. (c) A further exten-sion of the design that would be useful inmass properties calculations and for tech-nical illustrations is a shaded solid modelthat is constructed from the three-dimen-sional wireframe model.

(b)

Altv.e FASTwe-1 CHASSIS

U51.14"ri sctews4 msazi4 )4),r,

111 CHASSIS

(c)

For each competing vendor to have anequal chance in completing the benchmarkproject, all the pertinent data (general guide-lines and requirements, a parts list, a sketch,and component specification sheets) wereaccumulated into one package and thor-oughly reviewed for completeness. Thissame package of information was given to

all vendors involved. Every effort was madeto make available information comparableto that which is available to a designer atGCS.

One specific contact from GCS wasassigned to follow the benchmark projectand to field any questions that arose con-cerning the design. This simulated the engi-

neer/designer relationship that normallyexists on a project and allowed a betterview of how projects would actually pro-ceed on a normal basis using the system.

As much of the actual work was ob-served as was deemed practical. Each ven-dor usually did some preliminary workahead of time, such as creating patterns


Geometric construction capabilities

The following is a list of features that contribute tothe construction capabilities of a system. A briefexplanation is given of each for those unfamiliar withthe language of CAD.

Separate Drawing/Model Modes-This allows sep-aration of the three-dimensional model from drawingentities such as notes and dimensions.

Independent Construction Plane-Also referred toas working coordinate system, this allows entities tobe added using more than one coordinate system.

Number of User -Defined Views-A system will usu-ally define a certain number of standard views, butwill allow the user to define additional ones, hence,user -defined.

Number of Working Views-This is the number ofviews that may be displayed at one time.

Limit on Model Size-Most systems allow the filesize to be limited only by system disk space, butsome have smaller limits.

Use of Grid-The use of a grid causes entities to"snap" to grid points and therefore adds a guide forlocating entities quickly, yet accurately.

Number of Levels-Levels can be likened to sheets

of Mylar, containing entities, which can be viewed inany combination.

Color Definition-Color usually is defined by levels,component type, or entity number, or a combinationof these.

Offset Capability-This provides a quick method ofcopying geometry that needs to be duplicated aspecified distance from the original entities.

Surface Generation-Surfaces are an extendedapplication of three-dimensional modelling. It is par-ticularly useful in checking interferences.

Cross-sectioning-If a particular piece of geometryis surfaced, a spline can be generated automaticallyto represent its cross-section at a certain depth.

Automatic Hidden -line Removal-Used in three-dimensional modelling to make views of objects thatlook like the solid object would look.

Dynamic Capabilities-Dynamics allows the user tomove objects interactively on the screen.

Edit Versus Reference Levels-A method of identify-ing a set of entities on specified levels, which caneither be tagged as changeable or not.

and storing them, and practicing the tech-niques to be used in the design. Most ofthe actual design work in each case wasobserved live.

To compare the vendors on a commonscale, all the circumstances that could affectthe completion times and quality of thedesign work were documented. Such cir-cumstances included the experience of theapplication engineer executing the design,his (or her) background, the type of ter-minal used in completing the design (ver-sus the type of terminal that would bepurchased), the load on the system duringthe benchmark work, and the amount ofactual work completed. Some of the ven-dors did not complete all of the requestedwork, but were able to complete enoughto allow a realistic determination of theactual time it would have taken. Othersdid more than the required design, such asadding extra notes or providing an isomet-ric view of the design. All these factorswere noted and taken into account. Aneffort was made to adjust the completiontimes to reflect the effects of these factors.Although this gives a somewhat subjectiveresult, it is more realistic than ignoring theunequal circumstances altogether.

DemosAnother method of seeing what a particu-lar system can do, one which is a favoriteof most CAD vendors, is that of demon-strations, or demos. Most likely the com-panies will have certain "canned" demosready to show. These are usually impres-sive, but they sometimes will give a falseimpression of the ease of accomplishingcertain tasks with the system. Companiesrarely show any demo that would revealsystem limitations or awkwardness. To seea more realistic view of the particular sys-tem, a method was employed to force thevendor into demonstrating the system live.

Several small, typical design tasks, whichnormally could be done in an hour or twoon the drafting board, were outlined. Timeswere arranged to have these demonstrated,but the outlines were not given to thevendors ahead of time. This prevented anyadvance preparation and forced a live, real-istic demonstration of the system. Somesoftware bugs, which were convenientlyavoided in a canned demo, surfaced in alive demo such as this. All CAD softwarepackages have some bugs, but it is impor-tant to make sure they are limited, areones that can be worked around, and that

the vendor, when made aware of them,will respond to correct them. When suchbugs surfaced, the vendor was asked tocheck them out and tell us whether theywere bugs they were aware of and whattheir action concerning them would be.This allowed a first-hand view of the ven-dor's customer response.

Talking to usersWe talked to users of the different sys-tems-this proved to be an interesting andsurprising part of the system investigtion.First contacts can be made through thevendors, but by virtue of what they aretrying to accomplish, they are only goingto supply the names of contented custo-mers. The best source of vendor users is acomplete list of the user's group membersfrom either the vendor or the user thevendor referred. Then, by a random selec-tion of users, a more honest impression ofthe vendor can be obtained. A lot of timesthe users will be divided into groups catego-rized according to industry. This allowsidentification and contact with users withsimilar applications.

Questions posed to the users included:


Drawing/dimensioning capabilities

An important aspect of any CAD system is the easewith which drawings can be made. Listed below arefeatures that contribute to that ease.

Automatic Dimensioning-Entities can be identifiedfor dimensioning and the system automatically mea-sures the distance and inserts it in the dimension.

Automatic Update of Dimensions-If the entities thatare dimensioned to are moved or changed, thedimension will automatically be updated.

English -to -Metric Conversion-This is the ability tochange a drawing that was dimensioned in one sys-tem of units to the other.

Geometric Tolerancing-The capability to dimension

using geometric tolerancing methods is becomingincreasingly important.

Calculation of Tolerance Buildup-Automated rou-tines can be developed to calculate the tolerancebuildup in a certain area.

Exploded Views-A certain area oftentimes will needto be blown up in an exploded view to show detail.Automatic capability helps greatly.

Text Fonts-The kinds and number available help toenhance the appearance of a drawing.

Number of Line Weights-Different thicknesses oflines often need to be shown. It is a convenience tobe able to assign them automatically.

Enhanced productivity features

Features that add to the speed and efficiency of thesystem are the ones which allow the system to payfor itself. The features that GCS noted as enhancingdesign work with CAD were the following.

Menu-driven-Menu-driven systems allow the userto make choices from a displayed menu and there-fore are very easy to use.

Display Tolerance-Even though curves and sur-faces may be defined exactly, rough approximationscan be used for display to speed up repaint of thescreen.

Interruption of Commands-If a command takes along time to execute, and is initiated by mistake, it isbeneficial to be able to interrupt it.

Measuring Capability-This is the ability to have thesystem measure distances as would be done on adrawing.

Calculator Mode-Most systems will allow the userto calculate values as on a calculator.

Sketch Mode-Control of the cursor to sketch freelyon the screen adds flexibility to the system.

Command Log-A record of all operations executedare stored in a file and can be "replayed" as with atape recorder.

Programming-The ability to interface to thegraphics capabilities through programs adds toautomating certain routine procedures.

On-line Documentation-Time is saved when com-mand usage information is provided on the systemas well as in manuals.

Plot Queue-A plot queue stores plot files so that agraphics terminal is not tied up when plotting isbeing done.

Tablet Symbol Recognition-This is a type of com-mand execution where the system will recognizesymbols drawn with the cursor and execute the cor-responding command.

Why did you select this particular vendor?

What is the response of the vendor to ser-vice calls?

Have a significant number of bugs beenencountered in the software?

How big a system do you have?

How long have you had the system?

What types of applications do you runwith the system?

Most users that were contacted were wil-

ling to discuss any issues concerning theirsystem. A few questions were all that werenecessary to start the flow of information.CAD/CAM is such a new field that thoseinvolved are usually excited and proud ofwhat they are doing. Much insight wasgained by talking to users of the compet-ing vendors' systems.

Some of the best sources of CAD usersare within RCA itself. Different divisionsof RCA are using systems supplied byApplicon, Auto -Trot, Medusa, Computer-

vision, and Calma among others. The workdone by one division in an area, such asCAD, can and should be used by otherdivisions for the advancement of the com-pany. Some of the other CAD users ofRCA are Astro-Electronics Division, Mis-sile and Surface Radar, Broadcast SystemsDivision, and the Laboratories. Usefulinformation was obtained from the otherCAD users in the company, resulting in asignificantly less difficult decision -makingprocess.


Hardware features

Most of the features readily looked at on a CAD sys-tem are software -related. To ensure the best imple-mentation of the software, good hardware is also amust.

Screen Resolution-This parameter dictates howaccurately lines, arcs, and splines will look on thescreen.

Number of Colors-Most systems can only display acertain number of colors at once.

Size of Buffer-The buffer stores commands if theoperator types faster than the system can executeand executes them in the order entered.

Number of User -defined Menus-Menus are sets ofuser -defined macros (sequences of commands).Usually a menu is created for each type ofapplication.

Remote Diagnostics-Some systems provide diag-nosis of system problems via a phone modem, sav-ing the number of service calls.

Intelligence of Terminal-Some terminals have

computation power that offloads the CPU andenhances performance.

Type of Cursor Control-Cursor control is usuallyaccomplished with a joystick, thumbwheels, adigitizing pen, a light pen, a mouse, or a touchpad.

Separate Numeric Keyboard-A separate numerickeyboard provides for easy entry of numeric values.

Screen Adjustment-To provide a comfortableworking environment, a screen that swivels oradjusts in height is a nice feature.

Separate Alphanumeric Screen-To display thecommands executed, some systems provide aseparate alphanumeric screen, whereas some com-bine it with graphics.

Turnkey-A total system package-hardware, soft-ware, and support-is called a turnkey system.

Stand-alone-A stand-alone system is self-suffi-cient. It has its own CPU and does not affect thework of any other terminal.

Analyzing featuresOnce the actual data collection began, ob-jectivity was the most important qualityfor the investigator to have. Many timesthe opinions of management or the viewsof the company on specific concerns willchange in the course of the investigation.Collecting the information as objectivelyas possible allowed more flexibility whenthe decision process actually began.

The procedure used by GCS in accu-mulating decision -making data on all thevendors was to make note of every featureeach system had. This entailed time-con-suming visits to each vendor, to watchwork in process, or to work on the systempersonally. The benchmark project itselfprovided an excellent mechanism for gain-ing knowledge about a system. During theGCS benchmark, approximately two work-ing days were spent with each vendorobserving the benchmark design in prog-ress. Extensive notes were taken on every-thing imaginable-design approach, hard-ware features, line -font capabilities, cursor -control mechanism, number of colors fordisplay, dynamic capabilities, and so on.After visiting all the vendors the first time,an accumulative list of features was devel-oped. The major features were dividedinto four categories: geometric constructioncapabilities, drawing creation/dimensioning

Cindy Burton is a Mechanical Engineerwho has been with RCA almost two years.She began her work at RCA in theApplicon printed wiring board design facil-ity, working as a printed circuit boarddesign consultant and as a computer gra-phics development engineer. As a result ofher benchmarking effort and the acquisi-tion of three Auto-Trol mechanical designterminals, she has taken on the role ofLead Development Engineer in theMechanical CAD area.

Ms. Burton received her BS in Mechani-cal Engineering from Ohio Northern Uni-versity in 1980. Prior to coming to RCA,she held a position of Mechanical DesignEngineer with Westinghouse Corporationin Lima, Ohio, where she worked on thedesign and analysis of generator andmotor housings for aircraft electrical sys-tems and related applications.Contact her at:RCA Government Communications SystemsCamden, N.J.TACNET: 222-3134

capabilities, enhanced productivity features,and hardware features. Some of the fea-tures noted about one vendor may nothave been investigated with another. There-fore, by accumulating the list of features,second contacts with each vendor weremade in order to clear up all issues withall vendors.

After all the information had been col-

lected and tabulated, it was condensed toa number of issues considered important.GCS had ten major issues that it felt wereimportant in choosing a CAD system. Thethree most important issues were speed ofthe system, geometric construction capabil-ities, and reliability of the system. The nexttwo areas of concern were dimensioningcapabilities, and the ability to customize


the system. Of next importance were sur-facing capabilities, the ease of use of thesystem for casual operators, and hardwarefeatures. The final group of issues involvedavailable training and the aesthetics of theterminal layout. Each group of issues weregiven a different weight between one andfour according to their importance to GCSmanagement in deciding on a CAD vendor.

Each vendor was given a rating fromone to ten for each issue (such as geome-tric construction capabilities, hardware fea-tures, and so on). According to how theywere judged in each compared to the othervendors. These ratings were based on thelist of features accumulated and tabulatedduring the benchmark and demos. Then

the vendors were each given a weightedscore for each issue, which was obtainedby multiplying the one -to -ten ratings bythe weight of importance (a number be-tween one and four) for each issue. Totalweighted scores were then calculated foreach vendor and used as a major indica-tion in determining the best system forGCS's use.

By analyzing the systems in this manner,the information that was collected can beused, even if views or intentions change,merely by choosing different major issuesor assigning different ratings for them. Itmay also be desirable to look at the dataunder several different priority arrangementsfor the issues in order to get a broader pic-ture of how the systems compare.

ConclusionOne of the key factors in being satisfiedwith a decision or a purchase is assurancethat the investigation was thorough, sothat the buyer does not see a better systemtwo months down the line. The wholebenchmark and investigation procedure wasan eye-opening experience. GCS obtaineda broad picture of what is available fromthe turnkey CAD industry market. Al-though the technology itself is changingrapidly, GCS is satisfied that the systemchosen was the best available, at the time,to fit their specific needs.

Q. What do these winners of the 1984 RCA LaboratoriesAchievement Award share in common?

Robert A. DuschlKrishnamurthy JonnalagaddaEdward H. AdelsonCurtis R. CarlsonAlbert P. PicaVictor AuerbachThomas Y. ChenWalter G. Gibson

Thomas F. LenihanWilliam E. BabcockWilliam E. RoddaWerner F. WedamWalter F. KosonockyFrank V. ShallcrossGlenn A. ReitmeierNorman D. Winarsky

A. They all contributed to articles published in the past two yearsby the RCA Engineer.

That's just part of the reason why in the past two years the RCAEngineer has won four awards from the New York Chapter of theSociety for Technical Communication.

We at the RCA Engineer are proud to present RCA's very best engi-neers and writers from throughout the company, and not just at theLabs. Many divisional Technical Excellence Award winners are RCAEngineer authors, too.

Join the winners and write an article for the RCA Engineer.

Upcoming issue themes for 1984 include "Technical excellence- inJuly/August, "Materials science and applications" in September/October, "Imaging technology" in November/December, and"RCA's communications businesses and technologies- in January/February 1985.


W. Carey

The MSR local area network

A local area network of computers, terminals, and peripheralsnow in place at Moorestown has greatly amplified the service ofthe Missile and Surface Radar Development Facilityand the skills of its users.

This article describes the origins of thepresent computer network system associatedwith the Missile and Surface Radar (MSR)Software Development Facility at Moores-town, N.J. The present configuration ofthat facility was described in a recent arti-cle by Liggett.' A systematic, integratedcombination of hardware, software, andfirmware provides a cohesive, reliable, data-

communication network that meets engi-neering communications requirements. Thisnetwork is based on three principles: net-work transparency, virtual -circuit switch-ing, and centralized network control.

BackgroundIn 1980, RCA Moorestown established asoftware development facility (SDF) de-signed to house all the processors involvedin software development (Fig. 1). The com-puting power represented by the machines

Abstract: A growing organization withan increasing number of buildings in theMoorestown area, RCA Missile andSurface Radar faced the problem ofmaking its computer facilities available tothe greatest number of users with the leastamount of access effort. The local areanetwork precludes the need for acquiringadditional computers by making moreefficient use of existing systems.

Terminals operating at 2400 baud areconnected via lines and trunks from thenetwork processor and network controlconsole to the VAX 11/78s, PDP 115,Eclipse, and Novas in the SoftwareDevelopment Facility.

©1984 RCA CorporationFinal manuscript received April 19, 1984.Reprint RE -29-3-6

in the SDF was at that time sufficient tomeet all MSR requirements.

A terminal room that could house asmany as 20 terminals was set up adjacentto the SDF. The number was chosen tominimize both the expense involved in set-ting up the linkages and the waiting timerequired for access to any particular sys-tem. Although the number 20 still holds,unanticipated requirements soon createddoubt as to the advisability of having asingle terminal room servicing all program-mers.

First of all, as the number and varietyof projects increased, locating all involved

-Fig. 1. MSR's Software Development Facility is dedicated to supporting the devel-opment and test of software systems.

people within reasonable walking distanceof the terminal room became impossible.MSR is a widely dispersed organization,and programmers and engineers who usethe system for development work are sta-tioned in all parts of the complex. Someusers would find themselves walking asmuch as a quarter mile through inclementweather to get to a point from where theycould access the computers.

Furthermore, many in-house projects re-quire operations to be performed on dif-ferent systems. A user might need oneterminal for part of a particular applica-tion, and then a different, and frequently


VAX 11/78032 PORTSNODE 14NODE 28

PDP 11/7020 PORTSNODE 26

PERKIN ELMER32208 PORTS

AUTO DIAL3 PORTS

ES

NETWORK

CONTROL

NODE 100

CONSOLE

MULTIPLEXERBLDG 10110 TERMINALSNODE 15

TRUNK 29MULTIPLEXERBLDG 1054 TERMINALSNODE 29

MULTIPLEXERBLDG 10824 TERMINALSNODE 12

Fig.2. This network configuration handles MSR's tremendous variety of digitalprocessing tasks.

busy, terminal to complete his task. Thefrustration factor could be excessive.

In some cases a program must be enteredand compiled on one host, then transferredto a different host for debugging and/orexecution. Transferring of the programsbetween hosts was done by magnetic tapeor floppy discs, and necessarily involvedgoing into the SDF room and physicallyhandling the hardware, sometimes withcatastrophic results when attempted by per-sonnel unfamiliar with the system. Thegrowth of the number of personal comput-ers and home terminals had also spurredrequests for access to the mainframes overphone lines. With the dedicated terminalapproach, such access had to be denied.

By 1983, the use of digital processingon various Missile and Surface Radar proj-ects had multiplied. Not only had the num-ber and variety of computing systems-micro, mini, and midi-expanded dramati-cally, but hardware and software documen-tation was now being included on thediscs and tapes associated with the sys-tems. Thus, in addition to the conventionalusers, secretaries and stenographers wereextending use of the system for word andtext processing. MSR faced the choice ofeither greatly expanding the number ofprocessing systems, or making far moreefficient use of the existing systems. Net-working was an attempt to implement thelatter approach.

Selection criteriaMSR issued a request for proposal de-scribing our problem and a general ap-proach to the solution to a dozen potentialsuppliers. Our hardest task was to decidewhich of a number of technically excellentreplies was the best. To that end, MSRestablished a set of selection criteria thateach proposal had to meet to qualify forfurther consideration.

Basically the proposed arrangement hadto provide

A means for accessing computers in theSDF from anywhere in the MSR com-plex.

A means whereby a single terminal couldaccess any of the resources in the SDF.

These capabilities in a manner that madethe network transparent. The fact thatthe switch was between the terminal andthe computer should be undetectable.

These capabilities without the use of mech-anical switches. The network must estab-lish "virtual" circuits; that is, circuitswhich have no physical existence, butwhich exist only in the memory of thenetwork.

Control over the network from one cen-tral location, which could permit or denyaccess of specific terminals to selectedresources.

The ability for data to be interchanged

between computers without the use ofexternal media, and at the command ofthe user (this capability might be limitedto ASCII data with no penalty).

The capability of dumping data to remoteprinters.

Some communications capability via com-mon carrier, as opposed to leased tele-phone lines.

The ability to dial out from a terminal ora computer and to treat incoming callsas coming from terminals.

Selected systemAlthough none of the proposals fulfilledevery requirement in every detail, the sys-tem made by Digital Communications Asso-ciates, Inc. of Atlanta was selected. Theoriginal configuration of the system isshown in Fig. 2.

The network processor and the networkcontrol console are physically located inthe SDF, near the computing systems withwhich they interface. The role of the net-work processor is to accept traffic fromlines and trunks and, by examining desti-nation tags on the incoming information,send it to the correct line or trunk.

In the current context, a trunk is acommunications link that can carry infor-mation destined for two or more ports.Hence, the link between the network pro-cessor and the PDP-11/70 is a trunk, sinceit can carry information for up to 16 portson the PDP-11%70. The links between thePerkin-Elmer and the network processorare not trunks, since each is dedicated to aspecific port on the Perkin-Elmer.

