mcoa · The astute reader might point out that a single package could do the job by building the MEVDP model from scratch, ... are published in a document titled "Manufacturing Design

mcoa EngineerA technical journal published by theRCA Technical Excellence Center El 13 Roszel Road 0 P. 0. Box 432 0 Princeton, NJ 08540-0432 0 Tacnet: 226-3090

RCA Engineer Staff

Tom King Editor

Mike Lynch Associate Editor

Louise Carr Art Editor

Rosemarie Cruz Editorial Assistant/Composition Specialist

Maria Pragliola Secretary

Editorial Advisory Board

Tony Bianculli Manager, Engineering Information, TechnicalExcellence Center 0 Jay Brandinger Staff Vice -President,Systems Engineering, RCA Laboratories 0 Jim Carnes DivisionVice -President, Engineering, Consumer Electronics': JohnChristopher Vice -President, Technical Operations, RCAAmericom 0 Jim Feller Division Vice -President, Engineering,

Aerospace and Defense 0 Mahlon Fisher Division Vice -President, Engineering, Video Component & Display Division0 Larry Gallace Division Vice -President, Product Assuranceand Environmental Engineering, Solid State Division 0 ArchLuther Senior Staff Scientist, RCA Laboratories HowieRosenthal Staff Vice -President, Engineering 0 Bill UnderwoodDirector, Technical Excellence Center 0 Bill Webster Vice -President and Senior Technical Advisor, Electronic Products andTechnology

Consulting Editors

Ed Burke Administrator, Marketing Information and Commun-ications, Aerospace and Defense 0 Walt Dennen Manager,Naval Systems Department Communications and Information,Missile and Surface Radar 0 John Phillips Director, BusinessDevelopment and Services, RCA Service Company

Cover illustration by Joan Ogden,RCA Laboratories.

Cover design by RCA Engineer Staff.

Our cover: This random fractal sunrise over a 'self -similar' seawas created by Joan Ogden using the multiresolution pyramidtechnique described in the article on page 4.

Cloud -like structures at four different scales ranging from fineto coarse were used to construct the sun and its corona. Theshaded spherical shape was achieved by using different fuzzycircular masks at each scale. The sea and sky were added, andthe whole scene was pseudo -colored. The red bar at the top ofthe cover is not part of Joan's graphic.

0 To serve as a medium of interchange of technicalinformation among various groups at RCA 0 To create acommunity of engineering interest within the company bystressing the interrelated nature of all contributions 0 Todisseminate to RCA engineers technical information ofprofessional value 0 To publish in an appropriate mannerimportant technical developments at RCA, and the role of theengineer 0 To help publicize engineering achievements in amanner that will promote the interests and reputation of RCAin the engineering field 0 To provide a convenient means bywhich the RCA engineer may review professional work beforeassociates and engineering management 0 To announceoutstanding and unusual achievements of RCA engineers ina manner most likely to enhance their prestige andprofessional status.

?noir= kllis/10A picture isAworth a thettirandwive#44.--Complex designs, from submicroscopicmolecular arrangements to very large assembliessuch as the Space Station, can be achievedrapidly and accurately because of the capabilitiesof modern computer systems. Today anenormous amount of data can be assimilated andprocessed on a near real-time basis bycomputers linked by modern communicationsnetworks.

Advancement in computer graphics hasprovided a means for presenting these largeamounts of data to our most efficient data transfersensor, the eye. Displays have been designed toenhance the transfer of data, thus enabling us torapidly perceive relationships, trends, andinsights, thereby eliminating the need to view tensof thousands of data words.

Business units throughout RCA are findingmore and more uses for computer graphics, fromcomplex engineering drawings to ergonomicallydesigned screens and displays. In Astro-Electronics Division, computer graphics hasincreased both productivity and quality of design,

manufacture, test, and operation of spacesystems. For instance, systems engineers usethree-dimensional computer graphic displays todesign the various modules of the Space Stationthat will be assembled in space. The microwavedesign engineer can view a complex combinerassembly in a three-dimensional display beforetransmitting the design data to a computer-

controlled milling machine to make the part.And, as you will find in this issue, throughcolor -coded displays of a spacecraft system,an operator can very quickly assess the statusof a complex spacecraft system.

Tomorrow, through the application of artificialintelligence, we can look forward to even moreefficient and productive systems to design,manufacture, and operate our products.

114j --)42,1,

Robert MillerDivision Vice -President, EngineeringAstro-Electronics Division

alma EngineerVol. 30 No. 5 Sept. Oct.l 1985

computer graphics 4 Pyramid -based computer graphicsJ.M. Ogden I E.H. Adelson IJ.R. Bergen I P.J. Burt

16 Graphics for the definition and visualization of physical systems-some case studiesR.D. Scott

. as an engineering tool 24 Spooled -graphics laser printing from CMSD.P. Barton

34 Accurate layouts for automatic component insertionR.A. Monat

41 The operator display emulation facilityUsing computers to design system display/control consolesR.A. Faust, Jr. IW.L. Mays

47 An improved man/machine interface for spacecraft controlR.H. Harten IT. Crowe IA. Anton

. . . in RCA products 56 Graphics for election '84: An introductionand services D. Rabinowitz

57 NBC election '84: Results by designT.C. Alfieri

62 Innovative uses of computer graphics at NBC-Election day, 1984C. Barton

67 Techniques for enhancement of interactive color graphicsfor Cl systemsL. Arlan

general interest 75 Plasma etching for integrated circuit fabricationM. Leahy

83 How to evaluate and shop for computer graphics softwareJ. Warner IJ. Slitz

on the job/off the job 89 Engineering an electrically -powered automobileS. Kreitzberg IJ. Krupa

departments News and Highlights, 93 'Erratum, 46

Copyright c 1985 RCA CorporationAll rights reserved

in this issue . . .

computer graphics

Ogden et al.: "In synthesizing and combining images, we would liketo imitate the artist's ability to see an image on both large and smallscales by representing the image mathematically on many differentspatial scales."

Scott: ". . . a primary goal is to reduce the drudgery of data entry,minimize errors, and allow the computer user to interact more andmore in "natural" language, in conversational terms, or by menu."

16

Barton: "From a systems design viewpoint, our laser printer is notan end, but a beginning."

Monat: "Rather than show the footprint, we chose to show theexclusion area. The exclusion area shows how much space is taken upby components already placed."

Faust/Mays: "The mission of the ODEF team is to design, develop,and test operator interfaces for military systems under development atRCA."

Harten et al.: "The main design concept we followed was that theoperator should be made aware of the problem and led to it, not forcedto find it."

24 Footprint 34

SOLAR ARRAYS47

NORTH SOUTH

41

4,1-

Alfieri: "A graphic artist's ability to enhance our daily newsproductions with timely, unique, and innovative artwork was a qualitywe wanted in an automated system."

Barton: "The system developed by NBC was greatly influenced bythe requirements of the live production environment in which thesystem was finally going to be used."

Arlan: "The analyst operating in today's tactical environment has acomplex array of data resources that provide information regarding theenemy's movers, shooters, and emitters."

57

VOTE2%in

62

PANIC

SIGNALHANDLER

Leahy: "Not only is this technology new, but we know it is absolutelyessential to RCA's advanced circuit fabrication plans."

Warner/Slitz: "Choosing a graphics software system for yourapplication consists of more than finding software features thatcorrespond to your discrete functional requirements."

Kreitzberg/Krupa: "This 75 -pound monster came equipped withroller bearings on both sides of the armature, a series/parallel(compound) field, and the capability to accommodate a dc input . . ."

in future issues . . .

mechanical engineeringtechnology for the 1990s

89

/. 83

J.M. Ogden E.H. Adelson I J.R. Bergen P.J. Burt

Pyramid -based computer graphics

Pyramid -based graphics techniques can provide realisticcomputer graphics on small systems without the complexitiesof physics simulations.

Human beings have an intuitive feel for graphics. Graphicsproblems such as blending two images smoothly, interpolatingto fill in missing image data, or creating realistic looking imagesare routinely solved by artists using traditional media. It is

much more difficult to perform these tasks mathematically. Amajor step in solving a computer graphics problem is choosingan appropriate numerical representation for the image, onewhich allows us to use our visual and artistic intuition in anatural way.

Artists often tend to separate the spatial scales of an imagewhen creating or altering a picture. When an artist paints alandscape, the coarse scale (low spatial frequency) informationis filled in first, as a wash of color in a large region. Inter-mediate -size details can be added next with a medium -sizebrush. As a last step, the artist draws the fine details with asmall brush. The image can be considered a sum of overlays ofincreasingly fine detail. When an artist touches up a damagedpicture, both the large and small scale variations are consideredin filling in the missing pieces. In synthesizing and combiningimages, we would like to imitate the artist's ability to see animage on both large and small scales by representing the imagemathematically on many different spatial scales simultaneously.

The simplest way to represent an image is to set the numericalvalue of each pixel proportional to the image intensity. Thisrepresentation is useful when we want to paint or draw directly,one pixel at a time, or when simple manipulations of contrastor color are desired. But it becomes cumbersome when wewant to look at an image at several spatial resolutions.

Abstract: This paper describes pyramid solutions to graphicsproblems that have proven difficult in other imagerepresentations. The "physics simulation" approach grows moreout of the physics and mathematical modelling traditions.Greater realism can be achieved by using the physics simulationapproach, but the complexity and computation time are vastlyincreased over the multiresolution pyramid approachesdescribed here.

©1985 RCA Corporation.Final manuscript received October 21, 1985Reprint RE -30-5-1

A Fourier transform representation can be used to separatethe various spatial scales of an image. Unfortunately, when weleave the familiar spatial domain for the spatial frequencydomain our intuitive feel for the problem is lost. Operating onthe Fourier transform of an image, we can no longer "see" localspatial features in a recognizable form. What is really needed isa representation that describes an image at multiple spatialresolutions, and also preserves the local spatial structure thatallows us to "see" the picture at each scale. Pyramid representa-tions are ideal for this class of problems.t'2'3

The pyramid representation and computer graphicsThe pyramid representation expresses an image as a sum ofspatially bandpassed images while retaining local spatial informa-tion in each band. A pyramid is created by lowpass-filtering animage Go with a compact two-dimensional filter. The filteredimage is then subsampled by removing every other pixel andevery other row to obtain a reduced image G1. This process is

repeated to form a Gaussian pyramid Go, GI, G2, G3-G,,(Fig. 1).

Gk(i,j)=XGk-1(2i+m,2±n),k=1,N

Expanding GI to the same size as Go and subtracting yields thebandpassed image Lo. A Laplacian pyramid Lo, L 1, L2,-L,1 canbe built containing bandpassed images of decreasing size andspatial frequency.

Lk = Gk - 1,k = 0,N-1

where the expanded image Gk,i is given by

G map= 4 XI G 10-1[(2i- m)/2, nY2)] f(m,n)

The original image can be reconstructed from the expandedbandpass images:

Go=Lo+LI,1 +L2,2 +....LN-1,N-1 + GN,N

4 RCA Engineer 30-5 Sept. / Oct.1985

Figure 2 shows an image represented as a sum of several spatialfrequency bands.

The Gaussian pyramid contains lowpassed versions of theoriginal Go, at progressively lower spatial frequencies. Thiseffect is clearly seen when the Gaussian pyramid "levels" areexpanded to the same size as Go (Fig. 3a). The Laplacianpyramid consists of bandpassed copies of Go. Each Laplacianlevel contains the "edges" of a certain size, and spans approxi-mately an octave in spatial frequency (Fig. 3b).

This pyramid representation is useful for two important classesof computer graphics problems. First, tasks that involve analysisof existing images, such as merging images or interpolating to fillin missing data smoothly, become much more intuitive when wecan manipulate easily visible local image features at severalspatial resolutions. And second, when we are synthesizing images,the pyramid becomes a multiresolution sketch pad. We can fill inthe local spatial information at increasingly fine detail (as anartist does when painting) by specifying successive levels of apyramid.

In this paper, we will describe pyramid solutions to somegraphics problems that have proven difficult in other imagerepresentations:

1. Image analysis problems(a) Interpolation to fill missing pieces of an image(b) Smooth merging of several images to form mosaics

2. Creation of realistic looking images(a) Shadows and shading(b) Fast generation of natural looking textures and scenes

using fractals

(c) Real-time animation of fractals.

Multiresolution interpolation and extrapolationThe problem

The need to interpolate missing image data in a smooth, naturalway arises in a number of contexts. It can be used to removespots and scratches from photographs, to fill in transmittedimages that are incomplete, and to create interesting computergraphic effects.

A solution

Insight into the problem of interpolation is gained by consideringthe image as a sum of patterns of many scales. A typicalphotograph includes small scale fluctuations due to surfacetexture, superimposed on more gradual changes due to surfacecurvature or illumination variations. Similarly, a painting is acomposite made up of features of many scales, rendered with

G(1 L(

Fig. 2. The pyramid as a sum of spatial frequency bands.

REDUCE (filter+subsample)

g11

L1=g1-g21

LO=g0-g11

Fig. 1. Building the pyramid.

brushes of different sizes. In predicting the values of the missingpieces of an image, we need to consider intensity variations onboth small and large scales.

Figure 4 shows a one-dimensional representation of an imageGo that has some missing values. One method of interpolating tofind the unknown region is to use a Taylor series expansion. Ifthe size of the missing piece is comparable to the finest scalefeatures of the image, a good estimate of the unknown valuex+ dx can be obtained through a simple linear prediction basedon the first derivative Go' at the point x:

Go(x±dx)=-Go(x)-1-dx Go'(x).

If the missing piece is large compared to the fine scale features of

L2,2 L3,3

Ogden/Adelson/Bergen/Burt: Pyramid -based computer graphics 5

G

L0.0 L.Fig. 3. Expanded Gaussian and Laplacian pyramids.

the image, we need to examine variations on larger scales as well.We can compute such an estimate by fitting the function Go(x)to a Taylor polynominal of higher degree. This involvestaking higher order derivatives that represent the image variationover a larger number of pixels. One disadvantage of thisapproach for computer graphics is that it is computationallyexpensive. It is also difficult to adjust the degree of the interpo-lating polynomial to account for missing regions of differentsizes.

An alternative way to look at the missing information onmany scales is to build a Gaussian pyramid. This represents theimage Go at different spatial resolutions ranging from fine (Go) tocoarse (G). The unknown piece of Go is also missing from GI,G2..G Note that the size of the missing region is reduced in thereduced pyramid levels. Now, instead of fitting a Taylor poly-nomial of high degree at the finest spatial scale, we use linearinterpolation at multiple spatial resolutions to fill in the missinginformation.

In the example shown, the size of the missing piece is largecompared to the fine scale variations contained in Go and GI, butis comparable to the feature size in G,,1 and small comparedto the coarse features of G. Starting with Go, we use linearextrapolation to predict the values of unknown points with twoknown neighbors.

Go(i)=2 Go(i+1)-Go(i+ 2)

For this example, most of the unknown values of Go are "leftblank." A small border one pixel wide is extrapolated into theunknown region. Now we build GI and extrapolate again. Atthis reduced resolution, the extrapolated border corresponds to alarger proportion of the unknown region. When the extrapolatedGI is expanded to full size, the border also expands in size fromone pixel to several pixels. Continuing to lower spatial resolutions,we eventually reach a reduced pyramid level G, where theunknown region has shrunk to only one pixel. Extrapolation atG gives a pyramid level with all the values filled in.

An extrapolated image is built by reconstructing the extrapo-lated pyramid. Starting with G, we expand to form G,1. Wherea pixel in the next highest frequency band G_, is missing, thevalue from G,, is used. Continuing this process, we form anextrapolated image Go, with all unknown points filled in.

Examples

Figure 5a shows a portrait that has had ink spilled on it. Thelocations of the ink spots are indicated in a mask image, Fig.5b. When a two-dimensional multiresolution interpolation proce-dure is applied, the missing image points are filled in smoothly,as shown in Fig. 5c. In many cases, the result is so naturallooking that the flaw would not be detected except on close

6 RCA Engineer 30-5 Sept / Oct.1985

Intensity

G.

x

G3

G.

G22

G3,3

Extrapolated Image

Fig. 4. Multiresolution interpolation. (a) Build Gaussianextrapolated pyramid. (b) Expand extrapolated pyramid.

x


(a)

Fig. 5. 5. Interpolation to fill missing points in an image.

(a)

Fig.6. Extrapolation example.(b)

examination. Figure 6 illustrates extrapolation when knownimage points represent only a small island within the imagedomain.

Image mergingThe problem

It is frequently desirable to combine several source images intoa larger composite. Collages made up of multiple images areoften found in art and advertising, as well as in science (forexample, NASA's mosaic images of the planets). Images caneven be combined to extend such properties as depth -of -fieldand dynamic range.

The essential problem in image merging may be stated asone of "pattern conservation." Important details of the compo-nent images must be preserved in the composite, while nospurious pattern elements are introduced by the merging process.Simple approaches to merging often create visible edge artifactsbetween regions taken from different source images.

To illustrate the problems encountered in image merging,suppose we wish to construct a mosaic consisting of the left halfof an apple image, Fig. 7a, and the right half of an orange, Fig.7b. The most direct procedure is to simply join these imagesalong their center lines. However this results in a clearly visiblestep edge (Fig. 7c).

An alternative approach is to join image components smoothlyby averaging pixel values within a transition zone centered on

-V.

(a) (b)

(c)

(c) (d)

Fig. 7. Multi -resolution spline of apple an orange. (a) Apple,(b) orange, (c) cut and paste composite (d) multi -resolutionpyramid mosaic.

the join line.3'4 The width of the transition zone is then a criticalparameter of the merging process. If it is too narrow, thetransition will still be visible as a somewhat blurred step. If it istoo wide, features from both images will be visible within thetransition zone as in a photographic double exposure. Theblurred -edge effect is due to a mismatch of low frequenciesalong the mosaic boundary, while the double -exposure effect isdue to a mismatch in high frequencies. In general there is nochoice of transition zone width that can avoid both artifacts.

8RCA Engineer 30-5 Sept./ Oct.1985

Multiresolution spline

We can resolve the transition zone dilemma if the images aredecomposed into a set of bandpass components before they aremerged. A wide transition zone can then be used for the lowfrequency components, while a narrow zone is used for the highfrequency components. In order to have smooth blending, thewidth of the transition zone in a given band should be aboutone wavelength of the band's central frequency. The mergedbandpass components are then recombined to obtain the finalimage mosaic.

Let So, SI, S2,...Sk be a set of K source images. A set ofbinary mask images M0, -M1, -- 2,--kM determine how the source- - Mimages should be combined. Mk is "1" where source image Skis valid, and "0" elsewhere. Simply multiplying SkX Mk andsumming over k would give a "cut and paste" composite withstep edges. Instead, we build a Laplacian pyramid Lk, for eachsource image, and a Gaussian pyramid Mk, for each maskimage. A composite Laplacian pyramid Lk, is formed by "cuttingand pasting" at each spatial scale by weighting each sourcepyramid level by its corresponding mask:

L ciaD = A 1 k(i,j) Lki(I,j)

The final image is reconstructed from Lc by expanding eachlevel and summing. Smooth blending is achieved because thetransition zone in each pyramid level is comparable to a wave-length of the central frequency of that level. When this procedureis applied to the apple and orange images of Fig. 7, an "orple"is obtained with no visible seam (Fig. 7d).

Multifocus

When assembling information from multiple source images weneed not always proceed region by region guided by maskimages. Some types of information can be merged in thepyramid node by node and be guided by the node's own value.Here we show how this type of merging can be used to extendthe depth -of -field of an image or increase its dynamic range?

Figures 8a and 8b show two exposures of a circuit boardtaken with the camera focused at different depth planes. Wewish to construct a composite image in which all the componentsand the board surface are in focus. Let LA and LB be Laplacianpyramids for the two source images. The low frequency levelsof these pyramids should have nearly identical values, sincechanges in focus have little effect on the low frequency compo-nents of the image. On the other hand, changes in focus willaffect node values in the pyramid levels where high spatialfrequency information is encoded. However, corresponding nodesin the two images will generally represent the same feature ofthe scene, and will differ primarily in attenuation due to blur.The node with the largest amplitude will be in the image that ismost nearly in focus. Thus, "in focus" image components canbe selected node -by -node in the pyramids rather than region-by -region in the original images. A pyramid Lc is constructedfor the composite image by setting each node equal to thecorresponding node in LA or LB that has the larger absolutevalue:

If

then

LAl(i,i)>L di,j)

Lcia,D= L Ai(l,i)

(a) (b)

(c)

Fig. 8. Multi -focus composite. (a, b) Two images of thesame scene taken with different focuses, and (c) compositewith extended depth -of -field.

else L o(l,f) = L 810,D.

The composite image is then obtained simply by expanding andadding the levels of Lc. Figure 8c shows an extended depth -of-field image obtained in this way.

Creating realistic looking images: shadows andshadingThe problem

We will consider three different approaches to the problem ofcreating a realistic looking image. The first approach is to usethe computer as a paint box. No mathematical description ofthe scene is given. All the renderings, shading, shadows, andhighlights are done "by hand," as though the artist were using acanvas. The advantage of "paint box" approach is completeartistic control. The disadvantage is that it is time consuming tocreate an image and difficult to make major changes withoutredrawing completely.

A second approach is to create a three-dimensional mathe-matical "universe." The artist specifies the location of objects inthis new world, their shapes and physical properties, and thelocation of light sources. A two-dimensional, photograph -likeimage of the three-dimensional world is made by tracing a largenumber of light rays as they are reflected, refracted and absorbed.The advantage of this "physics simulition" approach is thatvery realistic looking images can be created. This approach isalso flexible in that the viewing angle and properties of thecomponent parts can be changed as input parameters. Thedisadvantage is that it requires a complex physical model and alot of computation time.

There is a third, "multiresolution," approach that lies some-where between the first two and combines some of the advantages


(a)

(d)

Fig. 9. Multiresolution shadowing and shading of flatshapes.

of each. As in the paint box approach, we are concernedprimarily with painting the two-dimensional image. The differ-ence lies in the type of information on the artist's palette. Weusually think of a palette as an array of colors that can beblended and applied to an image. Using the pyramid we canextend the definition of palette to include multiresolution shapeand edge information as well as color and intensity. Multi -resolution lowpass and bandpass copies of image features areextremely useful in creating special effects and in adding realismto an artifically generated shape. The advantage of this approachis that natural looking images can be generated quickly withoutresorting to an elaborate physics simulation. Artistic decisions aremade by viewing the two-dimensional image and the pyramidlevels and combining desired elements of each.

An example

As an example of how an artist might use the pyramid as aspatial frequency palette, consider the problem of making flatshapes (Fig. 9a) appear three-dimensional by adding realisticlooking shadows and shading. Both the shadows and shadingresemble blurred copies of the original shape. Building a Gaussianpyramid of Fig. 9a, we select a lowpass copy that resembles softshadows (Fig. 9b). Comparing Fig. 9a and a slightly displaced9b pixel by pixel, and taking the maximum value at each point,

(b) (c)

(e)

tIf)

gives an image of shadowed paper cutouts floating or sitting ona glass -topped table (Fig. 9c). Now we need to add dimensionto the white cutouts. It has long been known that filtering animage with a "gradient" filter (1,-1) gives an effect of sideillumination. When gradient filtering is done at multiple resolu-tions, there is a great improvement in this bas-relief effect.Performing a combination of gradient and lowpass filtering onFig. 9a gives us the low -frequency relief 9d. If the minimum ofFigs. 9d and 9a is taken, the result is an image of three-dimensional -looking droplets (Fig. 9e). Finally, we add Fig. 9eto 9c to form shadowed droplets in Fig. 9f.

Clearly, many other interesting graphic effects can be gener-ated by imaginative use of the pyramid. It provides the artistwith a convenient and efficient way of accessing certain impor-tant features of an image-shapes and edges-at multiple resolu-tions. For certain problems, especially when computation facil-ities are limited, the multiresolution approach offers considerablerealism for little computation time.

Pyramid generation of fractals

The problem

The computer graphics community has adopted fractals as aremarkably effective way of synthesizing natural looking tex-

10 RCA Engineer 30-5 Sept. / Oct. 1985

11I111rPir 11111111173

Fig. 10. Cloud as a sum of random circles.

1 cm= 100 km

1 cm = 10 kmFI

1 cm = 1 km

Fig. 11. Coastline on three different spatial scales.

irrirkir2

3

4

5

Fig. 12. Pyramid generation of fractals.

tures.5'6'''' The problem is to generate these textures quicklyusing multiresolution techniques.

Fractals and the pyramid

Traditional mathematics has relied on idealized models of thecomplicated and irregular forms of nature. Structures such asclouds, mountains, and coastlines are difficult to describe interms of continuous, differentiable functions. Recently,Mandelbrot devised a new class of functions called "fractals" toexpress these complex natural forms.'

A fractal function includes both the basic form inherent inthe object and its statistical or random properties. For example,a cloud can be visualized as a sum of random circles. The basiccircle shape is seen at randomly distributed sizes and positions(Fig. 10).

Fractals have the property of self -similarity over many differ-ent geometric scales. A fractal appears similar as the spatialscale is changed over many orders of magnitude. As an example,consider a very jagged, rocky coastline. The coastline looksqualitatively similar when plotted on different spacial scales (seeFig. 11).

We have devised a fast fractal generation technique based onthe pyramid algorithm, which takes advantage of the self-similarity of fractals. The pyramid breaks an image up into asum of bandpassed images plus a lowpass filtered image. If aninherently self -similar fractal image is decomposed into pyramidform, one would expect the bandpassed images to look similarat each spatial frequency scale. Conversely, if similar patternswere entered into each spatial band of a pyramid, the recon-structed image should look like a fractal.

Ogden/Adelson/ Bergen/Burt: Pyramid -based computer graphics11

Fig.13. Examples of fractals. (a) Clouds, (b) woodgrain, (c)galaxy, (d) papyrus, (e) granite, (f) waves, (g) fractal"tree," and (h) fractal gasket

Fractal generation methodThe method used involves replicating a basic form, or "gener-ator," at a succession of spatial scales, then summing to producea fractal image (Fig. 12):

1. A "seed image" is set. This can be a regular array or arandom pattern of dots.

2. A basic form, or "generator," is selected. This shape willappear at many spatial scales.

3. The seed image is convolved with a filter having the sameshape as the generator. Each dot in the seed image will reproduceas an image of the generator. The result is a sum of generatorswith amplitude and position determined by the seed image.The convolved image is entered into the various levels of thepyramid.

4. The pyramid is reconstructed by expanding the variouslevels and adding. Because we have included generators overa wide range of spatial frequencies, the reconstructed imageis a self -similar fractal.

5. The fractal image can be threshholded, colored, shaded,filtered, or otherwise manipulated to give pleasing effects.

Some examples of pyramid fractuals are given in Table I and inFig. 13.

Realtime fractal animation method

A moving fractal image (for example, clouds drifting by, orrippling waves) can be decomposed into a sum of spatialfrequency bands. The different spatial frequency componentsmove at different speeds. In an ocean wave, for example, low -frequency swells move more quickly than high -frequency rippleson the water's surface.

An initial fractal image is expressed as a sum of severalspatial frequency components. The image is advanced one timeframe by moving each spatial frequency component a specifiedamount and summing again. This creates a new fractal, whichis slightly changed from the previous frame. Continuing thisprocess gives an animated sequence of continuously evolvingfractal images. Each spatial frequency component can be periodicwith a fairly short repeat distance (say, half the image width),but the sum has a much longer period if the different componentsmove at various speeds that are not multiples of each other. Wehave demonstrated this technique using our DeAnza 1P8500image processor. Each new frame takes about 1/15th of asecond to compute, and the overall repeat time was severalminutes. Figure 14 shows an example of fractal animation for abeach scene (also see front cover).

ConclusionsPyramid representations have much in common with the waypeople see the world. Human beings like to look at things onmany spatial scales simultaneously. A strong analogy existsbetween an artist who paints images by adding progressivelyfiner details, and a computer scientist who constructs images byadding together the spatial frequency bands of a pyramid.Representing an image on multiple spatial scales allows us todo things like blending apples and oranges naturally. When weexpress a picture as a sum of pyramid bands and operate oneach band separately, we are taking an artistically familiar"painter's" approach. This contrasts with the "physics simulation"

12RCA Engineer 30-5 Sept. / Oct.1985

Table I. Fractal examples

Figure Seed Image Generator Shape

PyramidParameters:

Reconstructionfilter levels Enhancements

Clouds(13a)

Woodgrain(13b)

Galaxy(13c)

Papyrus(13d)

Granite(13e)

Waves(13f)

FractalTree(13g)

FractalGasket(13h)

Random noise

Random noise

Sparse randomnoise

Random noise

Random noise

Random noise

Geometricseries of

dots

Single dot

Circle

Tapered line

Dot

Tapered line

V-shapeconvolved w/gradient filter

Tapered lineconvolved w/gradient filter

Y -shapedbranch

5X5 filter Threshold(a=0.45)*

6 levels

3X3 filter(a=0.5)*5 levels

3X3 filter(a=0.5)5 levels

3X3 filter(a =0.5)*5 levels

3X3 filter(a =0.5)*5 levels

3X3 filter(a=0.5)*5 levels

2X2"zoom filter"

(1 1)

6 levels

3X2 filter(0 1 0)\1 016 levels

EnhanceContrast

Gradientfilter

Pyramid levelsmultiplied by rampto give perspective

*These filters are separable with the form f(x,y)=w(x)Xw(y) where w(x)=(1/4-a/2, 1/4, a, 1/4, 1/4-a/2)

approach, which grows more out of the physics and mathe-matical modelling traditions. Greater realism can be achievedby using the physics simulation approach, but the complexityand computation time are vastly increased over the multireso-lution pyramid approaches described. Considering how well wecan fake reality with the pyramid, it seems particularly wellsuited for making "realistic looking" computer graphic imageson small systems.

References1. P.J. Burt, "Fast Filter Transforms for Image Processing", Computer Graphics

and Image Processing, 16, pp. 20-51 (1981).

2. P.J. Burt and E.H. Adelson, "The Laplacian Pyramid as a Compact ImageCode", IEEE Transactions on Communication COM-31, pp. 532-540 (1983).

3. E.H. Adelson, C.H. Anderson, J.R. Bergen, P.J. Burt and J.M. Ogden, "Pyramidmethods in image processing", RCA Engineer, Vol. 29, No. 6, pp. 33-41(Nov./Dec. 1984).

4. P.J. Burt and E.H. Adelson, "A Multiresolution Spline with Application toImage Mosaics," ACM Transactions on Graphics, Vol. 2, pp. 217-236 (1983).

5. B. Mandelbrot, The Fractal Geometry of Nature, Freeman, San Francisco (1982).

6. A. Fournier, D. Fussell, and L. Carpenter, "Computer Rendering of StochasticModels," ACM Graphics and Image Processing Vol. 25, No. 6, p. 371 (1982).

7. A. Pentland, "Fractal-Based Description of Natural Scenes," IEEE Transactions,SMC-13 (1983).

8. J.M. Ogden, "Generation of Fractals Using the Burt Pyramid," presented at theOptical Society of America meeting (October 1985).


-+

LO

Fig. 14. Fractal animation method

LII

Moving Spatial Frequency Bands

Animated Fractal Sum

L22

Threshold Fractal

G33

Joan M. Ogden received a BS in Mathematics from theUniversity of Illinois in 1970, and a PhD in Physics from theUniversity of Maryland in 1977. After two years as a Postdoctoralresearch associate at Princeton Plasma Physics Laboratory, shestarted her own consulting company, working on a variety ofapplied physics problems.