Installed in the VAX and the PDP-11/70 are boards, built and supplied byDCA, which create 16 ports on the hostcomputer. That board communicates withthe network controller via a single 9.6kbps trunk. Two such boards are installedin the VAX and one in the PDP-11/70.

Three pairs of modems were includedin the system for long distance (up to aquarter mile, in this early system) com-munication with remote multiplexers. Thesemodems supply full -duplex communicationlinks over distances that digital signalingcould not handle. Each one accepts digitalinput at one port, converts it to phase -en-coded information, and ships it to the trunk.It also accepts phase -encoded informationfrom the trunk, converts it to digital levels,and sends it to its local host, which iseither the network processor or a statisticalmultiplexer.

The statistical multiplexer is a purelydigital device which interfaces with up to


32 terminals, and allots each one a portionof the multiplexer's total bandwidth, asdetermined by that terminal's immediateneeds relative to the rest of the terminals.Through modems, the network processordrives remote statistical multiplexers, eachof which interfaces with terminals that arerelatively local. The network processor alsointerfaces with the RACAL-VADIC tele-phone line interface unit which (1) acceptsincoming calls and routes them to thenetwork processor, and (2) provides anout -dialing capability so that a terminal ofa computing system may dial out to aremote location.

Installing the systemInstallation of the computer network sys-tem was performed by a team of bothRCA and DCA personnel. The systemwas installed in two steps-the first stepconsisting of testing each interface withoutactually "switching to traffic," and the sec-ond step consisting of the final installationsand a night-time switchover so that userswere relatively unaffected by the change.

As expected, most of the problems arosein the first step. The majority of theseproblems came about because of MSR'sunfamiliarity with switching systems of thisnature. We found for instance that, in

every case, the operating systems that wereintended to work with the switch wereimproperly generated and had to be redone.These problems and their solutions occu-pied the better part of a week but, all inall, the testing went rather smoothly. Theproblems associated with running all thenecessary cables occupied most of the nextweek, and these problems were of coursestrictly internal to RCA. DCA had nopart in them.

Operational experienceThe very first user reaction to the termi-nals on the switch was one of resistanceand grumbling because these terminals, inorder to increase their total number, werebeing operated at 2400 baud rather thanat 9600 baud.

However, the engineers rapidly discov-ered that the convenience of a terminal afew steps from their desk was far greaterthan the (mostly illusory) value of the fourtimes greater data rate, which could onlybe used in one direction and at infrequentintervals. The switched terminals were soonused far more heavily than those in theterminal room.

Whether productivity has measurably

increased is a debatable point, since wehave never had an adequate measure ofprogrammer or engineer productivity. Butthere can be no doubt that morale hasincreased, simply because use of the sys-tem is more convenient, and because frus-trating delays in gaining arrPss to a termi-nal are far less frequent.

Capabilities and expectationsin a product developmentEnvironment

The initial concept of the Software Devel-opment Facility was as a tool for the crea-tion and maintenance of deliverable soft-ware. An ancillary application was for thedevelopment and support of a limited num-ber of utility programs to be used primar-ily by software engineers for the purposeof expediting the development of software.Such programs as special-purpose editors,disc and tape utilities, and indexes of exist-ing software fall into this category. Theinitial concept expanded rapidly and almostfrom the first day of operation, additionalcomputing loads have been placed uponthe facility. The number and type of addi-tional tasks are listed below.

Computer -aided design

With the use of microcomputers in allfacets of MSR products, we find ourselvesfrequently tailoring a microcomputer instruc-tion set to a specific application. The hard-ware configuration designed to implementthe instruction set is fed into a software -implemented logic breadboard via a spe-cial algorithmic language, and the programtests the configuration against a standard.Any errors in the design are flagged, giv-ing the engineer an opportunity to correctthem before the hardware is built.

Cross -assemblers, compilers,and linkers

A variety of these have been made availa-ble. In general, they implement their as-signed function while running on one ofthe larger machines and create output tobe used by a small minicomputer that isnot a part of the SDF, or by a microcom-puter that is a part of deliverable equipment.

Simulation

An array of simulators has been incorpo-rated into the available software of theSDF, ranging from simple, but nontrivial

digital filter simulators to an expanded ver-sion of the General Purpose SimulationSystem.

Documentation

MSR is moving toward the preparation ofall documents on word or text processors.This includes proposals, reports, specifica-tions, and even this paper. Since dedicatedword processors are scarce, the tendencyhas been to obtain software which runs onone or more of the SDF machines, andcreate and edit the documents from a ter-minal. It is becoming more and more com-mon to see a secretary or a stenographersitting at a computer terminal and using itas a typewriter. By the summer of 1984,most secretaries will have desktop termi-nals that will be used far more often thantheir typewriters.

Education

Not the least of the SDF applications iseducating both MSR personnel and othersin the use of computers. In-house courseson programming are provided for MSRpersonnel, and the SDF computers arealso used for educating special groups suchas the Boy Scouts.

Each application has its own peculiarsoftware and user requirements. Secretariesdo not use complex operating procedures,simulation users do not need the detailedoperations involved in a compile and linktask, and we certainly cannot expect BoyScouts to be sophisticated users. Hence,support effort directed toward the SDFand its machines has increased along withthe work load on the system itself.

A measure of the success of the SDFand its underlying philosophy is the factthat as of now, software users of the SDFalmost always outnumber software devel-opers. The application of computers to thedesign of MSR products is still growing,but is already greater than the effort re-quired to design the software under use.With all the capability we now have, weare just beginning to comprehend that wehave made only one small step in thedirection of realizing the needs that willbecome apparent in a very few years. Theapplication of computers and digital logicis still increasing. If MSR is to maintain itsposition in its chosen field, we must con-tinue to expand our capabilities in digitalprocessing, both as it pertains to deliver-able products and to supporting our engi-neers in all of their manifold specialties.

One thing we are striving for, and prob-

Carey: The MSR local area network 37

VAX11/780

VAX11/780

POP

11/70

CK

SI M

5 9.6 KB I r P ST

I SI

9.6 KB I

355

05

is9.6 KB IZ

ST

ST

SI N..

9.6 KB

AUTO DIALAUTO POS

POP

11/24

CF -2

4

/ IST

AS

Sr

STpI

J/ AS

sT

M

I ST

ST

ILF2.NETWORK

DES. SYS.

TO BE ADDED BYSUMMER 1984

As

I

P ST

M ST

ST

M ST

ILF-4

Fig. 3. MSR Software Development Facility Local Area Network, Buildings 101, 108,138.

SYNCHMODEM

4 B KB I SYNCH

SYNCH 1 9.6 KBMODEM

SYNCHMODEM

9.6 KB

MODEM

SYNCHMODEM

SYNCHMODEM

9 6 KB

SYNCHMODEM

LDM4.8 KB

SYNCHMODEM

LDM9.6 KB

192 KB

LDM

SYSTEMCONSOLE

LDM9600

LDM

SYNCHMODEM

BAUD

8600

LDM

REMOTE LINEPRTG. SYS.

H120-1616STAT MAX

SYNCHMODEM

BAUD

9605

SYNCHMODEM

BAUDSYNCHMODEM

120-1616STAT MAX

120-1616STAT MAX

110-808STAT MAX

082 16

108216

101.2 16

138 3 8

1 120-1616STAT MAX

120-3232STAT MAX

120-1616STAT MUX

120-1606STAT MUX

H120-1606STAT FAUX

138.216

138.2 32

138-3

/6

FAATC

4

WASH. D.C.1 PRINTER

/4 2 TERMS

2 PRINTERS4 TERMS

1 PRINTER2 TERMS

ably the most important thing for the nearfuture, is a single operating system lan-guage that will be common to all ourcomputers. The two prime candidates atpresent are UNIX and Ada. Given a singleoperating system language, the next desid-eratum is a unified set of programminglanguages, compilers, assemblers, linkers,translators, and so on. Given that manyfeatures of these languages are stronglyhardware dependent, we have a great dealof work before us to even define the require-ments, let alone implement them.

At present we have, to a small extent,unified and centralized our hardware re-sources. We would like, some time in thefuture, to also unify and centralize thehuman resources required to make SDF atruly general purpose tool. It would bedesirable to have a pool of specialists avail-able for the purpose of entering large databases, of monitoring complex compilationsand linkages that require human supervi-sion, and of generally relieving the soft-ware engineers of most of their currenttime-consuming tasks.

A wide variety of currently nonexistenttools will be required to implement ourplans. More user-friendly software, espe-cially for the non -expert who currentlymakes up the majority of the users, is a

first requirement. In addition, general-pur-pose report generators, forms generators.code -to -flowchart translators, documenta-tion aids, indexes to tools, syntax checkingsoftware, parsers, and many other itemshave already been requested and are underconsideration.

Of prime importance in this category isa set of configuration management utili-ties, so-called programmer's workbench pro-grams. These will help us to maintain acomplete historical audit trail of all devel-opment of, and modifications to, softwareas it evolves. In a world where softwareavailable at a user's terminal can be mea-sured in millions of lines of code, such aservice will be indispensable.

Other and more exotic requirements thatwe hope to realize in the fairly near futureinclude electronic mail service and con-nection to a broadband net that will allowus to interconnect with other computersystems in geographically remote areas, in-cluding probably ARPA (Advanced Re-search Projects Agency in Washington,D.C.) and several RCA dedicated networks.

Conclusion

The network has proved its flexibility andreliability through four expansions and a

William Carey is a Senior Member of theEngineering Staff in the Special SoftwareDevelopment Test Systems group atMissile and Surface Radar. He receivedthe AB and MA degrees from ColumbiaUniversity. Since joining RCA in 1976 hehas been involved in the AEGIS SystemSubelement Test System (ASSETS) and theAEGIS Near Field Antenna System Testfacilities (ANFAST I and ANFAST II) devel-opment. He is also the responsible engi-neer for the MSR Local Area Network.Contact him at:Missile and Surface RadarMoorestown, N.J.TACNET: 253-7266

major relocation, which involved movingthe entire SDF to a building three milesfrom its original location. In this lattercase, the network was taken down, moved,reinstalled, reconfigured, and brought backup in less than twenty-four hours

The Summer 1984 configuration of thenetwork is as shown in Fig. 3. The FAAPomona, New Jersey and Washington,D.C. nodes are, respectively, 50 and 250miles from the network processor itself.The traffic is about four times that of theoriginal system, and still the network showsno sign of overloading. We are success-fully operating remote printers, and han-dling dial -in traffic on an almost continu-ous basis.

The only disappointment so far is thediscovery of a fundamental incompatibil-ity between the out -dial hardware and thenetwork as currently engineered. We havenot been able to successfully implementthe out -dial feature. Fortunately, this is afeature of very limited application.

Reference

1. C.K. Liggett, Jr., "MSR Centralizes SoftwareDevelopment Activities in New Computer Center,"RCA Engineer, Vol. 29, No. 1, pp. 37-39 (January/February 1984).

38 RCA Engineer 29-3 May/June1984

S.G. Handman

An attached processorprovides calculation speed and accuracy

This powerful new addition to the Corporate computer servicewill speed the calculation of complex and mammothmathematical problems.

Engineering and scientific computationsmay impose special requirements on a com-puter system-a high rate of throughputfor repetitive calculations, manipulationsof large arrays of data, precision in cascadedcalculations, accommodation of numbersover a large dynamic range, and reasonableease of programming. These classes of cal-culations have been limited by the tradi-tional, general-purpose computer architec-ture. Calculations on arrays of 1024 datapoints are routinely required and fast -Fourier transforms on 16,000 data pointsare not uncommon. To a large extent,algorithms for scientific calculations con-sist of long sequences of additions, multi-plications, and multiplication -additions in-volving real and complex numbers.

Abstract: Corporate InformationSystems and Services (CISS), Cherry HillNJ, has acquired a Floating PointSystems 164 attached processor to beintegrated with the host IBM systemrunning VM. The FPS 164 wasspecifically acquired to provide cost-effective processing of large-scale scientificand engineering applications, specificallyCPU -intensive vectorizable problems.Working as an attached processor, it canoperate in both batch and interactivemodes. Topics covered by the authorinclude computer accuracy and speedTwo methods for using the attachedprocessor-Single Job Entry (STE) andArray Processor Executive (APEX)-aredescribed with examples of pertinentapplications.

©1984 RCA CorporationFinal manuscript received March 22, 1984.Reprint RE -29-3-7

A computer must be able to manipulatethese calculations quickly and accuratelyto satisfy the needs of the RCA scientistsand engineers. In addition to speed andaccuracy, these calculations must be per-formed cost effectively. These calculationrequirements have led to the evolution ofsupercomputers, such as the Cray, and ofprogrammable array processors that canbe combined with a general-purpose hostto optimize throughput, precision, and dy-namic range. To address these needs, CISS(Corporate Information Systems and Ser-vices) in Cherry Hill, NJ., has acquired aFloating Point System 164 (FPS 164) at-tached processor to supplement its IBMgeneral-purpose computing facility.

AccuracyScientific computers are measured in bothaccuracy and speed. In scientific comput-ers, the speed is measured in megaflops(millions of floating-point operations persecond). Floating point refers to the binaryversion of scientific notation. In scientificnotation, a number is represented as theproduct of the mantissa, with a magnitudebetween 0.1 and 1, and the characteristic,an exact power of 10. Thus, 5068 is repre-sented as 0.5068 X 104. This method isnecessary in scientific calculations wherethe range of magnitudes may be very large.Computers represent this scheme with abinary version of floating-point notation inwhich the characteristic is an exact powerof 2, and the mantissa and exponent of thecharacteristic are expressed in strings of Osand is (binary notation).

A single floating-point operation can bethe addition, subtraction, multiplication, or

division of two floating-point numbers toget a floating-point result. Computers, suchas the FPS 164 or Cray, are designed for64 -bit (binary digits) processing, thus pro-viding up to 15 -decimal -digit accuracy.Such accuracy may be necessary to offsetthe inevitable small rounding errors thatcan occur during thousands of iterativecalculations.

Speed

The speed of some of the scientific comput-ers is increased where the hardware orsoftware can take advantage of vector pro-ri.sing. In conventional sequential proces-sors, only a limited number of instructionscan be performed in a single time slice.Each instruction depends on the comple-tion of a previous step-whether the twoprocesses are interrelated or not. Vectori-zation is used to increase the speed withwhich a large number of arithmetic pro-cesses are handled, thus allowing for thesimultaneous execution of a large numberof calculations. One method for such aspeed-up is computational pipelining.

Pipelining is analogous to industrial assem-bly -line processing where the manufacturedproduct moves through a series of stations.Each station executes one step in the manu-facturing process and all stations work simul-taneously on different units in differentstages of completion. Floating-point addi-tion, for example, calls for several stepsthat must be carried out in sequence. Apipelined floating-point adder is dividedinto segments, each of which does a por-tion of the work that takes one clockperiod. At the end of the clock period,each segment passes the result of its work

39RCA Engineer 29-3 May/June 1984

i = 1

FETCH X,

6 -CYCLESTARTUP

SEQUENCE

CYCLE

FETCH Y,

SAVE Y,

i = 2

i = 3

i = N

A -Xi FETCH X,

FETCH Y,

SAVE Y,

1 AX, + Y, A -Xi FETCH X,

2 N-1 FETCH Y,

3 STORE Y, LOOP

AX, + Y,

SAVE Y,2 N-1

3 STORE se, LOOP

AX, + Y,

2 N-1

3 STORE Y, LOOP

2

3

Fig. 1. By performing more than one task in a single cycle, pipelining reduces thenumber of steps required to perform long, repetitive processes such as matrixmultiplications.

ATTACHED PROCESSOR

HOST

TRANSFERCODE &

RESULTS

COMPLETE JOBEXECUTION

FORTRANSOURCE

CODE

COMPILER

LINKER

LOADMODULE

HOST 100%AVAILABLE FOR

OTHER TASKS

TRANSFERRESULTS

POST

PROCESSING

LewEXAMINERESULTS

Fig. 2. SJE (Single Job Executive) allows a computer user to submit the job, com-plete with data, to the FPS 164. This will allow the host IBM to process other tasks.

to the next segment and accepts the partialresults from the preceding segment. Hence,after a start-up time equal in clock periodsto the number of segments in the pipeline,the adder can produce one floating-pointsum per clock period as long as a newpair of operands is supplied to the firstsegment every clock period.

The application of pipelining to largelyiterative scientific calculations is a majorfactor in increasing the speed of their solu-tions (Fig. 1). Some of the applicationswhere engineers can take advantage ofpipelining include the solution of simul-taneous equations, image processing, FastFourier Transforms, and digital filtering, toname a few. At RCA, the engineeringcommunity can take advantage of pipe -lined calculations in such diverse applica-

dons as circuit analysis and structural anal-ysis (see sidebar section).

Attached processorsThere are several solutions to the need forhigh-speed, high -accuracy calculations.These solutions range from supercomput-ers (ranging in price up to $12 million) toattaching array processors to existing gener-al-purpose computers. Here are several fea-tures of an "attached" processor:

1. It is a peripheral for a conventionalhost computer and is intended to en-hance the performance of the host in"number -crunching" applications.

2. It achieves high performance throughparallelism or pipelining.

3. It can be programmed by the user toaccommodate a variety of arithmeticproblems.

How they workThe remaining part of the discussion onattached array processors will be devotedto the FPS 164 currently running attachedto the IBM 3081 in Cherry Hill. The FPSattached processor is nominally rated at12 megaflops and has a dynamic rangefrom 10-308 to 10+3°7 . It is connected tothe 3081 through an Input/Output (I/O)channel. Using the I/O channel, the hostsends and receives information to and fromthe FPS 164. The FPS 164 in Cherry Hillcurrently runs any job that can be com-piled and executed in ANSII StandardFortran 77. It has its own compiler withseveral levels of optimization that resideon the host machine. Advice in conversionof existing code to Fortran 77 can beobtained from Computer Services in CherryHill. Help can also be provided in thedefinition of jobs that would be candidatesfor vectorization. The FPS 164 also has amyriad of its own subroutines and func-tions, which were designed and written totake advantage of its pipelining capabilities.

There are two ways to use the attachedprocessor-to process an entire job, or toprocess a subsection of the calculations asa subroutine box. The first method de-scribed is termed SJE or Single Job Entry,while subroutine processing is accomplishedusing APEX (Array Processor Executive).

Single job entry (SJE)

The SJE Operating System accepts com-mands input into the 3081 via VM/CMS.The SJE job flow is straightforward.

1. Prepare source Fortran code using theCMS editor.

2. Include references to the FPS MathLibrary for additional performance im-provement (at user's option).

3. Compile the source using the FPS For-tran -77 compiler on the 3081.

4. Link the compiled code using the host -resident link editor to produce a loadmodule. At this stage all external refer-ences to Fortran functions and the FPSMath Library are resolved.

5. Run the job.

6. Transfer the results of the job back tothe host for output or analysis.

Using this approach, the host computerperforms processing for which it is best


Math libraries and more . . .

The FPS basic libraries include Fortran -callable rou-tines to perform:

Basic Math Routines

Vector addition, subtraction, multiplication, anddivision

Logarithms, trigonometric functions, random -number generation

Vector -to -scalar routines, such as the SVE (sum ofthe elements)

Vector comparisons

Complex vector arithmetic Matrix routines, such as multiplication, transposi-

tion, and inversion

Fast Fourier Transforms (FFTs) on real and com-plex vectors

Advanced Math Routines

Eigenvalue and eigenvector solutions

Transformation of complex Hermitian matrices

Revised simplex routine for the solution of linearprogramming problems

Image Processing Routines

Two-dimensional Fast Fourier transforms

Two-dimensional convolution and correlation

Spacious filtering Contrast enhancement

Simulation Routines

Single and multistep solutions for ordinary differen-tial equations

Runge-Kutta fourth -order integrator

In addition to these routines, other software, writtenespecially for the FPS architecture, can be obtained.

BSCLIB-a collection of subroutines developedby Boeing Computer Services

FMSLIB-Fast Matrix Solution Library

MSC NASTRAN-McNeal Schwendler Nastran ANSYS-Swanson, Inc., finite element package

QSPICE-circuit analysis by Quantitative Tech-nology, Inc.

suited, while the FPS -164 processes com-putationally -intensive jobs in parallel withthe host (Fig. 2).

Array processor executive (APEX)

A user can take advantage of the attachedprocessor's capabilities for only a subset ofhis tasks. Under this method the host 3081is called upon to perform scaler operationswhile the FPS 164 performs those calcula-tions best suited to pipelining. The APEXjob flow can lead to extremely efficientresults.

The host Fortran program calls an APEXroutine to transfer instructions and data tothe array processor and then continuesexecuting its portion of the task. Later, itcalls another APEX routine to verify com-pletion of the transfer, and then instructsthe array processor to start execution ofthe task. The host program once againresumes work on its portion of the task.

After that, the host program calls APEXroutines to verify completion of the arrayprocessor task, to return results to the host,and to ascertain that the results have beenreturned. At that point the host programcan perform any processing that dependson those results.

Using APEX calls, the host programcan also transfer data for a second task to

the array processor while the array proces-sor is completing the first task. The APEXlibrary thus gives the host/AP system amultitasking capability that results in ex-tremely efficient use of the system's hard-ware resources (Fig. 3).

The decision to use either the SJE orAPEX mode will depend on the nature of

the problem to be solved. Many problemswill clearly be more advantageously solvedby the use of APEX, because of theirlimited need for intensive calculations. Theymay depend heavily on interaction by anend user before performing calculations,returning intermediate results, and antici-pating further interaction before continu-

ARRAY PROCESSOR

DATA ANDCODE TO AP

APPROCESSING

(Host FORTRAN statement

RESULTS BACKTO HOST

API X

HAY

SUBROUTINE CALLTO AP

(Host FORTRAN statements)

HOST COMPUTER

HOSTPROCESSING

(Host FORTRAN statements)

APEX

El Atil

RETURN FROMAP SUBROUTINE

(AP FORTRAN statements)

Fig. 3. APEX (Array Processor Executive) mode will allow the computer user totake advantage of the capabilities of the FPS 164 when vectorizable code isappropriate. The user can run routines on the host IBM for other tasks, thus utiliz-ing the FPS 164 as a subroutine box.

Handman: An attached processor provides calculation speed and accuracy 41

ing to calculate. Such application areasmay use inquiries to set up a problem, andmay query the user for tolerance limitswhile calculating. In such a scenario, theAPEX mode is a good choice. Becausethere is no clear-cut definition of whereeach is optimal, the user must consider theunique nature of his problem.

The inclusion of the FPS 164 into thegeneral-purpose environment at the shareddata center in Cherry Hill will provideincreased capabilities to the RCA comput-er user at substantially reduced costs. The64 -bit capability of the FPS 164 will pro-vide the accompanying necessary accuracy.If you would like further information onattached processing, please contact theauthor.

Correction note:

March/April RCA Engineer"Image sampling and analysis technique for high -resolutionmeasurement of micrometer -sized features," an article by M.T.Gale and R.L. Covey published in the March/April 1984 RCAEngineer contains an error. Figure 10, p. 64, was printedincorrectly and should appear instead as shown here. You canobtain corrected copies of the entire article from Robert Covey,RCA Laboratories, TACNET: 226-3044.

Susan Handman is currently AccountManager at RCA Corporate InformationSystems and Services (CISS) in Cherry Hill,N.J. She is primarily responsible forengineering -related accounts. Shereceived a B.S. in Mathematics from theUniversity of Pittsburgh. After joining RCAin 1980, she provided technical support onmath and statistical packages, and alsoworked in customer services on engi-neering and scientific implementation.Ms. Handman is the new CISS EditorialRepresentative for the RCA Engineer.Contact her atRCA Corporate Information Systems

and ServicesCherry Hill, N.J.TACNET: 222-6242

12

11

EDGE 10SCANS g

8

7

6

54\32N1-

Z DIRECTION

1,14m

TIP SCANS

004 tan

11

TIME ( ELEMENT NUMBER)

12

8 m

Fig.10. The tip and edge scan measurement operationsshowing both an image plane schematic and the corres-ponding analog signals. All dimensions are referred back tothe stylus object plane. Only the photodiode array is movedduring these operations.


D. Zarodnansky

Adapting CMS for a varied user community

RCA Laboratories encourages scientists and engineers to beginand continue using CMS by providing site -specific helpfor them.

The VM/CMS timesharing system, which is used not only inthe RCA Cherry Hill computer installation, but also in manyother computing centers worldwide, is perhaps more suited forcomputer users with moderate to extensive experience than forsomeone who is just learning to interact with a computer for thevery first time. Although RCA's Corporate Information Systemsand Services (CISS) staff in Cherry Hill has made extensivemodification to the product in an attempt to make it more easilyused, most of their modifications have, of course, been in responseto the needs of general users of the system, regardless of theirlocation, as opposed to site -specific modifications.

Over the past several months, a number of steps have beentaken at the RCA Laboratories in Princeton, N.J., to addressuser needs that are specific to that site, and to enhance access toCMS for new users. These efforts have included a set of utilityprograms developed at the Labs to perform common functionssuch as formatting and printing documents on a high -qualitylaser printer, and doing file management. Wherever possible, theutilities are menu driven, so that users need not memorize com-mands. In addition, the installation at the Laboratories of data -stream protocol convertors has allowed our terminal users (mostlyon ASCII terminals) to access CMS in a full -screen mode, asopposed to the line mode, which was the only access to CMSavailable to most Labs users in the past.

A set of EXEC procedures, some of which provide site -specific utilities, and some of which provide alternatives to sys-tem commands, were written and tested at the Labs, and werereleased to the general user population as a utility package calledLABSUTIL. The LABSUTIL package is fully documented, andsupported by the Laboratories Computer Services activity.

In addition, another set of EXEC procedures that were

Abstract: A large fraction of the personnel at RCA's DavidSarno ff Research Center in Princeton use the CMS timesharingcomputer system provided by the RCA Corporate InformationSystems and Services activity in Cherry Hill, N.J. This paperdescribes some of the techniques employed at RCA Laboratoriesto make the CMS system easier for the "novice" to use, and alsodescribes some of the site -specific utilities developed to make thesystem more useful to the "veteran."

(D1984 RCA CorporationFinal manuscript received March 29, 1984.Reprint RE -29-3-8

designed to enhance CMS error recovery and avoidance, werewritten and tested on a small subset of novice users at theLaboratories, to determine if CMS usage could be made morepleasant for that group.

Access to this set of EXEC procedures is supplied to CMSusers only upon request. We felt that it might cause a problemfor the more experienced users to have the system respond inways other than what they were used to and what was docu-mented in the manuals supplied by IBM and Cherry Hill.

Preparing documents on the CMS systemOne of the first requests that users made was for easier access tothe IBM 6670 laser printer in the Labs Remote Job Entry facil-ity. The 6670 is a correspondence -quality printer for documentssuch as reports, letters, and internal memos. Since the 6670 pro-vides copy quality comparable to that available from an IBMSelectric typewriter, it was hoped that this facility would allowan increase in productivity by permitting users to use their ter-minals to view and edit documents, and to reprint the editedversion, rather than having to retype an entire document eachtime a revision was to be made.

The facility used for formatting documents within CMS isthe IBM Document Composition Facility (DCF), sometimesalso known as SCRIPT. It is available to all CMS system users.The 6670 laser printer was already in place as a result of itspurchase as part of an office automation pilot test of an IBMapplication program running under CMS.

IBM 6670 preprocessor software was installed on the sys-tem by CISS, to handle conversion of the DCF output into aform that could take advantage of the capabilities of the 6670,such as font switching, special symbols, and rotated printing. Thestandard means of invoking the preprocessor is to make a call toan EXEC procedure. Any options that the user wishes to usemust be supplied as part of the command that calls the EXEC.To make this process more user friendly, a driver EXEC wasdeveloped that presents all of the options in a menu format, andallows the user to select the desired ones. This driver EXEC thenconstructs the command that invokes the preprocessor, and exe-cutes this command.

Once the capability for processing documents for printingon the 6670 laser printer was developed, it soon became appar-ent that additional software for formatting equations on the 6670would be required as an adjunct to DCF. This took the form of

Zarodnansky: Adapting CMS for a varied user community 43

a DCF Application Programming Function (APF), calledDSRCPROF. Ultimately, this APF was expanded to providethe capability for automatic numbering and cross referencing ofthe equations within the body of the document, direct controlover font changes, and standard skeleton -outline formatter. Thissoftware has been tested, and is considered acceptable for clericalusers. Further development to provide additional flexibility andease of use, and to provide additional functions, is underway.

Another user request was that the documents be printableon special forms, specifically Internal Correspondence stationery.A system was developed whereby the application presents a se-ries of questions to be answered by the user, and the answers areentered into the proper places in a formatting framework. Thissystem has been expanded to cover forms such as External Cor-respondence stationery, legal -size paper, and special stationerythat the user may provide.

File management utility programsThe CMS user is aware that the CMS Virtual Machine containsa disk that is used to store files. The purpose of the files and themechanics of managing the disk space are concepts that the ca-sual CMS user may not understand, yet file management is animportant part of computer use. Users may tend to ignore theneed to occasionally clean off files that are no longer needed, inpart because they do not understand the reasons for, or meansof, doing so. To make file management easier for the user, LAB-SUTIL includes a number of file management utility programs.Some of these utilities are:

SCRUB, which presents the filenames to the user one at atime, along with a number of choices for the disposition of thefile. The user may KEEP, DELETE, TYPE, PRINT, or XEDITthe file, or QUIT.

FIND, which works like the LISTFILE command, except thatit then prompts the user for a disposition for the files found.This is useful for performing the same operation on a numberof related files, as well as for locating files on disks other thanone's own.

XCHECK, which cross-checks two disk areas looking for fileswith identical filenames and, upon finding a matching pair ofnames, matches the contents of the files and allows the user to

FUNKY WINKERBEAN

'Funky Winkerbean" by Tom Batiuk reprinted by permission from Field NewspaperSyndicate, Irvine, California. Copyright ® 1983 Field Newspaper Syndicate.

dispose of one or both of the files.This reduces redundancy andwasted disk space.

Documentation for the CMS userAll of the effort put into the utility programs described abovewould be useless if the users were unaware of the existence ofthe programs, or did not know how to use them. User supportand documentation comprise a major portion of the effort in-volved in making this system more user friendly. This supporttakes on a number of forms.

Written documentation

We have written manuals describing how to use these functions,which are distributed to users upon request. In addition, thosemanuals that we feel are of sufficiently general interest are dis-tributed to all CMS users at the Labs. These manuals areintended as an adjunct to, and not a replacement for, the CMSUser Guide, and other manuals supplied by the corporate CISSactivity. The manuals document the utilities as they apply to theLaboratories site, and also point up differences, where they exist,between system standard products, and those modified to bespecific for the Labs site. These manuals presently include a usermanual for LABSUTIL, one describing procedures for preparingdocuments using DCF and the 6670 laser printer, and one de-scribing the use of CMS in the full -screen mode through thedatastream protocol convertors available on the Labs' local areanetwork.

Internal documentation within programs

Everything on the LABSUTIL disk is written such that, ifinvoked with a question mark as the only argument, a briefHELP display of no more than one screen is shown. This displayexplains the syntax of the command and the valid options. Thisfacility will be extended to the other Labs Computer Services -generated software as time and manpower permit.

HELP files

Utilities on the LABSUTIL disk are all documented in standardCMS HELP formatted files. Thus, those who are familiar with

Tom BatiukIT KEEPS DOING WHAT I

TELL IT TO DO INSTEAD OFWHAT I WANT IT TO DO !


(FOR26670 V 9.1.0 03/01/84)

Settings in effect for the FORMAT67 processor are:

1) Bind = 12 , 12

2) Duplex = NO

3) Fonts are: ELITE 163 ELITE 193

4) Lines Per Inch is: 6.0 , Spacing is 1

5) Number of copies is: 1

6) Paper drawer selection is:

7) Forms type is PLAIN paper

Please enter the NUMBER of any options that you wish to change orexamine the settings of, or a carriage return to accept these settings. If youwish to change more than one setting, separate the numbers with aSPACE:

The local printer utility main menu.

Please do not print output over 10 pages long on local Dot -MatrixPrinter Thank You

Please select a printer from the following:

1 Printer in E-014 (Dot Matrix)2 Printer in S-209 (Dot Matrix)3 Printer in W -117C (Dot Matrix)4 Printer in SW -122 (Dot Matrix)RMT21 1403 Line Printer in Labs RJE roomLOCAL 3800 Line Printer at Cherryl-lill

Enter Printer number, or remote number, or "LOCAL"

To ABORT the print operation, enter only a CARRIAGE RETURN.

To print output on the 6670 LASER PRINTER in the RJE room, use theFOR26670 EXEC on the LABSUTIL disk.

The FOR26670 main menu. This utility allows use of the6670 laser printer.

this facility may receive help with these utilities in the samemanner that they get help with standard CMS commands.

The CMS HELP deskExperience has shown that if users can get immediate on -siteassistance with a problem, they are more likely to continue tomake use of the products provided on the system, and to ask forassistance when they need it. With this in mind, the Labs Com-puter Services group has a program to provide users with imme-diate assistance, including a helpline, which is staffed by a personwho primarily provides intensive user assistance, and one-on-oneinstruction when needed. Feedback from the users indicates thatthey are more comfortable asking questions in -this setting than ifthey are required to call off -site for assistance. In addition, theon -site personnel are trained in the site -specific software andhardware found at the Labs, and are thus able to provide assis-tance with them.

This also gives the the person providing assistance theopportunity to follow up on the original request if more help isneeded. A telephone -answering device monitors the helpline andtakes messages when the CMS help person is working withother users. For problems requiring immediate attention, therecorded message on the answering machine provides a numberthat users may call to have key members of the Labs ComputerServices staff paged to handle urgent problems. Initial userresponse to this service has been very positive.

Further attempts at user friendlinessWe have found that some of the nontechnical users of CMSrequire additional assistance with the system. Along this line, anumber of EXEC procedures have been developed to enhanceerror avoidance and recovery when executing some simple CMScommands. For example, a user attempting to type or edit adocument might forget to enter the filetype when naming thedocument. The standard CMS response to this omission wouldbe an error message saying, "INCOMPLETE FILEID SPECI-FIED." The EXEC procedure that these users now invoke inplace of the system command puts a different error message onthe screen, which tells the user exactly what is missing from thecommand. Because immediate feedback is provided as to thenature of the mistake, the user is much less likely to make thesame mistake in the future.

As another example, the default condition for the LIST -FILE command, which normally lists only the filename, filetype,and filemode, has been changed to show all of the informationassociated with the DATE option of that command, and assumesa "wildcard" for the filetype if not otherwise specified. TheEXEC procedures that implement these changes are containedon a disk that is accessed by the users only if they wish to usethe modified procedures.

Feedback from the users indicates that novice users find themodified procedures much more comfortable to use than stan-dard CMS. One might argue that the users will never learn"real" CMS this way but, on the other hand, some users feel thatthey have no more need to learn "real" CMS than they have tolearn assembly language, as long as the system does the job theyneed. The entire attempt to make CMS more "friendly" is

founded on this approach-the system exists as a tool for theusers. The system should be adapted to them; they should not beforced to adapt to the system.

David Zarodnansky attended Stevens Institute of Technology,and Trenton State College, where he received a Bachelor of Artsdegree in Psychology in 1975. He is attending Villzknova Univer-sity and expects to receive a Bachelor of Electrical Engineeringdegree in 1985. He joined RCA Laboratories in 1977 as aResearch Technician, and is presently an Associate Member ofthe Technical Staff with the Computer Services group, specializ-ing in support for Labs users of the CMS system, and the IBMPersonal Computer.Contact him at:RCA LaboratoriesPrinceton, N.J.TACNET: 226-2774

Zarodnansky: Adapting CMS for a varied user community 45

A. De Marco

Analyzing design alternativeswith the PRICE models

Complex engineering projects demand simplified assessmentsof cost tradeoffs, before the job begins.

Have you ever wondered how a concep-tual design change would affect the life -cycle cost of a system for years to come?Conventional methods of evaluating thecost impacts of such a change would con-sist of days and weeks of reviewing bills ofmaterial, labor tables, reliability records,inflation forefasts, and . . . what a mess!Faced with all of that effort you might bepersuaded to dismiss the cost considera-tions entirely unless, of course, there werean automated, streamlined procedure forreviewing the cost impacts of your con-cept. This procedure is the PRICE system.

...100%

JAN 85

The PRICE systemPRICE is a system of computerized cost -estimating models and supporting programs.The PRICE Hardware model was firstdeveloped in 1962 as an internal RCAprocedure to help automate the cost -esti-mating process. It was made commerciallyavailable for general use outside of RCAin 1975. Since that time, the PRICE fam-ily of models has multiplied to the numbersshown in Table I.

The PRICE models use the parametricapproach to estimating. Therefore, the costof a system is not estimated from parts

Abstract: The primary responsibility ofthe systems engineer is that he keeps thetask within bounds. Time, people, andmoney are limited resources. Conventionalmethods of estimating a program'sconsumption of these resources aredifficult and tedious. Parametric resourcemodeling can assist the systems engineer tobalance design margins against resourcelimitations by rapidly displaying the far-reaching impacts of design alternatives. Bycomparing the program cost and scheduleto the modeled cost and schedule throughthe execution and integration phases,engineering managers can helptroubleshoot unexpected problems that

may result in overruns in these areas.The PRICE cost- and schedule -

estimating models have been used in manydisciplines to progressively automate thecompilation of estimates of program -resource consumption. Theseconversational computer models, whichwere developed at RCA, are easy to useand provide many features that allowquick evaluation of trade-off studies. Anincreasingly important measurement ofsystem performance is its acquisition/life-cycle cost, and parametric cost modelingcan help to automate the complicatedmeasuring process.

01984 RCA CorporationFinal manuscript received April 30. 1984.Reprint RE -29-3-9

JUN 98

ONE COLUMN 3 MONTHS

02

lists and labor tables, but from the generalimpacts that certain descriptive measuresof that system have on its cost. This methodallows an analyst to describe a systemconcept that can be used in the models,without the analyst having to know thedetails of its "nuts and bolts."

A similar technique can be applied whenconsidering the construction and mainte-

Table I. The PRICE family.

PRICE Systems creates and maintains afamily of parametric models available toeach customer via network computer ser-vices. Briefly, the models are:

PRICE H-Cost estimates for hardwaredevelopment and production. Addressesassemblies, whole systems and systemsintegration.

PRICE HL-Cost estimates for equip-ment/system use -maintenance, repair, andother support functions to yield life cyclecosts.

PRICE M-Specific cost estimates formicrocircuit (chip) development andproduction.

PRICE S-Costs, schedules and man-power for design, implementation, test, andintegration of new computer software.

PRICE SL-Costs of maintaining a soft-ware system throughout its projectedoperational life.

PRICE PM-Prepares master schedulesfor interdependent development and pro-duction activities-a project planning andmanagement tool (a new model).


nance costs for a home. The number ofrooms in the house, and the type of homeare general indicators of the amount andquality of the required materials, and indi-cate the overall effort required for con-struction and maintenance. Pertinent laborrates depend upon time and geographiclocation. Therefore, the zip code of thehome and the year of the required taskscan be used to indicate the level of thelabor rates. Such a four -parameter model(Fig. 1) may appear simple at first glance,but consider the diverse ramifications ofthe addition of a single room. Not onlywould one aspect-such as carpentry cost-be affected, but masonry cost, roofing cost,wiring cost, schedules, and the ensuingmaintenance costs would all be affected invarying degrees. One can see that such amodel must be internally complex.

The PRICE models contain thousandsof integrated equations relating descriptiveparameters to costs and schedules. Theyhave played many roles and have beenused in numerous types of applications,including:

Evaluations of bids and proposals

Design -to -Unit -Production -Cost analysis

Estimates of cost to complete

Bid preparation and submittal

Estimates of cost to modify

"Should" cost analysis

Bid, no -bid decisions

Long-range planning

Procurement planning

Cross-checking of design concepts

Microelectronic cost estimating

Application to life -cycle cost analysis

Design cost trade-offs analysis

The key to flexible use of the PRICE sys-tem is model calibration. Each of thePRICE models operates in "reverse." Inthis mode, historical costs and schedules ofcompleted projects are entered as inputs tothe models. Indices are then generated,which fingerprint the peculiarities of a par-ticular product and a particular organiza-tion. These indices calibrate the general-ized PRICE system to a model of a specificapplication. The indices serve as the foun-dation for use in the "forward" mode toestimate the costs and schedules of futureprojects. It has been found that the PRICEsystem provides a universal method of gen-erating accurate cost and schedule estimatesin the space, defense, and commercial bus-iness environments.

MATERIALS

Number of RoomsClass of Home

Fig.1. The construction and maintenance costs for a house can be estimated witha simple parametric model using four inputs.

PRICE and the systems engineer

Resources available to any particular proj-ect are limited. Time, people, and moneyare all usually scarce. It is important thatthe design tradeoffs made during the con-cept phase of a program be balanced againstits resource limitations. The primary respon-sibility of the systems engineer is to boundthe task and to keep specifications withinbounds throughout program evolution.