In 1982 Dr. Ogden joined the Advanced Image ProcessingResearch group at RCA Laboratories, first as a consultant and,in 1984, as a part-time Member of the Technical Staff. Herresearch interests at RCA include applications of the pyramidalgorithm to noise reduction, data compression, and texturegeneration. She is the author or coauthor of more than 20articles, has one patent pending, and is coinventor of an issuedpatent.

Dr. Ogden was awarded an National Science FoundationVisiting Professorship at Princeton University in 1985. She is onleave from RCA to teach at Princeton during the 1985 academicyear.Contact her at:Princeton UniversityPrinceton, N.J.(609) 452-4966

14 RCA Engineer 30-5 Sept./ Oct.1985

Edward H. Adelson received the BA degree, summa cum laude,in Physics and Philosophy from Yale University in 1974, and aPhD degree in Experimental Psychology from the University ofMichigan in 1979. Dr. Adelson joined RCA Laboratories in 1981as a Member of the Technical Staff. As part of the AdvancedImage Processing group, in the Advanced Video SystemsResearch Laboratory, he has been involved in developingmodels of the human visual system, as well as image -processing algorithms for image enhancement and datacompression.

Dr. Adelson holds two U.S. patents and has published a dozenpapers on vision and image processing. His awards include theOptical Society of America's Adolph Lomb Medal (1984) and anRCA Laboratories Outstanding Achievement Award (1983). He isa member of the Association for Research in Vision andOphthalmology, the Optical Society of America, and Phi BetaKappa.Contact him at:RCA LaboratoriesPrinceton, N.J.Tacnet: 226-3036

James R. Bergen received the BA degree in Mathematics andPsychology from the University of California, at Berkeley, in1975, and the PhD degree in Biophysics and Theoretical Biologyfrom the University of Chicago in 1981. His interests concern thequantitative analysis of information processing in the humanvisual systems, and the application of this analysis to problemsin imaging system design and evaluation.

During 1981-82, Dr. Bergen was a Postdoctoral Member ofTechnical Staff at Bell Laboratories. In 1982 he joined RCALaboratories as a Member, Technical Staff. His researchactivities at RCA include theoretical and experimental work onhuman vision, as well as perceptual aspects of HDTV systemsand image processing techniques. In 1983, he shared an RCALaboratories Outstanding Achievement Award.

Dr. Bergen has published many papers on visualpsychophysics and perception, and is a frequent speaker atconferences on these subjects. He is a member of the OpticalSociety of America and the Association for Research in Visionand Ophthalmology.Contact him at:RCA LaboratoriesPrinceton, N.J.Tacnet: 226-3003

Peter J. Burt received the BA degree in Physics from HarvardUniversity in 1968, and the MS and PhD degrees, both inComputer Science, from the University of Massachusetts,Amherst, in 1974 and 1976, respectively. From 1968 to 1972 heconducted research in sonar, particularly in acoustic imagingdevices, at the U.S. Navy Underwater Systems Center, NewLondon, Connecticut, and in London, England. As a PostdoctoralFellow, he has studied both natural vision and computer imageunderstanding at New York University (1976-78), BellLaboratories (1978-79), and the University of Maryland (1979-80). He was a member of the engineering faculty at RensselaerPolytechnic Institute from 1980 to 1983.

In 1983 Dr. Burt joined RCA Laboratories as a Member of theTechnical Staff. He worked in the areas of computer vision,image data compression, and human perception, and in 1984 hebecame Head of the Advanced Image Processing group.

Dr. Burt has published more than 20 technical papers in hisfields of expertise and is a member of the IEEE.Contact him at:RCA LaboratoriesPrinceton, N.J.Tacnet: 226-2451

Ogden/Acelson/Bergen/Burt: Pyramid -based computer graphics 15

R.D. Scott

Graphics for the definition and visualizationof physical systems some case studies

PRSPEC: Computer graphics software that reduces the amountof engineering effort necessary to produce certain systemsmodels and facilitates error checking.

The concept of graphics pre- and post -processors is well established. Among theearliest users were structural analysts, whopreviously struggled with hundreds ofcoordinate sets and connection lists in theprocess of describing the geometry of aphysical system. They had to struggle againwith a mountain of paper output to inter-pret stresses, displacements, and otherparameters of interest in the analysis. Errorswere frequent, and often discovered onlywhen the answers made no sense.

The advent of two graphics processors,SUPERTAB from Structural DynamicsResearch Corp. (SDRC), and PATRANfrom PDA Engineering, Inc., reduced theanalyst's problem to that of constructingan image, on the screen, of the critical

Abstract: This article describes severalapplications of the graphics programPRSPEC, which include post -processingof data from mass properties andradiation shielding programs. PRSPEC isa graphics program that can be run ineither batch or interactive mode. It can beused both as a pre- or post -processor ofgeometric data, and provides several user -defined interfaces, among which is theability to execute user -written programsfor additional generation and conditioningof data.

The applications and softwaredescribed herein, including PRSPEC,were developed by the author at RCAAstro-Electronics Division. They aregraphics pre- and post -processors.

01985 RCA Corporation.Final manuscript received May 20, 1985Reprint RE -30-5-2

geometry of the structure (the process isnow interactive, using menus and cursors).Even the problem of constructing gridsand elements of a finite -element analysishas been software -automated for someclasses of problems. Another class of prob-lems is supported by SDRC's GEOMODsystem and related products that model"solid" objects and can preserve kine-matic coupling, both geometrically andgraphically.

Within the industry, a primary goal isto reduce the drudgery of data entry, mini-mize errors, and allow the computer userto interact more and more in "natural"language, in conversational terms, or bymenu.

As graphics packages are developed com-mercially, they tend to be specialized forparticular disciplines. Thus, finite -elementprocessors are broadly generalized to createmodels that look like finite -element models.With PATRAN and GEOMOD, solidimages of complex bodies-such asautomobiles, people, and the like-can berendered even to the extent of providingshadows. The process is, however, one inwhich the user defines the object as ageometric entity, and the software systemdoes the special rendering. While the com-mercial software systems do an outstandingjob for their applications, one often findsit difficult to do specialized operationsoutside the original scope because no usersoftware interfaces are provided.

Both the limitations of software flexibilityand restricted individual access to someresources has led to the development ofspecial applications of the graphics pro-cessor designated PRSPEC. This processorhas a general user interface, allowing batch

and interactive input, and especially theability to link user -written subroutines. Arecent application was the demonstrationof automatic boresighting of a satellitesensor model with respect to a given lineof sight, involving rotation about two gim-bal axes and automatic graphic display ofboresight verification and field of view.

The processor software is portablebetween IBM/CMS and other systems withFortran compilers.

Geometry visualization forradiation effects analysisAt Astro-Electronics, extensive use is madeof the Modified Electron -Volt Dose Pro-gram (MEVDP) software, developed bythe Kirtland Air Force Weapons Lab-oratory, for analysis of radiation damageof electronic components. The analysisconsists of determining the equivalentamount of mass shielding between a radia-tion source and a given component andinferring the reduction of dosage affordedby that shielding. To specify the equivalentmass, a complete geometric descriptionmust be made of all material surroundingthe object in question.

For the MEVDP program, the estab-lished method for providing input data isto divide the surrounding structure intosolid primitive elements (prims), much like3-D finite elements, whereby the elementsconsist of the predefined shapes of hexa-hedrons, spheres, cylinders, cones,hemispheres, and ellipsoids (see Fig. 1).Each shape must be described by certainkey points and/or parameters. Thus, acylinder primitive is defined by the twoendpoints of its axis and a radius. More -

16 RCA Engineer 30-5 Sept./ Oct. 1985

over, it must be located relative to thestructure being analyzed by the absolutecoordinates of the key points in a globalframe of reference.

Finally, for each element of the structure,there are associated parameters includinga sequence number, mass density, andpassible relations with other elements. (TheMEVDP admits Boolean operations, sothat the combined mass of two shapespartially occupying the same volume willbe treated as the mass of the union of thetwo shapes.)

The MEVDP makes no provision forthe generation of data. Its input protocolrequires that the data stream consist of asequence of entries for each element, eachsubset consisting of a header entry (withsequence number and physical properties),and a set of entries listing the coordinatesof the key points (see Table I). In theabsence of supporting software the analystwas required to build the model essentiallyby hand, scaling the drawings as necessary,visualizing rotations and translations ofthe elemental shapes, and keeping trackof the assembly by whatever means pos-sible. One unique concern was the handlingof "voids," wherein enclosed volumes arecreated by Boolean subtraction of onevolume from a slightly larger one. Thus,the void elements as well as the "real"structure had to be kept in mind.

Building a mass model by manual calcu-lations for a complex object buried insidea spacecraft (itself a complicated structure)is a formidable undertaking, and onefraught with errors. The question was posedto the author whether a way could bedeveloped to at least verify the geometryof the model before running the MEVDP.Inasmuch as the users were committed tothe use of the IBM/CMS system at thetime, and it had no finite -element method(FEM) processors installed, the availabilityof the PRSPEC graphics program led tothe implementation of a data streaminterpreter that yields graphic displays ofthe structure. Subsequent refinements haveprovided a fully interactive facility.

Data stream interpreter

Because each sequential subset of data inthe MEVDP input file describes a geo-metric entity, it is a fairly straightforwardprocess to capture each subset and createa 3-D image of the entry. The PRSPECprocessor was already equipped with theability to draw circles and ellipses(as well as straight lines), so the task wasto convert each MEVDP subset to a

HEXAHEDRON

8 POINTS AS LISTED

SPHERE

CENTER PLUS ANYSURFACE POINT

2

5

ACYLINDER

2 POINTS PLUSRADIUS

ELLIPSOID

CENTER PLUS3 SEMI.MAJORAXIS POINTS

HEMISPHERE

CENTER PLUSPOLAR POINT

Fig.1. MEVDP primitive geometry.

PRSPEC input data subset. It turns outthat there is essentially a one-to-onecorrespondence between the two. Forexample, for a hexahedron the MEVDPrequires the corner points to be specifiedin a particular order; therefore, a hexa-hedron can always be reformatted with aknown topology.

Similar considerations follow for theother primitives, but here the analyst needssome options. The graphics program oper-ates essentially in wireframe mode, butoptions have been developed that alloweach primitive (by itself) to be displayed

Table I. MEVDP source data format.

CONE

2 POINTS PLUSHALFANGLE

=

61 ,

TRUNCATED CONE

SAME AS CONE PLUSTRUNCATION POINT

TRUNCATED ELLIPSOID

SAME AS ELLIPSOID PLUSTRUNCATION POINT(S)

in either wireframe caricature, or withhidden lines removed. In this way, a com-plex body can be displayed with all theprimitives showing in their relative loca-tions, but with a significant amount ofclutter removed. The display is thus semi-transparent to the viewer. For the purposeof verifying the MEVDP input, this methodworks quite well, because errors in place-ment of the primitives are the primaryconcern, and the pictorial estheticssecondary.

As an example, consider the MEVDPrepresentation of an ice cream cone (a

RADIATION BODY DATA STREAM (02/01/85)3.000E+01 2.500E-01 1.000E-01ABFG 00 00 00 00 50 00

YES YES NO NO2***

STANDARD HEADER CARD: MEVDP DATA2000 9 1 0.2 0.0500 0 2 0

0.0 0.0 5.0 0.0 0.0 0.02001 7 1 1.0136 1.1000 0 2 0

0.0 0.0 5.0 0.0 0.0 6.0136

END

Scott: Graphics for the definition and visualization of physical systems 17

cone plus a hemisphere, of course). TheMEVDP data stream is shown in Table I,and the corresponding PRSPEC datastream in Table H.

The first five lines of Table I are control

Table II.

data entries for the graphics translator(named "RADBOD"). Usually the datafor an entire body are kept in a singlereference file, and graphics for part or allof the body are generated under control

PRSPEC data file generated from MEVDP data.

2WAY

YES NO 0 ONO -02N0

1.100E+01 8.500E+001.000E-01 1.000E-01 1.000E-01

NO NO YES NO NO NO NO NO NO

TEST SET1.000E+00

1.000E-01 3 3

1.000E+003 3

C: STANDARD HEADER CARD: MEVDP DATANEWP 71

PRIM CONE

2000 9 1 0.2 0.0500 0 2 0

CART 51 0.0 0.0 5.000E+00 52 0.0 0.0 0.0

CART 53 1.014E+00 0.0 5.000E+00 54-1.014E+00 0.0 5.000E+00

CART 55 0.0 1.014E+00 5.000E+00 56 0.0 -1.014E+00 5.000E+00

CART 57 0.0 0.0 -1.000E+00 58 0.0 0.0 1.000E+00

CIRC 51 1.014E+00 1.000E+01 0.0 51

B 51 51

JUMPF 51 52

5.000E-01 3

A 53 52 0 54 52 0 55 52 0 56 52

JUMPPRIM HEMI

2001 7 1 1.0136 1.1000 0 2 0

CART 61 0.0 0.0 5.000E+00 62 0.0 0.0 6.125E+00

CART 63 1.125E+00 0.0 5.000E+00 64-1.125E+00 0.0 5.000E+00

CART 65 0.0 1.125E+00 5.000E+00 66 0.0 -1.125E+00 5.000E+00

CART 67 1.000E+00 0.0 0.0 68 0.0 1.000E+00 0.0

CIRC 61 1.125E+00 1.000E+01 0.0 61

B 61 61

CIRC 62 1.125E+00 1.000E+01 0.0 63

B 61 68

CIRC 64 1.125E+00 1.000E+01 0.0 62

B 61 68CIRC 62 1.125E+00 1.000E+01 0.0 66

B 61 67

CIRC 65 1.125E+00 1.000E+01 0.0 62

B 61 67

PRIMENDEND75000 0.0 0.0 2.500E+00 5.600E+00 8.000E+00 6.000E+00

1.100E+01 8.500E+00 0.0 1.100E+01 8.500E+00 +2 0 0 0 0 0

END

STOP

(a)

Fig. 2. Options for the ice cream cone.

18

(b) (c) (d) (e)

of a "sieve," represented by line 4. By thismeans, multiple runs can be made, withdifferent graphic results, by selecting certaingroups of prims via the sieve. The sieveconsists of a list of serial numbers of theprims to be drawn, with "wild card" (*)options; for the above example all serialnumbers in the range 2000 to 2999 are tobe plotted.

The first and second lines of the file arethe file header and information controllingline styles. The third line of the file selectsgraphics options in terms of extra lines forvisualization. A cylinder, for example, iseasily visualized if a number of its gener-ators are shown, as well as its envelopetangents. Spheres and ellipsoids arerendered by a set of great circles. Theselines may be included or not, and hiddenlines shown as dashed, or removed, at thediscretion of the user. Envelope lines aredetermined by the program and shown, inany event.

Inspection shows that there is generallya one-to-one correspondence between thefiles represented by Tables I and II. Eachsubset of Table II starts with a controlentry that defines the shape (PRIM). Thenext line echoes the MEVDP header entryand will be stored for reference. Entrieslabelled CART are coordinate sets forboth key points and points generated forthe artwork. Entries for various circles(CIRC) follow. Once coordinate informa-tion has been accounted for, entries areadded to connect the various points withstraight lines. Without going into detailsof the meaning of the entries, it is simplynoted that sufficient information has beendefined to generate cartoons of the shapeslisted.

Figure 2 shows the several styles ofoutput that would be available from thisexercise. First, Fig. 2a is a simple wireframemodel of the cone and hemisphere com-bined. The objects have been renderedusing generators defined at four pointsequally spaced around the defining circles.The number of points chosen is, of course,arbitrary, but four are sufficient to give aworkable picture. A significant improve-ment in the cartoon occurs when envelopelines are added (Fig. 2b). These flesh outthe figures and help resolve ambiguities tosome degree. The next improvement inrendering is shown in Fig. 2c, in whichhidden lines have been removed and onlythe envelopes displayed. Fig. 2d restoresthe visible generators, and 2e shows hiddenlines as dashed.

A more serious example of an actualworking model is shown in Fig. 3. This

RCA Engineer 30-5 Sept. / Oct. 1985

figure represents a circuit board pack. Theaim is to assess the effective shieldingsurrounding a chip on one of the boards.Three levels of detail are shown, all derivedfrom a single database. Figure 3a modelsthe physical bodies surrounding a device(PIP) that contains a set of circuit boards.The model for the PIP is shown in Fig.3b, and the board level model (with chips)is shown in Figs. 3c and 3d.

Once the graphics data have been readfrom the input file, various views can becreated interactively for inspection. Uponverification of the total database, the entiremodel may then be submitted to theMEVDP for dose analysis.

Geometry visualization for massproperties analysisSubsequent to the creation of the radiationanalysis interpreter, a similar considerationbecame apparent with respect to theanalysis of spacecraft mass properties. Aprogram had been developed, by a numberof analysts at RCA Astro, for the deter-mination of overall mass, center of mass,and moments of inertia of a spacecraft,based on the elemental mass properties ofthe components. Elemental mass propertiesare found by examination of parts lists,drawings, and data from previous calcula-tions. Input data files are generated, byhand, for each subassembly on a spacecraftand considerations similar to the MEVDPinput protocol are required: coordinatesand rotations must be specified in an globalcoordinate frame, as well as elementalmass properties.

For the mass properties program, shapesare specified by type, length, width, radii(if appropriate), physical properties, andlocation of the center of mass, Table IIIshows a section of a typical input file,which may contain hundreds of shapesfor an actual vehicle.

Note that there is at least as muchdifficulty, if not more, in mentallyvisualizing the shape and orientation of anelemental object (in the author's opinion)compared to the MEVDP situation, andthe problems of verification are certainlyas severe.

As before, a PRSPEC-compatible trans-lator was written for graphic interpretationof the input data typified by Table III.The same set of prims used for MEVDPwas incorporated and found adequate, andFig. 4 shows a sample hardcopy plot froma mass properties run. It is interesting todiscover that the cartoons of Fig. 4 lookvery much like a spacecraft, even though

close examination reveals the abstractionsused to represent the components. Massproperties runs are not made interactively,although they could be. The size of thedatabase and its impact on cost is the

COMPUTER MODEL OF DIASP 50-2SPACECRA TPROGRAMMABLE INFORMATION PROCESSOR (PIP)

(PANEL NO.

(a)

DMU(c)

principal restriction.The mass properties program has been

installed in the Prime computer at Astro-Electronics, which also supports theMedusa CAD system. As a point of

Fig. 3. Radiation analysis graphics for a circuit board pack.

Table Ill. Mass properties program input data stream.

DMU

Id)

101030100-44.75 -5.2 -17.8 14.3 1.9 1.9101030100-44.75 -5.8 22.4 14.3 1.9 1.9101030100-44.75 -35.1 -10.2 14.3 1.9 1.9101030200-45.17 21.04 7.9 13.15 .38 .27101030200-45.17 21.04 11.9 13.15 .38 .27101030200-45.17 21.04 15.19 15.15 .38 .27101030300-45.08526.996 6.641 15.56 7.84 3.44101030300-45.08526.996 10.641 15.56 7.84 3.44101030300-45.08526.996 14.641 15.56 7.84 3.44101030400-44.75 16.0 12.5 14.0 1.9 1.9101030400-44.75 19.0 10.0 14.0 1.9 1.9101030400-44.75 19.0 14.0 14.0 1.9 1.9101030600-33.3 0.474 2.91 133943.932756.412945.934.20 -190.817.85 49101030700-44.06 26.44 -1.1 13191.7 104.5 291.2101030800-14.13 -18.9 -13.7 16.25 10.53 7.75

1

2010308000. 90. 90. 112.5101030800-14.13 -18.9 13.2 16.25 10.53 7.75

1

2010308000. 90. 90. 52.5101031001-44.095-43.01524.548 13171.54 167.44 20.23


1.100E+01 8.500E+002.500E-01 2.500E-01

C: X -Y -Z AXES:

1.000E+00 1.000E+002.500E-01 2.500E-01

CART 1 0.0 0.0 0.0

CART 3 0.0C: CONNECTIONSJUMP

1.000E+00 0.0

A 2

JUMPC: VIEWS

ENDEND

1 3 0 1 4

75000 0.0 0.0

1.100E+01 8.500E+00 0.0 1.100E+01

END

interest, the output from PRSPEC in thisexample was routed directly to a drawingfile, from which Fig. 4 was then plotted.With this capability, the analysis resultscan become part of the formal CADdatabase.

Interactive modification of thedatabaseWhile configuring the MEVDP interpreter,it was recognized that graphics support

would not alleviate the manual labor ofinitial preparation of data. AlthoughPRSPEC provides the ability to copyand/or transform (step -and -repeat) inputdata subsets, it was decided that a dedicatedmanipulator would be more convenientfor the analysis at that time. A processor(EXPAND) was created that does thefollowing operations on a data subsetrepresenting a prim: copy, translate, rotate,and stretch. It also defines a set of primsas a block subset and performs copy,

Fig. 4. Hardcopy output from mass -properties translator program.

Table IV. PRSPEC skeleton data file.

2WAY

YES NO 0 ONO -02N0 NO NO NO

NO NO NO NOTEST SET

3 3 3 3

2 1.000E+00 0.0 0.0

4 0.0 0.0 1.000E+00

2.500E+00 5.600E+00 8.000E+00 6.000E+008.500E+00 +2 0 0 0 0 0

NO NO NO NO YES

rotations, and translations for that set. Thisprovides a very convenient way to generatea data file that represents a complex bodymade of many similar parts, such as circuitboards and chips. The resulting data file isa standard MEVDP data file.

EXPAND is used in a batch mode,however, to aid in the preparation of theMEVDP data file. In a sense, it allows theanalyst to make mistakes more easily. Ifthe mistakes show up later on the graphics,the analyst must still go back to the sourcedata file and correct them.

We decided that the most cost-effectiveway to resolve the issue would be thefollowing:

Create an initial MEVDP model withEXPAND using the traditionalmethods.

Enter an interactive session withPRSPEC to visually examine the modelfor errors.Correct the errors in the same interactivesession by modifying the position, orien-tation, and size of shapes.

Add and/or delete certain prims ifdesired.

Save the results of the interactive session(now in PRSPEC format) in an outputfile.

Reinterpret the output file in terms ofa final MEVDP model file for analysis.

The astute reader might point out that asingle package could do the job by buildingthe MEVDP model from scratch, inter-actively, and retaining the MEVDP dataformat throughout. This is quite true, but(a) scratch is hard to come by, (b) themodel can indeed be built from scratchusing PRSPEC (see the following example),and (c) the final MEVDP data file is

usually generated as a one-shot operationafter the model has been interactivelyrefined. The current software set is theresult of blending existing software withnew, to address the needs of technique,schedule, and economics.

The MEVDP users were invited todefine a command set for the interactiveprocessing of the MEVDP model, and theset was then written into the PRSPECsoftware. The set allows the tracking ofany shape designated as a prim, by number,and provides the same transformations asdescribed above for the preprocessor.

With the interactive capability, one canstart with a skeleton input file for PRSPECand work the rest of a given exercise fromthere. Table IV shows a complete skeleton


Fig. 5a. The development of asimple cartoon for a spacecraftusing the available command set.(Graphics are taken from aTektronix 4014 terminal.) Startingat the top row and moving fromleft to right A unit cube isadded to the input axes set andscaled in the Z direction. Next,the scaled cube (solar panel) isshifted along the Y axis and thencopied next to itself. A cylinder(array boom) is then added,scaled, and placed at the firstpanel. Note that the prims arelabelled at their geometriccenters, for identification. Startingat the second row: A groupelement is defined, consisting ofelements 24, 25, and 26. Thegroup is copied and rotated tobecome the set 20, 21, and 22. Abox is then added to representthe spacecraft body, and a coneto represent the thrust motor.

The third row shows furthergroup operations, wherein the sixelements 2, 3, 4, 17, 18, 19 arerotated about the Y axis tosuggest rotation of the completesolar array. Finally, the entiremodel is rotated about the(arbitrary) vector (1,0,1).

Fig. 5b. The same set of figures (at aslightly lower viewing angle) with thelabels and hidden lines removed.Again, the hidden lines are processedat the element level, giving partialtransparency.


ARRAY NORMAL

Fig. 6. The same cartoon as in Fig. 5 can be used forconceptual analysis of orientation of a spacecraft withrespect to the earth and sun. The example is taken from astudy of the TOPEX spacecraft, in which one issue is tobe able to orient the spacecraft so that the solar array isalways pointing normal to the sun. This technique isknown as sun -nadir pointing, and involves rotations of

Fig. 7. A final example using the dataof Fig. 6 to create the illusion oflooking toward the sun from thespacecraft's reference, just at thebeginning of an eclipse.

the spacecraft about the local normal to the earth androtations of the array about its axis. Accurate graphics areindispensable for understanding the geometry of thisexercise. Note that both the spacecraft body and the arrayare oriented in different ways in each position, to align thevector representing the array normal parallel to the vectorpointing to the sun.


file that provides the system parametersand a set of X, Y, and Z axes. The filealso defines the initial orientation of theviewpoint for the 3-D data set.

Summary

This paper has described several applica-tions of the RCA general graphics programPRSPEC, in conjunction with dedicatedtranslators, whereby previously establisheddata streams are translated into a visualpresentation of the physical model impliedby the data. For established programsdeveloped before the widespread avail-ability of computer graphics, this techniqueis a means of enhancing those programswithout costly rewrites or reformatting ofthe data protocols.

Also shown is the value of the abilityto modify the model data while in aninteractive graphics phase wherein errorsare corrected, conceptual changes made,and an updated source data file is obtainedon a round-trip basis.

Because PRSPEC was written as ageneral 3-D graphics package, withinterfaces for user -written subroutines, it

Richard D. Scott has been active in theengineering graphics discipline for anumber of years. He participated in thesetup of the Astro-Electronics CADsystem (PRIME/MEDUSA), and wrote anumber of supporting software routinesfor the system. His background is inmechanical engineering-in both designand analysis-at Astro-Electronics andRCA's Advanced TechnologyLaboratories. He is Associate Fellow ofthe AIAA and a member of the InteractiveGraphics Committee.

Contact him at:RCA Astro-Electronics DivisionPrinceton, N.J.Tacnet: 229-2406

provides substantial flexibility for themanipulation and presentation of data, bothphysical and abstract. Applications havebeen made for vector maps and contourgeneration, kinematic studies, and thermalmodeling.

AcknowledgmentsThe mass -properties preprocessor waswritten by Robert Lutz, RCA Astro-

Electronics Division, who also generatedand provided the drawing of Fig. 4.

A post -processor (DERADBOD) waswritten by Amarjit Grewal, RCA Astro-Electronics Division, for the reconstructionof MEVDP data format from PRSPECinteractive sessions. He also created andprovided the shielding model of Fig. 3.


D.P. Barton

Spooled -graphics laser printing from CMS

CMS users at RCA Laboratories can now "spool" graphics filesto a laser printer, much as they would text or data files.

For some years, computer graphics hasbeen an expensive tool, restricted to theprivileged few whose applications de-manded, and whose funds permitted, itsuse. Its hardware costs have been higher,and it has consumed large amounts ofcomputing resources.

Now, because of the explosive growthin computer technology and the corres-ponding decrease in component costs,graphics hardware is becoming more andmore affordable. As is common in thecomputer industry, the hardware precedesthe software needed to use it. When com-puter users at RCA use the corporatemainframe or a local minicomputer toedit files or to run programs that producedata, they may indicate files or data to be

Abstract: When computer users at RCAuse the corporate mainframe or a localminicomputer to edit files or to runprograms that produce data, they mayindicate files or data to be printed atsome point in any session, either at theirlocation or at the corporate computercenter. This takes only a few simplecommands (it is called "spooling"), andthey can then proceed with their work.When the file or data is tex4 such aninterruption takes a few seconds at most.Until recently, however, graphics has beena different story. This paper describes asystem designed to provide spooling to ahigh -quality black and white graphicslaser printer.

C)1985 RCA Corporation.Final manuscript received May 20, 1985Reprint RE -30-5-3

printed, at some point in any session, eitherat their location or at the corporate com-puter center. This is called "spooling,"and takes only a few simple commands(see the sidebar on spooling). They canthen proceed with their work. When thefile or data is text, such an interruptiontakes a few seconds at most. Until recently,however, graphics has been a differentstory. The user has required a locally -connected printer or plotter, and has hadto tend it. There have been few choices:

Pen plotter-medium- to high -quality,slow, minutes for a typical plot, stillrequired for large-scale graphics outputsuch as engineering drawings.Graphics copier or electrostatic printer-low- to high -quality, fast, seconds perplot, expensive.Dot-matrix impact or ink -jet printer:lowest quality, medium -speed, speed,seconds to minutes for a typical plot,jagged diagonal lines and curves.Laser printers: highest -quality black -and -white, fast, seconds per plot,expensive.

In essence, we have been faced with eithergreat expense, low speed, or poor quality.A slow -speed pen plotter is not inexpensiveto RCA because of the long periods oftime an engineer or scientist spends waitingfor it to finish a plot.

At the David Sarnoff Research Center,our users have long recognized the needfor fast, convenient, high -quality, black -and -white hard copy graphics outputdevices used in the same way that we useour local and corporate computer centerline printers. Recently, appropriately -priced

graphics laser printers appeared, and weelected to develop the software necessaryto integrate them into our computingenvironment. Since the DSRC's businessis supporting RCA's other divisions, wedesigned this software to be usable fromother divisions.

From a systems design viewpoint, ourlaser printer is not an end, but a beginning.

Of course, we needed the spooled outputcapability possible with a shared printer,but while we can save the users time bytending the printer for them, we maynegate that saving if they must spendminutes walking across a large complexto retrieve plots. As laser printers becomeless and less expensive, we hope to movethem closer and closer to the user. Itwould be nice to have them as convenientlylocated as copy machines. Our "Labnet"Ungerman-Bass local area network canbe the vehicle for this, but the softwarearchitecture must allow for it.

Just as black -and -white laser printershave declined in price, other devices willfollow. Our software must allow for otherpossibilities, such as color slides, pen plots(for those users who must have them, sothat babysitting can be by operator, ratherthan by engineer or scientist). The arrivalof reasonably -priced, suitable -quality, colorlaser printers can be only a matter of time(if past experience dictates).

After we had considered the DSRC'susers on the corporate mainframe, we stillhad users of local computers with similarneeds to think about. Our softwarearchitecture had to be able to accommodatethem as well. Again, Labnet could providea data path.


Spooling

Different computer devices havealways had different rates of datatransfer and different capacitiesfor storing information (with thetwo characteristics usuallyvarying inversely). Smaller, fasterdevices have been placed closerto the actual computing (at themore abstract level), where theaction is, and slower deviceswith larger capacities have beenplaced farther away at the inputor output ends. The fastestdevices with the least capacityare the registers in the CPU(central processing unit). Nextare the RAM (random accessmemory) boards, some of which(called cache) are smaller andfaster than others. After these arediscs. After the disks are tapes.After the tapes are peripheraldevices such as terminals (hardcopy and video display), printers,and plotters. It may seem strangeto think of these peripherals ashaving the greatest storagecapacities, but you could type ona video display terminal untilyour fingers wore out withoutrunning out of virtual capacity,since it sends everything youtype to the computer. Likewise,

as long as the system designallows for pausing to changepaper, printers, plotters, and hardcopy terminals are only limited bythe amount of paper you can getyour hands on.