The scenario

The following example, based on an actualrace study, illustrates the role of the PRICEsystem in assisting the systems engineerwith his complex job. The scenario involvesthe insertion of advanced electronic tech-nology into a United States weapons sys-tem. The particular device in question isgenerally known as a Programmable Sig-nal Processor (PSP). The baseline configu-ration of the PSP, which is currently infull-scale development, contains large-scaleintegrated (LSI) circuitry. It is suspectedthat the insertion of very large scale inte-grated (VLSI) circuitry into the PSP wouldimprove reliability, enhance performance,and reduce supply and maintenance costs.The reduction in the supply and mainte-nance costs would be a result of reducednumber of part types, reduced circuit -boardspace, and longer times between PSP repair.

Two alternatives are proposed by thecustomer for the design and insertion ofthe VLSI technology: Concept A and Con-cept B, described in Table II. Concept Akeeps the PSP's overall architecture intact.The total PSP design is translated to its

VLSI equivalent and repartitioned for mini-mal size and integrated circuit types. InConcept B, enhancements are permitted inthe redesign to optimally use the VLSIarchitecture. Built -in -test functions are im-plemented, and operational throughput isincreased. It is claimed that Concept Bwill reap the benefits of Concept A whileenhancing the PSP's overall capability. How-ever, a high price will have to be paid dur-ing development.

It is the job of the systems engineer toexplore these concepts and determine whichis most appropriate to meet the customer'sspecifications. The validity of the seem-ingly attractive improvement claims mustbe carefully studied before embarking onthe long and costly process of developingthe VLSI technology.

The analysisAn analysis to further specify each alterna-tive concept and determine which willprove more cost effective can be performedwith the PRICE system. The analysis cov-ers all program phases from design througha ten-year deployment, and accounts foroperation and support.

The method used to estimate the var-ious costs and schedules is shown in Fig.2. Initially, data on each concept must becollected and scrutinized. The projects arecompared to similar projects that precededthem. The customer expresses uncertaintyin the deployment quantities of the device.Thus, the analysis must cover both a 500 -unit and a 2,000 -unit deployment scheme.PRICE inputs are formulated that will accu-

De Marco: Analyzing design alternatives with the PRICE models 47

Table II. Two design concepts for the insertion of VLSI circuitry into the PSP'sbaseline configuration are proposed.

Description

Baseline configuration.Currently infull-scaleproduction.

Concept A. The totalbaseline design istranslated to itsVLSI equivalentand repartitionedfor minimal sizeand IC types.

Concept B. The totalbaseline architec-ture is redesignedand enhanced tofully use VLSIcapability.

Picture of unitCOOLING AIR INLET

Capability(millions ofcomplexoperations

Weight (lbs) per second)

100 16

49 16

43 25

Integratedcircuits

Number ofboard pairs

Number ofsoftwareinstructions

5950 31 120,000

950 13 120,000

850 11 130,000

DESCRIPTIVEPROGRAM

DATA

PRICE

- _____ -;PR ICE

Fig. 2. The method used to estimate the life cycle costof each PSP design concept involved five of the PRICEmodels. Trade-off studies were performed along theway to further specify each concept. The PRICE modelsare each independent systems and can be used sepa-rately or in unison.

PR ICE

HL

PRICE

PRICE

SL

PRICE

A

LEGEND

INFORMATION FLOW

IN TRADEOFF STUDY


Computer Graphics: A design, feedback,and display aid for the PRICE engineer . . .

The PRICE modeling process involves systematicdata gathering, analysis, model formulation, and avery active feedback mechanism. Statistical tech-niques supported by graphical plots of data fits andtrends automatically transform raw data into informa-tion. Graphical illustrations identify data patterns thatare fitted to produce the PRICE parametric formulas.

Key input and output factors of the PRICE method-ology, such as cost and schedule data, are easilyconverted into scatter diagrams and trend plots thatare the basis of correlated algorithms. At this modeldesign stage, computer graphics saves the engineertime and effort. Erroneous data points and obvioustrends are spontaneously discovered.

As shown by the illustration below, a typicalmodeling cycle involves constant data consolidationand analysis. A number of graphics devices are cur-rently available to the PRICE engineer. A quick pre-view of the data points on a Tektronix 4105 terminalusing Tellagraf provides a general idea of data pat-terns and identifies clusters and outliers. Similarly,an IBM PC can act as the graphics monitor onceequipped with the proper interface. A hardcopy canalso be obtained by a screen dump of the plot to aTektronix 4695 ink -jet plotter. For a plot requiringexcellent quality, a Zeta 8 desktop pen plotter isused.

User-friendly graphics software and smart devicesautomate the data -to -information transition effec-tively. More time is available for critical questioningof assumptions and procedures. This results infaster turn -around of hypothesis formulation andtesting. As more data is fed from the PRICE Models'user, and as the scope of the models' technologicalapplications is broadened, the regular use ofcomputer graphics as a design and display aidincreases in value.

-Carmen Rueda,PRICE Systems

RAW DATA $

-;c211041...

FROM DATA TO INFORMATION

4PROsr,PLOT

HUMANHENGINEERING

PRICE HREADY

rately describe the concepts to the models.It becomes apparent that each of the

proposed concepts would entail the acqui-sition of thousands of custom and semicus-tom integrated circuits (ICs). However, thespecific design approach to employ in thedevelopment of these devices is not clear.Three options face the systems engineer inthis case: The full -custom handcrafted ap-proach, the standard -cell approach, andthe universal -array appraoch.

The handcrafted approach involves adetailed design of the IC at the transistorlevel. Design lead times are long and diffi-cult to predict. In many cases the cost andschedule risks entirely prohibit the hand-crafted approach.

The standard -cell approach abbreviatesthe handcrafted approach. The foundationof this approach is an extensive, well -doc-umented library of circuit building blocks,called cells. Each cell is a group of circuitelements combined to perform a certainfunction. The cells are very adaptable tocomputer -aided design because of their gen-erality. The designer now can constructthe IC in terms of these cells, obviating themanipulation of tens of thousands of indi-vidual transistors and circuit elements. De-velopment costs are much lower than thatof the handcrafted approach, yet the result-ing ICs are more costly to manufacture.

Universal arrays take this standardiza-tion a step further. In this approach, chips

are preprocessed to yield regular arrays ofuncommitted circuit elements. The designerconstructs the circuit in terms of cells thatcan easily be created by interconnectionsof these circuit elements. Thus, when thedesign is complete, the circuit elements onthe chip are merely linked together toform the IC. The development effort issimilar to that of the standard -cell approach.The turnaround times for fabricated unitsare much shorter than that of each of theother approaches, but the manufacturingcost per unit is higher.

The question before the systems engi-neer is: When does it become cost effec-tive to forgo the popular universal arrayapproach and opt for either the standard

DeMarco: Analyzing design alternatives with the PRICE models 49

cell or the handcrafted approach? Atradeoff analysis using PRICE M, whichestimates the development and productioncosts for ICs, can help to rescue him fromhis dilemma. The particular device chosenfor illustration is a MilSpec, 2500 -gate VLSIchip required for Concept A. The IC has64 pins and is fabricated using a CMOSprocess. The critical factors to be consi-dered for each approach are shown inTable III.

These, along with other descriptive pa-rameters, are entered into PRICE M and asensitivity analysis on production quantityis performed. The results are shown in Fig.3. The intersecting curves reveal the pointwhere each approach becomes more costeffective.

Although the handcrafted approach mayseem appealing, money is not the onlyresource to be balanced. Time is also aconsideration. Development schedules esti-mated by PRICE M and presented byPRICE PM showed that a large amountof time would be required to develop the

10 -

9 -

-

7

6

5

4

3

2 -

1 -

Table Ill. Some PRICE M descriptive parameters for each IC design approach.

Parameter Hand-crafted Std. cell ULA Explanation

SIZE (mis2) 1762 2252 3052 Surface area of chip

ECMPLX 1.1 1 0.3 Scope of work versusengineer's experience

NEWCEL 0.8 0.2 0.05 Percent of newlydesigned cells

DESRPT 0.2 0.25 0.5 Percent of design repeat

CADFAC 0.8 1 1.2 Degree of computer -aided design

ITERAT 2 1 0 Design/proto/testiterations

handcrafted approach as opposed to otherapproaches.

The decision becomes clear. It is esti-mated that either 30,000 or 100,000 ofthese devices will be needed, dependentupon the final production quantity of thePSP. Also, the maximum development timefor this project is 18 months. Therefore,the standard cell approach is chosen forthe IC.

HAND-CRAFTED

STANDARD CELL

UNIVERSAL ARRAY

V\.......\

The information produced by PRICEM on the custom and semicustom inte-grated circuits assists in describing eachsystem concept to the PRICE H model.PRICE H will estimate the PSP develop-ment and production costs as well as vitalreliability information. These data, alongwith an equipment deployment scheme,are entered into PRICE HL. The deploy-ment scheme defines equipment disburse -

This graph shows amortized unit costnormalized around the high valueversus the cumulative number ofproduction units. Break-even points forthe three design approaches can beseen.

0 10 20 30 40 50 60 70 80

1 I I

90 100 110

CUMULATIVE PRODUCTION UNITS (K)

120 130 140 150

Fig.3. The results of a PRICE M analysis can be used to decide which IC designapproach to implement.


ment, maintenance and supply locations,equipment operating hours, and the life ofthe program. Here, the benefits of the built-in -test functions in Concept B are realized.It is determined that the units' own abilityto determine which circuit board is mal-functioning in the field will allow a two -level maintenance concept to be imple-mented.

The two -level concept involves the re-placement of the boards at the equipmentlevel, with their repair at the depot. Themore traditional three -level maintenanceconcept must be used for both the baselineconfiguration and Concept A. In the three -level concept, the entire PSP must be re-placed at the equipment level. The defec-tive PSP must then be tested with specialtest equipment at an intermediate level todetermine the defective board. The boardis then sent to the depot for repair. Thisfactor will easily be analyzed since theabove are but 2 of 28 maintenance con-cepts PRICE HL considers with every run.PRICE HL then estimates the PSP sup-port costs and operational availability. Aconsolidated report displays the hardwaredevelopment, production, and support costs.

A similar procedure is used to estimatethe cost of software development, support,and maintenance. PRICE S and PRICESL are used for this process. For eachdesign concept, care must be taken whenevaluating the number of instructions, acritical system descriptor used by both mod-els. The baseline configuration consists of120,000 lines of developed microcode. Con-cept A requires that 10 percent of thesoftware be redesigned. Concept B requires10,000 lines of new software to imple-ment the built-in test equipment, plus aredesign of 25 percent of the baseline sys-tem's software. This complex scheme ismodeled with the PRICE input parameters.

PRICE S produces a report showingcosts and schedules for the various stagesof software development. It becomes appar-ent that the software development effortrequired for Concept B looms large. Apenalty must be paid if such a big job is tobe done in the proposed time. Inevitably,more people than normal will have to bebrought onto the job. PRICE S outputdisplays the cost penalty to be expectedwhen the accelerated schedule dictated isimposed on the project. The software datafrom PRICE S is then entered into PRICESL to generate software costs for the 10 -year support period.

Next, the hardware and software costinformation for each concept is consoli-dated and described to PRICE A. PRICE

260

2/0 -

220 -

200

ISO

160

140

120

100

(a)

PROGRAMMABLE SIGNAL PROCESSOR COST ESTIMATES500 PRODUCTION UNITS

A r/A

D

.70.0

600

500

400

300

200

100

(b)

BASELINE CONCEPT A CONCEPT B

DESIGN ALTERNATIVES

PROGRAMMABLE SIGNAL PROCESSOR COS' ESTIMATES2000 PRODUCTION UNITS

BASELINE CONCEPT A CONCEPT 8

DESIGN ALTERNATIVES

Legend¢21 1984$

INFLATED5

LegendZ21 1984$

INFLAfCD$

Fig. 4. The PSP cost estimates generated by the PRICE System can be comparedto determine the most cost-effective design alternative. The impact of inflationmagnifies the benefits of Concept B.

A applies economic escalation to the costsbased on specified spending profiles. Themodel combines results and presents a uni-fied picture of each concept.

The results of the cost analysis are shownin Figs. 4 (a) and (b) and 5. Both redesignconcepts show a reduction in total life -cy-cle cost, as was suspected. Examining thequantity tradeoffs, it appears that the casefor a redesign of the baseline configuration

is stronger in the 2000 -unit case than the500 -unit case.

Concept B proves to be the cost winnerof the two design approaches in the out -years. The additional initial investment de-manded by Concept B to further developthe VLSI technology greatly increases sys-tem reliability and maintainability, thus,decreasing support costs. This is due to thebuilt-in test functions and the resulting stream-


PROGRAMMABLE SIGNAL PROCESSOR COST ESTIMATES

CONCEPT A

BASELINE

41--1DS".77.

SAVINGS FROM BASEUNEOF 23.07.

CONCEPT B

DEV7.0%

SAVINGS FROM BASELINEOF 35.7%

Fig.5. The breakdown of the PSP cost estimates shows theeffects additional development spending has on support costs.

expected beneficial

Table IV. The evolution of the PSP as it progresses through various programphases.

PhaseConcept

explorationFull-scale

developmentProduction

and deployment

Weight 100, 49, 43 43, 40, 39 39

Number of prototypes 2, 5, 20 5 N/A

Production quantity 500, 2000 500, 2000 2094

Capability (MCOPS) 16, 15 25 25

Number of ICs 5950, 950, 850 850, 825 825

Number of circuit board pairs 31, 13, 11 11, 10 10

Number of software instructions 120K, 130K 130K, 131K 131K

Table V. A comparison of the typical resources required to perform the PSP costanalysis shows the large time savings provided by the PRICE System.

Cost analysis function Resources required

With conven- With thetional methods PRICE Systems

Secure descriptive information on thescope of work.

Select and analyze comparative costexperiences for estimating purposes.

Adjust historic cost data for differencesin scope of work, technology, economics,resources, scheduling, tooling, and so on.

Develop the estimate including potentialengineering changes, schedule factors,inflation, and so on, through a ten-yearsupport period.

Total for basic estimate above.

Perform sensitivity analysis andtechnology trade-offs.

Continue with design -to -cost tracking,throughout full-scale development

80 man-hours

20 man-hours

40 man-hours

12 man-hours

5 man-hours

Included in secondcategory by auto-mated models

150 man-hours 8 man-hours

290 man-hours

100-200 man-hours (but timeseldom permits)

60 man -months

25 man-hours

4 man-hours

6 man -months

lined two -level maintenance concept. Thebenefits of Concept B are magnified whenPRICE A inflates the dollars as they areexpected to be spent throughout the sche-dule. Though the rase appears strong forConcept B, the size of the additional invest-ment in development is a factor that mustbe presented to the customer.

The systems engineer has now accumu-lated a wealth of information from a rela-tively small amount of data. Armed withthis information, he can address the alter-natives confident that the design/cost/sched-ule tradeoffs have been thoroughly syn-thesized for each task. The systems engineerwas also able to uncover potential prob-lem areas, such as software development,that would result in cost and scheduleoverruns, if not addressed. Concept B waschosen in the scenario. The developmentfunds and manpower will be provided toaccomplish the design, because the custo-mer wishes to realize the huge supportsavings predicted in the out -years of theprogram.

Tracking design -to -costThe applicability of the PRICE systemdoes not stop at the concept explorationphase of a program. The systems engineermust orchestrate a program through detaileddesign, execution, integration, test, and oper-ation and support. He must also ensurethat the program stays within its proposedbounds throughout its life. This is a veryformidable task, since it is inevitable that asystem will change in size, shape, and func-tion as the program progresses from phaseto phase. The particular evolution of thePSP is shown in Table IV.

Once the PRICE descriptive parametersof, in this case, the processor, are estab-lished, it is a simple matter to update themmonthly, or even weekly, as the systemevolves. Parameters that were estimates dur-ing the concept phase can now be updatedto reflect internal and subcontractor actu-alities. Other physical parameters becomemeasurable as prototypes are built. ThePRICE system becomes a tracking machinewith this type of periodic parametric main-tenance. Such a machine produces monthlycost -allocation reports that can detect poten-tial overruns before they propagate. Also,the far-reaching life -cycle cost impacts ofproposed engineering changes can be imme-diately viewed and presented to the design-ers. This tracking procedure ensures thatthe cost and schedule consequences of eachdesign change are thoroughly analyzed. Pe -


riodic analysis is critical to keeping thetask within its bounds.

Time savings due to automationThe conceptual analysis and design -to -costtracking of the PSP were accomplished infar less time than conventional bottom -upestimating methods. Table V shows theestimated man-hours that were saved byusing the PRICE system. This savings wasrealized due to the high degree of automa-tion the PRICE system provides.

The "PRICE" of successThe results and the time savings presentedin this scenario are impressive. However,the road to the level of automation thatcan be achieved with the PRICE systemmust be initially cleared. Model calibra-tion is the key. Organizations that havecalibrated the PRICE products typicallyachieve savings by reducing estimating cy-cles to 8 to 12 times less than that of con-ventional methods. To realize this savingsrequires a significant commitment by thePRICE user. Special training is required,and there is the preliminary model valida-tion and acceptance phase that all organi-zations must experience. Model calibrationnever really ends. There is always the needto substantiate the PRICE performance, asthere is with the performance of otherestimating methods, including conventionalones. The most prominent characteristic ofan organization that successfully uses thePRICE system is an energetic commitmenttoward making the PRICE system workfor them.

Access to the PRICE system

The PRICE system is available to allPRICE -trained RCA personnel. Over 100RCA personnel have been trained in thePRICE system. Anyone desiring trainingmust attend one- to two-week courses inMt. Laurel, New Jersey. Each attendeereceives a computer -based work area anda complete set of reference material. Train-

ees learn the theory of the models, tech-niques to develop parametric data, andmethods of analyzing their results. Contactwith PRICE personnel does not stop whenthe courses are over. Trained PRICE usersare supported by a fully staffed Operationsgroup as well as periodic newsletters andtechnical bulletins which keep them abreastof current methods and events. For detailson PRICE training enrollment, contact Ms.Geri Devlin, (609) 338-5215.

ConclusionThe PRICE parametric system of estimat-ing saves time. The potential use of thePRICE system throughout the life of aprogram was illustrated by the Program-mable Signal Processor study. This andother cases prove that the PRICE systemis a valuable tool for systems engineering,as well as many other disciplines.

As an organization matures, managementidentifies opportunities to mechanize cer-tain tasks and reduce costs. This phenom-enon creates a steady rate of technologicalimprovement within the industry. However,dramatic productivity increases come aboutnot by mechanizing existing tasks, but ratherby using new methods that stimulatechanges in an organization and in the wayinformation is handled.

The PRICE system offers a means totruly automate the estimating process. Astrong commitment to the PRICE metho-dology will enhance the flow of informa-tion within an organization and facilitatefaster and more informed decisions.

AcknowledgmentsThe author wishes to acknowledge theguidance and support of the entire PRICESystems Operations and Engineering staffin the development of this paper.

References

1. J.T. Threston, "Managing the Future-The Expand-ing Role of Systems Engineering in a High Technol-ogy Society," RCA Engineer, Vol. 28, No. 4, pp.13-22 (July/August 1983).

Anthony DeMarco is a member of theengineering staff of RCA PRICE Systems.In this position, he researches, designs,and implements computer -based paramet-ric estimating models and supports theexisting family of PRICE computer models.His most recent assignments include thedevelopment of the PRICE microcircuitmodel, PRICE M, and on -going enhance-ments to the PRICE hardware model,PRICE H. Mr. DeMarco received a BSdegree in Mathematics from St. Joseph'sUniversity of Philadelphia, Pa., in 1979. Hejoined RCA PRICE Systems in 1981 afterteaching high school mathematics andcomputer science for two years. He is cur-rently pursuing a Masters of Science inComputer Science from the New JerseyInstitute of Technology.Contact him at:RCA PRICE SystemsMount Laurel, N.J.TACNET: 222-5836

2. No Author. PR/CE-An Executive Guide.

3. H.A. Freedman, "Information Technology Manage-ment," RCA Engineer, Vol. 28, No. 4, pp. 87-92(July/August 1983).

4. A. DeMarco, "The Economics of Choosing a Cus-tom Microcircuit Approach," Proceedings of theIEEE 1983 National Aerospace and Electronics Con-ference.

5. M.H. Burmeister, "RCA PRICE System-The Engi-neer's Flexible Toot," RCA Engineer, Vol. 26, No.5, pp. 21-27 (March/April 1981).

6. C. Hopkins, "Using PRICE to Track the Design-to-Cost/Life-Cycle-Cost Progress," Proceedings of thePRICE Los Angeles Symposium (February 1984).


H.G. Patterson J.L. Magos

CMOS technologymoves towards IC leadership

Semiconductor processing advances are opening new vistasfor CMOS technology. In microprocessor applications, memory,and logic devices, CMOS appears ready to displace two of itslong-time competitors, NMOS and TTL.