For some time, computersystems have had distributedprocessors (called controllers)for the slower devices. This hasmeant that as results wereproduced by fast CPUcomputations, they could bepassed to intermediate deviceswith greater capacities and lowerspeeds for buffering. Early in thegame, programmers discoveredthat information could be sent offto tape (spooling to spools oftape), freeing the CPU for faster,more abstract computation, whilean output controller transferred itout to a slow printer. Ashardware has declined in priceand increased in capacity andspeed, spooling moved to disc(in larger systems) and specialRAM buffers (for microcomputers).The principle remains the same:the waiter or waitress deliversthe food to the table while thechef cooks.

Our corporate mainframe's timesharingoperating system, IBM's CMS (Conver-sational Monitoring System'), provides twopaths from a user's account to remote jobentry (RJE) devices: (1) the "virtual"printer, and (2) the "virtual" punch.

For several reasons, we chose to makethe virtual punch our output path forgraphics hard copy, and with that choicewe had to consider the following:

(a) We would have to treat graphics hardcopy a little differently from text, sinceRJE station translation tables to trans-late from EBCDIC to ASCII (seesidebar) were developed long beforespooled graphics were even considered,and did not handle some "control"characters correctly for our laserprinter. We could avoid conflicts withthe needs of text printers by using thevirtual punch (and doing our owntranslation).

(b) Not many people are using the virtualpunch to punch Hollerith cards anymore.

(c) The existence of a separate path forgraphics would mean that a user couldeasily alternate between printing textand plotting graphics without needingto issue extra commands to redirectoutput from text printer to graphicsprinter.

Part of choosing a laser printer is choosingits graphics command set. All operate in a"native" mode (a language pecular to eachprinter, and which takes maximum advan-tage of its features, such as multiple fonts(character styles), but which may be com-patible with only a small portion of thegraphics "packages" (see sidebar, page 27)

in use. Most also operate in one or more"emulation" modes (imitating other olderdevices such as video terminals or pen

plotters, limited in features, but compatiblewith more software). An emulation modewould make our printer useful to themaximum possible number of users, whilethe native mode would make it moresophisticated for a few of our users. Emu-lation mode would also make it possiblefor our software to be used with morethan one brand of printer, giving us someprotection against a less than optimalchoice of brand and model, and also givingus the flexibility to acommodate newproduct offerings in a very volatile laserprinter market. We decided to start withemulation -mode compatible software forwidest possible initial use, and to addnative mode compatible software for moreelaborate applications later. Two emulationmodes are more popular than any others:

(1) Tektronix PLOT 10, emulating suchgraphics display terminals as the 4010and 4014.2 Tektronix made the firstwidely -used graphics display terminals.

(2) Calcomp plotter compatibility (Cal -comp made the first widely -used penplotters).

More current brands of terminals canemulate the Tektronix 4010 than any othertype. We chose 4010 emulation to givethe most DSRC users the capability toview plots directly on their terminals beforesending them on to the laser printer. Thisleft us with a choice of four brands oflaser printers with Tektronix emulation(this choice has since widened considerablyand continues to do so). We chose theQMS Lasergrafix 1200, using an originalequipment manufacturer's version of theXerox 2700 "engine" (printer mechanism)with a controller (actually a rasterizingcomputer) by QMS. The controller addednot only both native -mode and Tektronix4010/4014 graphics, but also a variety ofbuilt-in fonts. In addition, the QMS con-troller rasterizes in the printer (see sidebar,page 28), removing that computing loadfrom the host computer (CMS first, andlater others). 4010 graphics, at 1024 dotsby 768 dots, have "jaggeties" while 4014and QMS graphics (which cost more toproduce), do not. Here, there's no freelunch, but at least there's more than onemenu selection.

While we were choosing our printer,we were also designing the software tosend graphics to it from CMS (see Fig. 1).We had to do two things to plot on aremote graphics printer:

(1) Direct the graphics into disk files.(2) "Spool" those files to the printer.

Barton: Spooled -graphics laser printing from CMS 25

ASCII and EBCDIC

In computing's earlier days,industry members developedASCII (American Standard Codefor Information Interchange), astandard way of representingcharacters in computers. Sometime later, IBM developed alarger code called EBCDIC(Extended Binary -Coded -Decimal Interchange Code), nowused by IBM in its medium -andlarger -sized computers (thesmallest ones use ASCII).EBCDIC is naturally used by the"plug -compatible" manufacturersof IBM -like medium- and larger -sized computers. It is designed insuch a way that even though thecharacters are represented bynumbers in both sets, there is noformula that can be used tocompute one from the other.Translation must be done by atable that relates one set,character by character, to theother. Our corporate mainframe'sCMS operating system usesEBCDIC, and our QMSLasergrafix printer uses ASCII.

Directing plots into files can be accom-plished in one of two ways:

(1) By using the CMS FILDEF (file defini-tion) command' to channel a pro-gram's output into a file, rather thanto an output device such as a terminal.

(2) By using the RCA CISS-suppliedsystem enhancement ROUTERM(route terminal) command' to foolthe operating system into putting theoutput it thought was going to one'sterminal into an output file instead.

Where possible, we used the formermethod, since it is a more straight -forwardapproach to the problem, and reallyinvolves specifying a procedure for theuser to follow, rather than supplyingcustom software. There were cases, how-ever, in which we were forced to resort tothe latter. All involved using Tektronixemulation with the printer. Some graphicspackages used IBM assembly language (forspeed) to feed their output directly to theterminal, and could not be redirected witha FILEDEF. So far, three families ofcurrently -installed packages can produce

( FORMSCOPYFONT

(EBCDIC)

Commands.Address

(EBCDIC)

AddressFile

Commands,Address

(EBCDIC)

Fig. 1. Software architecture.

GraphicsPackages

Commands,Address(ASCII)

Commands,Address(ASCII)

output in the native language of the QMSLasergrafix. We FILDEF with these.

Spooling files to an output device requiresthree steps:

(1)

(2)

(3)

Starting spooling. For text printers thisis done by using the RCA CISS-supplied system enhancement ROUTEcommand,' which identifies the des-tination and provides options for mul-tiple copies, special forms, etc. Wemodeled our PLOTSTO commandon ROUTE.Spooling one or more files. For textprinters the CMS PRINT commandis used. We modeled our PLOTS com-mand on PRINT. (We used the namePLOTS instead of PLOT because therewas a program named PLOT alreadyin use by DSRC users).Stopping spooling. This step actually

PlotFile

(PLOTS (EBCDIC))

( PLOTS (ASCII) )

VirtualPunch

Remote.lob Entry

Scstem

QLaser

Printer

causes the files to be sent to the device.Instead of requiring the user to issuea SPOOL4 command to the virtualpunch, we provided the ENDPLOTScommand.

Our approach was to try to extend theuser's environment in a logical and con-sistent way to make it appear as if graphicshad been a part of the spooling environ-ment from the beginning.

PLOTSTO and ENDPLOTS also havea few "housekeeping" chores. When wespool the virtual punch to a remotedestination, punch some files and closespooling, the remote operator sees thepunch job identified on the remote stationconsole, and with commands, directs thejob to an appropriate device, such as anavailable card punch, which is not exactlywhat we had in mind. Normally, theFORMS option of the ROUTE command


Graphics packages

If there were no graphics"packages," graphics use wouldbe limited to people actuallytrained in making computersdraw pictures (a small group,indeed), which is the equivalentof limiting the use of cars toautomotive design engineers.Packages are simply programs,collections of programs, orcollections of pieces ofprograms. Graphics packagesavailable to the corporate CMSuser include:

(1) Tellagraf: a program used tograph data (scientific orbusiness) that uses English -likecommands such as, "make thex-axis 6 inches long." Tellagrafcan produce almost any type of2 -dimensional graph of data thatone might want, includingmultiple graphs on a page,allowing the user to takeadvantage of laser printerresolution for comparison.(2) Cuechart: a program that letsthe user select the graph he orshe wants from a book of cannedgraphs, and which produces thefile of commands that is used byTellagraf to produce that graph.Cuechart is limited to certainsubsets of graphs that can beproduced by Tellagraf (althoughone familiar with Tellagraf can"doctor" a graph produced byCuechart). A user familiar withTellagraf is able to "can" othercustom graph types forCuechhart users.(3) Disspla: a series of FORTRAN

subroutines to be called fromwithin the user's program (writtenin FORTRAN, PL/I, COBOL, etc.)to produce graphs. Disspla canproduce anything Tellagraf canproduce, and can also be usedfor 3 -dimensional graphs.Tellagraf, Cuechart, and Dissplaare products of ISSCO, Inc.(4) SAS/Graph: a data -graphingpackage built in to the statisticalpackage SAS. Between Tellagrafand Disspla in ease of use, ittakes advantage of SAS' ability tomanipulate the data statistically,and it graphs the results. It canproduce 3 -dimensional graphs,but not multiple graphs per page.SAS/Graph is a product of SASInstitute, Inc.(5) Focus: part of the Focusdatabase system, betweenTellagraph and SAS/Graph inease of use, with less richnessthan Cuechart, it benefits from itsinclusion within the Focusdatabase package (the user doesnot have to stop accessing datato graph it, and vice versa).Focus is a product of InformationBuilders, Inc.(6) DI -3000: a series ofFORTRAN subroutines to becalled from within the user'sprogram to produce graphicalobjects (lines, polygons, etc.).Unlike Disspla, DI -3000 permitsthe user to perform geometricmodelling and graphic input.(7) Grafmaker: a series ofFORTRAN subroutines to becalled from within the user's

program, which in turn call DI -3000 subroutines to producegraphs of data (including axes,tick marks, etc.). This packagegives DI -3000 graphingcapabilities similar to those ofDisspla.

(8) Contouring: a series ofFORTRAN subroutines to becalled from within the user'sprogram, which in turn call DI -3000 subroutines to producegraphs of 3 -dimensional data.This package gives DI -3000 full3-D data graphing capabilities.

(9) DI-Textpro: a series ofFORTRAN subroutines to becalled from within the user'sprogram, which in turn call DI -3000 subroutines to provide textcapabilities, including a variety offonts with controls over theirattributes. DI -3000, Grafmaker,Contouring, and DI-Textpro areproducts of Precision Visuals,Inc. These packages arecurrently installed on CMS on atrial basis.

Other programs on CMS includethe ability to graph output:SUPRA and SUPREME, used forsemiconductor processsimulation, are only twoexamples of programs eitheracquired from outside sources orwritten by RCA users.

Tellagraf (and thus Cuechart),Disspla, SAS/Graph, DI -3000,Grafmaker, Contour, DI-Textpro,SUPRA, and SUPREME workwith our laser printing software.

is our means of informing the operatorthat we want special paper in a printer,for example. Since this option can be anyfour characters, we chose "LASR" as theform for the QMS Lasergrafix. We antici-pate using "SLID" for a film recorder and"PENS" for a plotter.

Once we had designed a softwarearchitecture capable of providing thenecessities, we considered ways of takingadvantage of the special features of theQMS Lasergrafix printer.' This meantmaking PLOTSTO and ENDPLOTS more

sophisticated (if PLOTSTO turned on aspecial feature, ENDPLOTS should turnit off to avoid surprising the next user).

The ROUTE command allows a multi-ple copy option (the operating systemactually sends the job to the remote devicemultiple times, which means that the copiesare collated). We designed PLOTSTO toaccept an optional argument pair, COPYn, where n is the number of copies, andpass it on (to ROUTE).

The ROUTE command also permitsthe use of an address distribution (the user

may create files containing different addressblocks for distribution of output, assigneach a four -character distribution code,and select the one desired with the code).We designed PLOTSTO to accept anoptional argument pair, DIST xxxx, wherexxxx is the distribution code, and pass iton (to ROUTE).

The QMS Lasergrafix printer has twomodes, portrait (the long dimension ofthe paper is vertical, as in a portraitpainting), and landscape (the long dimen-sion of the paper is horizonal, as in a


Vectors, rasterization, and "jaggeties"

Rasterization

In the beginning, computergraphics was all done withvectors, or straight lines drawnbetween two points. The earliestgraphics devices were CRTdisplays and pen plotters onwhich images were drawn as aseries of straight lines, either bymoving an electron beam around,turning it on for lines and offbetween them, or by moving apen around, letting it draw onpaper for lines and lifting itbetween them.

Note that a point can berepresented by a line so shortthat we cannot tell it has anylength at all. We can drawanything with straight lines if wecan draw the lines short enough.We can draw almost anything ifwe can draw the lines almostshort enough. If we give"addresses" for the horizontal (x)and vertical (y) distances of apoint from the lower left-handcorner of our plotting surface(CRT screen or paper), the moreaddresses we have, the closertogether we can put the points,and the shorter the lines we canmake. Hence, more "addressable"points equals finer detail.

On high -quality vector CRTdisplays, the number of address-able points is high enough tomake points or lines drawn onthe screen overlap (because of"dot bloom"). There are actuallytwo types of vector CRT displays:

Jaggeties

:

-,-,

refresh, with short -persistencephosphors on which the imagefades unless it is redrawn manytimes a second, and storage, onwhich the vector is drawn once,and the phosphors are keptenergized by a "flood" gun whichfloods the whole screen withelectrons of lower energy thanthe drawing beam (early TektronixCRT display terminals such asthe 4010 and 4014 used storagetube technology developed byTektronix to retain waveforms onoscilloscopes).

On high -quality pen plottersthe points are closer togetherthan the width of the finest pen inorder to produce the desiredeffect.

The Tektronix 4010 used 1024points horizontally and 768 pointsvertically (it actually addressed1024 points vertically, but sincethe screen was not square, onecould only see or print 768). Fora 7.5X10 inch plotting area on an8.5X11 inch sheet of paper, thiscorresponds to approximately100 points per inch. The later4014 used 4096 points horizontallyby 3072 points vertically for anequivalent addressablity onpaper of approximately 400points per inch. The effect of lowresolution on vector devices isthe apparent use of a "broadbrush" to draw the image.

Later (and generally lessexpensive) devices such as

Broad Brush

raster CRT displays, dot-matriximpact, thermal and ink -jet aswell as laser (less expensiverelative to its resolutioncapabilities) used "rasterscanning" (just as a televisiondoes. Early raster -scan terminalswere little more than TV sets withkeyboards and some electronicsadded, which explains theirmuch lower cost.

In a raster scan device, the plotarea is scanned as a series oflines, usually from left to right,and line by line (usually from topto bottom). As each point on theplotting surface is scanned, if it ispart of the image, it is "turnedon", by brightening or darkeningthe electron beam (raster -scanCRT), firing a solenoid that slamsan ink -impregnated ribbon intothe paper (impact dot-matrix),spraying droplets of ink at thepaper (ink -jet), or brightening ordarkening a laser beam striking a"chargeable" drum. Dependingon the particular device, thepoints (and raster -scan lines)might or might not overlap. Thepoints are usually referred to as"pixels," or picture elements. Ifthe points do not overlap, lines,especially those that are notstrictly horizontal or vertical, havea stair -step effect known as the"jaggeties". Jaggeties are aproduct of low-cost, low -qualitycomputer graphics, and brandthe image as being "mechanical"


in appearance and generallyinferior to that which could beproduced by other means (manualdrafting, for example). It is myexperience that automation ismuch easier to sell to a client,particularly in view of its usualcost, if it does not produce aresult that is judged inferior tothat previously produced byother (usually manual) means.One would prefer then thatpoints overlap to eliminatejaggeties, and that they be smallenough to eliminate too much ofa "broad brush" effect.

To express an image as a setof vectors, one may simply givethe endpoints, coupled with"pen-up/pen-down" or "beam-off/beam-on" commands. Toexpress it as a raster image, onemust say whether each of the

dots is on or off (one might use adata compression code to saythe next n dots are on or off).Images without large solid filled -in areas are generally expressedmore compactly as vectors. Theprocess of translating the vectorcommands for an image into araster version is known as"rasterization." If one wishes touse a representation such asTektronix vectors for an image tobe displayed on a raster devicesuch as a raster -scan CRT orlaser printer, one must rasterize itfirst. For a complex image withhigh resolution (say the 300 dotsper inch for a 7.5X10 inch plotarea, which equals 6.75 milliondots to be on or off), thecomputations for rasterizationare not trivial! Depending uponthe device, rasterization may be

done in the device itself in"hardware" (actually by a built-incomputer designed for the purpose),or on the computer producingthe graphics in software. In theformer case, one pays the cost ofrasterization at the time ofpurchase, and in the latter case,one pays it every time a graph isproduced. Since the vectorrepresentation is generally morecompact, and the connectionbetween the device and thecomputer has a speed limit,graphs may be delivered to thedevice faster in vector form thanin raster form, and may, ingeneral, produce graphs morequickly. In sum, we are usuallymuch better off paying the "upfront" price of device orhardware rasterization.

landscape painting), and twelve characterfont (styles), nine portrait and three land-scape. We designed the PLOTSTO com-mand to accept the argument pair FONTxn, where x is either P or L (portrait orlandscape) and n is the number of thefont.

The system supplies leading and trailingbanner pages, but these are in IBM'sEBCDIC character set, not accessible toour software for translation to ASCII, andnot translated by the RJE station, sincewe needed to supress that translation andreplace it with our own for proper graphics.Fortunately, the result is relatively benign,because each of these has no ASCII linefeed or carriage return characters, and eachcomes out of the QMS Lasergrafix as asingle line of typed -over jibberish. Tocorrect for this, we made PLOTSTO createa banner page with the user's addressblock by arressing the RCA CISS-suppliedADDRESS command from withinPLOTSTO, which sends the address block(correctly translated) and saves it in a diskfile for ENDPLOTS to send as a trailingbanner page. Before the banner page, wemade PLOTSTO send commands selectinglandscape mode, and a default characterfont, character spacing and line spacing,to make all banner pages look the sameand to restore the printer to a knownstate. These are also stored in the filepassed to ENDPLOTS for the trailing

banner page. After the banner page, if theuser has requested a mode/font combina-tion, PLOTSTO sends commands request-ing the mode, the font, and appropriatecharacter and line spacing for the latter(not all fonts are the same size; see sidebar,page 30).

We wrote PLOTSTO, PLOTS andENDPLOTS in IBM's Pascal/VS,''' build-ing on Kernighan and Plauger's Software

Tools in Pascal,' after writing the necessary"primitives," or low-level modules to portthose tools to CMS.

Now that we have provided a practicalmeans of spooling high -quality hard copygraphics from the CMS environment, wewill extend the capability to DSRC usersof other computing environments, and toother shared graphics output devices. RCAAstro-Electronics Division now has a QMS

Prior to receiving his MS in OperationsResearch (with a minor in systems design)from Rutgers University, Don Bartonworked for nine years a a field engineeron U.S. Air Force flight simulators, mostlyin Japan. After receiving his degree, heworked as a hardware design engineer indata communications and computer peri-pherals, and later he worked for RCACamden on a distributed computing sys-tem. He joined RCA Laboratories in 1980,where he developed a testing informationsystem for VideoDisc research using adatabase and graphics on an IBM main-frame. He joined the Information SystemsPlanning and Computer Services Group in1983, and now supports scientific comput-ing for RCA Laboratories. His primaryinterests are computer graphics and theuse of the computer as a tool for theresearcher.Contact him at:RCA LaboratoriesPrinceton, N.J.Tacnet: 226-2995


Landscape font, portrait font, and sample QMS bullet list

THIS IS FONT P328.(PORTRAIT COMPLEX ROMAN BOLD 15 POINT)ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789 !(q #$%&`()_+-={}I(]\:";'<>?,./This is one of the nine portrait fonts available.

THIS IS FONT L328.(LANDSCAPE COMPLEX ROMAN BOLD 15 POINT)ABCDEFGHIJKLMNOPQRSTUVWXYZabedefghijklmnopqrstuvwxyz0123456789 !@#$%&*() -f---={}10\:";'<>?,./

This is one of the three landscape fonts available.

Use this format to produce a bullet chart

The 1 in column 1 of the first line says to start a new page.

' The 0 in column 1 of the second line says to skip a line(use a blank if you don't want to).

You may simply copy this file (QMS BULLET) off Lablib andreplace this text with yours.

Your file may contain many such charts.

When you are done, type:PLOTSTO 21 (FONT L328PLOTSCC filename filetypeENDPLOTS

where filename and filetype are for your chart file(s)(you may use PLOTSCC between PLOTSTO and ENDPLOTS more than oncefor multiple chart files)

These characters are proportionally spaced, so columnar text will be aproblem.

Example of QMS bullet list.


Satellite contour map

0

-3.5

RCA Continentc

-1.0 Db

0.0 Db

ViewBor Lat= 99.99 +Bor Lon= 99.99 +Sat Lon= 99.99W

-_ - - - - -

--------

. Ar l' AM °' '411141ft 'allignallettla% -.44174Wilk"AfIb.

inallir IF 1r7tlik A /

-2.5 -1.5 -0.5 0. 0.5

X Deg Az

We obtained graphic documentsfrom the Federal CommunicationsCommission (FCC) and from theRCA Astro-Electronics Divisionshowing contours of constantsatellite -to -ground received -antenna power (EIRP), or thepower density from the groundstation needed to saturate thesatellite amplifiers (SFD). Weplaced the contour information indata files by digitizing a largenumber of points on the contours,using a Tektronix 4954 digitizingtablet with a program written byK. W. Coffee and R. E. Enstrom.

In some cases, we recognizedthe maps as one of the standard

1.5 2.5 3.5

projections. In others, we couldnot recognize the projections. Aprogram written by A. Guidaconverted all of the projectionsto mercator.

We produced the maps (withtheir associated contours andannotation) using a Disspla-based program written by L.Kihn, and directed them to theQMS laser printer.

The example plot shows:

1 A title containing the type ofsatellite, its position relative tothe earth's surface, and theview of interest (not shown).

ly

-7.0 Db

-6.0 Db

-5.0 Db

-4.0 Db

- 3.0 Db

-2.0 Db

2. A projected map of the U.S.overlaid by the contours, withlabels.

3. Information about the "boresight"or equivalent aim point of the centerof the satellite's antenna beamaxis on the earth's surface.

The QMS laser printer furnishedplots with clear, readable textand well -delineated contours foranalysis on standard -size plainpaper for easy inclusion inreports.

-Les Kihn,RCA Labs

Lasergrafix 1200 printer using our softwarefor RJE printing.

AcknowledgmentsI wish to thank John Lee of RCALaboratories for his help with adaptingour existing Remote Job Entry station toaccommodate a laser printer; Sue Johnson

of CISS, Cherry Hill for her help withmaking a "new" IBM product, Pascal/VS,communicate with the CMS operatingsystem; Frank Zonis and Linda Rusnakfor modifying ROUTERM to enable usto direct Tektronix -compatible plots intoa disk file; Ray Davis, CISS for help withISSCO's products and the printer; Dr.

James Matey, RCA Laboratories for estab-lishing the SAS/Graph approach to plottingon the printer; Dr. Ron Sverdlove andJohn Fields, both of RCA Laboratories,for configuring the ISSCO QMS driver;Dr. Russ Barton for the "windowpane"example plots in his sidebar; and Les Kihnfor his plot of satellite coverage.


Windowpane plots

TUBE 4

TEKTRONIX 4014 EXMIPLE

TUBE 3 YO 3-__TITBE 8 YOR

Ideally, the red, blue, and greenbeams of a color picture tubeshould coincide (converge)everywhere on the screen. Thisis so, for example, the image of apurple hat doesn't exhibit red orblue fringes around the edges.Practical requirements of designand manufacturing do result insmall errors, however. By studyingthe nature of these errors, it ispossible to identify their causesand make corrections thatimprove picture quality.

The nature of misconvergenceerrors for color picture tubes areeasy to view using windowpane

plots. Each plot summarizes 100measurements of misconvergence;4 measures at each point of 25points on the screen. Red -beamposition relative to green, andblue -beam position relative togreen are measured both in thehorizontal and vertical directions.These errors are typicallyfractions of a millimeter, and arenot easily detected in a plot thatis to scale. Exaggerating theerrors by about 50 times makesthe pattern of misconvergencevisible. The relative positions ateach point on the screen areconnected by smooth curves; a

MTUBE 18 1 8

el

thick line for blue and a thin linefor red.

Because the information isreduced to shapes, it is possibleto view a group of many plotstogether to look for commonfeatures. The usefulness ofmultiple windowpane plotsdepends on the availability offast, high -resolution plottinghardware. The QMS laser printerprovides the speed and resolutionwe need, as shown in this exampleof the QMS "QUIC mode."

-Russ Barton,RCA Labs

References1. CMS Command and Macro Reference, IBM Corp.,

Endicott, N.Y. (1983).

2. 4014 and 4014-1 Computer Display Terminal User'sManual, Tektronix, Inc., Beaverton, Or. (1982).

3. Computer Services User Guide, Vol. 2, RCA Corp.,Cherry Hill, N.J. (1984).

4. CP Command Reference for General Users, IBMCorp., Endicott, N.Y. (1983).

5. QUIC Programming Manual, Quality Micro Systems,Mobile, Ala. (1984).

6. Pascal/VS Language Reference Manual, IBM Corp.,San Jose, Ca. (1981).

7. Pascal/VS Programmer's Guide, IBM Corp.,San Jose, Ca. (1981).

8. Kernighan and Plauger, Software Tools in Pascal,Addison Wesley, Reading, Mass. (1981).


An information networkwith over 7500 portsand no passwords

-tde°;e

.ntette%631fesao

so

wary

gronpof engineer:OS

as,.0).ye

'stn dents

co o\ ocl at;t,foo:

X:teoU.

ecentC,ci

lechnicalArticles

0.

1-:ts.Mnti\

10f oar Ccoint,r

aatO,cols at\

eVesstosiiastiosgot

Mary

..psCAtctioniscanntiocations

satelliitte:gertawoo:_se4-:lOrcx,07::

16-4/1/ci,

16

1 processcontrols

for 1.51manniactnrio

41--rt

14.0

.11ka.s ssTseocner%te

of cami date

s ,so o

Wherecan

1 getto

on

* Organicmaterials

and

*4:::::e3t-o:-::::::Tts9gocecrait

bair_ttrivieovs

4----Al011(Die

Pli

i,4 t'n OtOs

* ElectroniclioaqlOO

II'(banks, HarrJ.

$.1'AV

Aa AC

li \i, /,\ YEA

f),,,\,J`' ..4),\:/\

Vre

'1. AVP\kcP°9 )(V foc.,.

d,

At)W1

k\,Co

110

.

0.",

7 Joi kuiti).t,\

r\e-

The RCA Era nee r in-ormation networkcpfNates informally, but effectively. Memos,phone calls and wo'd-of-mouth are itsprctocols.

RCA Engineer art c es are a reusablerescurce. Reprints are valuable tools forrecruitment, marketing, seminars, andmeetings wi-h colleagues.

Reprints of single articles take about twoweeks to male. If you want to put togethera booklet of reprints from several issues, it

II

will take from four to eight wee<s. 'Ne can itfact, give you pretty much wtat you wantand the prices are lower than the originalcost of printing the articles.

For more information and a corno.lete prceschedule, call or write to:

Louise CarrRCA EngineerM.S. RR 13Princeton LabsTacnet: 226-3092

Pae Engineer 3C-5 Sect Oct 1c85 33

R.A. Monat

Accurate layouts for automatic componentinsertion

A unique program has been developed to show the spacingrequirements for automatic insertion of axial and radial leadedcomponents into printed circuit boards.

Many of today's printed circuit boards(PCBs) are dominated by digital integratedcircuits. In contrast, the boards in atelevision receiver are still very muchanalog with many discrete components; amaster board (see Fig. 1) may containmore than 400 components, of which aboutsix are ICs. Production runs for typicaldigital boards may be much less than100; runs for television boards may beseveral hundred thousand. The large vol-umes and cost pressures dictate that theboards be designed for automaticcomponent insertion (ACI) when feasible.

The ACI design rules are a function ofthe insertion machine, the component beinginserted, and the neighboring components.Conforming to these rules while designingthe PCB is time-consuming and frustrating.Now, a program has been implementedon a computer aided design (CAD) systemto graphically show the ACI constraintsas exclusion areas.

Abstract: Layout of a printed circuitboard for automatic component insertioncan be a tedious and time-consumingtask For each insertion machine andcomponenI up to 13 different clearancesmust be checked for each component.This paper describes a program thatgraphically displays the exclusion areaassociated with neighboring components.The major benefit achieved is a morecompact and accurate layout designed ina shorter time.

©1985 RCA Corporation.Final manuscript received August 6, 1985Reprint RE -30-5-4

Fig.1. The CTC-131 chassis was the first board designed using the "Footprint"program.

BackgroundPCB designers at Consumer Electronicscreate a layout based upon electrical,mechanical, and manufacturing constraints.These constraints often conflict with oneanother. The designer acts as a referee increating a layout that is acceptable to allparties. Electrical and mechanical con-straints are communicated to the designeron schematics, on drawings, and orally.Manufacturing constraints, for the mostpart, are published in a document titled

"Manufacturing Design Guidelines"(MDG).

One portion of the MDG describes ACIdesign rules for axial (e.g., resistor) andradial (e.g., transistor) leaded components.The ACI design rules attempt to reduce athree-dimensional dynamic problem to atwo-dimensional static representation.These rules fill three 2 -in. binders.

Most of the boards used in a televisionreceiver are designed interactively usingan Applicon AGS/870-I 2D CAD system.

34RCA Engineer 30-5 Sept. / Oct. 1985

Tools to further automate the layoutprocess have therefore been developed forthe Applicon system.

Using the ACI design guidelinesPCB designers have estimated that to checkeach pair of components can take up tothree minutes. Of course, for commonlyused parts the time required is much less.The complexity of the guidelines and thetime required to follow them are bestillustrated by the following typical example.

1. First, determine the insertion machinesto be used to populate the board. Vari-ous manufacturing plants and produc-tion lines within a plant have differentequipment complements. For this exam-ple, we will use a USM 5mm axialinserter. Data for each machine is in adifferent section of the MDG. USM5mm axial data is in section 03-02.

2. Determine the automatic insertionmethod -axial or radial -based uponthe part geometry.

3. Using the drawing part number on theschematic or Drawing List (DL), deter-mine the part's classification. FromFig. 2, we see that a 1/4 -watt resistoron drawing 99206 is assigned a classifi-cation of 1B2. The classification is basedon the component type and geometry.Data for this component is in subsection03-02-13, page 1B2 (see Fig. 3).

4. Assume that a 1/8 -watt resistor ondrawing 990697 is to be the neighboringcomponent in the layout. It has alreadybeen placed in the layout and duringmanufacturing it is to be inserted beforethe 1/4 -watt resistor. Again, use Fig. 2to determine a classification of 1 A 1.

5. Figure 4 specifies the dimensions to beused from Fig. 3 based upon the relativecomponent orientation. If the com-ponents are inserted parallel to eachother, then dimensions P, A2, B2, Cl,C3, D1, and D3 must be considered;otherwise A1, A3, B1, B3, C2, andD2 are used. Note that all dimensionsare relative to the centers of the holesfor the component leads. The actualdimensions to be used are in the firstline of Fig. 3, since we are using acomponent classified as 1 A 1.

6. Apply the appropriate dimensions tothe part to be placed in the layout atits proposed location. Figure 5 showsthe dimensions for the example chosen.Determine if any rules have been vio-lated. If not, consider the relationshipbetween the component to be placed

RCAO MANUFACTURING DESIGN GUIDELINES= Manufacturing Technology

Indianapolis, Indiana

03-01-03

tt. Page 7 of 3

SUBJECT AUTOMATICALLY INSERTABLE RESISTORS Date 5/13/85

1. SCOPE - The following list of resistors are automatically insertable andshould be specified wherever possible.

2. DEVIATIONS - Any deviation from these guidelines have to be approved by tieManufacturing Technology section.

1/8 Watt

Classification Drawing i Part

Minimum InsertionAxial SpanHole

Size

1A1

1A2

1A4

990691

2815583

1/5 Watt

ALL

ALL

ALL

PARTS

PARTS

PARTS

H

S2874082

1/4 Watt

2821551 ALL PARTS

181 993218 ALL PARTS993272 K

99328428169252874072

182 9920699291

993152727832735729

* * 2819320

* 183 2819434 ALL PARTS H

184 990401 ALL PARTS K

990413 ' K

993113 .K

Indicates Change, Addition or Deletion

Axial Radial

USM

.250 .200

.250 .200

.250 .200

.400 .200

.400 .200

.400 .200

.400 .200

.400 .200

.200

.350 .200

.350 .200

.350 .200

.350 .200

.350 .200

.375 .200

.250 .200

.375 .200

.375 .200

.375 .200

Supersedes 6:18/S4

Fig. 2. A major use of this table is to determine the classification given thecomponent's drawing/part number.

and other neighboring components. Ifa violation has occurred, move the pro-posed placement and perform this stepagain. Note that any move may createviolations for components that werepreviously checked.

This example omitted mention of manyproblems routinely encountered during alayout. For example, the placement orderduring the layout is different than theinsertion order. Three ACI passes aretypically used: (1) axial parts parallel tothe x-axis, (2) axial parts parallel to they-axis, and (3) radial parts. Within eachpass, the components are inserted in asnake -like pattern.

Designers initially attempted to ease theproblem of designing to the ACI guidelinesby building templates. Typically, one ormore cells were used to convey the data.The cells contained data for the mostfrequently encountered situations, but theycould not account for the dynamic insertionorder. Some of the cells were difficult toread. Guideline cells were constantly beingadded to and deleted from the layoutdrawings.

Automating the ACI guidelinesA decision was made to implement aprototype program to provide the designerswith some assistance in adhering to the

Monat: Layout for automatic component insertion 35

S

- SUBJECT

IRG/1 MANUFACTURING DESIGN GUIDELINES 03-02-13

Manufacturing TechnologyIndianapolis, Indiana

AXIAL INSERTIONPage /8V -

COMPONENT AND TOOLING CLEARANCES TO REsi mast:we nu g r3,1

FOR USM 5MM EQUIPMENT

BBIet RESIVON 6E250 &&&&&

81.121114.51In Om.M hM2.

9920699291993152727832735729

All pert$

COMPONENT ON BOARD

P Al A2 A3 81 82 B3 C1 C2 C3 D2 D3

U MO Ile 128 158

1.2 n M. 1211 128 IM183 03 199 III 128 201

161 MO 117 118 In MOn2 07 III 117 127 175

n2 95 10. IV nO 184

164 07 III 117 127 175

101 MI In 117 127 210

MI 03 151 III 129 201

M3 133 ne M 137 201

104 179 IM 117 n7 207

10 137 09 117 120 201

MO 172 151 118 125 200

107 172 In III 127 nOMS MS 70. U7 127 224

MS 117 122 III 121 191

MO 127 178 In Ill 195

MU 137 160 HV 126 205

MU 121 00 117 127 Ill

In 06 204 116 126 228

02 172 08 114 124 200 12

03 87

-2111------71-110-173-nr-i-59 es es

201301303 n M. 123 131 164 II IV

742 103 112 120 120 171 n 83

303 10 114 19 1211 M. 03 97

702 110 24 MI 65

30 1MO

204

83

5.2 1.1503 127In 139586 123561 157582 174583 187-txr-171682 119

4d1 1 /

1086

7E1 133Oil 1487A2 153701 158

129292671964819282SIS I

9162128097O f

IWV97

8082

102115

85IIIS1031021011011021021011079119100II10810280

112117150130121117128119158158161151157157182175131163144112143

53

00 83 101 80 95

14 100 113 54 100

205 111 151 105 115

IV 05 121 17 205

II 106 292 88 MO8. 100 113 54 10098 104 125 18 104

101 110 159 101 110

101 115 151 105 215

101 110 151 101 110

101 110 152 101 210

105 118 151 105 118

105 110 150 105 III

105 116 150 105 118

RI 101 178 55 108

17 105 131 17 105

17 105 145 07 103

97 105 155 57 105

101 iao 179 101 '710105 115 178 105 II!

104 117 ISO 104 113

87 5 8110 67 *05

7I 1309 e0

7

xi -Ili -Hi -1: 101117 84 104 II 004

123 40 100

1

146118 1311 107

107 III :813 107 118

124 17 231 106 117 202 104 117

5,--nr-n 9 100 isr-9T7a5-pf--rr-al-211 115 126 23. 87 109 200 67 10! 191 67 107

T -7 1T-112 102 171 111 in Ill 5I-r09"177117 127 212 VI 102 141 II 108 138 II 108

132 117 127 104 86 08 1.6 89 106 141 88 108

172 117 127 209 III 101 184 86 108 I51 .4 105

122 123 131 184 1111 83 I3! 88 *04 13; 10 304

117 117 127 20 84 105 183 OS 108 171 06 308

275 117 127 251 101 113 219 81 106 183 86 101

216 117 127 240 00 110 207 68 106 185 98 06

121 119 111--11r---15---11 125 40 *00 01 87 105

127 111 120 181 80 17 III 17 405 136 87 205

117 11! 120 101 83 84 130 94 wi: ' i:11/T3 131 191 87 BA MI nri:158 122 131 206 80 85 151 101 110 131 101 110

189 122 131 221 84 103 176 101 110 166 101 110

187 122 131 217 63 105 185 101 110 171 101 110

204 122 121 226 87 1011 192 101 110 176 101 110

IndicaleS Change, Addition or Decion SupersedesfiCii 3 Li,32

Fig. 3. This chart specifies the clearances from parts of classification 182 to allother classifications which can be inserted earlier.

ACI guidelines. As mentioned above, PCBlayout was being performed interactivelyusing the Applicon CAD system. Thisenvironment dictated user interface andperformance requirements. The user com-munity named the main and supplementalprograms "footprint" programs. The devel-opers consider the version that has beenused in production for more than twoyears a prototype.

Footprint vs. exclusion area

The traditional method of consideringrestrictions similar to the ACI guidelinesis by looking at the footprint-how muchspace is taken up by the component to beinserted and its associated insertion machin-ery. From the example above, one cannote that each time the subject componentis moved, its footprint changes. This impliesthat the program would be executed eachtime the component is moved. Further-more, it is difficult to assess the contributionof each neighboring component. This isan important consideration when a viola-

tion exists and the part must be moved.Which are the interfering parts? In whatdirection should the parts be moved? Byhow much?

Rather than show the footprint, we choseto show the exclusion area. The exclusionarea shows how much space is taken upby components already placed. If a pre-viously placed component is to be insertedafter the subject component, the dimensionsassociated with the previously placed com-ponent determine the exclusion area.

There are several advantages to usingthe exclusion area instead of the footprint.A major benefit of drawing the exclusionarea for each previously placed componentis that the designer can see the effect ofthe neighboring components and then candecide how to move the subject componentto remove violations or achieve a tighterlayout.

Alternatively, the designer may chooseto move a previously placed component.Once the exclusion area is calculated fora subject component, it remains constantno matter where the part is placed as long

as it is not rotated. The program to calcu-late the exclusion area is run once persubject component.

User interface

The design objectives for the user interfacewere to make it easy to use and to have itfit in with the Applicon supplied user inter-face. The steps involved are: (1) fit theneighboring components in the current win-dow; (2) select the subject component;and (3) make a tablet stroke which inturn executes a user command that invokesthe footprint program. Exclusion area datais then displayed on the graphics terminal.To eliminate the exclusion areas, justrepaint the screen.

From the display of the exclusion areas,the designer decides if the current place-ment is correct. In general, the placementis acceptable if the line segment joiningthe subject component's holes is on oroutside the composite exclusion area (seeFig. 6). In some instances it is acceptableto have either the holes or the body traversethe exclusion area; these two cases areindicated with arrows on the graphicdisplay.

To provide the most help to the designer,the footprint must keep track of the inser-tion order and not the placement order.When a new component is placed, it is

appended to the appropriate insertion orderlist. The designer can manipulate the com-ponent's position on the list.

Program design

The basic premise underlying the programdesign was that the execution time mustbe fast enough to make it an interactivetool. A 3 -second response time on asingle -user system was the design goal.Actual response times for this configurationare between 3 and 6 seconds.

The Applicon CAD system is imple-mented on a PDP-11/34 computer runningthe RSX-11 operating system. This com-puter has a 16 -bit address, which meansthat programs must be less than 32k inlength. The Applicon program is veryheavily overlayed, leaving very little spacefor user -written programs and data. Forthis reason we decided to implement theprogram in a separate task, which allowedus to use the 32k as we saw fit.

When the Applicon system loads a fileto be worked on interactively, it copiesthe data from an Applicon file structureinto an RSX-11 file. Fortunately, RSX-11 allows a second task to open a file for


read access but not for write access. Ourtask reads the open drawing file to deter-mine the selected component and thosecurrently in the display window. A singleread command reads two disk blocks (1block = 512 bytes).

The amount of ACI guideline datademanded fast data retrieval. The file isorganized in sections of six disk blocks;each read retrieves six blocks. The firstsection is an index for the remaining sec-tions. For each classification and insertionmethod, the index contains the address ofthe data within the file. For many cases,only two disk accesses are necessary.

Data to draw the exclusion areas isplaced in an RSX-11 dynamic storagearea. After all of the exclusion areas arecalculated, the user command running inthe Applicon environment is resumed, andit then draws the exclusion areas directlyto the graphic display, bypassing thedatabase.

Additional performance is gained byextracting information contained in theApplicon cells pertaining to ACI whenthe program is first started. Extracted dataare the hole locations, insertion method(axial, radial, etc.), and classification.Component body sizes are extracted froman auxiliary data file. This data is writtento a small, well -organized RSX file.

Using the footprint programBefore reading the following example, itwill be helpful to examine the display of asingle component and its exclusion areashown in Fig. 6.

The following procedure illustrates howa designer might use the footprint programin an original design:

1. Components in the center of Fig. 7ahave been placed correctly using the foot-print program. The designer desires toplace capacitor C982 above C981.

2. C982 is selected. Footprint is then calledon to draw the exclusion areas as shownin Fig. 7b.

3. Using normal Applicon move com-mands, the designer moves C982 intoits final position as in Fig. 7c.

4. A screen repaint is done showing thecomponents in their final position (Fig.7d).

Now suppose that an engineer decidesC981 should be replaced with anothercapacitor with a larger body. The designerdecides whether components must bemoved by performing the following steps:

RCR MANUFACTURING DESIGN GUIDELINESManufacturing TechnologyIndianapolis. Indiana

SUBJECT GENERAL INSTRUCTIONS

03-02-02

Page 3 of 3

Date 9/14/81

8.2 AXIAL INSERTION DIAGRAMS - All the data given in Sections 03-02-03 to 03-02-32are based on the following diagrams:

1-o+

-I i-

12-0E2E421EE

°"

a-I --'I CI-

42Ow

MOM

V/L41,LeaWan

IMIN

Fig. 4. Clearances between two axial parts are drawn in this figure. Similarfigures exist for radial -axial and radial -radial components. The designer mustdetermine which clearances are to be checked. The component being insertedis crosshatched. Rectangles at either end of the component represent theinsertion machine.

DI D3

91 108 95

Fig. 5. Clearances are shown for the parts described in the example. Theplacement satisfies the P dimension (91 mils) but violates the D1 dimension (108mils).


1. C981 is selected and footprint drawsthe exclusion areas (Fig. 8a). Note thatC981 is inside the exclusion area asso-ciated with C982 by 25 mils. All other

EXCLUSION AREA

component locations are still valid. Thedesigner chooses to move C982 up.

2. The designer unselects C981, selectsC982, repaints the screen and executes

COMPONENT BODY

COPPER PAD

11

HOLE LOCATIONHOLE SIZE (L =0 045-)

Fig. 6. The diagram is labelled to show the various elements of the componentand the exclusion area as seen by the drafting designer while he is laying outthe board at the Applicon terminal.

Fig. 7. A scenario of how the designer might use the "Footprint" program in anoriginal design. The text describes the operations involved.

3.

4.

footprint (Fig. 8b). This figure alsoshows the 25 -mil overlap violation.After C982 is moved up 25 mils, theexclusion areas are redrawn (Fig. 8c).No violations of ACI guidelines exist.

For a final check, the designer drawsthe exclusion areas now associated withC981 (Fig. 8d). Exclusion areas indicatethat the placement is still valid.

ConclusionThe project was successful because itdemonstrated that the AC1 guidelines couldbe automated. Some designers use theprogram whenever feasible. As mentionedpreviously, the current version is consideredto be a prototype; it does not handlemany special cases.

The program can be extended to performinteractive checking-it can tell thedesigner which components are producing


Fig. 8. An example of using "Footprint" for a component substitution. Note thatthe relative component insertion order is used to determine the exclusion areasby comparing 8a and 8b.

the violation. A further extension is toallow a batch check of the complete board.This would be useful as a final checkbefore the design is signed off. Also, theimpact of changing the insertion machinescould be evaluated. Another extension isadding more insertion methods such asintegrated circuits or surface -mount devices.

We achieved our goal of demonstratingthat interactive layout aids are viable andcan be used with complex rules.

AcknowledgmentsI would like to thank William E. Davisfor the time he spent explaining the opera-tion of the ACI machines and the use ofthe Manufacturing Design Guidelines. Heis also the author of the ACI portion ofthe MDG. Dr. Richard A. Sunshine pro-vided the inspiration and guidance neces-sary to complete the project. The other

team members involved in the programimplementation, Philip D. Deem and IwenKwo, deserve much credit for what was

accomplished. Al Riebe, our first user,spent many hours helping the implementa-tion team in debugging the software.

Bob Monat is Manager, ComputerIntegrated Manufacturing Systems in theConsumer Electronics Division. Bobreceived BSEE and M.Eng(EE) degreesfrom Rensselaer Polytechnic Institute in1971, and he started his career at RCAafter graduation. He has recently becomeresponsible for the development ofcomputer -based manufacturing systems.Upon transferring to Consumer Electronics,he had been involved with CAD/CAM inthe procurement of systems and thedevelopment of custom software to satisfyunique CE business requirements. Previouslyhe was at the Solid State TechnologyCenter (Somerville N.J.) and AutomatedSystem Division (Burlington, Mass., andWashington, D.C.). Contact him at:Consumer Electronics OperationsIndianapolis, Ind.Tacnet: 422-6682


To order your copies ofRCA A historical perspective . . .

Billing information(Purchase orders or checks must accompany multiple -copy orders.)

Name:

Color cover

88 pages RCA Business Unit:

Single copies of the History are free.

Multiple copies are $1.00 each,and the minimum order is $10.00.

Note: Distribution from theTechnical Excellence Centeris limited to RCA Corporation.

For single copies:Simply fill in your name and addresson the right and mail it in.

For multiple -copy orders:This order form must be accompaniedby a purchase order or a checkmade out to RCA Corporation.All the requested information must besupplied for multiple -copy orders.

Mail completed order forms to:Engineering PublicationRCA Technical Excellence CenterMail Stop RR13, Princeton Labs.

Please do not telephone.

Dept. Number:

Billing Address:

Charge Number:

Quantity ordered:

Mailing information(Use RCA internal mail for quantities up to 30 copiesand your organization's Shipping Address for larger quantities.)

Name:

Mail Stop:

Address:

R.A. Faust, Jr. W.L. Mays

The Operator Display Emulation FacilityUsing computers to design system display/control consoles

The evaluation of operator display and control combinationsearly in the system design/development process contributesto performance optimization and avoids expensivemodifications later.

The Operator Display Emulation Facility(ODEF) developed from a joint IR&Dventure by two Missile and Surface RadarDivision (MSRD) Engineering Departmentorganizations: Man -Machine Systems andDisplay Design. Prior to the existence ofthis facility, the performance characteristicsof a system's man -machine interface weretested only after design and prototypefabrication were completed. Revisions ormodifications identified late in system

Abstract: The purpose, organization,and system development activities of theOperator Display Emulation Facility(ODEF) are discussed in relation tospecified performance requirements. Thiscomputer laboratory facility focuses onthe application of human factorsengineering and display/controltechnology to the design, developmentand evaluation of man -machineinterfaces, commonly called consoles orworkstations. The marriage of the humanoperator to the system workstation,achieved via data analysis, modeldemonstration, and testing is describedWhen done early enough in the systemdevelopment process, the effort results inbetter, more timely and less costlyintegration of an operator and machinesystem.

©1985 RCA Corporation.Final manuscript received June 19,1985Reprint RE -30-5-5

development were difficult and costly tomake; this led to design compromises thatwould not have been nerfssary if the sameinformation had been available earlier inthe process. The ODEF allows evaluationof various display and control combinationsby display designers and human factorsspecialists long before prototype equipmentis fabricated and system software is final-ized. Achievement of an earlier appraisalof the operator/machine interface and therelated avoidance of late, expensive systemmodifications are the prime benefits offeredby ODEF.

The mission of the ODEF team is todesign, develop, and test operator interfacesfor military systems under development atRCA. They produce innovative combina-

tions of equipment and software that permitthe rapid evaluation of new display/controlconcepts via demonstration or man -in -the -loop experiments (Fig. 1). Two host com-puters, several high -resolution stroke -writing and raster -scan CRT displays, twoflat -panel displays (plasma and electro-luminescent), and a number of controldevices including touch screens, track balls,joy sticks, keyboards, and digital tablets,represent the major hardware componentsfor testing applications. Several softwarepackages have also been developed to assistin the design/evaluation process, including:

SKETCH-allows the rapid con-struction and magnetic storage offstatic display formats.

Fig.1. Some of the computer hardware resources of the ODEF.

Faust/Mays: The Operator Display Emulation Facility 41

MOVEX-can easily create and presentmoving -target scenarios with which anoperator can dynamically interact.

SIGX and ANIMAX-permit inter-active control and modification ofselected display elements.

The workstation developmentprocess

A flow diagram of the ODEF activitiesinvolved in a display/control developmentand evaluation project is shown in Fig. 2.The starting points-and the source of theoperating goals to be met-are the specifiedsystem performance requirements. Whileinput from ODEF personnel may berequested early in the system design pro-cess, Systems Engineering normally pro-vides these requirements, which are derivedfrom the system's mission objectives.

Operator role definition

System performance control functions areassigned to operators based on the skillsand knowledge of the anticipated usergroup, as well as on the capabilities of thegeneral human population. Human per-formance capabilities that influence oper-ator role definition include visual, anthropo-metric, and information promssing factors.

The ODEF workstation developmentprocess begins with a listing of statementsthat describe the various actions the humanoperator must take to enable the requiredsystem performance. Each statement iden-tifies a major operator function, such as"configure instrumentation radar for mis-sion," or "conduct system operability tests,"and is supported with directions specifyingthe individual tasks that complete the func-tion. A preliminary operations scenario isthen constructed, providing time -framesfor the functions and estimates of operatortask loading.

Display/control device selection

The role definitions and operations scenarioare the primary factors involved in theselection of display and control equipmentfor a workstation. The type of data to bedisplayed (graphic, alphanumeric, wordtext, etc.) and data update rates, togetherwith the speed and accuracy requirementsfor the proper operation of the controls,are also important. In addition to thesesystem -oriented factors, environmental vari-ables may have to be considered. Forexample, if the system is for use by combatinfantry, the light, noise, and temperature

SYSTEMREQUIREMENTS

PRELIMINARYOPERATORROLEDEFINITION

2

t

DISPLAY/CONTROLDEVICESELECTION

3

DISPLAY/CONTROLLOGICDEFINITION/DEVELOPMENT

4

DEVELOPDISPLAYFORMATS

I 5

ANALYZEOPERATORDISPLAY/CONTROLPERFORMANCE

7

CONFIGURE& TEST MMIOPERATION

6

Fig. 2. The Operator Display Emulation Facility approach to the man -machineinterface development process.

levels experienced by operators on thebattlefield will vary considerably fromthose experienced in a remote commandpost. The portability of a battlefield systemalso introduces the anthropometric (humanbody measurement) variables of thesoldier/operator. These variables narrowthe range of display and control hardwaresuitable for use in this system's workstationin terms of display brightness, contrastratio, size, weight, and reliability. Addi-tional hardware selection factors may be:(1) technical data on the performance,

size, weight, cost, and reliability; (2) resultsof trade studies that balance system perfor-mance capabilities against the performancecapabilities of other display/control deviceson the market (Table I); and (3) results oflaboratory experiments with high -resolutionCRTs, LCDs, plasma and thin-film electro-luminescent displays, and a variety of con-trol devices from trackballs and joysticksto touch screens.

Once display and control devices thatare generally suited for the system operatorfunctions are selected, ODEF personnel

Table I. Display Characteristics determined by technology.

Display Technology Characteristics

Sizes Resolution Brightness Power Weight Bulk

MonochromeCRT

large -to-small

highest high medium high high

Shadow MaskCRT (Color)

large -to-small

high medium medium high high

AC PlasmaPanel

very large-to -medium

medium medium low medium low

Thin -Film ACElectroluminescent

medium -to-small

medium medium low low low

Liquid Crystal medium -to-small

medium notemissive

minute low low

Light -Emitting small low high highest medium lowDiode


work with project design team membersto determine the specific configuration andnumber of these devices.

Display/control logic definition anddevelopment

ODEF Human Factors Engineers conductan analysis to identify critical operatortasks requiring the implementation of sup-portive control logic. Tasks with little orno margin for error that can greatly affectmission performance, tasks that must beperformed rapidly and accurately, and com-plex operator decision tasks typically qual-ify as critical. The "cause and effect" ofevery control action in the performanceof a critical task must be identified. Flowdiagrams, such as the one shown in Fig.3, are used to communicate these interac-tions and serve as a basis for staff discussionprior to program coding.

Display format development

Information needed by a system work-station operator to complete the assignedfunctions is identified in the task analysiseffort. The information, from simple go/no-go status to highly complex, multi-dimensional situation descriptions, isorganized into a unified display format.Human factor guidelines are available toaid in this design process, but creativityand trial -and -error are still needed toproduce an efficient display.

Factors to be considered include char-acter size, text arrangement (headings,subheadings, indentation, etc.), symbols,

Table II. Color display guidelines

1. Use color for identifying or reinforcingrelationships among spatially sepa-rated data or for alerting the operator.

2. Do not use color merely to make adisplay pretty; minimize the numberof colors used.

3. Use no more than nine colors on adisplay.

4. Maintain maximum spectral separa-tions between colors selected.

5. Do not use blue for small symbols ordetails.

6. Synchronize color usage with social/civil stereotypes (e.g., the traffic lightpattern).

7. Display luminance, resolution, and/orsymbol size must be increased asthe number of colors is increased.

8. Test the developed display to assurethat users perceive all the different kindsof data and data relationships present.

<PERFORMSPECIFICSYSTEMFUNCTION

ENTERCOMMAND

VIAKEYBOARD

DEPRESSFUNCTION

KEY

PRIOR NO: CONDI T

/ cOmmA NQ NOVALID?

. .

YES

ILLUMINATEFUNCTION

KEYBACKLIGHT

PERFORMSPECIFICSYSTEM

FUNCTION

'EXTINGUISHFUNCTION

KEYBACKL IGHT

.. REQUIRED .

..........

mE;ttVdt: 'NOTAVAILABLE.

MESSAGE: 'INVALIDREQUIRED COMMAND .........

Legend:OPERATOR ACTIONSYSTEM OPERATIONOR RESPONSE

Fig. 3. A representative flow diagram used to communicate the control logic forcritical operator tasks.

abbreviations, and proper use of color(Table II). Current CRT technology pro-vides the display designer with a largecolor palette; this is particularly useful inpresenting multi -dimensional information,overlays, target representations, and system -status levels. Once completed, the informa-tion display format is checked for coherencefrom an operator decision/action taskperspective.

At this point in the workstation develop-ment process, the laboratory's emulationequipment can be used to allow an operator

to test the displays, controls, and softwarelogic. Alternative designs are evaluatedand compared, and the one or two bestconcepts are submitted to the final twoprocesses: Configure and Test (Man -Machine Interface Operation, MMI), andOperator Display/Control PerformanceAnalysis.

Configure and testThe selected display and control devicesare arranged to stimulate the physical layout


Comparing operator performance onthree input devices-an ODEF casestudy

The speed and accuracy of operator controlresponses to stimuli presented in two differentscenarios were measured on three input devices:a CRT touch screen, a CRT and separate key pad,and a matrix of interactive key switches, each witha variable legend display. Twenty-four subjects,ranging in age from 23 to 63, participated in theexperiment as operators. Their task was to "pusha button" specified within a 4x4 matrix, presentedeither on the high -resolution (1024x1024 pixels)raster CRT used with touch screen or key pad, oron the interactive key -switch matrix.

Two operational scenarios were presented toeach operator: a randomly generated alphabetictask to be performed on each device, and asimulated radar operations task to be performedon the touch screen and key pad. Time pressurewas applied during certain trial runs to observe itseffect on operator performance.

In forced trials fewer incorrect key hits weremade with the interactive key -switch matrix thanwith the touch screen or key pad in the alphabeticscenario, but no significant differences were found

in either scenario in terms of the number of correctresponses (see Experiment Definitions below).

In unforced trials the touch screen proved to besignificantly faster (statistically speaking) than theother input devices in both operational scenarios.In the alphabetic scenario, the interactive key -switch matrix was more accurate than either thetouch screen or key pad. In the radar operationsscenario, the touch screen was more accuratethan the key pad.

Experiment definitions

Trial-The response to the presentation of a singlestimulus.Forced trials-The operator was allowed 2.5seconds in which to make a response.Correct (forced trials)-The desired response wasmade within 2.5 seconds.Incorrect (forced trials)-The desired responsewas not made within 2.5 seconds.Unforced trials-The operator was allowed asmuch time as necessary for the response.Error-The touch or depression of a wrong"button."

planned for the deployed system work-station. Here, careful attention is paid toanthropometric detail. Racks, tables, andchairs, as well as the control/display units,are arranged to achieve a "best fit" withrespect to the spatial envelopes representingoperator vision and reach. If an anthro-pometric problem is anticipated, a work-station mock-up is built to study thesituation in greater detail and determinethe best solution.

In the meantime, the interconnectionsfor the display and control devices (CRTs,CPUs, keyboards, etc.) are completed, andpreparations are made for the workstationdemonstration or test. If a demonstrationis required, a script containing the orderof presentation of system -specific designfeatures and benefits must be prepared. Aformal test requires more detailed prepara-tion. The experimental design for the testtypically includes identification of perfor-mance data to be collected (e.g., operatorresponse times), test subject type andnumber, the number and order of experi-mental tasks for the subject(s), and the

statistical methods for subsequent dataanalysis.

Applications software may have to bedeveloped to provide control over testtask introduction and/or to interact withthe workstation operator.

Operator display/controlperformance analysis

A number of the statistical analysis pack-ages for parametric and non -parametricdata were developed by ODEF personnel.The subjective evaluations and observationsof the test controllers and operators,together with the statistical data, are usedto interpret the results, identify potentialman -machine interface (MMI) design prob-lems, develop proposed solutions, andidentify areas for improvement.

Current developmentsPersonal computers are now commonlyused by engineers in the performance oftheir day-to-day tasks. There is, con-

sequently, a much greater appreciation forthe importance of keyboard layout, specialfunction keys, CRT screen display contrast(and the eyestrain that you get if it is

inadequate), and last but certainly not least,the interactive logic of the many differentsoftware packages available for everymachine. To a growing extent, the control/display operator is the performance -limitingfactor in today's high-speed, computer -controlled systems; the better the man -machine interfaces are designed, the betterthe systems will be able to accomplishtheir missions.

Current research efforts are directed at(1) achieving better physical and electricalintegration of available display and controldevices, and (2) developing better softwareto assist in the cost-effective design,development, and testing of man -machineinterfaces.

ReferencesI. McCormick, E.J., Human Factors in Engineering

and Design, Fourth Edition, McGraw-Hill BookCompany, New York, NY (1976).


When can the ODEF be involved in asystem design and development project?

The facility is of value whenever the capabilities ofman and machine are to be systematically combinedto achieve a desired result. The table below providesa summary of the relative abilities of men andmachines in performing system functions as aguideline for the development of man -machineinterfaces.

The typical use of the ODEF during the systemdefinition phase of a design project is the subject ofthis article. The computer -derived capability togenerate display images and emulate control/displayinteractions-long before the development ofsystem software or the production of componenthardware-is useful for system design reviews andvisual aid presentations. ODEF participates in IR&Dprojects, pre -proposal studies, proposal efforts, andsoftware module testing (by providing a hardwaretestbed environment).

Relative capabilities of humans andmachines'

Humans are generally better at:

Sensing very low levels of certain stimuli (visual,auditory, tactual, olfactory, and taste).Detecting stimuli against high "noise" levelbackground (e.g., blips on a CRT display withpoor reception).

u Recognizing patterns of complex stimuli whichmay vary from situation to situation (e.g., objectsin aerial photographs; speech sounds).Sensing unusual and unexpected events in theenvironment.

7 Storing large amounts of information over longperiods of time (better at remembering principlesand strategies than masses of detail).

- Retrieving (recalling) pertinent information fromstorage. (We frequently retrieve many relateditems of information, but the reliability of ourrecall is low.)

E Drawing upon varied experience in makingdecisions; adapting decisions to situationalrequirements; acting in emergencies (noprevious "programming" required).

7 Selecting alternative modes of operation whencertain modes fail.

El Reasoning inductively, generalizing fromobservations.

CI Applying principles to solutions of variedproblems.

E Making subjective estimates and evaluations.

n Developing entirely new solutions.

E Concentrating on the most important activitieswhen overload conditions exist (prioritizing).Adapting physical response to variations inoperational requirements (within reason).

Machines are generally better at:

Sensing stimuli that are outside the normal rangeof human sensitivity (e.g., x-rays, radarwavelengths, and ultrasonic vibrations).Applying deductive reasoning, such asrecognizing stimuli as belonging to a generalclass (where the characteristics of the class canbe specified).Monitoring for prespecified events, especiallywhen infrequent (but machines cannot improvisein case of unanticipated events).Storing coded information quickly and insubstantial quantity (e.g., large sets of numericalvalues).Retrieving coded information quickly andaccurately on specific request (when specificinstructions are provided for the informationtype).Processing quantitative information (followingspecified programs).

7 Making rapid and consistent responses to inputsignals.

7 Performing repetitive activities reliably.L Exerting considerable physical force in a highly

controlled manner.Maintaining performance over extended periodsof time (i.e., machines do not need "coffeebreaks").

1_ Counting or measuring physical quantities.Performing several (programmed) activitiessimultaneously.

- Maintaining efficient operations under conditionsof heavy load.

E Maintaining efficient operations despiteenvironmental distractions.