Ever since their development at RCA'sDavid Sarnoff Research Laboratories inPrinceton, N.J., in the early 1960s, Com-plementary -Symmetry Metal -Oxide Semi-conductor (CMOS) devices have movedsteadily to the forefront of the digital andlinear integrated -circuit technologies. Orig-inally conceived as a logic -device family,CMOS technology now spans the full rangeof digital circuitry from microprocessorsand memory chips to gate array and stan-dard -cell semicustom devices. CMOS useis growing in linear circuits such as opera-tional amplifiers and analog -to -digital con-verters, and it is fast becoming a key tech-nology for the emerging telecommunica-tions and data -communications markets.

Abstract: This tutorial on CMOStechnology reviews the fast-moving historyof metal -oxide semiconductors and showshow CMOS in particular holds the mostpromise for devices in the 1980s and1990s. The low power, the high noiseimmunity, and the wide operating -temperature range, all benefits of CMOS,are becoming increasingly important tocircuit designers. Enhancements, such asQMOS for higher speeds, and HCT forTTL logic functions, promise to broadenCMOS popularity. The authors concludewith a look at the other CMOS devices,such as static RAMs and peripheral chips.

©1984 RCA CorporationFinal manuscript received April 30. 1984.Reprint RE -29-3-10

In view of its wide acceptability, mostknowledgeable observers expect CMOS tobecome the dominant semiconductor tech-nology of the 1980s and beyond.

MOS and bipolar typesCMOS is one of three types of semicon-ductor circuit techniques that useMOSFETs (Metal -Oxide Field -EffectTransistors). The other two are PMOS(p -channel MOS) and NMOS (n -channelMOS). CMOS technology is unique inthat it combines both PMOS and NMOStransistors on a single silicon chip. Theprincipal advantage of this arrangement-but by no means the only one-is that aCMOS device consumes substantially lesspower than any other major form of inte-grated -circuit semiconductor.

In both semiconductor structure and per-formance, all MOS devices differ dramati-cally from the transistors of the other majortype of integrated -circuit technology, thatis, bipolar. The use of bipolar technologyis exemplified by the well-known digitalIC family known as Transistor -TransistorLogic or TTL. Although individual bipo-lar and MOS transistors can be describedideally as switches, their operating charac-teristics are exactly opposite.

A bipolar transistor is a current -con-trolled switch having a low input impe-dance, while an MOS transistor is a vol-tage -controlled switch with a high inputimpedance. An MOS switch is much likethe old-fashioned vacuum tube switch, be-

cause it is voltage -controlled with high inputimpedance. Bipolar technology is uniquelybuilt for high current and high speed-andit has high power consumption. CMOStechnology is characterized, in general, bylower current and speed-and it consumesvery little power. It is also perfectly suitedfor basic logic devices up through the densecircuitry required by the coming genera-tion of Very Large Scale Integrated Cir-cuits (VLSI).

On the other hand, bipolar logic circuit-ry consumes too much power to be usedfor VLSI-and the high power is usually ahandicap in its use in even the smallestscale integration. This fundamental differ-ence accounts for the diverse roles playedby each in the world of digital and linearICs. To understand how basic CMOS de-vices operate and are constructed, refer tothe box, "CMOS devices, top to bottom."

Four user benefits of CMOS

Low power

The ultra -low power consumption ofCMOS translates into a number of advan-tages for users as listed in Table I. Forexample, if an IC technology is powerconserving, designers need less expensiveand sophisticated power supplies, whichnaturally reduces system costs. Moreover,such supplies can operate without specialcooling equipment-fans, blowers, and soon-providing a savings in both the costand physical size of the equipment. Low -


power consumption also permits the designof battery -powered, portable products. Infact, CMOS is the only technology consid-ered by designers for products of this type.

Wide voltage range

A close second to the low -power advan-tage of CMOS is the advantage of thewide operating voltage range of CMOSICs. There are CMOS IC types that oper-ate at as high as 20 V. The widely usedCD4000B logic family operates from 3 Vto 18 V. Regulation is rarely needed andbattery life is extended due to this charac-teristic. In contrast, most TTL logic mustoperate at 5 V ±0.25 V; this means thatexpensive, high -current, and well -regulatedsupplies are a must with TTL, but notwith CMOS logic.

Switching -level stability

Another parameter favoring CMOS circuitsover TTL IC circuits is switching -level sta-bility over wide temperature ranges. Asthe table shows, plastic -packaged CMOSICs are specified to operate over a -40 to+85°C range. Plastic packages are the mostcommon and inexpensive types, filling mostindustrial and consumer applications. Ceram-ic -packaged CMOS devices are rated foroperation over the full military tempera-ture range of -55 to +125°C.

Plastic -packaged devices in other IC tech-nologies-NMOS, PMOS and bipolar-are usually rated for operation only over a0 to +70°C range. Because of this restric-tion, such technologies are a definite hand-icap in certain extended -temperature -rangeapplications. One of these, automobile -en-gine controls, is increasingly movingtowards the CMOS camp. If an enginecontroller is intended for under -the -hoodservice, CMOS becomes the favored tech-nology because its stable temperature char-acteristics can withstand the wide fluctua-tions experienced in the automotive environ-ment. There are many other "wide -temper-ature -range" applications of this type inspace, military, consumer, and industrialsystems that rely on very stable operationover extended temperatures.

Noise immunity

A fourth characteristic that sets CMOSatop all other IC technologies is its highimmunity to electrical disturbances (noise)generated within a system or arising fromexternal sources. In digital IC specifications,two parameters are defined-noise-im-

Table I. CMOS characteristics and benefits.Key characteristics Benefits

Very low power consumption:Typical values at 1 MHz/5V

Gate 1 mWCounter 3 mW8 -bit micro 5 mW4 K RAM 20 mW

Very low standby currente.g. 16K CMOS RAM15 µA/3 V/70°C

High noise immunitye.g. 1.5 Volts at VDD 5 Volts

Wide temperature range:Plastic -40 to +85°CCeramic -55 to +125°C

Low power supply costs

No cooling fans (lower cost/small size)

Battery operation/standby/portable

Allows backup storage withsmall/low-cost battery

Operation reliably in noise environmentswithout special filtering

e.g. process controllers

Reliable operation in hostile conditionse.g. automotive, engine controland military

munity voltage and noise margin. Noise -immunity voltage-either a logic high orlogic low-is that noise voltage at any oneinput to a logic device that does not causea false signal to propagate through the sys-

tem. Noise margin is the difference betweenthe specified input voltage of a device andits specified output voltage. Figure 1 illus-trates the concepts involved in noise defini-tions.

Noise margins are defined in terms ofinputs and outputs to gate elements, thebasic building blocks of all digital circuits.In Fig. 1, the input voltage to the loadgate is the output voltage of the driver

gate. Both gates in this example operatefrom a supply voltage of +5 V. For CMOSgates operating at a 5-V supply level, themaximum logic 0 input -voltage (Vu. max)to a gate can be 1.5 V. This means thatthe driver gate's logic -0 voltage can be ashigh as 1.5 V and still be recognized bythe load gate as a logic -0 voltage. Thus,the "logic -0 input region" in Fig. 1 extendsfrom 0 to 1.5 V. Similarly, the minimumlogic -1 input voltage (V,x min) can be 3.5V. That is, if the driver gate's output is aslow as 3.5 V, the load gate will still recog-nize this level as a logic -1 voltage. The"logic -1 input region" in Fig. 1 extends

DRIVER

-VOL

vim11-_

viL -

5 0 - OUTPUT CHARACTERISTICS

4.95 - - - - VDD0i=a1

LOGIC -IOUTPUT

,v0H

REGION3 5 - - - VIH MIN

I 5LOGIC -0 VIL MAXOUTPUTREGION /- --VOL

0.05- - - -I- J

VSS

LOAD

INPUT CHARACTERISTICS

LOGIC -IINPUT //,

/ REGION //////////INDETERMINATE

REGION

77/7.17.LOGIC -0 7/INPUT /

/ REGION.////////// /1

Fig.1. The noise margins of the CMOS CD4000B family are wider than those ofany other semiconductor technology. Because the logic -1 and logic -0 input regionsshown above are 1.5 V wide, CMOS devices are more immune to noise than otherlogic forms. In TTL logic, for example, the same input regions are just 0.4 V wide.

Patterson/ Magos: CMOS technology moves towards IC leadership 55

CMOS devices, top to bottom

From the most complex microprocessor circuits tothe simplest gates, deep within all CMOS devices isa basic circuit element called an inverter. As shownin Fig. A, a CMOS inverter is composed of a p -chan-nel MOS transistor at the top and an n -channel tran-sistor at the bottom. Because n -channel and p -channel transistors have diffusions of oppositepolarity dopings, they operate with voltages of differ-ent polarity. Thus, a positive input voltage turns onan n -channel device but turns off a p -channel. Con-versely, a zero voltage (ground as shown in Fig. A) atthe input to the inverter turns on the top (p -channel)transistor and turns off the bottom (n -channel) tran-sistor. Note at the bottom of Fig. A that both transis-tors can be represented as switches to simplify theexplanation.

This simple CMOS circuit performs the most basicof logic operations in a digital system, namely, that ofinversion. If the voltage applied to the input (In) ter-minal is positive, the p -channel switch turns off, dis-connecting the output (Out) terminal from the supplyvoltage (+). The Out terminal is therefore at groundpotential because the n -channel switch is turned onand connected to ground. Now if the voltage at the Interminal is zero, the p -channel switch turns on, con-necting the Out terminal to the positive supply vol-tage. This time the n -channel switch turns off. Thissequence describes the basic logical operation ofinversion; the output logic (voltage) level is alwaysopposite that of the input.

In an actual CMOS IC, p -channel and n -channeltransistors are fabricated on a silicon metal sub-strate, as shown in Fig. B. The contact terminal ofeither transistor is the gate. In the original CMOStechnology, the gate was made of metal, whichformed one plate of a capacitor. The other plate wasthe substrate of the device, and the two plates wereseparated by silicon dioxide (Si02), an insulating

In°

+VDD

H rig PMOS

Out

NMOS

CMOS vs NMOS

High

IDE)

0- -Low

CMOS INVERTER INVERTER OPERATIONNote:- !OD Flows When Input is High

Fig. A. The basic CMOS inverter.

METAL METALINTERCONNECTS INTERCONNECTSDRAIN SOURCE SOURCE DRAIN

GATE METAL

GATEOXIDE

PMOSTRANSISTOR

VDD

IN

NMOSTRANSISTOR

OUT

Fig. B. CMOS construction on a single chip.

from 3.5 V to the power -supply rail (+5V).

Note that both input regions are 1.5 Vwide, that is, they have noise margins of1.5 V. Assume, for example, that the drivergate output (voltage) is a logic -0 voltage,that is, a nominal VOL level of 0.05 V.Therefore, VOL can have 1.45 V of over-riding spurious noise and yet the load gatewill be able to recognize the signal as alogic -0 level. In effect, there is a margin oferror of 1.45 V against the effects of noise.The same holds true for the driver gateoutput logic -1 voltage of 4.95 V, that is,V 0 H. Noise can pull this level down to 3.5V, while the load gate continues to recog-nize a logic -1 signal. Obviously, the largerthe noise margin, the greater the device's

immunity to error from unwanted noise.CMOS provides the highest specified

noise immunity (or noise margin) of anylogic family. TTL logic, by contrast, offersa logic "0" ( VH, ,,,) of 0.8 V and an outvoltage of 0.4 V (VOL ,,). Thus, the noisemargin is only 0.4 V compared with 1.45for CMOS! Therefore, spurious noise is

more likely to introduce errors than is thecase with CMOS. Furthermore, CMOSlogic operating at 15 V is specified to pro-vide 5 V of noise immunity at both "high" -and "low" -logic levels. The foregoing dataillustrate the fact that CMOS can serve inapplications in which comparatively highlevels of electrical noise can be encoun-tered. Two notable examples are factoriesthat use heavy machinery and automobiles;

in the latter, solenoids, ignition, horn, andso on, generate lots of high-energy noise.

CMOS in perspectiveWith its multiplicity of technical advan-tages over other semiconductors, CMOS isalready becoming the leading IC technol-ogy, especially for digital circuits.

First, Fig. 2 illustrates the expected useof MOS technologies PMOS, NMOS,CMOS, HCMOS through the end of thisdecade. The projections indicate thatCMOS will experience the most dramaticgrowth of any technology. By 1990, accord-ing to Integrated Circuit Engineering,CMOS will command about 50 percentof the IC market, up from about 20 per -


material. Because the gate input is a capacitor, theinput current is extremely low (for descriptive pur-poses, it can be assumed to be zero).

When the output of one CMOS inverter drives thecapacitive input of another, no IR (resistive) voltagedrop occurs in the drain circuit as a result of theinput current (assumed to be zero). Therefore, theoutput voltages can rise or fall to their full positive(power supply voltage) or negative (ground) levels.Moreover, since one transistor is on and the other isoff, the net current in the drain circuit-called quies-cent current-is extremely low, typically in thenanoampere (10-9amps) range. For this reason,CMOS devices consume the least power of anysemiconductor technology.

Another observation from Fig. A and the operatingdescription given above is that this operation occursover a 3 V to 18 V supply voltage (CD4000B logic).Also, the ideal switching voltage is 50 percent of thepositive supply voltage (a 1.5 V switching level for aVDD of 3 V and a 9-V switching level fora VDD of 18V). Over a temperature range of -55°C to +125°C,this switching voltage varies less than ±1 percent ofthe supply voltage.

Some years ago, a significant processing break-through occurred in MOS technology. This was thedevelopment of the silicon gate as a replacement forthe metal gate. In silicon -gate technology, a con-ducting form of silicon serves as the gate materialinstead of metal. Not only does the silicon gateimprove the performance of CMOS devices, it allowsthe gate to be formed before the drain and sourceare diffused (see Fig. C). In this process, a patternedpolysilicon material is used to define the gate areaso that the following source and drain diffusions areself -aligned to the gate. The process, called self -aligned silicon gate, is now standard in all CMOSdigital devices.

Among the advantages of the self -aligned silicon -

Fig. C. CMOS self -aligned silicon -gate technology.

gate process compared to metal gate are loweroperating power consumption, higher speed, higherdensity, and lower operating voltages. Silicon -gateCMOS devices are characterized by low -powerconsumption and high density (many devices in agiven amount of chip area). By contrast, bipolar (TTL)devices are characterized by low density and highpower. Obviously, however, each technology has anappropriate role in digital system design.

Not content to stand still, RCA for example, forgedahead in processing technology with the develop-ment of self -aligned silicon gate CMOS on a sap-phire substrate. Called CMOS/SOS (CMOS/Silicon-on-Sapphire), the process produces the highestfrequency MOS integrated circuits. In the fabricationtechnique, a thin film of heteroepitaxial silicon isdeposited on a sapphire substrate. High-speeddevices are possible because the process elimi-nates many of the parasitic capacitances of conven-tional structures. Distributed parasitic capacitanceimposes frequency limitations on a circuit. In addi-tion to high speed, CMOS/SOS ICs have the highestresistance to nuclear radiation. This makes them thepreferred technology for ICs used in aerospace andcertain military systems.

cent at present. HCMOS is a new high-speed version of CMOS considered pri-marily for logic devices and is included inthe total of 50 percent. Over the sameperiod, the use of NMOS will remainabout the same as at present, while PMOSwill be on the decline as a digital tech-nology.

Historically, PMOS is the oldest of theMOS semiconductor processes and wasused in the early memory devices of theearly 1970s. PMOS was comparatively easyto build, but it had some major draw-backs, including: it requires two power -supply voltages for operation rather thanthe single supply voltage of all other tech-nologies, it's slow, it's very temperaturelimited, and it consumes appreciable power

when compared to CMOS. NMOS, whichis more difficult to fabricate, appeared onthe scene in the early 1970s, just in timeto catch the phenomenal growth of micro-processor and memory devices. The bigadvantage of NMOS over PMOS is thatits current carriers are electrons rather thanholes. The greater mobility of electronsmeans that the same size NMOS transistoris two to three times faster than its PMOScounterpart. Looking at it another way,for a particular speed a transistor in NMOSis smaller than the PMOS version andhence, NMOS chips are much smaller.NMOS also runs at lower threshold (turn -on) voltages and can operate from a single+5 V supply, making it compatible withother logic technologies.

As a result of these features NMOSbecame, and still is, the leading technologyfor the fabrication of microprocessor, mem-ory chips, and VLSI circuits in general.Interestingly, however, neither NMOS norPMOS was much of a factor in logic -fam-ily ICs-inverters, gates, flip flops, coun-ters, and other such devices classified assmall-scale integration or SSI. This vastarena is dominated by bipolar (TTL) tech-nologies. The only competition here camefrom CMOS, but the two logic technolo-gies split off into different application areas(TTL for high-speed computers and CMOSfor most other uses).

As is well known, computers will con-tinue to be not only the fastest -growingmarket in the electronics industry, but the

Patterson/Magos: CMOS technology moves towards IC leadership 57

major consumer of ICs as well. Of course,CMOS logic serves in computer systemsalso, but primarily in peripheral circuitsthat can tolerate its slower speed constraints.Nevertheless, the CD4000B series ofCMOS logic, manufactured by semicon-

ductor giants such as RCA, Motorola, andNational Semiconductor, has grown intoone of the most widely used SSI familiesin the electronics industry. As describedpreviously, CD4000B logic offers users sig-nificant advantagaes over TTL-lower pow -

Table II. CMOS logic versus two competitors.LS TTL NMOS CMOS CMOS advantage

Gate delay

Power (f/flop at 1 MHz)

Noise immunity

Temp. range (commercial)

Supply tolerance

8 nS 8 nS 8 nS

20 mW 8 mW 2 mW

0.4 V 0.4 V 1.5 V

-40° +85°C -40° +85°C -55° +125°C

±0.25 V ±0.25 V 2 V -18V

NMOS - 37 °'o

CMOS - 49 0/0

GaAs

Fig. 2. CMOS is emerging as the leading semiconductor technology of the decade.if the projections shown here are on target, CMOS technology will capture 50percent of the total digital chip market by 1990. Other technologies, by comparison,are either declining or remaining static.

904000B

-41CMOS

VCC 5 VCL = 15 pF

3020

10 HC/HCT LSTTL

5 ALS

FAST 10 KH3 AS S ECL2

1. I

0.001 0.1

10 KH ECL4

2 3 5 10 20

TYPICAL dc POWER DISSIPATION (mW)

Fig. 3. A dramatic increase in its speed capability has pushed CMOS into the samespeed range as bipolar (LSTTL) technology. But CMOS logic consumes far lesspower than bipolar, giving it a significant edge in all but the highest -speedapplications.

er, wider voltage range, wider temperaturerange, higher noise-margins-with the onlylimitation being a lower speed. Also, whencompared with the higher -powered TTLat 5 V operation, CMOS is very reliabledue to its lower IC chip -power dissipation.

Dramatic changes are occurring in thesemiconductor industry of the 1980s, includ-ing the reversal of the traditional roles ofdigital technologies. One of the most far-reaching is the emergence of a new classof CMOS logic, which runs at the samespeed as one of the most widely used TTLfamilies, LSTTL, for low -power SchottkyTTL. QMOS (Quick CMOS) by RCAoffers the best of both worlds-the speedof LSTTL and the low power consump-tion of CMOS.

Figure 3 shows the speed versus powercharacteristics of the major semiconductortechnologies used in logic devices. Notethe position of the HC/HCT CMOS fam-ily. It is about ten times faster than itsolder brother, 4000B CMOS, and just asfast as its principal TTL competitor, low -power Schottky TTL (LSTTL). AlthoughHC CMOS and LSTTL operate at thesame speed-about 9-ns average propaga-tion -delay time-the difference in powerdissipation overwhelmingly favors CMOS.Its dissipation is barely measurable (sev-eral orders of magnitude less) on the powerscale in Fig. 3, but that of LSTTL is about1.5 mW. Generally, a CMOS system dis-sipates about one-hundreth of the powerof LSTTL logic when the operating, ordynamic, power is considered along withdc power.

The implications of fast CMOS havesignificant importance to designers. For thefirst time, they can obtain bipolar perfor-mance along with all the power, voltage -range, temperature and noise -margin en-hancements of CMOS. In fact, a specialCMOS family called HCT has been createdto provide pin -for -pin and specification -for -specification replacements for all of theTTL logic functions. The basic family, HC,is intended primarily for new all-CMOSdesigns. Therefore, in all but the highest -speed applications, the HC/HCT family ofCMOS logic will be the successor to theTTL family. Since the majority of data -processor digital circuits operate in thespeed range of fast HC/HCT CMOS, thefamily is well positioned to replace TTLas the leader in digital IC logic technology.