Rick Faust is Unit Manager, HumanFactors Engineering at the Missile andSurface Radar Division. Since joining RCAin 1968, he has conducted human factorsengineering projects for a variety ofmilitary systems. Man -machine interfacedefinition and development is his principalarea of technical interest. A member of theHuman Factors Society and IEEE, Rickreceived a BS in Physics from LafayetteCollege and an MA in Sociology from theUniversity of Maryland.Contact him at:Missile and Surface Radar DivisionMoorestown. N.J.Tacnet: 253-6049

Bill Mays is Unit Manager, Displays, in theSignal Processor and Computer SystemsSection at the Missile and Surface RadarDivision. He received the BSEE degreefrom the University of Texas (Austin), andjoined RCA in 1966. Bill has worked ondigital system design and logic design andverification, primarily in computer interfaceequipment for various radar systems suchas the doppler tracker, laser range tracker,and dynamic target simulator. More recently,he has directed computer graphic displaydevelopment projects, producing systemsoftware, firmware, hardware and testfacilities, and conducted technologicalstudies of flat panel displays for militaryapplications.Contact him at:Missile and Surface Radar DivisionMoorestown, N.J.Tacnet: 224-3448

ErratumThe following team received a DavidSarnoff Award for Outstanding Achieve-ment in Science, 1969:

L. J. BertonH. A. FreedmanJ. A. GoodmanN. L. GordonJ. T. O'Neil, Jr.B. H. SamsA. H. SimonT. M Stiller

For the creation of a versatile, advanced,time-sharing system for the Spectra 70/45computer.

Their names were inadvertently omittedfrom the list of past winners in ourJuly/August issue.

Authors. Faust. left and Mays.


Battery Area

1. Battery Voltage ProblemIF battery voltage it too low then:a) Check the solar array area for maximumpower output <shunt ourrents..sun angle)

la) If power can not be inor eeee d then lookfor ways to deo eeeee the loads.

Consult with system engineer before adjustingor removing any loadsIF battery voltage is too high then:a) Check. the solar array area to see ifpower can be deo eeeee d (Inc eeeee shuntcurrents or Increase the Sun angle)

b) Check loads to sate IF all systems areoonneoted as expected.

Consult with the system engineer after theproblem has been focal lied.

IMMO IMMO I MOM 111.MEM MUM

Fig. 16. Level 6, diagnostics screen.

RCA Satellite Systems

SC 1 $C3

1/4$C 4 $C 5 $C 6

42.111.011111111110 111111111111 111111111111 MIME MINIMMIME MIMEO MEM

Fig. 17. Level 1, overview after the problem has been solved.

IMMI sample

To get an idea of how the IMMI system works,assume you are an operator. Figures 11 through 17contain a series of photos of actual IMMI monitoringscreens. Figure 11 contains an overview of anentire multi -spacecraft system. There is a satellitesymbol representing each spacecraft. One of thesesymbols is red, one is yellow and the rest are green.Red indicates a serious problem, while yellowindicates a warning situation. At glance, you knowthat there are problems in two spacecraft and therest are working normally. Associated with eachsymbol is an option at the bottom of the screen. Youselect option S/C2, since it is the spacecraft withthe most serious problem.

Within a second, the screen shown in Fig. 12appears. It contains an overview of the majorsubsystems of that particular spacecraft. There is asymbol associated with each major subsystem. Thebattery representing the power subsystem is red,indicating a problem in that subsystem. You selectpower option. This causes the screen shown in Fig.13 to appear. It contains an overview of the powersubsystem. There is a symbol representing eachmajor component, from solar arrays and shunt tobatteries. The battery box is red, indicating aproblem with one of the batteries. Selecting option5, associated with the batteries, causes the screenshown in Fig. 14 to appear.

This display contains a parallel designrepresenting the two batteries in the spacecraft.Within each battery's design are boxes representingthe main telemetry points such as voltage, current,and temperature. In addition are on/off statusindicators for components like heaters and

chargers. These provide the viewer with additionalinformation about the battery area. The parallelstructure allows the user to locate similarinformation about the other batteries quickly andeasily. The box associated with the voltage inbattery 1 is red, indicating an abnormal batteryvoltage. The other battery sensors have normalreadings. To obtain more detailed information,option 1 is selected. This causes the screen shownin Figure 15 to appear.

This screen contains a linear meter -like display ofthe actual telemetry values. Each telemetry pointhas a graph associated with it. Each graph containsa title, the actual value, the limit values, the units anda graphic representation of the value. The user cansee at a glance how close a value is to a limit.These types of displays can be used to monitorvalues thay may be approaching limit values. Inaddition, the name and bar are color -coded torepresent the limit condition of the telemetry point.In this case the entry for battery 1 voltage is red.

The final step in the process is obtained byselecting option 1. This causes the screen shown inFig. 16 to appear. This is a diagnostics screen. Itcontains some simple instructions to the operatorconcerning possible problems with the batteryvoltage. This forms the logical end to the operator'ssearch. After the necessary information has beenobtained, the system analyst can be called to helpidentify the problem. This entire search would takeless than 10 seconds on the IMMI system. Thereader may even have learned something about thesatellite's structure just by looking at the screendesigns. This is part of the goal of the IMMI design.Figure 17 shows the system status after theproblem has been corrected through commandingexternal to IMMI.


11111111111111.111M1 0 MEM

T721A

75

11

TT388

25

I0

T28875

ISOM NORMINI MUM= MEI= 11.7. a51 12,12tin'7,Lx"

NORTH PANEL Thermal Status

28 "20

11168

60awa

TT1S8

60sspa

5

20

TT198 7736A

57sspa ;: bat 48

20

II12I a 25

3

2,

TIM

bcn

11-11:6

76

sspa

7

123

1715A

5676

TT,188

111:

sspa

1

20

sspa11

11488

76

12

TT198

FE

1

1188

1146A

Fig.18. Display of temperatures on the north panel usingthermometers.

Screen drawing time

The screen drawing time was found to be a very importantfactor. Experiments were made in which the time it took todraw a screen was varied. The times ranged from under asecond to approximately ten seconds. In a monitoring situationwhere a user wants to rapidly flip from one screen to the next,

drawing time of two seconds or more was found to beunacceptable. Users became annoyed while waiting for screensto appear. After some experimentation, a screen drawing timeof about one second was found to be a reasonable compromise.The user is willing to accept the slight delay, while the time islong enough to prevent any severe restrictions on graphicscomplexity, resolution and hardware speed. A maximum ofabout two seconds was established as an upper limit on all butthe most complex analyst support screens.

Graphics complexity

The level of complexity of the graphics was also investigated. Abalance had to be struck between placing as much informationas possible on a screen and user confusion. In a monitoringsituation, a user should be able to glance at the screen anddetermine if there is a problem, and where. Typically, ourscreen designs limited the number of objects or items to between10 and 20 per screen. This permitted rapid recognition and didnot require the user to study the screen extensively.

The complexity varied somewhat with system structure level.At the top levels, the number of items per screen tended to beten or less. These levels provide the controller with the basicoverview of the system and its major components. A simplerdesign makes it easier for the user to assess the overall situation.The screens tended to be more complex at the lower levels.These screens contain the details of the different parts of thespacecraft subsystems. In many cases they contain parallelstructures representing multiple or redundant components.

If there is a flow involved in the subsystem, such as power orcommunications payload, a left to right or top to bottom designis used. If there is any parallelism in the equipment functionality,

Graphics symbols

The screen designs used in the IMMI system makeextensive use of symbols, or icons. In themonitoring screens, the symbols are basicallysimplified pictures that represent the differentcomponents. Figure 12 is a good example. Itrepresents the overview of the subsystems of asingle satellite. Little picture elements are used torepresent each major subsystem in the spacecraft.Power is represented by something meant to looklike a car battery. Thermal is represented by athermometer. The basic idea was to keep it simpleand easily recognizable. As the system matures,these symbols will probably be altered to reflectexperience gained in different applications.

Figure 18 contains a display of temperatures onthe north panel. The use of a thermometer -likerepresentation quickly tells the user that the valuesbeing displayed are temperatures. Again, color -coding is used to show limit and warning statusinformation.

The system analyst support screen designscontain more of the technical symbols found intechnical drawings. Since the expert is familiar withthese symbols it is unnecessary to usesimplifications. The top-level analyst supportscreens still make use of the block diagrams, etc.,necessary to display the functional information.

this is reflected in the graphical layout of the screen. Thesetechniques were used to help convey the functionability andinterrelationship of the different components.

Screen resolution

The screen resolution was determined by several factors. Themain factor was speed. The more resolution, the longer it takesto fill the pixels on the screen. A 1024X1024 screen has fourtimes as many pixels to fill as a 512X512 screen. If the graphicsdevice is fast enough, this may not be a problem; however, inmany devices it is.

Experiments showed that 512X512 point resolution is sufficientfor the level of detail needed for monitoring displays. In manycases, less resolution would have been permissible, except forthe text fonts. They were too large and coarse at the lowerresolutions. The only area where higher resolution is useful is atthe lower level subsystem displays or in the analyst supportdiagrams. There is a danger with the higher resolution, however,since a designer may be tempted to add too much detail andcomplexity in a screen design.

Additional factors were the display monitor characteristics. Ifone compares low- to high -resolution monitors, the problem offlicker can become quite noticeable. To minimize flicker, long -persistence phosphor screens or a non -interlaced display can beused. The long persistence phosphors are easier on the eye, butcan leave ghost images of the previous screen. The non -interleaveddisplays require a higher refresh rate, are more expensive, andare less available at the higher resolutions. The 512X512 resolu-tion was found to be satisfactory for monitoring displays.

R.H. Harten/T. Crowe/A. Anton: An improved man/machine interface for spacecraft control 53

Fig. 2. Detailed functional diagram of part of a communica-tions payload.

are currently operating multi -spacecraft systems, enhanced mon-itoring and control capabilities were needed. The ImprovedMan/Machine Interface (IMMI) project was begun in 1983 byEmil Dusio to investigate ways in which spacecraft monitoringand control and operator efficiency could be improved.

An initial analysis of existing satellite control systems showedfour areas where improvements would have a noticeable impacton system performance:

(1) The overall user interface to the system should be simplified.(2) The user interface must support the needs of the operator

and the analyst.(3) An overview of the total system must be provided.(4) A structure should be provided for the analysis of spacecraft

conditions.

The problem was how to make it easier for a spacecraftoperator to control and monitor multi -spacecraft systems. Theoperator must be able to recognize problem conditions beforethey become serious. At the same time the system must provideimproved support for the spacecraft analysts and engineers whoare responsible for normal spacecraft operation and trouble-shooting. The goal was to make these people more effective andto permit more complex systems to be controlled by fewer,less -skilled operators.

The main design concept we followed was that the operatorshould be made aware of the problem and led to it, not forcedto find it. Also, the scope and complexity of the problem shouldbe revealed to the user as it is identified and located. Thisapproach could also be extended to the more detailed supportrequired by the analyst studying normal and abnormal spacecraftconditions.

The implications of these goals were almost contradictory.To assist the operator, we wanted to provide a good overviewand simple structures and screen designs. To aid the analyst, wehad to provide detailed information and possibly complex screendesigns. The key to solving these design requirements was toprovide a structure that uses a top -down approach. The layeredapproach provides an overview at the upper levels and moredetails at the lower levels. This satisfies the requirements of bothclasses of users.

Several other requirements had to be considered in theIMMI design. The system must be able to interface with

Fig. 3. /MM/ diagram of the communications payload.

existing systems, making it attractive to existing customers. Inaddition, it should be able to evolve into a product that couldbe integrated into future ground system designs. Its use of high-speed color graphics and a structured approach make it a goodbasis for future system designs.

StructureThe approach adopted to satisfy various requirements was to setup a structure tree based on the subsystem functionality of thespacecraft. This structure defined the way the information wouldbe organized, how it would be grouped, and provided a naturaltop -down approach. To satisfy the dual requirements of systemmonitoring by an operator and detailed problem analysis by aspacecraft engineer, a parallel structure was adopted. Both partswere based on the same spacecraft functionality. The monitoringfunction uses simpler graphics designs and is designed for easyrecognition of problem situations. The analyst function providesmore complex information. Both are part of the same system,and it is possible to move from one part to the other. To furtherenhance this capability, IMMI allows a user to select and rapidlyflip between screens in both parts of the system. This permitsthe analyst to make full use of the monitoring and detailedinformation, and vice versa.An example of the different levels of complexity between the

monitoring and the analyst screens can be seen in Figs. 2 and 3.Figure 2 is an engineering diagram showing the part of acommunications satellite's payload. It shows all the possiblepaths, switches, and key components. The circles with twocurved lines are switches which, if set in the "other" position,would cause the circle to rotate 90 degrees and a different set ofconnections to be made. To figure out a route, a viewer mustliterally follow it with his finger. The numerous switches permitthe systems engineer to reconfigure the components or reroutethe signals around a faulty or substandard component.

Figure 3 is a sample IMMI screen design covering the samepart of the payload. It does not show any of the complexswitches. It only shows the resultant routes and the componentsinvolved. If there is a problem in channel 9, the operator


immediately knows that solid state parametric amplifier (SSPA)9 is involved. It is much easier for the satellite controller to obtainthe critical information using this screen. Figure 2 contains thetype of information a systems engineer would require to switchout the faulty components. Figure 3 is better suited to determiningwhich component is faulty. The IMMI structure must accom-modate both types of information.

Status displaysThe status displays are screen designs and a structure tree thatprovide the spacecraft monitoring function. The status displaysare organized in a top -down, hierarchically -structured manner.The top-level display contains an overview of the entire system,with very little detail. The second level contains an overview ofthe main subsystems of a single satellite. At each successivelylower level, the structure branches into the subsystems andsubassemblies. At each level, the screen displays a more detailed

SC1 SC2 SC3

III 11 DILLSC4 SC5

II II 11 II

ATTITUDE POWER THERMAL REACTION

III II I I I

SOLAR ARRAYS MOTORS

I 1 (VALUES)NORTH SOUTH

SHUNTS

(VALUES)

BATTERY

1-1-1BAT 1 BAT 2

(VALUES) (VALUES)

DIAG. DIAG.

POWER BUS

(VALUES)

Fig. 4. Structure tree.

SC 2Reaction Control Panel Sensors

ISIAH WM IZMIR NUNN OMNI I.MIMI IMMO NEM MEM =RIM

Fig. 6. Screen showing the thermal sensors in the reactionsubsystem.

diagram of one of the components shown in the next -higher -levelscreen. The number of levels and the degree of branching isdetermined by the complexity of the different spacecraft systems.Figure 4 shows an example of this type of structure tree.

The layout of each status display screen is designed toemphasize the main components and their interrelationship.The design goals are to make the user aware of the basicfunctionality of the subsystem or subassembly, indicate thestatus of the main components, and show where to look forfurther information or more detail.

The same basic screen layout is used for most of the displays.

NORTHSOLARARRAY

SC 1 POWER SUBSYSTEM

UNREGULATEDBUS

/N. MOTOR)]

\ S. MOTOR

SOUTHSOLARARRAY

SHUNT I AMPS1 OFF I 2 OFF

NORTHSHUNTS

SOUTHSHUNTS

CHARGECONTROLLER

BATTERIESBATTERY

BUS

5 EMCEE) 01333EED 011103133D I3 015E13333 0 1332113311 1t -7 f..---11.-

OPTIONS LINECORRESPONDS TO FUNCTION KE VS

6

POWER COMMUNICATIONS

\ THERMAL

A P R T C

SATELLITE REACTION \NO ATTITUDE COMMANDING

Fig. 5. Basic IMMI screen layout.

HEADERINFORMATION

MAINDIAGRAM

SYSTEMSTATUSOVERVIEW

ATTITUDE SUBSYSTEM

AttitudeErr_

.01.

Earth Sensor 1:

-.01.

ch 0.00All 0.80

Torquer

"T .M..NEM

I I I .

-.03-.02-.01 I .01 .02 .03PITCH

Attitude Processor

MEZEILII=11=

0.0000.000

ReactionF1,111.9

auto ape4

11

2

1014

11

1512IC

lioeumtua WEsel 1

Demand e ,pm0.00o dee

Milian IMMO ELMMN=

Fig. 7. Overview of the attitude subsystem.


SC 2 STRUCTURAL - North Panels Display

I II12 101301.

TT58 TT5A

II 52 I

OTOTOTOTOTC 11kILOICKOM

T138A

EPC 2

OTT198

NA 2aTrose

AB

2

B

CUM)

BPS4

T 88

2 [04AntennaPanel

94

B

BasePanel

PSE

KEY: esa . Mir 0 Comp mount - Panel mount

IBEX= .111111111111 IMBIBBB ZOBUCt=MI 1111111111111

Fig. 8. Structural layout of components on the northpanel of the spacecraft.

Power Subsystem Configuration SC 2

To

UNREGULATED BUS

rrom Unregula140 Bus

is UEw :1Uu E i :10113MOM merle MEW OMNI 111111=1. '111 MUM:2

Fig.10. Analyst overview of the power subsystem.

This layout is shown in Fig. 5. The top area contains adescription of the information being displayed. The bottomcontains the user options and prompt information. Slightlyabove the prompt information is the total system status informa-tion. The remainder of the screen is devoted to the status oranalyst support information.

Figures 6 and 7 are sample IMMI screens. Figure 6 showsthe main components of the reaction subsystem and their asso-ciated thermal sensors. The viewer can identify a sensor with aname and a component. At the same time an overview of themain components of the reaction subsystem is provided. Thecolors of the sensors indicate their status: green is normal,yellow is warning, red is serious error. Figure 7 is an overviewof the attitude subsystem. The central display indicates the rolland pitch errors and the location of the Sun and Moon. The

Help Battery 1 Charge Status SC 2

II Mint ITEM BM

Fig. 9. Circuit of a battery charger.

RCA Satellite Systems

SC 1

SC 4 CC 5

SC 3

SC 6

LIBUINNMI NOM 11111111111 111111M1111 MIMI! MIME 22r,

Fig. 11. Level 1, overview of the total system.

status of the other parts of the attitude subsystem, as well as thethruster status of the reaction subsystem, is shown. Color codesindicate thruster status (enabled, armed, firing). If the Sun isnear the Earth, the Earth sensor is disabled. These screensprovide an overview of a particular area and make it easy forthe operator to detect error conditions.

User options are indicated at the bottom of the screen. Thereis one associated with each major component shown on thescreen. Selecting one of these options causes a new screen toappear. The new screen will contain more details about thatcomponent. This procedure is repeated at each level until it isunreasonable to go into more detail. The bottom level is usuallydetermined by the actual telemetry sensor points on thespacecraft.

How does this structure aid the user? Starting at the toplevel, the user can quickly answer the basic question, "Are thereany problems?". If one of the items is colored yellow or red, the

50 RCA Engineer 30-5 Sept /Oct.1985

R.H. Harten T. Crowe A. Anton

An improved man/machine interfacefor spacecraft control

A new top -down approach to spacecraft monitoring systemsuses color graphics and a logical tree structure to improve thesystem capabilities and reduce the number of skilled operatorsnecessary.

Current trends in communication and other spacecraft systemsare toward more complex spacecraft, more telemetry per space-craft, and more spacecraft being controlled from a single location.This trend places increased demands on the existing methods ofspacecraft control and monitoring. The volume and complexityof the telemetry information make it increasingly difficult forsatellite ground station controllers to notice and respond toabnormal spacecraft conditions.

Current systems typically use a series of telemetry limitchecks and alarms to monitor and signal abnormal conditions.This works well for isolated or extreme conditions, but it is notas well suited for complex problems that may involve severalsubsystems. Nor is it particularly sensitive to spotting gradualtrends or correlated problems. In current systems, the systemperformance is strongly dependent on the ability, experience,and skills of the satellite controllers involved.

A typical spacecraft monitoring system consists of an alarmsystem and screens full of telemetry data. A typical layout for ascreen of telemetry data is shown in Fig. 1. The telemetry datacomes from sensors on the spacecraft, which monitor the differentfunctions. Limit checks are built into the ground station software

Abstract: The trend in communication and other spacecraftsystems is towards more complex spacecraft, more telemetry perspacecraft, and more spacecraft being controlled from a singlelocation. This trend places large demands on the existingmethods of spacecraft control and monitoring. The volume andcomplexity of the telemetry information make it increasinglydifficult for satellite ground station controllers to notice andrespond to abnormal spacecraft conditions. This articlediscusses an improved man/machine interface for spacecraftmonitoring and control one that reduces the number of skilledoperators needed while allowing improved control of multi -spacecraft systems.


B-BAT1IC13-fiAT2IC

B-BAT1I0B-BAT2ID

B-TT39B.739

a-TT12B-TT12

that monitors the telemetry information, and when a value goesoutside of a predetermined range, an alarm is sounded. If thereis an alarm condition, the alarm messages indicate the problemtelemetry points and the displays are used to show the actualvalue(s) causing the alarm. In less critical situations, operatorswill monitor numerous telemetry points by placing them on thesame display screen and examining them periodically to see ifthey are drifting or approaching critical values.

This method has worked successfully for many years, but itssuccess has depended on skilled and experienced operators. Asthe complexity of the spacecraft payloads increases and thenumber of spacecraft grows, a larger number of these experiencedoperators will be required. Since the typical lifetime of thenewer spacecraft can be as long as ten years, any means ofreducing the number of required operators, while maintaininghigh operational standards, will have a significant impact on theoperating costs of a multi -spacecraft system.

Since several of RCA Astro-Electronics Division's customers

Pali 401 BABY28 A5C BiBY sumo STATUS

ASC ASC

B-LOM 5 B -HIGH 5 - -65/09/19 14:35:29

6 -ARRAY ARRAY 0.11aREMT 20.471 Fff5 0-01 AZ CCP 241.65 UV13-4194UKT NORTH 94111 I 5.6471 WS 0-C1 EL ELEVATIO4 20.742 DM9-591141 SOUTH SHUNT I 5.6471 AIDSB-BUSVILTS BUS VOLTAGE 35.059 VOLTS B -PITCHER PITCH MCP .00000 0E0

a-RILER ROLL ERROR .35188 EEGCHARGE CURRENT 2.0196 WSCHARGE CURRENT 2.0196 AIDS B-SPEED1 NA 1 SPEED 6050.1 RPM

B-WSDEMR10 11I4 CONK 12.000 RPM

DISCHARGE 0.0 .00000 Aff5DISCHARGE CUR .00000 N1,5 B-TCPCUER TEPOLER CEHTREL NEGATIVE

Et.TORELIE 7160IJER STATUS .607E6 WSBATTERY 1B -.6473 GELS'BATTERY 28 -1.155 GELS!

SCLITH SH.HT 31.736 C615INORTH SAUHT 21.553 COSI

B-CPCOCE1 0:414k40 CPCILE 0

9-CPUCE2 CEPIINC CPCEDE 0

B-A6C1 RECEIVER 1 AOC 3.0392 va.Ts

1111111111= " ----Fig. 1. Typical page display of telemetry data.

RCA Engineer 30-5 Sept. / Oct 1985

Satellite Subsystem

SC 1

Reaction

Power

-Z I

Thermal

PC..

Command

Communications

Fig. 12. Level 2, overview of a spacecraft's subsystems.

Utu Josystem

Batter c-tnt,

primariC backuppullup pullup

on off

current

pressure

temperaturev/t curve 2

pri heater on

sec heater off

pr l Mary

nar ger

battery 2

coltageurrent

pressure

temperaturev/t curve 2

prt heater off

sec heater on

batters

OZWWW MANI MUNN 01410441 OZWWUMW 12444ZtLUZERIEWSPN

Fig. 14. Level 4, overview of the batteries.

operator knows that there is a problem. The same screen alsoindicates any other problems. By following the colored indicatorsand selecting the appropriate options, the operator can localizea problem in a matter of seconds. The operator, in locating theproblem, is gaining knowledge on the state of the rest of thesystem at the same time. Are there other problems? If so, arethey related? This provides the operator with a better understand-ing of the nature and scope of the problem.

The accompanying sidebar provides an example of how theIMMI displays could be used in a spacecraft monitoring situation.

An interesting offshoot of the structured status screen systemis that it can be used as an aid in familiarizing new engineersand managers with the system. By flipping through the screens,the viewer is presented with a visual display of the maincomponents in the system, their complexity, and how they areorganized. The status screens provide a basic description of the

Fig.13. Level 3, overview of the power subsystem.

Power Subsystem BATTERY' 1

Batte. t.94128. 2-.4 32.2 36.8

96.70 volts

6 ttttt y I Volt 4te53 C 92,6 3 S 5

96.46 volts

h4,9* Current=4.166 amps

wcharge Cur4.880 al4eS

essur.309.0 psia

1 SOC76

100.0

11114111 IMMO MINI 411154104=MI MIME

Fig.15. Level 5, meter display of telemetry values andlimits.

system. The analyst screens can be used to provide detailedinformation. In no case would they replace standard documenta-tion and manuals, but they could serve as a good introductionand aid.

Graphics

One of the basic assumptions of the Improved Man/MachineInterface (IMMI) project was that color graphics would be animportant tool. It remained to establish how it should be usedand what the requirements would be. Four main properties hadto be evaluated: screen drawing time, graphics complexity,resolution, and effective use of colors. While many of theseproperties are interrelated, it is instructive to consider themseparately.


The information transfer problem

One of the main problems encountered by the IMMIproject was the transfer of information from thesatellite subsystem experts into the IMMI designand software. Typically, this is done by projectmembers talking to the experts, obtaining anunderstanding of each subsystem and integratingthis into the software and graphics. This process isvery time-consuming and open to errors. It is an artto be able to ask the proper questions to obtain theinformation and understanding needed to create thesystem. Yet one requires a degree of understandingbefore one can ask the proper questions. Theexperts are concerned with the design, fabrication,and functionality of the subsystem. It is not alwayseasy for them to understand the type of informationa software designer requires. In addition, it issometimes difficult for the expert to be able to checkthe work of the software designer. A similarsituation exists in many artificial intelligence andexpert systems.

There does not appear to be an easy solution to

this problem. The success of many of the existingexpert or Al systems has depended on the ability ofthe computer expert to learn the problem area indetail or for the engineer or applications expert tolearn the computing techniques. Neither method isparticularly efficient in manpower.

In an attempt to address this problem, future IMMIsystems will contain design tools that will simplifythe job of the engineer trying to "tell" the computerwhat he or she knows. Extensive use of graphicsand simple editable lists will be used. The primarygoal is to minimize the amount of computing anexpert must learn. The key to the system will be astructured approach in which the experts will firstdefine the basic structure tree and then fill in thedetails at the specific levels. The structuralinformation will be used to guide and prompt theexpert. Using a top -down approach should make iteasier for the system to lead the experts to the areaappropriate to their subsystem. It should also givethem a feel for how their information fits into thesystem as a whole.

Colors

The use of colors in the graphics displays was a fundamentalpart of the IMMI approach. Three basic guidelines werefollowed:

(1) Standardize the meaning and use of the colors in the designs.(2) Limit the number of colors to avoid confusion.(3) Use more subdued colors to avoid confusion and eye strain.

A maximum of 16 different colors were used in the screendesigns. Approximately half of these were assigned to indicatesubsystem and component status (green for normal, yellow forwarning, red for severe error), ON/OFF, etc. The colors of thevarious boxes or symbols on a screen will change color dependingon the status of the component or subsystem they represent.The remaining colors were used to denote and enhance featuresin the more complex designs. By standardizing the color usage,it is easy for a user to quickly pick out the critical elements oneach screen. The number of colors was limited to preventconfusion and prevent the operator from having to remembertoo many color codes.

Although only 16 colors are used, each is custom -blended tosuit our needs. In general, the primary colors and their simplemixes (cyan, magenta, etc.) are too bright for prolonged viewing.To overcome this we either added grey or decreased the intensityof the color. Shades of brown, gold, and blue were used inmany of the more complex screen designs to delineate particularfeatures. While the color choices are always biased by personalpreferences, the more subdued shades are easier to view andseem to be quite effective, especially in low light situations.

Analyst support screensIn addition to the monitoring screen, the IMMI system alsocontains analyst support or help screens. These screens typicallycontain detailed diagrams, schematics and drawings of the space-

craft structure and its component subsystems. Typical examplesare shown in Figs. 8 and 9. These screens are designed to aidthe analyst. They can be quite complex and will be examined indetail by the analyst. There is a tendency to make these diagramstoo complex, as the initial attempts to design these screensshowed. Figure 10 is an example of such a screen, which mayhave to be redesigned. Some care must be taken to preserve thestructured approach in which the highest level of detail is

reserved for the lower level screens that cover a restricted partof a subassembly. Using this approach, we found a resolution of512X512 was satisfactory, although additional resolution wasjustified in some diagrams. As with the operator monitoringscreens, the size and readability of the descriptive text placed aneffective limitation on the resolution that could be used. Evenon the complex screens, the best guideline is to keep it as simpleas possible.

DiagnosticsThe bottom level of the IMMI structure tree is devoted todiagnostics. This was a logical outgrowth of the structuredapproach. Assuming a simple problem, the user has been led tothe appropriate area of a spacecraft subsystem. This area typicallydeals with a few components and a limited number of telemetrypoints. Since the number of possibilities is limited, it becomesattractive to add a diagnostic capability. Information of theform, "If voltage x is low, check settings y and z. If they arenormal, then do the following" becomes a possibility. Thestraightforward case is relatively easy to handle, but complexsituations involving more than one subsystem can not be handledin this manner. This type of situation might better be handledby an expert system. Considerable experience will have to begained before a true diagnostic system can be implemented.

A slightly different approach is to use the diagnostics as anoperator aid. For a given problem it contains instructions for


collecting and preparing the nerfssAry information for the systemsengineer, who will have to analyze the problem. In this case themessages would be more of the nature: "If x is wrong, thenobtain the following information." One could even have thecomputer obtain the information automatically at the operator'srequest. This approach is easier to implement and can beaugmented with suggested actions as experience is gained. Thediagnostic instructions could be updated by the systems engineerto reflect the current situation and previous experience. Futureimplementations of IMMI will adopt this latter approach. Theuse of knowledge -based and artificial intelligence systems is alsobeing investigated for possible future applications.

Conclusions

The IMMI project has demonstrated that significant improvementsin satellite monitoring can be achieved by using a highly -structured, top -down system that incorporates high-speed colorgraphics. The system is capable of supporting both the monitoringrequirements of a satellite controller and the detailed information

needs of a system analyst. The effectiveness of the system isdependent on careful graphics design criteria and a proper top -down structure that forms the relationship between the graphicsscreens. This structure is based on the actual structure of thesatellite and its subsystems. Rapid screen drawing times, simplescreen designs, and standardization of layout, symbols, andcolors were important factors in the success of IMMI. Thesystem will inevitably evolve as more experience is gained.

AcknowledgmentsThe IMMI team also included Rhonda Adamson, who workedon the database, and Joan Hennessey and Tina Segal, whodeveloped some of the subsystem and analyst support screens.We would also like to acknowledge the support of E. Dusio, S.Malyszka, R. Farag, R. Hrusovsky, W. Upson, E.C. Scholz, S.Fox, and J. Swale. The photographic work was done by LeoCashel. We also appreciate the expert input from variousmembers of the Spacecraft Systems and ASOC groups.

Authors, left to right: Harten, Crowe, and Anton.

Ronald Harten is a Principal Member of the Technical Staff,Astro-Electronics Division. He joined the Division in 1984,assuming the lead engineer role on the IMMI project. He is alsoinvolved in other aspects of ground system software. Prior tojoining RCA, he worked as a research radio astronomer in theNetherlands for over 12 years. Ron earned his AB in Physics atCornell University in 1966, and his PhD in Astrophysics at theUniversity of Maryland in 1971. He is a member of the AAS andIAU.Contact him at-Astro-Electronics DivisionPrinceton, N.J.Tacnet: 229-3551

Tim Crowe is an Associate Member, Technical Staff at Astro-Electronics Division. Since joining RCA in 1983, Tim has beeninvolved with various aspects of satellite ground station softwareand test equipment. His responsibilities on the IMMI projectinclude telemetry analysis, database organization, and space-craft subsystem display screens. Tim holds a BA degree inComputer Science from LaSalle University.Contact him at:Astro-Electronics DivisionPrinceton, N.J.Tacnet: 229-2682

Art Anton is Manager, Aerospace Ground Equipment SoftwareDevelopment, Electronic Systems, at Astro-Electronics Division.His responsibilities include software design and development ofspacecraft checkout stations and IR&D projects, such as IMMIand Artificial Intelligence Systems for Space Platforms. He joinedRCA in 1977 at RCA Laboratories, and worked in advancedtechnology software systems. In 1979, he joined RCA GlobalCommunications as Group Leader responsible for the develop-ment of software for various communications networks andstore -and -forward SWITCH systems. Prior to RCA, his workincluded design of electrochemical and automated process -con-trol systems. He holds a BA in chemistry from City University ofNew York, an MS in Physical Chemistry/Mathematics fromStevens Institute of Technology, and an MS in ComputerScience from Rutgers University.Contact him atAstro-Electronics DivisionPrinceton, N.J.Tacnet: 229-3537


Graphics for Election '84: An introduction

One goal of any television production isthe communication of information. At asporting event, the camera can tell uswho's on first, and an instant replay canprovide insight into how a touchdownwas scored. On a news program, thecamera is pointed at a visiting dignitary ora burning building. In each of these cases,the camera provides visual informationthat emphasizes the importance of theevent.

The Presidential Election is anespecially rich source of visualinformation. The primary races leadingup to election night, the issues, the locales,and the contrasting styles of thecandidates are all interesting andimportant points of focus for newscameras. But the finale, Election Day,poses a difficult problem: Where do wepoint the cameras?

The story on Election Night is one ofnumbers. How much of the vote is in?Who won in which states? By whatmargin? What is the new composition ofthe House and Senate? Solving theproblem of effectively communicating thisinformation was the purpose of the NBCElection '84 Graphics DevelopmentProject, begun in mid -1983.

There were three teams involved in theproject, which was under the overalldirection of Tom Wolzien, NBC NetworkNews. Roy Wetzel, Election InformationUnit, Network News lead the statisticaland data tabulation group. Bob Brandel,Network News Production and Design,headed the graphic design group. Iheaded the engineering developmentteam, Computer Imaging, responsible forthe technical implementation of theproject.

Concepts of the types of displays weredeveloped. We decided to use a new andunique approach that incorporated "real"images such as faces, buildings, and anti-aliased text. We also felt that animationwas important, because it would create adynamic sense of excitement. The keyissue in the approach, however, was thatthe displays must visually convey thequalitative aspects of the data: How close

is the race? Who is ahead? Howsignificant is a particular issue? Previouselection coverage had done little morethan display numbers. The drawbackwith that approach is that most of thesignificance is lost. The numbers arefactually correct, but they are dismallypoor at conveying information to thetelevision audience.

Production constraints had to beaddressed as well. Because the data wasrapidly changing as votes were counted,the animation displays had to reflect thelatest information. The Election Nightshow is live, so displays had to beavailable reliably and rapidly. Provisionshad to be made for redundancy, errormonitoring, and "disaster recovery." Andthe system had to be able to cover morethan 500 races nationwide.

Once the design and productioncriteria were established, the search beganfor a total system approach to meet them.Computer -synthesized animations, using ahigh-speed computer and general-purposeframe buffer, could provide the neededimage quality, but not the speed,flexibility, or usability necessary. Other,real-time animation systems wereinvestigated, but they were expensive,hardware -intensive, and could notprovide the image quality (complexity)that production design had requested.

At the time, we were starting to use aQuantel Paint Box in daily programming.While meeting the design criteria ofrealistic, anti-aliased images and text, ithad neither the facility for animation, northe ability to create an image as afunction of external data. Severaldiscussions followed with Quantel, and itseemed that software could be developedthat satisfied the data -driven animationcriteria. Essentially, we would create adual -functionality for the machine. Itwould be an artist -operated graphic workstation, and a data -driven real-timeanimation machine.

The following articles by Tom Alfieriand Christine Barton discuss the softwareand system development aspects of theproject. Let me add that I think that

people issues were just as important tothe overall surrfssful outcome: Bycombining the talent and insight ofprogrammers, engineers, systemspecialists, artists, and producers, we wereable to do much more than any oneperson initially envisioned. Tradeoffs,such as whether to perform a particularfunction in hardware or software (whichsoftware?), or to have the artist create animage, became easier to resolve as welearned each other's capabilities. Thedesign and engineering developmentteams learned to adjust the displays andthe software to accommodate the timedemands of a live television broadcast.

That's how our Election Nightpresentation used some of the mostsophisticated graphic displays ever toappear live on the air.

David RabinowitzDirector, Computer ImagingNBC

David Rabinowitz joined NBC as a SeniorStaff Engineer in Technical Developmentin 1982. Prior to that, he was with ChyronCorporation as a design and projectengineer. He holds one patent in digitalvideo switching, and is a member ofSMPTE, IEEE and ACM-Siggraph. Mr.Rabinowitz is an alumnus of (SUNY) StonyBrook.

He holds an Advanced Class AmateurRadio License (call sign: WA2CZO).

TIWNWIIIIII11:11111111M 111011.111=

56RCA Engineer 30-5 Sept /Oct. 1985

T.C. Alfieri

NBC election '84: Results by design

NBC engineers, working against the clock, developed a digitalgraphics system that was a major contributor to the success ofNBC's Election Night '84 broadcast.

Election '84: Results by Design was thetitle of NBC's commitment to present anew on -air look for its election nightautomatic display system. The system tobe replaced had been in existence sincethe 1976 Presidential election. It utilizedChyron II character generators driven byspecial interfaces for the automatic displayof numerical vote totals on a coloredbackground (Fig. 1). This system hadbecome cumbersome to maintain, but moreimportantly, it could not take advantageof the talents of graphic artists. A graphicartist's ability to enhance our daily newsproductions with timely, unique, and inno-vative artwork was a quality we wantedin an automated system. The Quantel Paint

Abstract: This article describes thegraphics system used by NBC during itsbroadcast of the 1984 Presidentialelection. The system it replaced had beenin existence since the 1976 Presidentialelection. It utilized Chyron II charactergenerators driven by special interfaces forthe automatic display of numerical votetotals on a colored background Thissystem had become cumbersome tomaintain, but more importantly, wasincapable of utilizing the talents ofgraphic artists. The Quantel Paint Box,because it had shown itself to be a reliableand operationally useful tool, was selectedas the graphic system. To meet the controland development requirements, a VAXcomputer system was selected

01985 RCA Corporation.Final manuscript received June 11, 1985Reprint RE -30-5-7

POPULAR VOTEPresident 2%inCarter

628.414 42%Reagan

795.223 53%Anderson

57.395 4 %Fig. 1. Example of the type of display used in the 1980 Presidential election.

Box, because it had shown itself to be areliable and operationally useful graphicstool, was selected as the graphic system.To meet the control and developmentrequirements, a VAX computer systemwas selected.

Accomplishing our goal required theclose working relationship of many andvarious talents. Programmers, graphicartists, and hardware engineers workedfor nine months developing software, creat-ing artwork, and implementing hardware.What follows is an overview of the system.

Scope of projectThe project entailed the installation ofeight Quantel DPB 7000 Paint Boxes in atemporary facility. These would be underthe external control of two Digital Equip-ment Corporation VAX 11/750 computers.This facility would have to suffice as asoftware development laboratory, graphicartist design area, and a graphics playbackarea for the Election '84 broadcast.

Due to a tight time frame, the followingcriteria had to be met after the installationof the first three systems:

Alfieri: NBC election '84: Results by design 57

Ira et M4.

1.-

free 4.4- A.

It 4

Fig. 2. Layout of the room that contained the first four Paint Box systems.

The facility had to allow simulta-neous use by programmer analystsand graphic design artists.All future equipment installationswould not cause down time.

Any system in the facility must beavailable for use by either a pro-grammer analyst or graphic artist.

The facility must be able to providedemonstration tapes for projectstatus review.

With the facility eventually going toair, distribution of program video,monitoring feeds, timing feeds, control,and communication links was required.Control considerations were alsoneeded for two additional Paint Boxsystems installed in the WN BC graphicsarea.

Facilities

Because a temporary facility would beused to house the Paint Boxes and theirsupport equipment, all construction changeshad to be minimizixl, so some compromiseshad to be made. Two considerations thatcould not be compromised, however, werepower and air conditioning. A vacatedoffice area adjacent to a computer roommet these requirements. It was selectedfor installation of the first four systems(Fig. 2). This area afforded the additionalbenefits of a raised computer floor and a

clean air environment. The raised floorallowed for easy cable routing, thus mini-mizing cables on the floor and the need toconstruct costly overhead cable troughs. Itmust be noted that with existing fire lawsonly plenum cables can be placed in thistype of raised floor. Also, cable bundlesize was limited so as not to restrict theflow of air conditioning. To maintain aclean air environment, direct proper airflow, and eliminate the possibility of staticelectricity, all carpeting was removed. This25x14 -foot area eventually contained sixracks for equipment, four design/develop-ment areas containing four Paint Box workstations, six DEC computer terminals, aU-matic videotape machine, and a colorcamera. Conditioned power was used forall technical equipment. To improve roomlighting, a track light on a dimmer controlwas installed.

The room size was not sufficient toseparate the people involved with softwareand graphic development from the noisegenerated by the twelve 330 -megabyte diskdrives. To minimize the ambient noise aremovable, sectionalized enclosure waserected around the six racks of electronicequipment. The enclosure was constructedwith acoustic wall cover mounted inaluminum channels and suspended from aceiling track. The acoustic wall coveringhas an unfinished side that exposes thefiberglass interior. To safeguard againstparticles of fiberglass being rubbed loose,

a very fine nylon mesh was fastened overthe exposed area. To allow the heat gener-ated by the six racks of equipment toflow from the enclosure, openings wereprovided at the top and bottom of allpanels.

An additional facility was needed toaccommodate the installation of the remain-ing Paint Boxes. Using power and airconditioning once again as the primaryconsiderations, an area called 9HC wasselected. 9HC, which had housed theoriginal Chyron II election graphics displaysystem, overlooks the floor of Studio 8H,our election night studio. With only aminimal amount of development timeremaining, acoustic enclosures were notinstalled. Existing technical power forequipment, air conditioning, and lightingwere used without modification.

System requirements-development periodGraphic design requirements for Election'84 consisted of developing unique artworkfor Presidential and Congressional contests.Original artwork was also required fordepicting the responses obtained to thevarious poll and analysis questionnaires(Fig. 3). Included in this design task wasdefining the look of the animation display.Approximately 500 displays were preparedfor use during the election night presentation.

Programmer analysts developed special-ized software for use with the Paint Boxesand VAX 11/750 computers. The softwaredevelopment included writing programsfor animation, creating new fonts, VAXcontrol programs, system diagnostics, andPROM programming.

Four Paint Box systems were used forthe software and graphic design require-ments of Election '84. Each system con-tained a Paint Box, synchronizing generator,safe -area generator, RGB monitor, andfive disk drives (see Fig. 4).

Video from each system, both encodedand RGB, was routed to a jackfield. Thejackfield allowed for easy routing of videosignals from one system's output to theinput of any other system. The design ofthis facility and the equipment used allowedfor all picture transfers from system tosystem to be accomplished in either adigital or RGB format. Maintaining picturetransfers in these formats eliminated anyconcern of picture degradation due toNTSC encoding or decoding processes.

Included in the equipment was a highquality, inexpensive color television cameracapable of producing both NTSC and RGB


video signals. This camera allowed thegraphic artist to shoot either 35mm slidesor 8x11 photographs of candidates fromthe numerous House and Senate contests.The video from the camera was then cap-tured on a Paint Box system in the RGBformat. To accommodate the requirementto shoot from a 35mm slide to an 8x11photograph a no. 2 close-up adapter wasused on the standard 10x1 zoom lens.

All encoded signals used within theroom were for either comparison or record-ing purposes. The comparisons allowedgraphic artists to evaluate the picture dis-played on the RGB design monitor withan NTSC monitor reproduction. The record-ing of pictures and animation sequencesallowed the review of design conceptsand progress of the project.

The RBG signals were used for thetransfer of analog video prior to recordingonto a disk in a digital format. Once indigital form, picture information was trans-ferred either by a floppy disk, removablestorage drive (RSD), or 330 -megabyte disk.An RSD, which holds approximately 70pictures, was the main means of transferringnew information onto systems.

When a Paint Box system was used forgraphic design development, no operationalchanges were required from any manufac-turer's specified procedures.

The hardware requirements for softwaredevelopment, in addition to the equipmentpreviously mentioned, were two VAX11/750 computers and six DEC computerterminals (Fig. 5). The six terminals wereconnected to the VAX via RS232 linksrunning at 9600 baud. These terminalswere used by the programmer analyst todevelop and execute all necessary pro-grams. During software development theseprograms were resident in the VAX11/750. The Paint Boxes used withsoftware development were connected tothe VAX 11/750 via an RS232 linkrunning at 19.2k baud. The program resid-ing in the VAX was downloaded to thePaint Box by the programmer analyst,using one of the six terminals.

System requirements-night-of-airThe ten Paint Box systems used for night -

of -air were divided into four configurations(Fig. 6). Each configuration was called a"user." The four users were National Vote,employing four systems, and Summary,Poll and Analysis, and WNBC, each usingtwo systems. Each user was unique in thetype of information it displayed. The

-fl= How they voted68%

Fig. 3. Graphics produced as a result of vote totals andquestionnaires.

Fig. 4. Configuration of a Paint Box system.

hardware for each user system wasidentical.

Each system consisted of a Paint Boxwith a varying number of drives. NationalVote had three on-line 330 -megabyte diskdrives, Summary had one, Poll and Anal-ysis had two, and WNBC had one. Eachsystem within each user's group was iden-tical. The redundancy was required toallow for back-to-back displays and toaccommodate the pace at which each user

RGH MON NINE MONI, ,val

was to be incorporated into the live airproduction.

With the exception of the Poll andAnalysis systems, two VAX 11/750 com-puters, VAX A and VAX B, controlledthe picture retrieval process of the eightPaint Boxes. The four National Vote PaintBoxes, referred to as Paint Boxes 1 through4, were on VAX B. The two SummaryPaint Boxes (5 and 6) and the WNBCsystems (9 and 10) were on VAX A. A


SV5 I

r--LOPEI 5175

HI31:213 SVS 3

LIPR SVS 4

X TERMINAL 11

TERMINAL 21

TERMINAL

HTERIVIIIVAL

iFERMINAL

TERMINAL 61

CAMERA

U-IfillATIC

Fig.5. Hardware requirements for software development.

Fig. 6. Configuration of the ten Paint Box systems.

passive 2x1 switch at the inputs to VAXA and VAX B allowed the VAX SystemManager to selectively redirect the controlof any Paint Box to either VAX. Controldata was transmitted at 19.2k baud be-tween the VAX Paint Boxes via RS232or RS422 EIA standards. The two stan-dards were necessary to accommodate thevarious distances between Paint Boxes andVAX systems. RS232 should not be usedat 19.2k baud beyond 50 feet. Since only

National Vote systems were within 50feet, an RS232-to-RS422 converter wasnecessary. The converter is an in -line, self -powered device that was installed in thePaint Box racks.

On the night of air, all Paint Boxesunder the control of the VAX had pre-determined data written on their 330 -megabyte disks. This data was eitherpictures of candidates, symbols, maps, orHouse and Senate layouts, depending on

the user. The VAX obtained current votecount statistical data of all races viadedicated telephone lines from the RCAComputer Center located in Cherry Hill,New Jersey. This data was used to updatethe various displays throughout the evening.The determination of which race was tobe called and subsequently displayedinvolved four terminals. The terminals,referred to as List, Release, Control, andAnimate, were connected to the VAX viaRS232 communication links running at4800 baud.

The List operator entered data thatdetermined what races were to be aired.The Release operator released the racesstored in the VAX on a first-in/first-outbasis to the next available Paint Box, whichbrought up a static background. TheAnimate operator initiated a commandthat then modified the existing displaywith realtime-generated dynamic displayssuch as "towers" (Fig. 7), "sliding faces,"or "mushrooms" (Fig. 8). If a Paint Boxsystem failed, the Control operator's func-tion was to eliminate that Paint Box fromthose available to the Release operator.

The two systems for Poll and Analysis,also referred to as Chancellor's StatisticalStoryboard Package (CSSP), were VAX -independent. These systems were operatedby graphic artists from the air studio duringthe election night telecast. For noise andaesthetic reasons the Paint Box mainframeand associated disk drives were located inroom 9HC. The two artist work stations,safe -area generators, and ancillary computerterminals were on the studio floor. Eachwork station was connected to its PaintBox control input via an RS422 link. Thecomputer terminals required for shiftingbetween the CSSP and normal paint modeswere connected to the Paint Box via RS232links.

The encoded video output of each PaintBox was routed to distribution amplifiers.The two amplifiers per system suppliedthe required pre -equalized program andlocal monitoring feeds. Paint Box videofor the eight network systems was routedto a central equipment area for appropriatedistribution. To meet timing requirements,eight synchronization processors were used.The synchronization processors were in-stalled in a rack on the floor of studio 8H.They sent a composite black -burst referencesignal to each Paint Box. The Paint Boxesin turn sent their video to the synchroniza-tion processors for video processing. Afterthe necessary initial level and timing adjust-ments were made at each Paint Box, allfuture rehearsal and night -of -air level and

60 RCA Engineer 30-5 Sept./Oct 1985

timing corrections were made by a trans-mission engineer at the synchronizationprocessor rack. With video from the PaintBox systems entering their appropriate pro-duction switchers, the air requirements forthe project were satisfied.

The graphics displayed during the elec-tion night telecast had the primary functionof disseminating information. The qualityand manner of presentation, however, hadthe goal of capturing the viewer's attention.As with all visual effects, the wanted impacton an audience has only a limited duration.This project was developed around a con-cept that addresses this problem. We nowhave hardware that allows the imaginationof talented people to be translated intodynamic, timely, and unique displays.

A

Tom Alfieri began his NBC engineeringcareer in 1964 as a television ElectronicMaintenance Engineer. From 1972 to 1982he worked on various assignments as aDevelopment Engineer in the TechnicalDevelopment Department. Since 1982 hehas been with the Computer ImagingDepartment, working on the developmentof digital graphics systems. This articledescribes one such system.Contact him atNBCNew York, N.Y.Tacnet: 324-2786

Fig. 7. Realtime-generated dynamic "tower" display.

Fig. 8. Realtime-generated dynamic display nicknamed"mushrooms."


C. Barton

Innovative uses of computer graphics at NBCElection Day, 1984

In this companion piece to Tom Alfieri's article, Christine Bartondiscusses NBC's Election '84 coverage from the perspective ofthe software developer.

NBC wanted to present its 1984 electionnight coverage in a way that would bemore interesting and comprehensible thanthe traditional two-dimensional graphs orlists of candidates' names followed by votetotals. Graphic artists developed severalinitial storyboards that used color and gra-phic symbols in addition to the more com-mon elements of candidate name and cur-rent vote to communicate the progress ofa race. The artists also wanted to animatethe graphics as a means of directing theviewer's attention to areas of the screenthat contained the most important data.

Abstract: An Election Graphics Systemwas developed by NBC to display electionresults graphically in response to cuesfrom a director during live television cov-erage of the election on November 6,1984. The goals for the system were toconvey information about the progress of468 races on the ballot across the UnitedStates on that Election Day in an easilyunderstood format. In order to meet thesegoals, NBC had to assemble a staff fromseveral of its departments, configure newdisplay devices with several computer pro-cessors, and develop computer software tocontrol its prototype display system. Thispaper discusses issues confronted duringthis project and presents the solutionschosen.

©1985 RCA Corporation.Final manuscript received June 18. 1985Reprint RE -30-5-8

An additional consideration at this stagewas the desire to have the graphics dis-played in response to a few simple key-board strokes.

Initial equipment selectionFollowing a survey of existing turnkeydisplay systems in use within the broad-cast industry, the artists selected QuantelPaint Boxes as display devices that wouldallow them to easily produce high -qualitygraphics, and which could incorporate sec-tions of photographs and conventional artwith computer generated artwork. Thisdevice's ability to include pictures of can-didates in the election displays was viewedas an extremely useful method of com-municating important election information,namely the candidates who were running.The device was also capable of animatingthe graphics by changing sections of thevideo frame within a vertical interval.Although the Paint Box hardware wascapable of creating the sort of displays theartists and their producers desired, the turn-key software delivered with the Paint Boxesdid not always allow access to the hard-ware features in a way required to pro-duce the storyboarded displays. New soft-ware would have to be written to drivethe Paint Box hardware.

In order to make the device displayaccurate vote totals for candidates in 468races in response to an operator typingtwo or three keys, a Digital EquipmentCorporation VAX 11/750 computer was

connected to the Paint Box using a serialinterface.The interfaced VAX also allowed the newsoftware for the Paint Box to be writtenand compiled on the VAX and loadedinto the Paint Box for testing.

System developmentAlthough the original storyboards includeddrawn icons to represent candidates fromeach party, discussions with the hard-ware manufacturer made it clear that thesystem could provide an even better wayof communicating the desired information:a photograph. Frequently during the pro-ject, a much better solution to a problemwas found than had been expected byhaving artists work routinely with thetechnical staff.

Because this sort of interaction is limitedwhen services are purchased from an out-side vendor, NBC preferred to have itsgraphics system developed in-house. Thisarrangement was complicated by thedepartmental organization of NBC. Thestaff working on the development of theElection Graphics System included eightartists from Network News Graphics, sixtechnical programmers and engineers fromEngineering's Computer Imaging Depart-ment, seven programmers from the Net-work News Election Unit, and more thannine managers from all these departments.Although in-house development wasviewed favorably on this project, the workwas aided by a good working relationship


between Quantel and NBC. Just as theinteraction of NBC artists with NBC pro-grammers and engineers produced betterresults, the successful involvement ofQuantel programmers and engineers withthe NBC technical staff greatly enhancedthe project.

The development work by artists andthe technical staff produced seven story -boards that were used on the night of air.Using these storyboards, over 550 animateddisplays were prepared for possible use onthe night of air. Each storyboard requiredPaint Box software subroutines to bewritten to display the desired animation.Each of the 550 displays using these story -boards required that files containing textfonts and pieces of artwork be availableon disk drives connected to the Paint Boxthat was going to be used to construct thedesired display. All Paint Boxes were notconfigured on the night of air to be ableto produce any of the 550 displays. EachPaint Box was loaded with only a subsetof the subroutines written for the story -boards, and had access to a set of disksthat contained only a subset of the totalnumber of artwork files. This was requiredbecause the total collection of artworkfiles was too large to be included on twohard disks, which was the disk configura-tion of some of the Paint Boxes.

The system developed by NBC wasgreatly influenced by the requirements ofthe live production environment in whichthe system was finally going to be used.The display devices obviously had to bequite reliable to prevent failures while theywere displaying race results on the air.Since the displays were going to be con-trolled from a graphics control room staffedby 8 people, the displays had to be calledup by simple commands. Control roomsare a very noisy environment; requests toeach staff member in the control roomare verbal, producing a cacophony ofsound. Such an environment is not con-ducive to the concentration required forentering complex keyboard sequences. Inaddition to simple commands, the pressuresof the control room suggested assigningeach staff member in the control roomonly one function; on a cue from an asso-ciate director he need only type one shortcommand on his keyboard, instead of selec-ting one command among many.

The graphics system on the nightof air

The hardware configuration used for NBC'sPaint Box graphics system on the night of

VIDEO JACKFIELD

Fig. 1. Hardware block diagram.

air is shown in Fig. 1. Eight Paint Boxeswere controlled by 2 VAX 11/750 proces-sors. An IBM mainframe that collectedelection results was connected to bothVaxes to allow automatic updating of votetotals. In addition, two Paint Boxes oper-ated on the night of air independently ofVAX control, using software developed atNBC. This hardware configuration doesnot include conventional turnkey displaydevices used by NBC on election night;these devices included Paint Boxes, DubnerCBG boxes, ADDA still stores and tapemachines that were run by conventionalsoftware and control methods.

The software used on election night isshown in Fig. 2. Four categories of soft-ware and disk files, also termed "users,"were created for the night of air. Eachuser category consisted of 68000 software,Paint Box disk files, VAX software, and

VMS disk files. The 68000 software thatcontrolled a Paint Box was written inConcurrent Pascal. The VAX software writ-ten under the VMS operating system inPascal also consisted of a number of inde-pendently scheduled processes. Directoriesfor disk files on both the Paint Box andthe VAX were maintained on the VAX,so that the VAX processes had access toall information required for control of thePaint Box displays. The software and diskfiles were organized so that additions ofnew displays or modifications of artworkfor existing displays impacted the diskfiles only; modifications to the motionsdefined by a storyboard were the onlychanges that required software changes.Data packets passed between the softwareprocesses running on the VAX and the68000 carried parameters that fleshed outthe generic storyboard into a display that

Barton: Innovative uses of computer graphics at NBC-Election Day. 1984 63

PAINT BOX 1

68000 PROCESSES

BACKGROUND

PROCESS

PANIC

SIGNALHANDLER

COMMANDMESSAGE

HANDLER

KEY :

- SERIAL CABLE

VMS MAILBOX

VMS GLOBAL

SECTION

DISK

TRANSFER

PROCESS

DIAGNOSTIC

PROCESS

COMMUNICATIONPACKET

HANDLER

PAINTBOX

CONTROL

PROCESS

1

PAINT BOX 2

I PROCESSES

SIMILAR TO

PAINTBOX 1 I

PAINT BOX 3

( PROCESSES

SIMILAR TO

PAINTBOX 11

PAINTBOX PAINTBOX PAINTBOX

CONTROL CONTROL CONTROL

PROCESS PROCESS PROCESS

2 3 4

CURRENT GRAPHICS

VOTE TOTALS CONTROL

FILE PROCESS

LIST

PROCESS

OPERATOR

RELEASE

PROCESS

OPERATOR

RUNTIME SOFTWARE FOR NATIONAL VOTE USER

ELECTION NIGHT 1984

OPERATOR

VAX I VMS

PROCESSES

CONTROL

PROCESS

OPERATOR

PAINT BOX 4

(PROCESSES

SIMILAR TO

PAINTBOX 1)

Fig. 2. Runtime software for national vote user, election night, 1984.

showed the results of a particular race.These data packets typically included cur-rent vote totals, disk file locations, anddisplay -related coordinates. An emphasisin the software development was on generalsoftware that was data -driven wheneverpossible. This was required because thesoftware coding, compilation, and loadphase was time consuming, while changesto disk files were in general easier. Anotheremphasis in the software was its organiza-tion into independent processes and mod-ules that implemented logically disjointfunctions. Hence the VAX software inFig. 2 for any user category consisted ofat least 6 VAX processes and 6 Paint Boxprocesses.

Within each user category the operator

FLUSH

PROCESS

OPERATOR

interface to the display system used fourprocesses: List, Release, Animate, and Con-trol. The List operator entered three -orfour -character codes on a VAX terminalfor races that were considered newsworthyand would soon be discussed by the newsstaff. The Release operator released thenext race code in the list just discussed toan idle Paint Box. The Control operatorwould remove a Paint Box that was failingfrom use until it had been repaired. TheAnimate operator would cause a PaintBox that had previously been preparedfor use by a Release to be animated onthe air.

The operational environment used onthe night of air is shown in Figs. 3, 4, and5. The control room for the network's

election coverage was located near thestudio. Local news used a separate controlroom near their studio. A single graphicscontrol room was located several floorsaway from both of these control rooms.The ten Paint Boxes, two Vaxes, and IBMdata gathering equipment used in the sys-tem described here were located in fiveareas at NBC, including the network studioand four equipment rooms separated byseveral floors. The IBM mainframe waslocated in Cherry Hill, N.J. The director,several associate directors, and the Listoperator for network coverage worked inthe main control room on the night of air;the Release, Animate and Control opera-tors for network coverage worked in thegraphics control room with another asso-ciate director. Cues for graphics were givenusing telephone lines between an associatedirector in the main control room and theassociate director in the graphics controlroom; the graphics control room associatedirector than passed the cues verbally tothe Release and Animate operators sittingnext to her. The Control operator in thegraphics control room was aided by twotechnical staff members, who were con-nected by phone lines to the equipmentrooms. These technical support personnelcoordinated repair procedures for failingequipment with staff in the equipmentrooms; when a device was again opera-tional, the support personnel informed theControl operator that it could be includedfor on -air use.

Precautions against failing displays onelection night included redundant equip-ment and support personnel trained toremove and repair failing equipment whileon air. Most of the hardware and softwareran quite reliably in practice, however.The election day use of the displays in-cluded hours of rehearsal and test, as wellas three broadcasts. During this 30 -hourperiod there were about a half dozen minorfailures, all of which were easily recover-able and did not produce undesirable on -air displays.

To aid in the general awareness of allthe graphics control room staff of theprogress of the show, two audio channelswere broadcast over speakers in the graph-ics control room. One channel carried theaudio from an open microphone in themain control room. This channel carried,among other voices, all cues from theshow's director, and the executive pro-ducer's instructions to the production staffas to the desired content of upcomingprogram segments. The second audio chan-nel carried the program as it was broadcast.


NET wafin

STUDIO(8,C

NETWORK

CONTROL ROOM

(8H CONTROL)

LOCAL

STUDIO

(6A)

LOCAL

CONTROL ROOM

)6A CONTROL)

GRAPHICS

CONTROL ROOM

(31( CONTROL)

NATIONAL VOTE

USER EQUIPMENT

P81, PB 2, PB 3, P84

(1700 PAINT BOX ROOM

SUMMARY USER

EQUIPMENT ROOM

PB 5,PB 6(9HC I

POLL AND ANALYSISUSER EQUIPMENT

PB 7,PB 8(8H)

CL

EcUIPMEN"

P8 9 PE 1(.

',9241

VAX

EQUIPMENT ROOM

(17001

IBM

EQUIPMENT(CHERRY HICL,N 1)

GRAPHICS OPERATIONS

COMMUNICATIONS LINES

ELECTION NIGHT 1984

Fig. 3. Graphics operations communications lines, election night 1984.

Other monitoring equipment includedvideo monitors for all ten paintboxes, twoADDA still stores, a Dubner CBG box,and three monitors carrying election cover-age by other broadcasters.

SchedulingThe following are project milestones ofThe Election Graphics System:

Project defined-April, 1983Equipment selected-July, 1983Development equipment and staff inplace-November, 1983

Initial software definition-December,1983

Initial software available for test-April,1984

Working storyboard definitions-June,1984

Initial Election Graphics System re-hearsals-October, 1984Finalized storyboards, artwork, soft-ware, hardware-November 3, 1984Night of air-November 6, 1984

Several displays produced by NBC's Elec-tion Graphics System are shown in Figs.