Table II summarizes the key parametersof CMOS, NMOS, and LSTTL. The speedenhancement made possible by QMOS-type processing advancements gives CMOS


a substantial advantage over the other de-vices.

CMOS-a completefamily of devices

While CMOS logic gradually overtakesTTL, the advantages of the technology arecontinuing to make an impact on micro-processor and memory ICs. In micropro-cessors, RCA introduced the first CMOS8 -bit chip, the CDP1802, in 1976. It hasbecome the most popular CMOS micro-processor. Moreover, the same technologi-cal advantages responsible for producingfast CMOS logic-processing, lithogra-phy-have spawned a variety of improvedversions of the CDP1802.

The family tree of Fig. 4 shows the evo-lution of RCA's CDP1800 series of micro-processors. Two trends are evident. First,the speed of CMOS microprocessors isincreasing. The latest versions run at clockfrequencies of 5 MHz compared to the 3.2MHz of the CDP1802A. Second, improvedsemiconductor processing allows a largernumber of system functions to be incorpo-rated on -chip. The CDP1804A, for exam-ple, contains 2 kbytes of read-only memory(ROM), 64 bytes of random-access mem-ory (RAM) and a timer/counter circuit inaddition to the basic microprocessor. Inaddition, users obtain all of the power,temperature, and noise -margin benefits asso-ciated with CMOS technology. Thus,CMOS microprocessors can be applied in

256 K

64 K

>-

16 K

Lt.)

4 K

1 K

1800 SERIES

180265 MHz HP

1806A5 MHz HP

TIMER/CNTR

Fig. 4. Beginning with the 8 -bit CDP1802A microprocessor, RCA's family has stead-ily increased in both speed and functionality. Together with the power savings,wider temperature range and noise -margin advantages of CMOS, these proces-sors are on the verge of industry leadership.

equipment in which it is virtually impos-sible to use either a bipolar or NMOSmicroprocessor.

As microprocessor -based systems becomemore memory -intensive, the need for high -density, low -power memory devices in-creases. CMOS technology's forte is instatic RAMS (rather than dynamic RAMs)and large ROMs. These two types of mem-ory chips are projected to experience thelargest growth increases of any memorydevices during the period to 1990. Onereason-applicable to RAMS-is that astatic RAM requires far less control circuit -

72 74 76 78

YEAR

80 82 84

1000 nS

500 nS

200 nS

100 nS

50 nS

(a) (b)

Fig. 5. Two reasons why CMOS static RAMs are ousting their NMOS competitionare (a) increased density and (b) faster access times. With their high speed andlow -power dissipation, CMOS RAMs are expected to be one of the leading memorytechnologies by the end of the 1980s.

ry (called overhead) than a dynamic RAM.Because the basic storage cell of a RAM

is more complex than that of a ROM, thetotal storage capacity of RAMs is muchsmaller than that of ROMs. At present,static RAMs are commonly available in16-kbit sizes with 64-kbit chips expectedshortly. By contrast, ROMs can be as largeas 256 kbits, four times the storage capac-ity of RAMS.

Users can look forward to two trends inmemory technology. The first is that densi-ties-the number of bits/chip-will con-tinue to grow larger. Just as important, the

76 78 80YEAR

82 84

Patterson/ Magos: CMOS technology moves towards IC leadership 59

speed of memory devices continues in anadvancing trend. Figure 5 shows these twodevelopments compared to NMOS tech-nology for static RAMs.

As most users are aware, a completesystem requires support devices to transferdata between the processor and memory,handle input/output (I/O) functions, per-form timing functions, and other such so-called housekeeping chores. A wide rangeof peripheral chips is available in CMOStechnology to support the microprocessorsand memories. In the arsenals of RCAand other manufacturers are chips such askeyboard interface devices, interrupt con-trollers, real-time clocks, UARTs (Univer-sal Asynchronous Receivers/Transmitters)and video controllers. These and other de-vices permit users to construct systems byassembling building blocks of completelycompatible components.

The decades of the 1980s and beyondhold great promise for CMOS technology.Because of its inherent characteristics, it

stands on the doorstep of technology lead-ership across the broad range of all inte-grated circuits.

H. Gene Patterson, Director of Memory/Microprocessor Operations, RCA SolidState Division, graduated with a BSEE fromthe University of Illinois, Champaign -Urbana, Ill. He received an MSIA degreefrom the Graduate School of Administra-tion, Carnegie-Mellon University, Pitts-burgh, Pa. Before coming to RCA in 1981,he worked at Sprague Electric, in Massa-chusetts, and Harris Semiconductor, inFlorida.Contact him at:RCA Solid State DivisionSomerville, N.J.TACNET: 325-7087

RCA Review available nowRCA Review is a technical journal publishedquarterly by RCA Laboratories in conjunctionwith the subsidiaries and divisions of RCACorporation. Copies of the issue are available(at a cost of $5.00) by contacting the RCAReview office at Princeton, TACNET: 226-3222.

Contents, March 1984PIN-FET Receiver for Fiber OpticsS.A. Siegel and D.J. Channin

The Effect of Carrier -Induced IndexDepressions on Fundamental -Transverse -ModeCharacteristics in DH Laser StructuresD. Botez

Low Expansion Porcelain Coated Copper -CladInvar SubstratesK.W. Hang, A.N. Prabhu, J. Andrus,S.M. Boardman, and L.S. Onyshkevych

Jim Magos earned his B.S. from LongIsland University and his M.B.A. fromColumbia University. He joined the SolidState Division in 1973 as a design engi-neer for power transistors. He transferredto the CMOS Logic Marketing departmentin 1975 and was promoted to Manager in1980. In December 1983, Jim becameDirector, Standard IC Products Marketingencompassing bipolar, CMOS, QMOS, andtelecommunications products.Contact him at:RCA Solid State DivisionSomerville, N.J.TACNET: 325-6717

A Novel Memory Device for VLSI E2PROMS.T. Hsu

A Study of the Etching Kinetics of Low CarbonSteel Using the Ferric Perchlorate-PerchloricAcid System as a ModelR.B. Maynard, J.J. Moscony, and M.H. Saunders

Ferric Chloride Etching of Low Carbon SteelsR.B. Maynard, J.J. Moscony, and M.H. Saunders

Interaction with Metal Films on Glass: A NewMethod for Evaluating Solder ParametersS. Tosima, S. Harada, and E.O. Johnson

Extending the Content and Expanding theUsefulness of the Simple Gaussian LensEquations-Part 3-AL.T. Sachtleben


The Engineer's Notebook

A day in the life of an information -efficient scientist

Paul SchnitzlerRCA Laboratories

Princeton, N.J.

My name is Thursday, Annie Thursday. I'm ascientist.

6:58 A.M. I just rolled out of bed; looked in the mirror;didn't like what I saw. Got moving anyway.

N:00 A.M. After the usual things, I found myself in mylaboratory. Punched the ON button on the terminaland went to hang up my coat. Coffee smelled good;somebody left some for me.

Logged on; checked mail. Three notes from theboss: "Need to see you, Thursday," "Need to seeyou, Thursday," and one flashing, "Where the hell areyou?" The next three items are memos, two on videoprocessing, the third on the new PAX machine. I seethis last one is written by Joe Friday.

Key in message to boss: "I'm back-when's a goodtime?"

Call up CALENDAR. I see it's my son's birthday

Courtesy IEEE Communications Magazine. Adapted from Vol. 20, No. 3, pp. 24-25

(May 1982).

tomorrow and my anniversary is coming up in twoweeks. I have a meeting tomorrow morning. Today isclear.

8:07 A.M. Now to work. Turn on the main power tothe bench and drink some coffee. Jack Kraft, thetechnician I work with, tells me that everything hasbeen running okay for the last three days. He wantsto know why I came in this morning. I throw aneraser at him.

Started measurements on the new video modulator/up converter. It's the old story, set a frequency,check the level, record the amplitudes, do it again.Key each value into the terminal, one at a time.Called up PLOT. Well, that helps; at least the curve issmooth, but something's wrong near point ninemegahertz. I'll take more points.

Enter points, PLOT. That's better; save this stuff. I tellKraft to continue the run; it looks like it's going well.

Message -waiting signal. The boss again: "Forget it."Fine.


The Engineer's Notebook

8:48 A.M. Let's see what Friday's latest report has tosay.

Spend the next hour reading reports on the tube.Take some notes; I want copies of a couple of pageshere and there. Send a particularly interesting figureto the color printer.

9:51 A.M. Jack says the data is done. I call up thegraphs on the terminal one after another. Seven lookinteresting and I request copies.

Jack will pick up the figures and the graphs while Icheck the shop schedule, on the tube again, for thestatus of the new modulator/up converter assembly.

The model shop schedule shows a delay. No reasonentered.

10:14 A.M. It appears some personal reconnais-sance is called for. I key in a note to Jack that I'mgoing down to the shop.

I meet Sam Waterford in the hall. He is working onfiber optics these days. He suggests a cup of coffeeand I say fine. Sam tells me he has new data onsome low -loss crystal material. I must remember tolook it up. I step into the next office and find a freeterminal; I key in a quick note to myself to check outWaterford's work.

10:38 A.M. The shop. After nosing around a fewminutes, I find my job has been delayed because of anew high -visibility program that got in ahead of me. Italk to the administrator and with a little sweet talk Iget a new, firm date. Well, that cashed in one favor.

11:00 A.M. I'm back at the lab. I find Jack has beengoing strong. The graphs are on my desk and theyneed some study. They don't make much sense;there is some inconsistency between them. I can'ttell whether there's noise in the measurements orreal artifacts are present. Perhaps some additionaldata is required. But Jack is busy right now.

11:58 A.M. Len Awsom looks in. Suggests lunch. Thesalad and yogurt are okay. The argument about theprice of butter in the Third World countries makes nosense at all and lasts 45 minutes. Not okay.

1:01 P.M. I see Jack; he tells me his other experi-ment is performing strangely and he wants me tocheck it out. I agree. We call up the statistical -analy-sis package to check the significance of the data.The difference of the means shows the confidence isjust over 50%, not nearly good enough. We agreemore data is necessary and, on the terminal, I enterchanges in the experimental plan to reflect thesenew requirements. The tube responds with sug-gested revised milestone dates for the project. Thenew dates look acceptable and I okay them.

2:13 P.M. I have been thinking about that technicalreport I'm writing and I have some new ideas. I call

up my current version and type in some new text. Ina few minutes, I have placed it in the correct locationin the body of the report. I realize that a figure wouldbe helpful at this point. I check my project experi-mental data file and I find just the figure I want. Inotice the address in memory and "send" a copy ofit to the report.

I read the whole report at this point; it reads well. I'lllet it stew in my mind overnight just to be certain andI'll release it for distribution tomorrow.

2:39 P.M. I call Ellen Strong, the hot -shot divisionprogram manager. The recent manufacturingchanges are going well, but she tells me of a newproblem. The schematics are in file and I look themup on the tube. After some discussion, we agree onthe problems and some changes that might help. Ipropose some tests to check out the changes and,with some modifications, she agrees. The data willbe recorded at the division and will be accessiblethrough my file.

3:29 P.M. More coffee. Jack talks of a date he hadlast night. I listen; I guess I'm a voyeur at heart.

3:41 P.M. Back to those graphs. I still don't knowwhat to do with them. Awsom had provided an anal-ysis of the problem and I applied his techniques tothe data. There are some serious discrepancies.

The only way I'll resolve this is if I meet with Len andJoe. I use the CALENDAR routine to find some opentimes on all our schedules. 11:00 A.M. tomorrowlooks good. I send a message to each of them fortheir approval of that meeting time.

4:16 P.M. I finally have some time to see Waterford'scrystal -loss data. There is a lot of data and it takesme some time to look it over. There are a number ofinteresting facets. I need to talk to him more about it.4:50 P.M. Uh, oh, late again. Before leaving, I key upthe TV listings. Looks like the only thing on tonight isa new film, "Through the Kinescope." I guess I justcan't get away from it.

Better type in a note to my husband that I'm going tobe late. He'll get it when he gets home.

6:12 P.M. Dinner. Same old stuff, but tasty. My eight -year -old Jill comes in. She shows me some whitematerial in her hand and wants to know what it is. Itell her it's paper and she asks what the funny marksare. I tell her that's writing. I explain that that's theway we used to send each other messages. I alsotell her that I did all my homework on that stuff whenI was her age. She looks at me like I'm an escapeefrom a looney bin.

Kids!

Contact: P. Schnitzler TACNET: 226-2199


Pen and Podium Recent RCA technical papers and presentations

To obtain copies of papers, check your library or contact the author or his divisionalTechnical Publications Administrator (listed on back cover) for a reprint.

AdvancedTechnology Laboratories

G. AmmonPerformance of an Optical Disk Jukebox-Presented at the Topical Meeting onOptical Data Storage, Monterey, Calif., andpublished in the Proceedings (4/18-20/84)

M. BeackenEfficient Implementation of Highly VariableBandwidth Filter Banks with HighlyDecimated Output Channels-Presentedat IEEE 1984 Intl Conf. on Acoustics,Speech, and Signal Processing, SanDiego, Calif., and published in theProceedings (3/19-21/84)

M. CrouthamelSpacecraft Thermal Design UsingInteractive Graphics-Presented at theAIAA 22nd Aerospace Sciences Meeting,Reno, Nev. (1 /9-12/84)

A. FellerCustom LSI Devices Using the StandardCell Approach-CAD/CAM Series of Textsto be published by Auerbach Publishers

J. Gaev P. Kleinosky1J. McAdamsTest Programs That Teach-Presented atthe IEEE Workshop on Automated TestProgram Generation, Crystal City, Md., andpublished in the Proceedings (2/8-9/84)

R. HackenbergComputer Processing of Phase StructureRules-Presented at CALICO (ComputerAssociated Language and LearningConsortium) Symposium, Baltimore, Md.,and published in the Proceedings(1/14/84)

W.A. HelbigHardware Descriptive Languages-Presented at the 17th Hawaii InternationalConference on System Sciences,Honolulu, Hawaii, and published in theProceedings (1/4-6/84)

R.A. McClain W.B. SchamingThe RCA Multifeature Statistical Tracker-Presented at ARO Workshop on AnalyticMethods for Target Acquisition andTracking, Ft. Belvoir, Va. (2/13/84)

Astro-Electronics

G. BeckThe ACTS Flight Segment: Cost -EffectiveAdvanced Communications Technology-

Presented at the 10th Annual AIAACommun. Sat. System Conf., Orlando, Fla.(3/18-22/84)

D. ChuHarmonic Responses Reanalysis ofModified Structures-Presented at theAIAA 25th SDM Conference, Palm Springs,Calif. (5/14/84)

S. Dhillon1H. Goldberg P. GoldgeierR. Sudarsanam17/12 GHz Communication Receiver forDBS-Presented at the Intl MicrowaveSymposium, San Francisco, Calif. 1984IEEE MTT-S (5/30/84)

J. EngelProduction of the SOOS CradleStructure-Presented at the Fifth StrategicSystems Project Office Mgf. Tech. Mtg.,Ontario Canada (3/28/84)

R. GounderChaired two sessions on SpacecraftStructures and Materials -29th NationalSAMPE Symposium and Exhibition, Reno,Nev. (4/3-5/84)

R GounderStructures and Material TechnologyInterfaces in Satellite Systems-Presentedat New York Chapter of SAMPE,Woodbury, Long Island (3/20/84)

S. Moochalla1D. Aubert20-GHz Lumped Element GaAs FET DriverAmplifier-Presented at the 84 IEEEMTT-S Conference, San Francisco, Calif.(3/84)

S. Moochalla1E. Kohut1D. AubertS. Dhillon1H. ZelenDual Mode Driver Automatic Level ControlAPM for Direct Broadcast SatelliteTransponder-Presented at the AIAA 10thComm. Sat. Systems Conf., Orlando, Fla.(3/84)

C. Profera1H. Soule 1J. RosenJ. Dumas 1J. MacGahanShaped Beam Antenna DBS-Presentedat the AIAA Comm. Sat. Systems Conf.,Orlando, Fla. (3/18/84)

A. Weinrich1M. FreelingRCA Advanced Satcom - The First AllSolid State Communications Satellite-Presented at the AIAA 10th Comm. Sat.Conf., Orlando, Fla. (3/19-22/84)

Automated Systems

J.W. BetzPerformance of the Deskewed Short -TimeCorrelator-Presented at the IEEEConference on Acoustics, Speech, andSignal Processing, San Diego, Calif. (3/84)

M.J. CantellaID.F. DionA Real -Time Histogram EqualizationProcessor for Dynamic RangeCompression of IR Video Images-Presented at the National Aerospace andElectronics Conference, Dayton, Ohio(5/84)

M.J. Cantella I F.F. MartinPerformance and Application of a160 x 244 Element Schottky Barrier IRFocal Plane Array-Presented at the 32ndNational Infra Red Information Symposium,San Francisco, Calif. (5/84)

G.R. EdgarLocation of Multiple Faults by DiagnosticExpert Systems-Presented at SPIE'sTechnical Symposium East '84,Washington, D.C. (4/84)

H.W. GrunbaumLearning Curves: The Theory andImplementation-Presented at the NCMAWorkshop 1984, Cambridge, Mass. (3/84)

J.D. McCreadyGraphic Design Industry-Presented atthe Regional Career Fair '84, University ofLowell North Campus (3/84)

E.M. MelendezChapter Motivation Workshop Talk-TauBeta Pi, Northeast Regional Convention,Northeastern University, Boston, Mass.(4/84)

S.B. MesnickPost Award Audit-Presented at theNCMA Workshop 1984, Cambridge, Mass.(3/84)

E.H. MillerMilitary Applications of CommercialComputer Technology-Presented at theCommand, Control, and CommunicationsTMSA Conference, San Francisco, Calif.(3/84)

K.I. PressmanComparison of FAR (Federal AcquisitionRegulations) with DAR (DefenseAcquisition Regulations)-Presented at theNCMA Seminar, Buffalo, N.Y. (3/84)


GovernmentCommunications Systems

D.E. BrittonFormal Verification of a Secure Networkwith End -to -End Encryption-Presented atthe 1984 Symposium on Security andPrivacy, Oakland, Calif., and published inthe Proceedings (4/30/84)

J.H. HooverA 10,3 -Bit Optical Disk Jukebox System-Presented at the Tape Head InterfaceComm. Colony 7, Maryland (3/84)

Laboratories

D.E. Ackley ID. Botez I B. BognerPhase -Locked Injection Laser Arrays withIntegrated Phase Shifters-Published RCAReview, Vol. 44 (12/83)

R.C. AligPicture Tubes for Color Television-Presented at the Virginia PolytechnicInstitute, Blacksburg, Va. (5/8/84)

R.E. AskewA Cooled Low -Noise GaAs FETAmplifier-Published RCA Review, Vol. 44(12/84)

I. Balberg IN. Binenbaum I C.H. AndersonCritical Behavior of the Two -DimensionalSticks System-Published Physical ReviewLetters, Vol. 51, No. 18 (10/31/83)

R.R. BartonDefect Location Clustering Schemes-Published North -Holland EuropeanJournal of Operational Research, Vol. 15,pp. 203-211 (1983)

D. Botez1J.C. ConnollyM. Ettenberg D.B. GilbertVery High CW Output Power and PowerConversion Efficiency from Current -Confined CDH-LOC Diode Lasers-Reprinted from Electronics Letters, Vol. 19,No. 21, pp. 882-883 (10/13/83)

D. BotezIJ.C. ConnollyHigh -Power Phase -Locked Arrays ofIndex -Guided Diode Lasers-PublishedAppl. Phys. Lett., Vol. 43, No. 12 (12/15/83)

M. Ettenberg IG.H. Olson II. LadanyP. Webb1N.J. DiGiuseppe IT.J. ZamerowskiJ. AppertOn the Reliability of 1.3 -pm Lasers and1.0- 1.7 -um Detectors Grown by VaporPhase Epitaxy-Presented at the IntlSemiconductor Laser Conf., Rio deJaneiro, Brazil (8/7-10/84)

T.J. Faith IC.P. WuElimination of Hillocks on AlSiMetallization by Fast -Heat -PulseAlloying-Published in Appl. Phys. Lett.