6 and 7. The displays were used in NBC'scoverage of the Presidential race onNovember 6, 1984.

ConclusionsThe Election Graphics System was veryeffective for several reasons. There was anemphasis throughout the graphic designphases of the project on graphics thatcommunicate specific election data, insteadof on graphics intended only to grab atten-tion or showcase graphics techniques. Pro-duction of some graphics storyboarded bydesigners could only have been done withthe type of system we had. The success ofthe graphics produced was greatly enhancedby frequent communication between gra-

Christine Barton is a member of NBC'sComputer Imaging Department, and wasactively involved in software developmentas project manager for the 1984 electionproject. Prior to work at NBC, she workedas a systems and computer graphicsprogrammer in several research groups,including the NASA/Ames ILLIAC IVparallel computer project, three Evans andSutherland Computer Corporationcomputer -generated imagery projects,and New York Institute of Technology'sComputer Graphics Laboratory.

Ms. Barton received a BS inMathematics and Computer Science fromthe University of Illinois in 1971 and an MSin Mathematics from New York Universityin 1981. She is a member of ACM, IEEE,and ACM Siggraph.Contact her at:NBCNew York, N.Y.Tacnet: 324-5390

Barton: Innovative uses of computer graphics at NBC-Election Day, 1984 65

phic designers and techncial personnel,even when these groups worked in differ-ent corporate departments or for outsidevendors. An additional consideration thatcontributed to the success of the night -of -air operation of the system was an empha-sis on a simple operator interface appro-priate for the election night operationalenvironment.

AcknowledgmentsThe following people were involved onthe Election Graphics System at NBC:Network News design artists wereJohn Brainard, Anthony Franqueira, StevenGiangrasso, Kevin Hale, Michael Katz,Jeff Maltz, and Larry Wasserman; theComputer Imaging staff was DavidRabinowitz, Christine Barton, ThomasAlfieri, Diane Piencinski, Andrew Siegel,and Maria Cantatore; the Network NewsElection Unit data collection staff wasJoel High, William Niemi, Patricia King,Richard Ho, Nieba Jones, Ellen Mendelson,and Theresa Chan; Quantel programmerswere Robert Long and Simon Carter;Management on the project includedJoseph Alicastro, Joel Blumenthal, JosephBranch, Robert Brandel, Ralph Famiglietta,Sheldon Hoffman, Richard Passentino,David Schmerler, Roy Wetzel, and ThomasWolzien.

Fig.6. NBC national vote display 1984.

COMMUNI -

CATIONS

ASSOCIATE

DIRECTOR

DISPLAY

ASSOCIATE

DIRECTOR

PROGRAM

ASSOCIATE

DIRECTOR

DIRECTORTECHNICAL

DIRECTOR

LIGHTING

DIRECTOR

TECHNICAL

DIRECTOR

RELIEF

TECHNICAL SUPPORT SUPPORT EXECUTIVE SUPPORT REMOTE

SUPERVISOR PERSONNEL PERSONNEL PRODUCERPRODUCER

PERSONNEL PRODUCER

GRAPHICS

LIST

OPERATOR

Fig. 4. Network control room seating, election night 1984.

SUPPORT

PERSONNEL

RELIEF

GRAPHICS

ANIMATE

OPERATOR

GRAPHICS

RELEASE

OPERATOR

ASSOCIATE

DIRECTOR

TECHNICAL

DIRECTOR

TECHNICAL

SUPPORT

GRAPHICS

CONTROL

OPERATOR

3t(

MANAGER

TECHNICAL

SUPPORTRELIEF

Fig. 5. Graphics control room seating, election night 1984.

Fig. 7. NBC electoral vote display 1984.

66 RCA Engineer 30-5 Sept. / Oct 1985

L. Arlan

Techniques for enhancement of interactivecolor graphics for CI systems

Today's state-of-the-art battlefield intelligence gathering toolsnecessitate the use of recent developments in video processingto enhance the man/machine interface of colorgraphicworkstations.

The battlefield of today contains an ever-increasing proliferation of sophisticated sen-sors and jammers. Old methods of handlingand processing battlefield intelligenceinformation-paper records and mapboardoverlays with grease pencils-cannot copewith today's environment in which there

Abstract: The use of automated supportfor command control and communicationsintelligence C31 systems is rapidlyexpanding with the increased use ofcomputer technology, colorgraphicworkstations, and electronic storagemedia Newly emerging technologiesprovide practical solutions to supportapplications such as tactical intelligenceanalysis through the use of interactivecolorgraphics and enhancement of theman/machine interface. This paperdiscusses tactical intelligence require-ments, the overall analysis approach, andthe new technologies available forenhanced operational support. Theenhancement of the man/machineinterface through effective high -definitiondisplay of color maps is also addressed

©1985 RCA Corporation.Final manuscript received May 24, 1985Reprint RE -30-5-9

are thousands of sensor and human intelli-gence reports per hour. The command,control, and communications intelligence(0) system of today must provide ahighly automated management systeminvolving automatic data processing (ADP)to assist in processing highly dynamic tasks.The system must receive many types ofsensor reports and assist the intelligenceanalyst in the correlation and fusion of realtime data. It is vital that the analyst commu-nicate with this system in a highly efficientmanner through elective use of tools basedon the feasible and practical application ofemerging technologies.

In order to adequately define these toolsand maximize the utilization of these tech-nologies, it is first necessary to determinethe intelligence requirements and understandthe analysis approach.

System requirements

The principal requirement of a tactical C3Isystem is that it provide battlefield analysisin a timely and accurate manner by integrat-ing all sources of intelligence. These sourcesinclude signal intelligence (SIGINT), whichis the fusion of electronic and communica-tion intelligence (ELINT and COMINT);image intelligence (IMINT), which includes

imagery from sources such as radar, photo,infrared (RADINT, PHOTINT, SLAR,IR); and human intelligence (HUMINT)from foward observers, reconnaissance, andprisoner interrogation. The function of thesystem is to provide tactical commanderswith essential elements of information formission planning by locating, identifying,and tracking the enemy as quickly and asaccurately as possible. This informationallows the intelligence team to analyze theenemy in the context of the battlefield withall its constraints such as weather, visibility,and surface conditions. It is then possibleto estimate the enemy's capability andsituation so that predictions can be madeas to his intentions. Tactical reports areprovided in near real time to the supportedcommander so that he can "see" the battle-field more effectively and respond appropri-ately with electronic warfare options. Intel-ligence analysis must also identify andwarn of threats that require immediateresponse.

The use of highly -automated support inreporting, displaying, and disseminatingspecific location, time, and activity informa-tion from a large amount of data meansthat fewer people are in the loop, therebyproviding useful intelligence in a moretimely and accurate manner.

RCA Engineer 30-5 Sept. / Oct. 1985 67

Fig. 1. RCA military colorgraphics workstation. This is an upgraded version ofan AN/TSQ-130(V), showing alphanumeric database information on the rightscreen and electronically -stored maps with colorgraphic overlay displayed onthe left screen.

HE AO SETvOLUMEAFT

0EMT COMM

BUSv BUS, Busv BD S v

042

BuSV

u 0

1 COM SW8 OPR RW1 OPR

7?:

INASEPANT

IBC AVSVID.0 MARA

W A

BoSv

LL

Bo S v EILLSv

NIONHOLD RELEASE

RFC

0 0 0 0

CACI 2MIL 1,3B n"..1

0 0

ABORTLOC OR

LIN "LIZ!RUMCNA.

SCROLMOM.

MirSCRIOL

PRINTMAPSEL

fMCNXNN YOREMODE

RILEWORD

2 3

4 5 6

9

0

Fig. 2. RCA colorgraphic keyboard. Secure and non -secure communicationcontrols are on the upper portion. Special function keys are in the middle tworows, and the standard keyboard is at the bottom, with commonly used word -select keys indicated.

Analysis approachThe analyst operating in today's tacticalenvironment has a complex array of dataresources that provide information regard-ing the enemy's movers, shooters, andemitters. Examples of these resources areTEAMPACK, TRAILBLAZER, AN/TRQ-32, QUICKFIX, and other directsupport systems and sensors. It is theanalyst's objective, through use of anautomated system, to correlate the data

from this myriad of collection resourcesand present this data in a manner thatdepicts the battlefield situation. This canbe best accomplished by plotting the enemycommand and control nodes on a map sothat a pattern can be formed from theenemy's electronic, infrared, acoustic, andother emissions. In addition, the analystcan quickly retrieve and review historicaltrends about the enemy from a database,which may help to predict the enemy's

intentions when compared to the presentscenario. The commander and his staff canthen plan an appropriate response anddeal in a timely manner with immediatethreats.

Colorgraphic workstationThe primary tools of the modern C3I analystare communications, an automated data-base, and a military colorgraphicworkstation.

The communication subsystem cantypically provide the analyst with an arrayof voice, teletype, and data links in either asecure or nonsecure mode. Each of thesefacilities can be via wire or radio, and theanalyst can communicate with otheroperators by voice or display screen.

The colorgraphic workstation typicallycontains a microcomputer that is loadedwith the software necessary to support thetype of functions required, a memory tobuffer incoming and outgoing messages(and is linked to a host computer), andmass memory system containing the data-base. Figure 1 shows an adaptation of anAN/TSQ-130(V) workstation that has beenupgraded with high -resolution color mon-itors and a color graphics generator. Theoperator has complete access to all com-munications, database, and algorithmicsupport through the control keyboard andassociated headset. Color graphics areavailable on either or both monitors andelectronically stored maps can be retrieved,displayed, and overlayed with color graphics.

The colorgraphic keyboard, shown inFig. 2, illustrates the controls available tothe intelligence analyst. The upper portionincludes the selection and control of bothsecure and non -secure communications. Atouch-tone dial set is shown on the right.The next two rows are the functionalmode select pushbuttons with lights thatindicate the selected mode. A trackballgraphic cursor control is on the right andthe cluster of eight controls next to it arethose included for interactive colorgraphicfunctions. The lower portion is a standardASCII control keyboard including numer-ical entry pad on the right, cursor controlson the left, and some unique word -selectkeys for frequently used words such asunit, location, date/time of intercept,activity, frequency, call sign, and casenotation.

Figure 3 illustrates some typical militarysymbols that are used to overlay standardArmy topographic maps. Under control ofthe operator or by graphic command fromthe host computer interface, a variety of


military symbols, alphanumeric annotations,and conic sections may be selected bymenus from display lists and non-destruc-tively overlaid on standard map back-grounds in a prioritized manner. Variousattributes such as color, shape, size andblinking can be employed for rapid andaccurate identification. The operator cantypically draw a forward line of troops(FLOT) or place selected symbols usingthe cursor that is controlled by a trackballor other equivalent device such as a graphicstablet or mouse.

Map backgroundsTwo basic types of map backgrounds canbe stored, retrieved, and displayed elec-tronically to replace the old paper mapwith grease pencil overlays. One type is adigital -based decluttered map shown inFig. 4, which can be used for navigationpurposes and is useful for showing majorroadways, rivers, and boundaries. Thesecond type is the standard Army topo-graphic map that contains a much higherlevel of information content for detailedCI analysis. In addition, analysts are moreaccustomed to working with topographicpaper maps. The information content ofthese maps, however, requires mass storagecapacity and high -resolution display/processing for legibility. Figure 5 is ahypothetical example of a topographic mapwith simulated graphic overlay showingelements of the enemy's forces positionedalong the Schmalkalden River, with keybridges indicated where river crossings areanticipated. Friendly forces are positionedon the high ground in the upper left cornerof the map. Analysis has been accomplishedthrough reception of enemy emanationsand visual observation of the enemy force'skey -element positional moves.

Map storage media

In order to select the appropriate mapstorage media using existing technology, itis first necessary to analyze the capacityrequirements of standard topographic mapswhile considering the man -machine interfacerequirements of resolution and informationtransfer.

Figure 6a illustrates a typical 18 -inchsquare map containing minimum line widthsof .004 inch. If an approximation is madethat at least one pixel is required to supportand reproduce this minimum line width,then the entire map consists of 2X107 pixels.Secondly, if a high -resolution display isused that displays 480 active lines with a

Fig. 3. Typical military symbols. These are used as graphic overlay elements ondisplayed maps to depict battlefield situations.

Fig. 4. Digital map. This is a sample of a digitized map stored in a standardcomputer memory and displayed on a CRT. Maps of this type are useful indepicting uncluttered scenes consisting of boundaries, roads, and rivers.

horizontal displayed -resolution of 640 pixelsper line, a single displayed image willcontain a maximum of 3 X 105 pixels, asillustrated in Fig. 6b. The total number ofindividual images required to represent theoriginal map, then, is approximately 64.The optical magnification from paper mapto displayed image is approximately 4:1with full quality reproduction, and eachmap segment is presented to the operatorwith high legibility at a normal viewingdistance of la inches from the display.

If a typical map with a scale of 1:50,000is employed, the displayed map segmentrepresents a tactical area of approximately

eight square kilometers based on the 4:1optical magnification. In reality, a highlylegible image covering an area of 3X4kilometers, as shown in Fig. 5, is practicalwith enhanced video processing (opticalmagnification 3:1), which results in a mapconsisting of 37 map segments.

The storage requirements for a usefuloperational area can be illustrated byconsidering an area covering central Europe,which covers 696,000 square miles(1,200X580 miles). Approximately 3,500maps of 1:50,000 scale are required, result-ing in a total storage requirement of 129,500images. If a 50 -percent overlap is required

L. Arlan: Techniques for enhancement of interactive color graphics for CI systems 69

to permit any point on the original mapto be reasonably centered, the total require-ment becomes 518,000 images, or approxi-mately 1.6X101' pixels. This represents avery high storage capacity requirementcoupled with a requirement for rapidrandom access. A quick review of storagedensities available in state-of-the-art tech-nologies is shown in Table I. As illustrated,the only digital storage media thatapproaches the requirement of 1.6X10"pixels is the newly emerging optical datadisk. Assuming 8 bits/pixel for a digitizedmap in each of three primary colors, atotal of 19 disks would be required. Incomparison, an analog storage media of

either tape or laser videodisc containingNTSC encoded color imagery has a muchhigher capacity. Tape is not consideredbecause of its long random access time.The laser videodisc is the obvious choice,and could supply the required capacitywith six discs. In addition, any map can berecalled from a given disc within 5 secondsusing commercially available, low-costplayers employing a computer interfacefor automated control.

Enhanced video processing

To maximize the information transfer andfunctional efficiency of the intelligence

Fig. 5. Topographic map with graphic overlay. This is a photograph of atopographic map containing high information content and displayed on a high -resolution display. Graphic overlays are used to depict enemy elements and keypositional forces.

18"

1:50.000 MAP

18" 0.004"/LINE

I PIXEL/LINE

> 2 X 107 PIXELS/MAP

Fig. 6A. Map pixel density. This illustrates the total numberof pixels required to produce a typical cartographic mapof approximately 18x18 inches.

operation, it is important to consider theman -machine interface. One of the mostvital workstation factors is the resolutionpresented to the operator by the display. Ahighly legible image can reduce fatigue,provide a high degree of informationtransfer to the analyst, and maximize theresponse time through rapid absorptionand comprehension of information.

As previously described, the standardvideodisc system provides a currently avail-able source of rapid access cartographicinformation for the C'l analyst. TheNational Television Systems Committee(NTSC) format was defined nearly 30years ago, but until recently has not beenprocessed to extract all of the inherentcolor resolution.

New techniques involving comb filterscan substantially improve the image qualityand legibility of cartographic maps retrievedfrom videodisc storage and displayed atthe colorgraphic workstation.

Table I. Recording densities ofdifferent media.

DigitalCassette -106 to 10' bitsFloppy disk -2X107 bitsHigh -density 9200 -foot tape -3X10'° bits

Optical data disk-2X10" bits(two sides, 12 -inch diameter)

Analog9200 -foot tape-3X10" pixels (8

bits/pixel)

Laser video disk-2.8X10wpixels (8 bits/pixel)

10.6"

13" HI RESOLUTION CRT

525 TVL NTSC 1480 ACTIVE)

0.016" LINE SPACING

640 HORIZ. PIXELS

NUMBER OF SUBIMAGES -TOTAL DISPLAY PIXELS

TOTAL MAP PIXELS

OPTICAL MAGNIFICATION = 4:1

3 X 105 PIXELS/DISPLAY

- 64

Fig. 6B. Displayed map pixel density. The number of pixelsresolvable on a 13 -inch diagonal display is shown here.This determines the number of sub -images required tolegibly display the full map shown in Fig. 6a.


The fundamentals of color televisionsignals are described in the literature, 1'2and will not be repeated here except for afew brief concepts that will help to describethe signal processing employed for enhancedmap display. The NTSC encoded signalconsists of two components; the luminancesignal (Y) and the chrominance or colorsignal (C). The luminance signal representsthe black -and -white content with full band-width information out to 4.5 MHz. Thechrominance signal consists of two phase -modulated components; I (in -phase), andQ (quadrature), which are referenced to acolor sub -carrier frequency. The color sub -carrier frequency (3.579545 MHz) waschosen to be an odd multiple of the half-line frequency (455xhalf-line). Since a line -scanned television system is a two-dimen-sional sampled system of an image, theluminance signal contains 60 -Hz (verticalsample rate) power sidebands that areclustered at the horizontal line -rate sampleand at all of the line -rate harmonics withan amplitude characteristic shown in Fig.7. The chrominance portion of the com-posite signal also produces power clustersin the same fashion, but these are interleavedwith the luminance clusters because themodulation reference frequency was chosenas an odd half harmonic of the horizontalrate. The relative amplitude and interlacedstructure of the Y and C components areshown throughout the entire bandwidth,as illustrated in Fig. 7. The specific band-widths allocated for Y, I, and Q in theNTSC standards are shown in Fig. 8. Ascan be seen, the luminance signal occupiesthe full 4.5 -MHz bandwidth, but the I andQ components occupy limited bandwidthsof approximately 1.5 MHz and 0.5 MHz,respectively, after demodulation.

Until recently, separation of the chromi-nance from the luminance signal was accom-plished by bandwidth filtering. A 3.58 -MHz trap is used to take the chrominancesignal out of the composite signal, whichleaves the luminance. The chrominancesignal is developed by using a 1 -MHzbandwidth filter that is bandpass-centeredaround 3.58 MHz. Unfortunately, this doesnot extract the interleaved luminance signaltrapped in the bandpass chrominance signaland the chrominance in the bandrejectedluminance. The result is the familiar dotcrawl on vertical edges and color aliasingeffects seen on home television receiversresulting from coherent high -frequencyspatial patterns. In addition, bandrejectionof the luminance signal substantially reducesthe black -and -white picture resolution.

To more ideally separate the luminance

4

RELATIVE

AMPLITUDE

IH

-f

VCOMPONENT

WITH 30& 613 Hz

SIOEBANDS

"4- 2

IH

T-/ 3.58iI 1H -41 MHz

C \COMPONENT

WITH 30& 60 Hz I

SIDEBANDS

Fig. 7. Relative amplitudes of NTSC component signals. These are the relativeamplitudes of the luminance and chrominance components of the NTSC signal.The exploded view illustrates the nature of the interleaved energy clusters ofthese components.

LUMINANCE

4.2 MHzI SIGNAL

I) SIGNAL 13 58 MHz

I COLOR

SUB CARRIER

2 3

FREQUENCY IN MHzSPECTRUM OF NTSC VIDED SIGNAL

Fig. 8. NTSC Y, 1, & Q bandwidths. This shows the allocated bandwidths of theluminance (Y), the in -phase (I), and the quadrature (Q) components of thechrominance signal. The cross -hatched areas illustrate the overlapping natureof the component spectra, which makes separation by simple bandwidth filteringundesirable.

and chrominance interleaved signals, a filterhaving peaks and nulls at the various energyclusters throughout the entire bandwidthand having a linear phase characteristic isrequired. The overall filter response has a"comb" shape, and is thus referred to as acomb filter. For the past several years ithas been possible to make practical combfilters by implementing charge -coupleddevice (CCD) delay lines, and this hasresulted in higher resolution and reducedcross -color effects. Figure 9 illustrates apractical filter circuit using a CCD delayline to achieve separation of the luminanceand chrominance components throughcomb filtering. If the input to the circuit is

an applied unit -impulse function S(t), itcan be shown by Laplace transformationthat the output of the luminance channel1/1(w) and the output of the chrominancechannel H2(w) can be expressed by

I Hi (0))1 = 21 cos T/21

1/2(0.)1 = 21 sin T/21

where T = the delay time

Therefore, If/1(01 and 1 H2(w)1 have theshape of full -wave -rectified sinusoids. If Tis made equal to 1/fH, where fH is thehorizontal scanning frequency, the 1 Hi (f)1has a maxima at the harmonics of fh and

and


COMPOSITE

VIDEO

INPUT

TIME DELAY ELEMENT

T , 4

INVERTER

SUMMATIONNETWORKS

1 0 +2 0 -3 0+

OUTPUT I

COMBED4 0 -

LUMINANCE 5 0+6 0 -

OUTPUT 2

COMBED

CHROMINANCE

262 0Fig. 9. Basic comb filter for separation of luminance and chrominance. Thissimplified functional diagram shows how a time delay element is used toimplement comb filtering of the luminance and chrominance signals. Adding thein -phase luminance components separates the luminance from the compositesignal by cancelling the chrominance signal. Inversion and addition separatethe chrominance out by cancellation of the luminance component.

I 0+

2 0-

3 0+

4 O-

S

1 535 00-

63 0+

0-

0+

o-

0+

I FRAME REQUIRED FOR COLOR DECODE

BETTER RESOLUTION THAN FILTER DECODE

LOSS OF VERTICAL RESOLUTION BUT BETTER THAN LINE COMB

Fig. 11. Field combing. This shows the phase relationship of each scan linecolor subcarrier frequency in two corresponding interlace fields of a completeframe. Field combing of luminance and chrominance is achieved by using a fullfield delay.

IH2W1 has maxima at harmonics betweenharmonics of fh, thus forming the desiredcomb filter.

Another simplified way of graphicallyillustrating the line combing previouslydescribed is shown in Fig. 10. This diagramrepresents the scan lines in a single field ashypothetically viewed from the side edgeof the scanned raster. The phase relationshipof the subcarrier frequency is shown foreach scan line. If each scan line is delayedby a full line time and then compared tothe next line in real time according to thecircuit in Fig. 9, one can see that combingwill take place by cancellation-the com-ponents in one line are 180 degrees out ofphase with components in the next line. Ifboth scan lines have identical information,perfect combing takes place and full resolu-

tion with no cross -color components willbe realized. This process of line -combfiltering has been employed in commercialTV receivers for several years.

Line -comb filtering is not ideal, however,when the two scan lines being comparedhave different information content. Thisresults in imperfect combing and a loss invertical color resolution. Techniques havebeen employed to partially restore the lossof vertical resolution, but averaging twolines from two different geometric locationsof the image is not ideal.

A better approach, depicted in Fig. 11,is to perform field combing by using a fullfield delay and then comparing scan linesfrom two interlaced fields that are geo-graphically close together. Finally, the idealconfiguration is a full -frame comb that

SINGLE FIELD

Fig. 10. Single -field phase relationshipof scan lines. This diagram illustratesthe relative phase relationship of thecolor subcarrier frequency in eachscan line of a single NTSC field, andhelps to explain the process of theline -comb filtering shown in Fig. 9when the delay time is one line time.

now allows the combing process to use thesame geographically -located lines withidentical information content as shown inFig. 12. Frame combing can be accom-plished by digitizing a full frame ofstoring the frame in a memory, and thenreading this frame out of memory insynchronization with the next frame beinggenerated. If motion in the scene exists,then the full -frame delay will obviouslyproduce undesirable effects by comparingtwo dissimilar frames. In this application,however, the maps stored on videodisc arestationary images, and two identical framescan be stored that will allow full -framecombing of the composite signal. In additionto full color resolution with no cross -colorproducts, the signal-to-noise ratio is

improved over that from a single frame,and defects in the videodisc such as localizeddropouts are reduced.

Figure 13 is a photograph of an NTSC-encoded map that has been frame -combedand displayed on a 13 -inch diagonal, fine -pitch -mask CRT. The color graphics depictthe locations of elements of the 39thMotorized Rifle Battalion and the forwardline of troops (FLOT). The graphics areoverlayed on the map background in aprioritized, non-destructive manner. Thedirect -viewed quality of this high -resolutioncolor image is much greater than can berepresented by the reduced -size black -and -white photograph adjacent to it. Thisimagery was produced using a frame -combsystem developed at RCA Laboratories.


Colorgraphic workstationThe full colorgraphic workstation for ClIanalysis is illustrated by the simplified blockdiagram in Fig. 14. The dual -screen work-station can be utilized to present graphic-overlayed maps on either or both screens.A typical mode, shown in Fig. 1, employsa map with a tactical operational scenariooverlay on one screen and alphanumericson the other screen that provide informationfrom the database pertaining to operator -selected inquiries.

A particular map is act sled from thevideodisc player by a given lattitude/longitudinal coordinate and map scale factorcommand. The video signal is digitized byan 8 -bit A/D converter sampling the NTSCvideo at a 14.3 -MHz clock rate. The clockis derived from a circuit that phase locksto four times the 3.58 -MHz color burst carrier,providing precise registration accuracy. Italso provides approximately 750 samplesper active line time to more than supportthe 4.5 -MHz bandwidth, as well as quadra-ture sampling of the encoded chroma forease in subsequent digital decoding. Inaddition, horizontal and vertical sync signalsare stripped from the video and used bythe address control logic for writing thedigitized video into the two -frame (four -field) memory. Digital frame grabbingallows a single videodisc player to serveeither a single or group of display proces-sors, and provides the two frames requiredfor frame combing. Once loaded, thememory readout is synchronized to thegraphic display generator. Readout of thetwo frames requires two lines of videobeing clocked out in accordance with framecombing, as shown in Fig. 12. The readoutcan be selected to form a standard 525 -line, 2:1 -interlaced format with a 60 -Hzfield and 30 -Hz frame rate, or a progressivescan can be performed, resulting in a 60 -Hz non -interlaced frame rate.

Simple digital addition and differencingof the two frames provides luminance andchrominance signals, as described in thecombing operation. Further decoding ofthe 3.58 -MHz chrominance signal into itsin -phase (I) and quadrature (Q) componentsis accomplished digitally. The I signal isobtained by selecting the chrominance sam-ples with the 4x color subcarrier clockand low-pass filtering (2.5 MHz) the output.The Q signal is similarly obtained by select-ing the chrominance samples with the sameclock shifted by one cycle (correspondingto 90 -degree phase shift for the 3.5 -MHzcolor subcarrier), and low-pass filtering(2.0 MHz) the output. The signals arethen converted to analog, and simple linear

263 0+ o-I 0+ 0-

264 0_ 0+? 0-

3 0+0+

0+O NTSC

0-LUMAAID IBM MID 0

0- 0.0- 0.

0+ o -0+ 0- OCHROMA

525 O

262 O

F 13 F3 14

2 FRAMES REQUIRED FOR FULL COLOR OECOOE

FULL BANDWIDTH HORIZONTAL RESOLUTION

NO LOSS OF VERTICAL RESOLUTION

Fig. 12. Frame combing. The scan line phase relationship in four fields is shownhere to illustrate the effectiveness of frame combing by comparison of scanlines from identical geometric locations of the image. Two full frames of thesame image information are required for frame combing.

Fig. 13. Frame -combed map with graphic overlay. This photograph was takenfrom a high -resolution color display of a comb -processed map image that wasretrieved from videodisc storage. Color resolution of high quality is achievedwith no artifacts or dot crawl due to cross -color products. Graphic overlayedsymbols are used to depict a military situation.

matrixing of the Y, I, and Q signals pro-duces the required R, G, and B signals.

The output of the colorgraphic generatoris synchronized to the readout from theframe -comb memory system, and is mergedwith the analog map video in the videokeyer circuit. This circuit prioritizes thegraphic overlay in a non-destructive mannerso that changes can be made in the graphicoverlay as required without having to recallthe map from the videodisc each time achange is made.

Any one of many graphic generators

existing on the market today can be utilized,but it must be capable of gen-locking toexternal sync and providing a control I/Ofor the videodisc player. It must also containsoftware primitives for graphic generationsuch as circles and vectors, provide a displaylist of required graphic symbols, and bedual -ported for the left and right screendisplays. Other features such as GraphicKernel System (GKS) software implementa-tion, resolution, processing speed, localmemory, DMA, color pallette, and architec-tural structure are application -dependent.


KEYBOARD

GRAPHICS

TABLET

/ RS 232

VIDEO DISC

PLAYER

HOST

COMPUTER

V

NTSC VIDEOCOLOR

BURSTFEVER

NTSC VIC0

LP FILTER

PHASE

DETECTOR

SYNCSEPARATOR

VERT CALSYNC

COLOR BURSTGATE CONTROL

HORIZONTAL

SYNC

RAOIN OUT

,PROCESSOR

LOCAL

MEMORY

DISPLAY

CONTROLLER

ADDRESS

CONTROL

LOGIC

A/0

V.:

: I

143 MHz CLOCK

PHASE

,'ER

liA7E

FIELD NO I

240 750

FIELD NO 2240 750

-11

FIELD NO 3240 750

FIELD NO 4240 750

2 FRAME RAMREFRESH MEMORY

DIGITIZERCONTROL

MEMORY

REA OUTCON ROL

SYNC

MASTERSYNC

GEN

GRAPHIC

DISPLAY

LUMA

AOD

SUBTRACT

8/

iCIIROMA

DECODE

VIOE1

I VIDEO

0/A GAIN

8 CONTROL

MAT HIS

lbw

IMP

COMPOSITE SYNC

VIDEO NEVER

Fig. 14. Block diagram of colorgraphic workstation. Thissimplified block diagram of a colorgraphic workstationshows the techniques required to phase -lock video froma videodisc player for proper registration of the digitizedvideo. Subsequent frame comb processing is achieved

ConclusionThe use of automated tools for CI analysisis imperative in today's highly complexbattlefield environment. Automation mustbe applied to the maximum extent possible,and is exemplified by the use of interactivegraphics tools and automated databases.Techniques available to improve the man -machine interface have been described andare available with existing technology andoff -the -shelf components. These tools, whensupplemented with training, exercises, tests,and creative analysis utilizing artificial intel-ligence, can facilitate modern C I systemsfor effective tactical intelligence operation.

ReferencesI. D.H. Pritchard, "U.S. color television fundamentals-

a review," RCA Engineer, Vol.29, No. 6 (Nov./Dec.1984).

2. R.T. Keen, "Encoding and decoding WidebandIchroma signals," RCA Engineer, Vol. 29, No. 6

(Nov./Dec. 1984).

IGH RESOLUTION

COLOR

DISR,AY

from the memory, which refreshes the display in syn-chronism with the microprocessor -controlled colorgraphicgeneration. Non-destructive and prioritized graphic overlayis achieved via the video keyer circuit.

Lionel Arlan is Manager of DesignEngineering, Command and ControlEngineering at Automated SystemsDivision. He received his BEE fromClarkson College of Technology in 1954,and his MS in Electrical Engineering fromDrexel Institute of Technology in 1961. Mr.Arlan has been involved with electro-optics and related weapon systems sincejoining RCA in 1954. His responsibilitieshave included design, development, andevaluation of special television cameraapplication including slow -scan and lowlight level tv systems. For the past severalyears, Mr. Arlan has been responsible fordevelopment of techniques for enhancingcolorgraphic display systems for retrievingand processing high -resolution cartographicmaps from videodiscs.