P.J. Gale I B.L. Bentz 'WI. HarringtonC.W. Magee1H.A. WeakliemAnalysis of Silanes for Hydrogenated

Amorphous Silicon (,,-Si:H) Production byGas Chromatography/Mass Spectrometry(GC/MS)-Presented at the 32nd AnnualConf. on Mass Spectrometry and AlliedTopics, San Antonio, Tex. (5/27/84)

J. GibsonEffects of Transmission Impairments onthe Quality and Coverage of Multi -ChannelSound for Television-Presented at theNAB Convention, Las Vegas, Nev.(4/30/84)

A.M. Goodman J.P. RussellL.A. Goodman I C.J. Neuse IJ.M. NeilsonImproved COMFETs with Fast SwitchingSpeed and High -Current Capability-Reprinted from Proceedings of the IEEEInternational Electron Devices Meeting(12/83)

J.M. Hammer1C.C. NeilAdjustable Modules for High -Power (Over7.5-mW CW) Coupling of Diode Lasers toSingle -Mode Fibers-Journal of Light waveTechnology, Vol. LT -1, No. 3 (9/83)

E.F. Hockings1S. Bloom1D.J. TamutusA High -Transmission Focus Mask forColor Picture Tubes-Published in RCAReview, Vol. 44 (9/83)

S.T. HsuA Simple Method to Determine SeriesResistance and k Factor of an MOS FieldEffect Transistor-Printed in RCA Review,Vol. 44 (9/83)

G. Kaganowicz1J.W. RobinsonRoom Temperature Glow DischargeDeposition of Silicon Nitrides from Sift,and NH3-Presented at a Session ofPlasma Technology, 34th Canadian Engi-neering Conf., Quebec, Canada

M. KaplanX -Ray Photochemistry: w-Chloro OlefinSulfones-Reprinted from PolymerEngineering and Science, Vol. 23, No. 17(Mid -December 83)

H. Keiss V. Augelli1R. MurriCarrier Lifetime from TransientPhotoconductivity Measurements onSilicon Films-Presented at Intl Conf. onPhysics of Semicond., San Francisco,Calif. (8/6-10/83)

H. Keiss1G. Wieners IR. KellerPhotoconductivity in trans-(CH),,-Presented at Intl Conf. Physics andChemistry of Low -Dimensional SyntheticMetals, Abano, Italy (6/17-22/84)

M. Keith IR. SiracusaA Complete Software Implementation ofNABTS Teletext Decoder-Presented atConf. Digest of Intl Conf. on ConsumerElectronics

H.P. Kleinknecht I H. MeierOptical Profilometer for Monitoring SurfaceContours of Si Power Devices-Reprintedfrom SPIE Vol. 398 -Industrial Applicationsof Laser Technology (1983)

W.F. KosonockyVisible and Infrared Solid -State ImageSensors-Published in the Proceedings ofIEDM 83, - 1 (1983)

M.E. Labib R. WilliamsThe Use of Zeta -Potential Measurementsin Organic Solvents to Determine theDonor -Acceptor Properties of SolidSurfaces-Published in the Journal ofColloid and Interface Science, Vol. 97, No.2 (2/84)

H.W. Lehmann1R. Widmer1M. EbnoetherA. WokaunIM. Meier IS.K. MillerFabrication of Submicron Crossed SquareWave Gratings by Dry Etching andThermoplastic Replication Techniques-Published in the J. Vac. Sci. Technol.,B 1(4) (10/83-12/83)

H.W. LehmannDry Etching for High ResolutionMicrofabrication-Published inMicrocircuit Engineering 83

B.J. LechnerHigh -Definition Television-Presented atthe EE Department Seminar, ColumbiaUniversity, New York, N.Y. (4/6/84)

S.G. Liu S.Y. NarayanRapid Capless Annealing of 28Si, 84Zn, and9Be Implants in GaAs-Published inJournal of Electronic Materials

P.A. Longeway I R.D. Estes I H.A. WeakliemDecomposition Kinetics of a Static DirectCurrent Silane Glow Discharge-Published in Journal of Phys. Chem. (1984)

J.R. MateyUser -Defined Functions for SAS-Published in SAS Communications

D.D. MawhinneyMicrowave Tag Identification Systems-Published in RCA Review, Vol. 44 (12/83)

D. Meyerhofer L.K. WhiteImage Formation in the Sublayer of aMultilayer Resist Structure-Presented atthe SPIE Advances in Resist TechnologyConf., Santa Clara, Calif. (3/12-13/84)

R.W. PaglioneMiniature Microwave Antennas forInducing Localized Hyperthermia inHuman Malignancies-Published in RCAReview, Vol. 44 (12/83)

J.I. Pankove I D.E. CarlsonJ.E. Berkeyheiser I R.O. WanceNeutralization of Shallow Acceptor Levelsin Silicon by Atomic Hydrogen-Publishedin Physical Review Letters, Vol. 51, No. 24(12/12/83)

F. Sechi I R. Paglione I B. Perlman IJ. BrownA Computer -Controlled Microwave Tunerfor Automated Load Pull-Published inRCA Review, Vol. 44 (12/83)


S.A. Siegel1D.J. ChanninPIN-FET Receiver for Fiber Optics-Published in RCA Review

R.K. Smeltzer I C.W. Benyon, Jr.Dielectric Integrity of Gate Oxides in SOSDevices-Published in RCA Review

H.S. Sommers, Jr.Experimental Study of the Lasing P/NJunction as an Electro-OpticalTransducer-Published in the J. Appl.Phys., Vol. 55, No. 5 (3/1/84)

P.J. Stabile IA. RosenW.M. Janton IA. GombarMillimeter Wave Silicon Device andIntegrated Circuit Technology-Presentedat the IEEE MTTS Symposium, SanFrancisco, Calif., and published inSymposium Digest (5/29-6/1/84)

P. SteinApplying Statistics to Real -LifeProblems-Presented at AntiochUniversity Colloquium, Yellow Springs,Ohio (4/27/84)

F. Sterzer1R. PaglioneA. Winter IJ. Laing1E. FriedenthalJ. Mendecki IC. BotsteinEnhancing the Efficacy of LocalizedThermotherapy by Monitoring Changes inTumor Blood Flow-Published in theInternational Journal of RadiationOncology/Biology/Physics

F. Sterzer R. Paglione IJ. MendeckiE. Friedenthal IC. BotsteinHeating Patterns During Cancer HeatTherapy as a Function of Blood Flow-Presented at the 4th Annual Meeting of theNorth American Hyperthermia Group,Orlando, Fla. (3/24-29/84)

J.H. Thomas, Ill I R.V. D'AielloP.H. RobinsonA Scanning Auger Electron SpectroscopicStudy of Particulate Defects inMetallurgical -Grade Silicon-Reprintedfrom Journal of the ElectrochemicalSociety, Vol. 131, No. 1 (1/84)

J.H. Thomas, IllImproved Sensitivity of BackscatteredElectron Detection Using Beam BrightnessModulation with Phase Sensitive Detectionin Scanning Auger Instruments-Published in the J. Vac. Sci. Technol.,A2 (1) (1 /84-3/84)

S. TosimaSurface Acoustic Wave Stylus: Part 2 -Relationship Between Rectangular andFan -Shaped Interdigital Transducers-Published in the RCA Review, Vol. 44(9/83)

S. Tosima1M. NishikawaSurface Acoustic Wave Stylus: Part 3 -Optimum Tip Shape for Pickup Devices-Published in RCA Review, Vol. 44 (9/83)

S. Tosima1M. NishikawaT. Isawa I E.O. JohnsonSurface Acoustic Wave Stylus: Part 1 -Pickup and Recording Devices-Published in RCA Review, Vol. 44 (9/83)

S. TosimaIM. NishikawaSurface Acoustic Wave Stylus: Part 4 -Pyramid -Shaped Surface Acoustic WaveTransducer for Signal RecordingCutterheads-Published in RCA Review,Vol. 44 (9/83)

J.L. VossenThin Film Technology-Presented in theFrontiers in Chemistry Lecture Series, NewPaltz, N.Y. (5/3/84)

C.C. WangHeteroepitaxial Growth andCharacterization of CompoundSemiconductors-Presented at ChemistryColloquium, State University of New York,Stony Brook, N.Y. (5/4/84)

L.K. WhiteAdvanced Optical Lithography for ICFabrication-Published in RCA TREND

L.K. WhitePlanarization Phenomena in MultilayerResist Processing-Published in the J.Vac. Sci. Technol., B1 (4) 1235 (1983)

C.A. WhybarkThe Making of a VideoDisc-Presented atthe Northeastern Christian Jr. CollegeScience Fair, Villanova, Pa. (4/7/84)

O.M. WoodwardBroadband Balun-Published in RCAReview, Vol. 44 (12/83)

C.P. Wu1G.L. SchnableB.W. Lee I R. StrickerImproved Conductivity in Polysilicon Filmsby Pre-annealing-Reprinted from Journalof the Electrochemical Society, Vol. 131,No. 1 (1/84)

B.S. YarmanA Dynamic CAD Technique for DesigningBroadband Microwave Amplifiers-Published in RCA Review, Vol. 44 (12/84)

Missile and Surface Radar

K. AbendSeminar Coordinator and Editor ofSeminar Lecture Notes-Presented atIEEE Philadelphia Section EducationalSeminar: Techniques of Modern SpectralEstimation, University of Pennsylvania,Phila., Pa. (4/7/84)

A. Afrashteh IV. StachejkoL -Band Phased Array T/R Module-Presented at IEEE AP/ MTT-S, Phila., Pa.(3/22/84)

J.A. Bauer1P.N. BroneckeF.R. Kolc1R.L. SchelhornOverview and Application of Surface

Mounting Technology - Parts 1 and 2-Part 1, March 1984; Part 2, April 1984.Electri Onics

R.M. BlasewitzAda As A Program Design Language-Presented at the Second AnnualConference on Ada Technology, Hampton,Va., and published in Proceedings(3/27-28/84)

D.F. BowmanImpedance Matching and Broadbanding,Chapter 43-Published in AntennaEngineering Handbook, Second Edition,McGraw Hill

F.J. BuckleyA Guide to Standards Development for theSoftware Engineering StandardsSubcommittee-Published in Computersand Standards, Vol. 2, No. 4 (1983)

F.J. BuckleyIEEE Standard 830-1984-IEEE Guide forSoftware Requirement Specifications-Printed February 1984

F.J. BuckleySoftware Quality Assurance-Published inthe IEEE Transactions on SoftwareEngineering (1/84)

M.W. Buckley, Jr.1W.C. Grubb, Jr.Electro-Optics, Fiber Optics, and Lasersfor Non -Electrical Engineers-Presentedat George Washington University,Washington, D.C. (4/23-25/84)

C. DiMariaThe Robot That Does Windows-Published in RCA TREND (2/84)

J.R. FoglebochA High -Performance Multiple AlgorithmProgrammable Processor-Presented atthe IEEE National Radar Conference,Atlanta, Ga. (3/14/84)

W.C. Grubb, Jr.1M.JW. Buckley, Jr.Minicomputers, Microcomputers, andMicroprocessors for Non -ElectricalEngineers-Presented at GeorgeWashington University, Washington, D.C.(2/6-7/84)

D.R. HiggsSoftware Documentation-Presented atEast Stroudsburg University ComputerScience Seminar, East Stroudsburg, Pa.(3/13/84)

M. NiemeyerEstimation of Coordinate Misalignments-Presented at Modeling and SimulationConf., University of Pittsburgh, Pittsburgh,Pa. (4/19/84)

F.E. OlivetoFault Tolerant Design-Presented at theFifteenth Annual Reliability Symposium,Phila., Pa. (4/25/84)


W.J. Paterson I C.T. JorgensenProblems in Designing a VLSI BasedComputer to Meet a Specified ConcurrentFault Detection Capability-Presented atthe IEEE Built -In Self Test Workshop,Kiawah Island, S.C. (2/28-3/1/84)

W.T. Patton J.T. NessmithAntenna Engineering Handbook, SecondEdition, McGraw Hill

M.D. RauchwerkA Microprocessor -Based Floating PointLibrary-Presented at IEEE Southeastcon'84, Louisville, Ky., and published in theProceedings (4/11/84)

S.M. ShermanMonopulse Principles and Techniques-Published by Artech House, Inc., Dedham,Mass.

M. WeissModal Testing in the PerformanceVerification Process of ShipboardEquipment-Presented at the 2ndInternational Modal Analysis Conf.,Orlando, Fla., and published inProceedings (2/5-10/84)

F.E. WuebkerDesign Experiences from the MultilevelSecure MCF Operating System-Presented at IEEE Symposium onComputer Security, Oakland, Calif.(4/23-24-25/84)

F.E. WuebkerMilitary Computer Family OperatingSystem: An Ada Application-Presented atSecond Annual Conf. on Ada Technology,Hampton, Va., and published inProceedings (3/27-28/84)

F.E. WuebkerlJ.O. NeilsonDesign Experiences from the Multi -LevelSecure MCFOS-Presented at IEEE 1984Symposium on Security and Privacy,Oakland, Calif., and published inProceedings (4/84)

NBC

J. GibbingsStereo Audio Production for Television-National Association of BroadcastersAnnual Convention and InternationalExposition, Las Vegas, Nev. (4/28-5/3/84)

P. SmithThe Status of Teletext in North America-Special session of the North AmericanNational Broadcasters Association(NANBA) Technical Committee, Las Vegasand Los Angeles (5/2-5/84)

Solid State Division

G. DolneyUsing the SPICE Computer Program as aCircuit Design Aid-A twelve -week coursepresented at Wilkes College

H.R. Ronan, Jr. I C.F. Wheatley, Jr.Power MOSFET Switching Waveforms: ANew Insight-Presented at EleventhAnnual International Power ElectronicsConf., Dallas Tex., and published inProceedings (4/10/84)

Engineering News and Highlights

Steven Bonica becomes Vice -President,Engineering, NBC

Steven Bonica has been named Vice -Pres-ident, Engineering, it was announced byMichael J. Sherlock, Executive Vice -Pres-

ident, Operations and Technical Services,NBC. Bonica will report to Sherlock. In hisnew position, Bonica will act as Chief Engi-neer for the NBC Television Network. Hisareas of responsibility will include systemsengineering, technical development, videoimaging systems, and frequency coordina-tion of allocations.

Most recently, he had been Director, Broad-cast Systems, Engineering, NBC, since July,1982. In that post he was responsible forthe design and implementation of all televi-sion broadcast facilities for the NationalBroadcasting Company in New York. Beforethat he was Director, Videotape Operations,Network News.

Bonica began his professional career atNBC in January 1969, in the News FilmDepartment as an Editing Room Assistant.He later became an Assistant Film Editor,then Film Editor. During his seven years inNews Film, his editing assignments took

him to Europe, South America, Russia,Africa, and Southeast Asia.

In 1976, he became a Videotape Engi-neer in the Operations Department. Laterthat year, he moved to the Electronic Jour-nalism Department as an editor of video-tape. He was a supervisor when he left EJin 1980 to rejoin the News Department as aTechnical Planning Manager. His respon-sibilities in that position included the plan-ning and development of future news -edit-ing systems and the organization and super-vision of videotape operations for coverageof the Republican and Democratic NationalConventions.

In 1981, he became Manager, VideotapeOperations, Network News, and subse-quently, Director. Born and -raised in NewJersey, Bonica studied electronic engineer-ing at Fairleigh Dickinson University.


Herman Heads Solid State Technology Center

John M. Herman, Ill has been appointedDivision Vice -President in charge of theSolid State Technology Center in Somerville,N.J. The technology center is part of RCAGovernment Systems Division and primar-ily designs and manufactures major sys-tems for the Department of Defense andother federal agencies. Dr. Herman reportsto James B. Feller, Division Vice -Presidentof Engineering for GSD.

Dr. Herman's operation engages in thedevelopment of technology and the fabri-cation of integrated circuits for the busi-ness units in the Government Systems andthe Solid State Divisions.

Herman will also oversee SSTC's otheroperations, which include custom and semi -custom design, program management, pack-age/assembly, test, computer -aided design,design automation, and photomask develop-ment.

Prior to his appointment, Dr. Herman wasSSTC's Director of VLSI Manufacturing Tech-nology. Dr. Herman joined RCA in 1982 asthe Director of Integrated Circuit Designand Process Development. Before joiningRCA, Dr. Herman held several engineeringand management positions with Texas In-struments, Inc. He was elected a seniormember of that company's technical staffin 1980. Dr. Herman holds two patents forfabrication methods on high-performance,bipolar logic circuits using Schottky diodes.Dr. Herman received his bachelor and mas-ter of science degrees from Southern Metho-dist University, and his doctorate degreefrom the University of Illinois.

Elliot heads newAmericom marketing groupfor business networks

RCA American Communications, Inc., an-nounced the formation of a new group forthe development and marketing of businessnetworks via its Satcom satellite system.The new group will be headed by DennisW. Elliott, Vice -President, Business Net-works. His organization will include sales,technical and business personnel respon-sible for marketing, design, installation andservice of specialized networks for busi-ness.

Mr. Elliott brings to the new service offer-

ing an extensive background in engineer-ing, business analysis, marketing, and fi-nance. According to Dr. James J. Tietjen,President of RCA Americom, Business Net-works will focus primarily on digital net-work offerings ideally suited for satellitetransmission, and falling within transmissionrates from 9.6 kilobits per second to 1.544megabits per second. The offering will alsofocus on and expand RCA Americom's 56Plus Service (56 kbps), which has beentransferred to the new group from Com-mercial Communications Services.

Miller named Chief Engineerat RCA GovernmentCommunications Systems

The appointment of Donald D. Miller asChief Engineer was announced by Law-rence J. Schipper, Division Vice -Pres-ident and General Manager at RCA Govern-ment Communications Systems. Mr. Millercame to RCA from the ITT Defense Com-munications Division in Nutley, N.J., wherehe was Vice -President and Director, Stra-tegic Communications Systems, and hadbeen employed since 1975.

Previously, Mr. Miller worked for six yearsat General Electric, Pittsfield, Mass., as agroup leader in mechanical engineering.Born in Los Angeles, Mr. Miller received abachelor's degree in mechanical engineer-ing from the California State PolytechnicInstitute at San Luis Obispo, and master'sdegrees in mechanical engineering and man-agement from Rensselaer Polytechnic Insti-tute.

Staff announcements

Executive Vice -President Staff

Roy H. Pollack, Executive Vice -President,announces the appointment of Jay J.Brandinger as Staff Vice -President, SystemsEngineering. In this capacity Dr. Brandin-ger will report to Roy H. Pollack organiza-tionally and to Dr. William M. Webster, Vice -President, RCA Laboratories, on a functionalbasis.

Roy H. Pollack, Executive Vice -President,announces the following appointments onhis staff: Carl R. Turner, Division Vice -Presi-dent and General Manager, Solid State Di-vision; Robert S. Pepper, Vice -President,

Business Development; and Gordon W.Bricker, Staff Vice -President, Planning.

Americom

James J. Tietjen, President and Chief Oper-ating Officer, RCA American Communica-tions, Inc., announces his organization asfollows: John Christopher, Vice -President,Technical Operations; Dennis W. Elliot, Vice -President, Business Networks; Harold W.Rice, Vice -President, Video and Audio Ser-vices; Robert E. Smylie, Vice -President, Gov-ernment Communications Services; JamesJ. Tietjen, Acting, Employee Relations;


James J. Tietjen, Acting, Finance; and JackF. Underwood, Vice -President, CommercialCommunications Services.

Astro-Electronics

H. Soule has been appointed Manager, Anten-na Design.

Consumer Electronics

Larry A. Olson, Manager, ManufacturingTechnology Center, announces the appoint-ment of Allen L. Collier as Manager, Auto-mated Process Development.

James R. Arvin, Manager, Quality Control,announces the appointment of Robert A.Straub as Manager, Production QualityControl.

Globcom

William A. Klatt, Director, Network Opera-tions, announces the organization of Net-work Operations as follows: Joseph D.Ciaccia, Manager, KCC TOCC and TroubleReporting Center; John A. Kruk, Manager,Gateway Operations; Eugene B. Stanley,Manager, KCC Plant Services; and John P.

McIntyre, Administrator, Data ManagementProjects.

John P. Shields, Director, Network Engi-neering, announces that the EngineeringSystems Development function has beentransferred from Leased Facilities and Sys-tems Operations to the staff of the Director,Network Engineering.

RCA Service Company

Donald M. Cook, President, RCA ServiceCompany, announces the following appoint-ments in his organization: Earle A. Maim,II, is appointed Division Vice -President, Auto-mated Business Communications Services.In this capacity, Mr. MaIm will be respon-sible for planning, developing, evaluating,and test marketing all new voice and/ordata communication systems and will con-tinue to be responsible for the Data Servi-ces business. Raymond J. Sokolowski con-tinues as Division Vice -President, Con-sumer and Commercial Services, and willalso assume responsibility for the Tele-phone Systems business.

Donald M. Cook, President, RCA ServiceCompany, announces the appointment ofMichael F. Camardo as Division Vice -Pres-ident, Government Services.

Solid State Division

Jon A. Shroyer, Division Vice -President, LSIand Technology Development, announcesthe appointment of Martin A. Blumenfeldas Principal Member of the Technical Staff.Mr. Blumenfeld will report to Ian Arnott,Manager, Wafer Fab Operations.

Stephen C. Ahrens, Director, Engineering-Standard Integrated Circuit Products Engi-neering, announces his organization as fol-lows: Sushi! K. Chawla, Manager, Telecom-munications System Design; CharlesEngelberg, Manager, Test Engineering;Merle V. Hoover, Manager, Engineering-Computer, Telecommunications and Indus-trial Products; Lewis A. Jacobus, Manager,Engineering-Logic Products; Sterling H.Middings, Section Manager-Layout Ser-vices; and Bruno J. Walmsley, Manager,Engineering-Automotive and ConsumerProducts.

VideoDisc Division

Arnold T. Valencia, Division Vice -Presidentand General Manager, VideoDisc Division,announces the appointment of HarryAnderson as Division Vice -President, DiscOperations.

Professional activities

RCA Laboratoriesachievement awards given

Dr. William M. Webster, Vice -President,RCA Laboratories, Princeton, N.J., an-nounced that the following engineers andscientists have been given RCA Laborato-ries Outstanding Achievement Awards forcontributions to electronics research andengineering during 1983. Recipients of indi-vidual awards are:

Robert A. Duschl, for outstanding perfor-mance in conception, development, and im-plementation of a manufacturing inspectionsystem for in -process automated evaluationof color picture tubes.

Lorenzo Faraone, for contributions to theunderstanding and control of tunneling cur-rent through thin thermally grown silicondioxide layers of silicon devices.

Krishnamurthy Jonnalagadda, for analysiswork to accurately predict the performanceof voice and voice -band data circuits inthe RCA single-sideband system.

John H. Thomas, Ill, for contributions inthe areas of Auger electron spectroscopyand x-ray photoelectron spectroscopy and

their use in wide areas of materials charac-terization research.

Edward H. Adelson, Charles H. Anderson,James R. Bergen, Curtis R. Carlson, andAlbert P. Pica, for contributions to the devel-opment of advanced signal -processing con-cepts that match properties of the humanvisual system.

Victor Auerbach, Thomas Y. Chen, WalterG. Gibson, Thomas F. Lenihan, andCharles M. Wine, for contributions to thedevelopment of the first CED random-ac-cess interactive VideoDisc player.

William E. Babcock, William E. Rodda,Walter Truskalo, and Werner F. Wedam,for contributions to the development of a13 -inch -diagonal high -resolution computermonitor prototype with switchable scanrates.

William J. Bachman, Robert R. Demers,Nitin V. Desai, Robert W. Jebens, Frank R.Reed, and Gerard Samuels, for contribu-tions to the development of concepts thathave led to an innovative VideoDisc playerdesigned for automated manufacturing at alower cost.

Donald F. Battson, Walter F. Kosonocky,Peter A. Levine, and Frank V. Shallcross,for contributions to the design and devel-opment of a high -sensitivity, low -noise CCDimager.

Rodney P. Borchardt, Glenn A. Reitmeier,Terrence R. Smith,and Christopher H.Strolle, for contributions to the -design andimplementation of an advanced computer -based television simulation facility.

Richard Brown, Henry C. Johnson, AdolphPresser,and Franco N. Sechi, for contribu-tions to the development and application ofa new batch -fabrication process of minia-ture microwave circuits.

Michael T. Duffy, Michael F. Leahy, andJer-Shen Maa, for contributions in devel-oping methods for the reactive sputter etch-ing and plasma etching of insulating, semi-conducting, and conducting elements of inte-grated circuits.

John R. Fields, Ronald Sverdlove, andNorman D. Winarsky, for contributions tothe development of a computer programthat fully models the electron beam in apicture tube.


Joseph H. McCusker, Sidney S. Seffren,Alan Sussman, Barry J. Thaler, andThomas J. Ward, for contributions leadingto innovations in the manufacturability, reli-ability, and design of high -voltage trans-formers for TV receivers.

RCA publicationssweep competition

The New York Chapter of the Society forTechnical Communication honored several1983 RCA technical and news publicationsin several categories for their excellence incompetition with those from, for example,AT&T, IBM, Western Electric, Ford Founda-tion, Springer-Verlag, American Physical So-ciety, and more. These awards are continu-ing signs of the high professionalism ofall RCA engineers who write for publication.

Periodicals-RCA Engineer magazine (Dis-tinguished Technical CommunicationsAward)

Training Manuals-"Guide for RCA Engi-

neer Authors" (Distinguished TechnicalCommunications Award)

House Organ-MSR's RCA Family (Awardof Excellence)

Newsletter-AEGIS Excellence Newslet-ter (Award of Merit)

Reports-AEGIS EDM4 DevelopmentAward Fee Progress Report (Award ofExcellence)

Reports-AEGIS Production ProgramsStatus Report No. 63 (Award of Merit)

Publication Graphics-Illustrative-MSR's"47/51 Parallel Course" (Award of Merit)

Single Sheet Designs-MSR's "ChemicalSafety" (Award of Achievement)

Astro professionals serve AIAA

The following members of RCA Astro-Elec-tronics held appointments to American Insti-tute of Aeronautics and Aerospace Tech-nical Committees:

J.N. La Prade-Associate Member of Com-mittee on Communications Systems

P. Pierce, Principal Member TechnicalStaff-Aerospace Power Systems Commit-tee

R. Joshi, Senior Member Technical Staff-Astrodynamics Committee

P.G. Goodwin, Manager, Electronic Sub-systems Engineering-Computer SystemsCommittee

R.A. Amadio, Manager, Financial Control-Economics Committee

S.J. Arkuszewski, Member Technical Staff-Ground Testing Committee

R.D. Scott, Staff Engineer-Interactive Com-puter Graphics Committee

L. Gomberg, Director, Satellite Programs-Management Committee

R.T. Feconda, Member Technical Staff-Solid Rockets Committee

Technical excellence

Indianapolis Consumer Electronics annual awards

Nortrup

The 1983 Indianapolis Technical ExcellenceCommittee banquet was held on February,1984. The keynote speaker was Mr. R.H.Pollack, Executive Vice -President of RCA.The banquet was attended by over threehundred members of the CED technicalcommunity. Dr. J. Donahue, CED Vice -Pres-ident and General Manager, presented theannual awards. Dr. J.E. Carnes, DivisionVice -President, Engineering, was an after -dinner speaker. The recipients are:

Kevin Nortrup-for the definition and designof a menu -based customer interface to theCTC 141 digital receiver. Specifically, in

Becker Benson

recognition of outstanding engineering con-tributions made toward the completion ofthe engineering prototype system in timefor use in a marketing study and for engi-neering evaluation.

Basab Dasgupta-for the development of abetter theoretical understanding of yoke pa-rameters. He has developed a mathemati-cal basis to describe yoke phenomena thathas advanced the state-of-the-art of CEyoke design.

Hollis Becker-for the development of agraphics printing process combining inks

Shah

and a topcoat that produces a highly reli-able product. It is used on the unified remotetransmitter for J -Line television and Video -Disc products. The wear performance ofthis process is presently considered supe-rior to CE competition.

Walt Benson and Kashyap Shah-for out-standing effort in the design, development,and implementation of CE's first fully auto-mated robotic spray -painting system. Thisrobot is used on portable television cabi-nets.


RCA Somerville 1983 third- and fourth-quarter award recipients

Larry Rosenberg, Manager, Design Automation, Solid StateTechnology Center (left) with (from left to right) MikeGianfagna, Rich Gopstein, and Joe Mastroianni.

Bob Geshner, with (from left to right) Steve Preston, BernieOstrowsk!, Mary Evancho, Marie Wendt, and Joe Mitchell.

Nick Kucharewski, Manager, Design Engineering, Solid StateDivision, (left) with (from left to right) Bob Pollachek, DebbiePetryna, and Dale Barker.

D. Barker, R. Pollachek, and D. Petryna, for outstanding teameffort in achieving, on schedule, the design of a family of64K/128K/256K ROMs. This family far excels competitive prod-uct in die size and performance.

I. Wacyk and D. Alessandrini, for the design, layout, and testingof a radiation -hard (1 -?.200 Rad Si) 16K SOS RAM, which includesfuse switching of redundant columns for increased yield capabil-ity. The RAM was designed to be tolerant of degraded operationalparameters so that full functionality can be maintained as long aspossible in destructive radiation environments.

Bob Geshner, Manager, Mask Tooling Operations, SolidState Technology Center (left) with (from left to right) RonnieToland, Al Shoemaker, Jon Butcher, and George Durante.

IC Products Engineering,(second from left) with (from left to right) Bob Dawson, TomFreeman, and Tom Deegan.

M. Gianfagna, T. Deegan, T. Freeman, R. Gopstein, and J.Mastroianni, for interdepartmental team work, over a two-yearperiod, that resulted in the specification and development of asymbolic layout system to layout and verify bipolar integrated cir-cuits. The intent of this work was to reduce the number of siliconcuts required to obtain a working bipolar IC. Key elements of thissystem include the following: symbolic layout and layout compac-tion computer aids, a parameterized cell library that has beencharacterized, automatic detection of shorts or opens in a layout,and generation of a circuit description for the layout that includesparasitic devices and is suitable for simulation.

M. Evancho, J. Butcher,S. Shoemaker, R. Hilton, S. Preston, R.Toland, A Frattali, B. Ostrowski, M. Wendt, G. Durante, and J.Mitchell, for consistent pursuit of major manufacturing goals ofhigh yield and effective cycle times. Throughout the first andsecond quarters of 1983, this group exceeded all goals set for it,allowing production of high -quality masks for the factories andpilot lines that has allowed RCA to respond to customer demandsand allowed the Engineering design community to test designsand complete difficult reiterative mask sets to prove designs suchas AT/MAC, Jelly Bean, "ROM - Express." In June, they pro-duced a global line yield of 99 percent against an aggressiveplan of 77 percent, with cycle times of 1.3 to 1.7 days against acommitment of 2.0. On some ROM jobs, cycle times of as little as12 hours were realized.


GCS announces team technical excellenceaward

The Technical Excellence Award has been made to the Comput-er Controlled Equipment Engineering team responsible for thedesign and development of a 1 -MHz Digital Acquisition Systemunder the Unit's Independent Research and Development pro-gram. The team members are: Thomas J. Fritsch; Joseph F.Hoey; Edward A. Timar; and Stanley A. Tomkiel.

This equipment can digitally channelize a 1 -MHz bandwidthsignal into 160, 8 -kHz wide channels (on 6 -kHz centers), recog-nize signals at a 4 -dB S/N ratio in an 8 -kHz bandwidth, and pro-vide cross -correlation of a multi -pattern signal at a 10 -Megabitrate. The project included the development of a special massstorage, interactive controller capable of sustaining a 2.4-Mbitread/write rate in association with a Winchester disc drive.

The team's performance in conceiving a two -stage coarse andfine-grain processing architecture to handle a wide range of sig-nal characteristics, and developing an overlap seek and read/write of data in order to record 16 channels and read 2 channelsof 125 Kbps real-time data simultaneously was an outstandingexample of engineering creativity.

This effort has led to an in-house resource for sophisticatedsignal -processing equipment, digital filtering and correlation. RCAGovernment Communications Systems has submitted a sole -

GCS Chief Engineer, D.D. Miller (far left), with TechnicalExcellence Award winners (from left to right): Joseph F.Hoey, Thomas J. Fritsch. Stanley A. Tomkiel, and EdwardA. Timar.

source proposal for classified applications of the system that isnow included in the customer's 1984 budget and is expected tolead to substantial Navy business in the future.

Proud of your hobby?Why not share your hobby with others? Perhaps their interest will make yourhobby more satisfying. Or maybe you'll find others who already share your hobby

and who can help make your own efforts more rewarding.

The RCA Engineer likes to give credit toengineers who use their technical

knowledge away from the job. We'vepublished articles about subjects as

diverse as a satellite weather station,model aircraft and railroading, solar

heating, and an electronic fish finder.

For more information on how you canparticipate in this feature of the RCA

Engineer, call your local EdRep (listedon the inside back cover of the Engineer)or contact Frank Strobl.


Dr. Harold B. Law died on April 6, 1984, inPrinceton Medical Center. Dr. Law was 72years old. He was recognized throughoutthe electronics industry for developing fab-rication techniques that led to the first prac-tical shadow -mask color picture tube de-monstrated by RCA in 1950. Among his

Obituary

Dr. Harold B. Law, RCA color television pioneerkey contributions were the "lighthouse" tosimulate the shadowing of electron beamson the tube's faceplate and the correspond-ing photodeposition of a mosiac of tiny phos-phor dots to produce the color picture. Stillused today, Law's techniques made pos-sible the hundreds of millions of color TVreceivers produced in the last 30 years.

Dr. Law received many honors for hiswork, bcith from RCA and from outside pro-fessional groups. In 1979, he was electedto membership in the National Academy ofEngineering of the United States of Amer-ica, the highest professional distinction thatcan be conferred on an engineer.

Born in Douds, Iowa, but raised in Kent,Ohio, Dr. Law received B.S. degrees in Lib-eral Arts and in Education, both in 1934,from Kent State University. He earned M.S.and Ph.D. degrees in Physics from OhioState University in 1936 and 1941, respec-tively. In 1959, Dr. Law received a citationfrom Kent State as an outstanding gradu-ate. He was to receive an Honorary Docto-rate from Kent State at the 1984 graduationceremonies.

Dr. Law joined the RCA Corporation inCamden, N.J., in 1941, working on televi-

sion camera tubes. He transferred to thenewly established RCA Laboratories inPrinceton in 1942, and was one of threeRCA researchers honored by the Televi-sion Broadcasters Association in 1946 forthe development of the image orthicon cam-era tube, the "technical accomplishment ofthe year."

He was named a Fellow of the TechnicalStaff of RCA Laboratories in 1960 and twoyears later, was appointed Director of theRCA Electronic Components Materials andDisplay Device Laboratory. He retired in1976.

In 1955, the Institute of Electrical andElectronics Engineers (IEEE) awarded Dr.Law the Vladimir K. Zworykin TelevisionPrize and in 1975, the Institute presentedhim the Lamme Medal "for outstanding con-tributions in developing color picture tubes,including the fabrication techniques whichmade color television practical." Dr. Lawwas a Fellow of the Society for InformationDisplay (SID) and in 1975, received the SIDFrances Rice Darne Memorial Award forhis work in color picture tube development.

Harold B. Law:Inventor/Engineer/Manager: 1911-1984

The sudden and unexpected death of Dr. Harold B. Law wasa great blow to all of his many friends and former colleaguesat RCA. Although he was formally retired since 1976, he was afrequent visitor at RCA Laboratories and active in social andcommunity affairs up to the time of his death.

Few engineers have done work that has been seen directlyand regularly by more than a billion people; Harold is amongthem. How many know what's inside a color television set, orhow the programs are generated and transmitted? But everycolor television viewer looks directly at Harold Law's creation,the shadow -mask picture tube screen, and every viewer isaffected by those beautiful colors that are displayed on it.

Several publications tell the story of the color picture tube,but the best is the one Harold wrote himself.' He tells therehow he made the two key inventions-the method of photo -depositing a pattern of color phosphors, and the invention of a"lighthouse" for producing the correct light and shadow areasso that these phosphors are positioned in exactly the rightplace. To Harold, and to all of us who were involved, it hasalways been an amazing story. First, the early success tookplace in less than 6 months between 1949 and 1950, andsince then the shadow -mask principle and Law's basic tech -

1. Harold B. Law, "The Shadow -Mask Color Picture Tube: How It Began,"IEEE Trans. on Electron Devices, Vol. ED -23, pp. 752-759 (July 1976).Also in RCA Engineer, Vol. 22, No. 1, pp. 88-94 (June/July 1976).

niques have remained dominant for all these 34 years. It'slikely that more than 200 million color tubes have been made,with a total economic value far greater than any other singleelectron device.

Dr. Law's career was well under way at the time of his colortube work. A graduate of Kent State University, with a Ph.D.from Ohio State University, Harold joined RCA in 1941. Hisearly work was on television camera tubes, and he was amember of the team that developed the image orthicon duringthe war. From the start, Harold showed the characteristics thatled to his achievements. He lived by the motto "if at first youdon't succeed " and he was one of the most dedicatedworkers in the electron device field. At RCA, he was promotedfrom the technical staff to Fellow and then to LaboratoryDirector.

Law received many honors but he was a modest man, notat all changed by this imposing array of accolades. There is,however, one more award, which Harold was told about, buthad not yet received. Kent State was to bestow an honorarydoctorate on him at their 1984 graduation ceremonies. Of allhis honors, this one pleased him most; Ruth, his devoted wifeand helpmate for 42 years, will accept the award for him.

Those of us who saw Harold after retirement know that hisactivities on a small farm were exactly what he wanted to do.He worked tirelessly on garden, fruit, and vegetable plots; heused his technical expertise to solve problems at home, at hischurch, in the community, and among neighbors. His lastyears were happy and contented ones.

-Edward W. Herold


Editorial RepresentativesContact your Editorial Representative at the TACNETnumbers listed here to schedule technical papersand announce your professional activities.

Broadcast Systems Division (BSD) TACNET*Jack Dearing Camden, New Jersey 222-4688

Consumer Electronics (CE)

*Eugene Janson Indianapolis, Indiana 422-5208John Hopkins Indianapolis, Indiana 422-5201Larry Olson Indianapolis, Indiana 422-5117Byron Taylor Indianapolis, Indiana 426-3367Don Willis Indianapolis, Indiana 422-5883

Corporate Information Systems & Services

Patent Operations

George Haas Princeton, New Jersey

RCA Communications

American Communications*Murray Rosenthal Princeton, New Jersey

Carolyn Powell Princeton, New JerseyGlobal Communications*Dorothy Unger Piscataway, New Jersey

TACNET

226-2888

258-4192258-4194

335-4358

Sue Handman Cherry Hill, New Jersey 222-6242 RCA LaboratoriesCorporate Technology Eva Dukes Princeton, New Jersey 226-2882

*Tony Bianculli Princeton, New Jersey 226-2111RCA Records

Government Systems Division (GSD)*Greg Bogantz Indianapolis, Indiana 424-6141

Advanced Technology Laboratories*Merle Pietz Camden, New Jersey 222-2161 RCA Service Company

Camden, New Jersey 222-2731

Astro-Electronics *Murray Kaminsky Cherry Hill, New Jersey 222-6247*Frank Yannotti Princeton, New JerseyCarol Coleman Princeton, New Jersey

229-2544229-2919

Dick Dombrosky Cherry Hill, New JerseyRay MacWilliams Cherry Hill, New Jersey

222-4414222-5986

Automated Systems*Paul Seeley Burlington, Massachusetts 326-3095 "SelectaVision" VideoDisc Operations

Dale Sherman Burlington, Massachusetts 326-3403

Government Communications Systems *Nelson Crooks Indianapolis, Indiana 426-3164

*Dan Tannenbaum Camden, New Jersey 222-3081Thomas Altgilbers Camden, New Jersey 222-3351 Solid State Division (SSD)

GSD Staff

*Peter Hahn Cherry Hill, New Jersey 222-5319 *John Schoen Somerville, New Jersey 325-6467

Susan Suchy Solid State Technology Center Power Devices

Somerville, New Jersey 325-7492 Harold Ronan Mountaintop, Pennsylvania 327-1473Missile and Surface Radar or 327-1264

*Don Higgs Moorestown, New Jersey 224-2836 Integrated CircuitsGraham Boose Moorestown, New Jersey 253-6062 Dick Morey Palm Beach Gardens, Florida 722-1262Jack Friedman Moorestown, New Jersey 224-2112 Sy Silverstein Somerville, New Jersey 325-6168

John Young Findlay, Ohio 425-1307

National Broadcasting Company (NBC)Video Component and Display Division

*Bob Mausler New York, New York 324-4869

*Ed Madenford Lancaster, Pennsylvania 227-6444New Products Division Nick Meena Circleville, Ohio 432-1228

J.R. Reece Marion, Indiana 427-5566*Art Sweet Lancaster, Pennsylvania 227-6878

Bob Reed Lancaster, Pennsylvania 227-6845Bob McIntyre Ste Anne de Bellevue 514-457-9000

*Technical Publications Administrators, responsible for review and approvalof papers and presentations, are indicated here with asterisks before their names.

RCA EngineerA technical journal published bythe RCA Technical Excellence Center"by and for the RCA engineer"

13 Roszel Road, P.O. Box 432. Princeton, NJ 08540

Address correction requested

EJ NUSSEN203 ELKINS RDCHERRY HILLNJ U8034

Printed in the U.S.A.Form No. RE -29-3

6SDGC1

BULK RATEUS Postage

PAID

Phila., Pa.Permit No. 2906

Documents

RCA Engineer - WorldRadioHistory.Com · 5/6/1984 · RCA authors describe today's realities-the ongoing and successful efforts to automate the engineering workplace.-MRS Illustrations