Mr. Arlan is currently responsible forlaser rangefinder programs, the MinutemanDrums program, and IR&D directed towardfurther enhancement of CI colorgraphicdisplay terminals.Contact him at:Automated Systems DivisionBurlington. Mass.Tacnet: 229-3729


M. Leahy

Plasma etching for integrated circuit fabrication

The demand for smaller, more complex integrated circuits isspurring increased interest in dry etching, a process with muchpotential and an intriguing set of challenges.

Ahuman hair is a huge object, if youare a microlithographer who prepares pat-terns of integrated circuits on semicon-ductor materials. A design as narrow as a100 -micrometer -wide human hair wouldbe simply too bulky to be of practical usein a typical integrated circuit.

The semiconductor industry routinelyproduces patterns with features muchsmaller than the merely invisible. Theseobjects, so far removed from everydayexperience, must nevertheless be made ingreat volume to exacting specifications.To accomplish this, many techniques arebrought to bear on a thin slice of silicon,in a hundred or more steps, to make thechips found in more and more of today's

Abstract: One of the newest methods formaking integrated -circuit chips is dryetching-exposing wafers to an excited -gas plasma to selectively remove materialfrom the surface. Not only is thistechnology new, but we know it isabsolutely essential to RCA's advancedcircuit fabrication plans. Most marketstudies predict that over the next fiveyears, dry etching equipment will accountfor 10 to 12 percent of the waferprocessing equipment budgets of allsemiconductor houses. This articlediscusses this new technology, and someof the reasons for the recent intenseinterest in it


products. One of the newest methods formaking these chips is dry etching-expos-ing wafers to an excited -gas plasma toselectively remove material from thesurface.

Not only is this technology new, butwe know it is absolutely essential to RCA'sadvanced circuit fabrication plans. Mostmarket studies predict that over the nextfive years, dry etching equipment willaccount for 10 to 12 percent of the waferprocessing equipment budgets of all semi-conductor houses. That percentage willamount to about 2.5 billion dollars from1984 to 1988 inclusively, and points tomajor efforts in this technology by almostevery manufacturer. What are the reasonsfor this intense activity in the field?

To answer that question, we shouldexamine the history of the integrated circuit.New designs are created by increasing thenumber of devices on a chip and decreasingthe sizes of the individual devices. Wherethe smallest object in a 1 -kilobit randomaccess memory (1 k RAM) might havebeen 10 micrometers across, the latestgeneration of 1 megabit RAMs uses 1.25 -micrometer features. As new discoveriesin microlithography are made, the circuitdesigners are provided with new techniquesfor smaller, faster, and more complexdevices.

Microlithography is a generic term forthe fabrication of minute objects by form-ing patterns on surfaces. In simplest terms,a film is formed, certain areas of the filmare covered with an etch -resistant material,and the uncovered sections are etched

away. As the size of the critical features inintegrated circuits has decreased, the natureof the etching process has undergone afundamental change away from a chemicaltreatment in some liquid etchant. Thereason for this change is that a liquidetchant simply cannot produce the geometriesnecessary for advanced circuit designs.Before explaining why this is the case, amore basic question should be answered.

If circuits are becoming more complex,why is it necessary for the size of theindividual features to decrease? Why notmake the overall circuit larger? If thetechnology for making 10 -micrometerdevices is well developed, why not use itfor making the newest circuits? The answeris two -fold. First, speed -of -operation is animportant consideration in most circuitdesigns, especially in microprocessors. Thepropagation delays of signals within a cir-cuit become longer as the circuit wiringbecomes longer. The fastest devices mustbe optimized for minimum interconnectionlength.

The second factor is, however, the moreimportant limit in circuit fabrication. In atypical IC, there may be ten masking andetching operations, as well as dozens ofother fabrication steps. For a CMOS mem-ory, 14 masking levels and well over 100processing steps would be used. At eachof these steps, contamination is presentand some defect may be introduced, caus-ing a failure when the circuit is tested.With random defects, the probability of acircuit having such a defect is proportionalto its area. Reducing the area of the circuit

RCA Engineer 30-5 Sept. / Oct.1985 75

Fig.la. Isotropic Etching occurs inmost wet etching solutions. Materialis removed from under the edges ofthe mask.

MIIIIIMmEl1111111INIFig.lb. If the linewidth is less thantwice the film thickness, the under-cutting will destroy the pattern.

Fig.1c. An anisotropic etch will notundercut, and the pattern will bepreserved.

will reduce the number of circuits thatwill have a defect and will therefore in-crease the yield. Making the circuit ele-ments smaller also allows the designer toadd increased functionality and complexityto a device without paying a penalty inincreased area and lower yield.

Characteristics of etchingprocessesThere is a basic phenomenon that occursin a liquid etchant that precludes its use inmaking small devices. Figure 1 a illustratesthe etching of a film that has been maskedwith some pattern. Liquid etching is iso-tropic, meaning that it has no preferreddirection. Material is removed from underthe edges of the mask. This is known asundercutting, and it becomes a problemwhen the mask is so narrow relative tothe film thickness that the two undercutsmeet and the line disappears. (Figure 1B)This occurs when the linewidth is lessthan twice the thickness of the etchingfilm. Since the layer thickness is for themost part constrained by the electricalparameters of the devices, little can begained by thinning the material. It is es-sential to minimize the loss in line size. Ifthe undercutting can be eliminated com-

Fig. 2a. Barrell or Tubular Reactor. Thewafers are placed in a holder andinserted into the chamber. There isoften a shield to prevent the actualplasma glow from touching the waferitself.

Fig. 2b. Parallel Plate Plasma Etcher.The wafers are placed on the groundedplate with the opposing plate poweredby a radiofrequency (RF) source.

Fig. 2c. Relative Ion Etcher. The wafersare placed on the powered electrode.The chamber serves as the opposing,grounded electrode.

pletely, the constraint of line width versusline thickness is removed. (Fig. lc) Themethods for doing this will be discussedlater.

Any etching process can be describedby three characteristics: directionality, selec-tivity, and uniformity. Directionality, alsoknown as anisotropy, is a measure of theetch rate perpendicular to a surface versusthe etch rate parallel to that surface. Asmentioned, liquid etchants usually do nothave any directionality. (An exception to

this is the etching of single -crystal siliconwhere certain crystal planes etch muchfaster than others, leading to a directionaletch.) Selectivity is the ability to etch onematerial faster than other materials. Liquidetchants are ideal in this characteristic sincetheir action is simply chemical dissolution.By choosing the etch bath compositioncorrectly, only the target material will dis-solve, giving infinite selectivity. Uniformitycan be expressed relative to any etch pa-rameter, although the usual usage is thatof an etch rate uniformity across a singlewafer or batch of many wafers. Liquidetchants can be very uniform if correcttechniques are used.

Dry processes have the same character-istics, but in most cases the balance issomewhat different. Directionality can bevery good, leading to no undercutting.Selectivity is often not perfect, due to theenergetic processes occurring in the plasmathat tend to etch all materials to somedegree. The control of uniformity is asmuch a function of the reactor design as itis related to the etching chemistry chosen.

One of the challenges of the technologyis to balance the factors to fit a particularapplication, since in most cases, maximizingall three is not only difficult but often

will be giventhat should illustrate why, for example,extreme selectivity or perfect directionalitycan lower yields.

Historical backgroundThe development of dry processing toolsbegan in the late 1960's with the use ofoxygen plasmas to remove organic mate-rials. This was known as plasma ashing.When it was found that silicon and siliconnitride could be removed in CF4/02plamas, the semiconductor industry beganto take notice. The equipment being usedat that point was commonly known asthe barrel or tubular reactor (Fig. 2a). Aradio -frequency (rt) signal is applied toexternal electrodes surrounding a quartztube that is filled with gas at a pressure ofabout 1 torr. (Metal chambers with dif-ferent excitation schemes are also in use.)The discharge causes the dissociation ofthe gas into reactive species, which thenetch the wafer. These tools etch isotro-pically and their advantage is not in theability to etch smaller features, but toreduce the water consumption and chem-ical disposal problems associated with wetetching.

For many years it was known thatmaterials could be removed in high vacuum


by accelerating non -reactive atoms (suchas argon) in a dc or rf discharge. Theimpact of these atoms removes materialballistically, and is called sputtering. Thisprocess is highly directional, but suffersfrom many disadvantages. All materialsare eroded to some degree and selectivityis very low. Materials tend to redepositon the wafer surface, causing many un-desirable effects.

It was recognized that by using a sput-tering configuration with a gas that wouldetch the desired material rapidly, higherselectivity was possible. Directionalitycould still be obtained due to the bom-bardment of the wafer surface. This led tothe development of two types of etchingtools, the Parallel -Plate Reactor and theReactive Ion Etcher.

The Parallel Plate reactor (Fig. 2b) con-tains two electrodes. The internal pressuremay be an order of magnitude lower thanin the tubular reactor (typically 50 to 500millitorr). An rf signal is applied to theupper plate and the wafers are placed onthe lower plate, which is grounded. Thiscreates a potential difference between thewafer surface and the bulk of the plasmadischarge. This potential drop of a fewtens to a few hundreds of volts acceleratescharged reactive species towards the wafer.The etching reactions can be enhancedwhere there is bombardment, leading todirectional etching and decreased under-cutting of the mask. This configuration iscalled plasma etching (PE).

The reactive ion etching (RIE) configura-tion is similar to the parallel plate reactor.The upper electrode is grounded, and usu-ally made larger than the lower electrode.The rf power is applied to the lowerelectrode. In some cases, the upper plateis removed and the etching chamber itselfbecomes the opposing electrode (Fig. 2c).In this case the accelerating potentials canbe very large, on the order of 500 volts.The etcher is operated at even lower pres-sures, 10 to 50 millitorr or so, and veryhigh directionality can be obtained. Becausethis configuration does generate a substan-tial sputtering component, these tools aresometimes called reactive sputter etchers.

The number of variations on these basicthemes is increasing rapidly. Many of theinnovations are concerned with automationof the process, allowing humans to beremoved from direct contact with thewafers. This can prevent some contamina-tion and its associated yield loss. Computercontrol of all etching parameters is becom-ing available, as well as in -line integrationof etching equipment with other processing

New techniques for dryprocessing technology

The basic principle used in dryetching is the enhancement ofchemical reactions by theaddition of energy. This energybreaks a molecule apart intoactive species, which then formcompounds with the substrate.These compounds are volatileand evaporate, leaving an etchedsurface behind. The methodused to add this energy to thechemical reaction is not limitedto radio frequency excitation.

The dissociation of etch gasmolecules occurs primarily byelectron impact. If more electronscan be generated, etch rates canbe increased. This can be doneby magnetic confinement of thedischarge. This technique iswidely used in magnetronsputtering systems. The sametechnique can be used inreactive ion etching.

One of the difficulties withreactive ion etching (and high -power plasma etching) is thecreation of both ions andelectrons with very high kineticenergies. While this is useful forhighly directional etching, itcreates the possibility of damageto the wafer surface. The impactcreates damage sites which,while necessary to the etchingprocess, may destroy theelectrical integrity of theintegrated circuit. There aretechniques that can minimize thedamage or remove it, butavoiding its creation is moredesirable. Lowering the input rfpower will lower the bias voltagebut will also lower the etch rates.Magnetic confinement can retain

high etch rates under theseconditions by increasing theefficiency of the electron -impactionization of the reactive gas. Infact, high input power can beused since the low impedance ofthese discharges leads to lowbias potentials.

The use of magnetron -enhanced reactive ion etchingwas first reported by Horiike, andan etch tool is availablecommercially fromTylan/Tokuda. It contains aplanar magnetron and moves themagnet assembly to achieveuniformity. Materials ResearchCorporation has changed theconfiguration of the magnets toachieve uniformity withoutmotion. The MIE-720 is availableas a cassette -to -cassette,automated etcher, and the MIE-710 is a research tool without theautomation features.

Another development on thehorizon of dry processing avoidsthe generation of bias voltagesand plasmas completely. Siliconcan be made to react withchlorine by focussing anultraviolet laser on the wafer inthe presence of the gas. Thisenergy input serves to generatean etching reaction. Silicontetrachloride is formed andevaporates from the surface. Thistechnique has the intriguingproperty of not requiring a resist.The laser beam can be focussedonly in the areas to be etched,yielding reasonably highresolution, good directionality,and very high selectivity. The firstreports of this technique areappearing in the literature, andcommercial development shouldbe a few years away.

tools. However, there is still room forfundamental changes in the design of adry processing machine. (See the sidebar.)

Sources of informationBecause of the critical importance of thisfield to semiconductor fabrication, all ICmanufacturers are involved in dry -processdevelopment to some degree. Not surpris-

ingly, many of the innovations in the fieldare considered proprietary and do not reachthe scientific literature rapidly, if at all.

Most other semiconductor companiesdo not have RCA's advantage of a centralresearch laboratory. In many cases, theadvanced work is performed very close tothe actual production line, with scientistsand engineers responsible for both researchand developmental functions. Only the

M. Leahy: Plasma etching for integrated circuit fabrication 77

largest manufacturers can support centralresearch facilities, and there are only afew universities that can fully address therather complex issues of semiconductorfabrication. This means that crucial dis-coveries do not follow the usual path topublication, making it difficult to pinpointwhere the important work is being done.

In a limited space, it is impossible tochoose those contributions that can beconsidered most important to the field ofdry processing. It would be better to list afew names that can serve as introductionsto the literature. By using their papers as astarting point, you will come upon mostof the important researchers in the field.(The reverse searching technique of findingthose papers that refer to the early impor-tant work is very useful. Check with yourRCA librarian for information about usingthe Citation Index.)

Richard Bersin at the InternationalPlasma Corporation, Alan Reinberg atTexas Instruments (both now at PerkinElmer), and Rudolf Heinecke at the Stan-dard Telecommunications Laboratory (nowat ASM) can certainly be considered pio-neers, as well as Adir Jacob at the LFECorporation. Dan Maydan and DavidWang of Bell Laboratories (both now atApplied Materials) are best known for thedevelopment of useful etching machines.John Coburn and Harold Winters at IBMand Daniel Flamm and Vincent Donnellyat Bell Laboratories are well known forbasic studies of reaction mechanisms andphysical measurements of discharges. C.J. Mogab at Bell Laboratories andGeraldine Schwartz and Paul Schaible atIBM have made many contributions tothe understanding of the physics andchemistry of dry processing.

RCA scientists have made importantcontributions in this field. The work ofHans Lehmann, Bernard Curtis, andRoland Widmer of the Zurich Labora-tories, and John Vossen, Jer-Shen Maa,Bernard HaIon, Michael Duffy, andMichael Leahy of RCA Laboratories hasbeen presented at many international con-ferences and published in various journals.An even larger body of knowledge hasbeen gained by RCA through the work ofmany process engineers in our factoriesand pilot lines, and those contributionsare critical to RCA's future in the micro-electronics industry.

Using the technologyThe ability to use dry etching depends onmany factors, but two areas are particularly

important. The first can be called ProcessMonitoring and Control. There are measur-able quantities that can be used as monitorsof the conditions within the etching tool.These can be used, for example, to detectthe end of an etch step or the degradationof the etching characteristics. There aremany methods used to accomplish thisobjective, and some will be reviewed indetail.

The second area of importance is closelyrelated to the particular circuits being made.It is essentially the art of making intelligenttradeoffs between process characteristicsand circuit designs. With geometries ap-proaching 1 micrometer and critical filmthicknesses of a few nanometers, the pro-cessing conditions and the three-dimen-sional design of the device are closelycoupled, and failing to recognize this oftenmeans the difference between a successfulproduct and disaster.

Before discussing some of the ways inwhich dry processes are controlled, thecommon features of all dry etching toolsshould be reviewed, as well as some detailsof the chemistry and physics of etching.

Design of an etcherWhile the design of a reactor is still some-thing of an art and many approaches areused, there are a few basic principles thatare followed. A dry etcher requires a waferhandling mechanism, a gas handling system,vacuum pumping, and electrical control.The choices made in each of these cate-gories interact with the other requirements.

Wafer handling may be fully manual.The earliest laboratory systems were some-times placed in production lines withoutautomated wafer transfer, and good yieldscould be obtained. However, due to par-ticulate contamination caused by manualhandling, the direction is toward fully -automated transfer of wafers in and outof the etcher. Currently, wafers are trans-ported around processing lines in batchesof 25 in clean carriers of one sort oranother. (There have been some ratherambitious attempts in the industry at waferhandling across a whole production line,with tracking of individual wafers movingon air tracks. For the most part, batchedhandling is still the rule.) Wafers can beloaded into the etching chamber in twoways. Either the etching chamber is broughtto atmospheric pressure, opened, loaded,and re -evacuated, or an intermediate vac-uum load lock (as in an air lock) is used.There are advantages and disadvantagesto each approach.

A vented -chamber system is simpler todesign, which usually translates into in-creased reliability. However, it is usuallyloaded manually, adding to particulatecontamination of the wafers. It is moresusceptible to contamination from atmo-spheric water vapor and some etch pro-cesses (notably aluminum etching) aredifficult to control under such conditions.The operator must also be protected fromthe etching chemistry, since an open cham-ber may emit noxious fumes. The pumpingsystem must be capable of repeated cyclingfrom room pressure to high vacuum andshould pump rapidly to provide adequatethroughput.

In a load -locked system, the main etch-ing chamber remains at either high vacuumor at the processing pressure. In theory, itis never opened to the surroundings. Oper-ator safety is easier to insure, and usuallya relatively small load -lock is cycled inpressure. Most load -locked etchers aresingle -wafer tools, although some exist thathold full 25 -wafer cassettes in the locks.The disadvantages are primarily mechan-ical. First, the handling must be automatedand is usually complicated. Many sealingsurfaces are used for the lock and chamberdoors, leading to decreased vacuum integ-rity. There must be many sensors alongthe path of the wafers as they move inand out of the chambers, leading to com-plexity. This complexity, in a productionenvironment, means decreased reliabilityand increased cost.

A reactive gas must be introduced intothe etcher in some controlled fashion. Thiscan be done simply by allowing gas toflow through a restriction. In some etchers,there is no other control of the pressureexcept by gas flow, leading to pressureinstabilities as the etching proceeds. Fur-ther, it means that the pressure and massflow are not independent of each other.This can be a problem when a pressure ischosen to provide a certain directionality,for example, and a gas flow must bechosen to provide uniformity. The abilityto throttle the vacuum pumps to provideconstant pressure over a wide range ofinput gas flows can be a distinct advantage.Closed -loop mass -flow controllers andthrottle valves are now used on mostcommercially available systems.

The vacuum pumping method can takemany forms, depending on the operatingconditions. Plasma etch tools operate atrelatively high pressures and mechanicalpumps are sufficient in most cases, althoughgas flows may be very high and largepumps may be required. RIE systems


operating at a few millitorr use diffusionpumps, turbomolecular pumps, and cryo-pumps to achieve their low base pressures.For vented -chamber etchers it is desirableto have a low base pressure to minimizecontamination, even though the pressureduring the etch may be high. In the caseof etching with toxic and corrosive gasessuch as chlorine (C12), hydrogen chloride(HC1), or boron trichloride (BCI3), extremeprecautions must be taken with the vacuumpumps, especially when cryopumps areused. In almost all cases, special pumpoils are required to prevent contaminationof the etching chemistry and dangerouspump conditions.

Electrical control takes many forms. Therf generator should be stabilized with feed-back control. It should be mated to theetching chamber to minimize losses andradiated power. The rf power level shouldbe controlled accurately, and the optionof regulating the power in response to thedeveloped dc self -bias should be available.There must be interlocks to prevent theapplication of rf power when the chamberis open, as well as interlocks for operatorsafety. Full microprocessor control is nowbecoming the normal situation in mostdry etch tools, to the degree that manysystems only require specification of asimple recipe to etch wafers. The nextgeneration of etchers will be integratedinto a centralized wafer management com-puter, which will provide an even higher -level protocol to production line engineers.

Basics of the etching processUsing silicon as an example, etching isbased on the formation of a volatile com-pound of silicon. Silicon tetrafluoride (SiF4)and silicon tetrachloride (SiC14) both havehigh enough vapor pressures to be pumpedaway during the etch process. Taking themost studied case of carbon tetrafluoride/oxygen mixtures (CF4/02), the rf dischargecauses the release of free fluorine from thegas mixture. This fluorine will adsorb onthe silicon surface, react with the silicon,and form SiF4, which then evaporates.The same process can occur using thechlorine analogue, carbon tetrachloride(CC14), to form SiC14. In fact, the menuof choices of etchant materials is steadilyincreasing, although the end productsremain the same.

The basic chemistry is simple, and if itwere only chemistry that determined theetching characteristics, there would be littlebenefit in using dry processing. As men-tioned above, liquids etch isotropically.

Explanation of dc self -biasformation in rf discharges.

When a dc potential is imposedacross two electrodes in a low-pressure gas, a plasma candevelop. This plasma will becomposed of charged andneutral species. In the types ofdischarges used insemiconductor fabrication, theneutrals far outnumber thecharged species, by three or fourorders of magnitude. If anelectrically -isolated disk isplaced in the discharge, a flux ofboth positive and negativecharges will arrive at the surface.The negative charges are carriedby electrons, which move muchfaster than the ions carrying thepositive charges. Since moreelectrons arrive per unit timethan ions, the disk will becomenegatively charged. Electrons willbegin to be repelled and positiveions will be attracted, forming aregion of higher positive chargenear the disk. The electron fluxwill be reduced due to therepulsion and will eventuallybalance the ion flux, leaving thedisk charged to some negativepotential. Positive ions are nowaccelerated to the disk and mostof the electrons are repelled.

The potential developed at thisdisk in a dc discharge is rathersmall, a few tens of volts, and theacceleration of the ions is rathermodest. While this added energywill enhance a chemical reactionsomewhat, a few hundred voltsof potential drop will do better. By

using an ac discharge with anexcitation frequency greater thana few tens of kilohertz, the dcself -bias can be greater than 500volts. This phenomenon is againrelated to the relative speeds ofelectrons versus ions and thecapacitive coupling of thedischarge.

Using a low -frequency acexcitation, the voltage measuredacross the blocking capacitorwill follow the excitation voltage,with only small lags due to thefact that the insulator cannotcharge instantaneously. At highfrequencies, the effect of relativecharge -carrier speed takes over.Because the positive ions movemuch more slowly than theelectrons, the sample never fullyreaches the full positive potentialduring its half -cycle. During thenegative half -cycle, the samplecharges rapidly. The resultant acvoltage waveform measuredacross the capacitor will beapproximately equal to theexcitation waveform but will bedisplaced negatively at someroot -mean -square voltage. Thisis the self -bias voltagesometimes referred to as the dcbias. This bias governs theenergy acquired by positive ionsbombarding the sample, and canaffect the directionality,selectivity, and etching rate of aparticular plasma.

For a detailed discussion ofthis topic, see Glow DischargeProcesses, Brian Chapman,Wiley-Interscience, New York(1980).

There are no accelerating fields. Barreletchers do not produce any fields at thewafer surface, and also etch isotropically.The etching is primarily a chemical effect.In the geometry of a parallel -plate or RIEconfiguration, a dc self -bias due to the rfexcitation develops (see sidebar), and aphysical etching effect can be generated.The added energy can change the chemistryby increasing the reaction rate at the pointof bombardment. This imparts a direction-ality with little undercutting, since the side -walls are protected by the resist and cannot

"see" the impinging reactive species. Com-plications arise when there are sufficientuncharged reactive species in the plasmadischarge. These can drift to the sidewallsand cause undercutting. In some cases,reactive species can migrate along thesurface and accomplish the same effect.

A second mechanism for producing di-rectional etching relies on a self -maskingor passivating phenomenon. Many gasesused in etching can also form polymers.These polymers can recombine with theactive etchants, lowering the etch rate at


RELATIVEINTENSITY

TIME

Fig.3. Emission at 703.7 nanometers. Monitoring an etching process usingoptical emission. In this case, the emission at 703.7 nanometers is due tofluorine, which is consumed by the etching reactions. When the silicon film isetched through, the fluorine increases, and the etcher can be turned off.

the point of recombination, or can simplymask the etching by coating the surface.If competitive etching and masking reac-tions occur on the wafer surface, directionaletching can be obtained. If the passivationis removed only where there is bombard-ment, the sidewalls of the defined featureswill be protected from the etching. In thisway, an etching chemistry that would nor-mally undercut can produce very direc-tional etching. Aluminum can be etchedin boron trichloride/chlorine mixtures andcan be made to undercut severely. Theaddition of small amounts of chloroform,CHCI3, under otherwise identical condi-tions, will cause the sidewalls to be pas-sivated, allowing vertical line profiles tobe obtained.

Process monitoring and controlThese examples hint at some of the tech-niques that can be used to monitor dryprocesses. The gas is dissociated intoreactive species when the plasma dischargeis initiated. These species usually emitradiation as they lose energy to their sur-roundings. The wavelengths and intensitiesof the emissions can be used to identifythe nature and concentrations of the var-ious components of the reactive gas mix-ture. Figure 3 depicts the emission intensityfrom fluorine atoms in a CF4/02 plasmaduring a silicon etching process. As theplasma is ignited, the fluorine concentrationincreases. When the etching begins, thefluorine is consumed by the reaction, whichforms silicon tetrafluoride. Its emissionintensity will drop. At the completion ofetching, the amount of exposed silicondrops rapidly (only the sidewalls may stillbe etching) and the fluorine emission

increases again. This is a method forendpoint detection, and has been in usefor many years. Monitoring a reactionproduct, rather than a reactant, is alsofeasible. Etching of silicon dioxide inCF4/H2 can produce carbon dioxide,which will increase during the etch andthen decrease at the endpoint. For someparticularly difficult cases, multiple speciesmay be monitored to provide an accurateendpoint.

This same technique can be used as adiagnostic tool to judge the health of anetcher. Hydrogen emissions are often thesign of a water leak, while nitrogen almostalways means an air leak. In a highlyautomated etcher, a computer may obtainthe optical emission signature and compareit to some known reference before allowingany wafers to be etched.

Similar monitoring of the chemicalcomposition of plasma discharges can bedone by mass spectrometry, infraredabsorption, and laser -induced fluorescence,to name a few techniques. The details ofthese methods are documented in theliterature.

There are also electrical parameters thatare indicative of processes occurring inplasma discharge. The rf-induced dc self -bias can be a sensitive indicator. Most rfgenerators are capacitively -coupled to theetch electrodes, and the medium betweenthe plates will affect the electrical charac-teristics. As the plasma chemistry changes,at endpoint for example, the dc bias willalso change. Even if the dc bias is notused as an endpoint indicator, it shouldbe monitored, and if possible, controlleddirectly. The magnitude of the bias willoften be the controlling factor in the direc-tionality of the etch.

Building an integrated circuit

The fabrication of an integrated circuit isa complex process. Many rather sophis-ticated procedures are used to make aworking device out of silicon or someother semiconductor material. The role ofdry processing in this sequence can beseen by examining the way in which atypical device is made.

By the time the process is complete, thewafer will have thin films of silicon, silicondioxide, silicon nitride, and aluminum onit. There will be various amounts ofdopants such as phosphorus and boron inboth the silicon and in some of the silicondioxide. The thickest films are 1 to 2micrometers thick, while the thinnest maybe only 5 nanometers (50 Angstroms).The dopants are introduced into very spe-cific places in extremely small quantitiesto modify the electrical behavior. Alu-minum is used as the "wires" that inter-connect various elements of the circuits.

These thin films are deposited over thewhole wafer surface and then removed incertain areas. In every case, the removalcan be done with a dry etching procedure.The interactions that can occur during theprocess can be illustrated by looking atone particular step.

Consider the case of a polycrystallinesilicon (polysilicon) film used as an inter-connection. The structure is shown in Fig.4. In many circuits, there are two or morelevels of wiring. Usually one of these layersis aluminum, but there may be one ormore levels of highly -doped silicon actingas connectors, much like a multi -layeredcircuit board. Some insulator, such as sil-icon dioxide, will be in place over thewafer surface, and holes will have beenmade in it to allow contact to the under-lying levels. Contact will typically be madeto the wafer surface itself and to a layerof polysilicon already formed.

The second polysilicon layer is deposited.It covers the first layer conformally, fol-lowing the hills and valleys already presenton the wafer surface from previous process-ing. It is now doped so that its conductivityis high enough to carry enough current.Photoresist is patterned over it, and it isready to be etched.

There are a number of areas in thestructure that can cause problems. At A,the polysilicon passes over a step formedby the layers below. Measured normal tothe wafer surface, the polysilicon is sub-stantially thicker. (It is approximately equalto the step height plus the thickness of thedeposited film.) If the etching is highly


directional, this area will require a muchlonger time to clear. This means that theselectivity of the etchant must be veryhigh, or the insulating oxide at B will bedestroyed during this overetching time.Lowering the directionality will help theproblem at A but will undercut the narrowline at C, causing high resistance or anopen circuit. Raising the selectivity of theetch towards oxide is usually desirablebut can cause second -order complications.Oxides form on top of the polysiliconnaturally in air, and also during the dopingstep. If these oxides are not completelyremoved, the polysilicon will be uninten-tionally masked and short circuits willresult.

These difficulties are related to the struc-ture as it exists before etching. Similarproblems may be created for subsequentoperations at any etch step. Later in theprocess, aluminum is deposited over thewafer and it must cover the steps on thesurface created by all the previous process-ing. If the sidewalls of the etched polysili-con are vertical and no smoothing occurswhen the next insulator is deposited, themetal may not be able to pass over thesteps without breaking.

Sometimes the problems that develophave even more pragmatic causes. A casein point was the early development of alarge -area charge -coupled device at RCALaboratories. This CCD was very large(about 0.5x0.8 inches) and contained agroove cut into a polysilicon layer. Thegroove was 3 micrometers wide and 10meters long, and had to be made with nobridging across it. (That is the equivalentof building a sidewalk from New York toLas Vegas and keeping it free of fallenleaves.) When the process was changedfrom a wet etch to a dry etch, the yieldwent down. Examination revealed smallbridges in the photoresist. Since the dryprocess did not undercut, these smallbridges were resolved. In the wet etch, theundercutting etched away the silicon underthe bridges and the CCD functionedproperly.

It must be kept in mind that integratedcircuits are three-dimensional structures.The structure affects the electrical param-eters, and indiscriminate changes in theprocess can result in transistors that donot perform as the circuit designers in-tended. Speed requirements may not bemet in a memory, timing may not besynchronized in a microprocessor, and ingeneral, a working device will not beobtained. It takes a concerted effort byboth processing engineers and circuit de-

PHOTORESIST OXIDE

POLYSILICON in SUBSTRATE

Fig.4 Etch Problems in Real Circuits. (A) Areas where the vertical thickness ofthe silicon is greater due to an underlying step. (B) Thin oxide under the silicon,requiring highly selective etching. (C) Narrow photoresist lines that cannot bedefined properly if excessive undercutting exists.

signers to reach a compromise that yieldsa product that can be manufactured at aprofit.

Directions for the future

There are some developments that shouldattract attention in the next few years.Reactive Ion -Beam Etching (RIBE) andMagnetically -enhanced Ion Etching (MIE)are certainly noteworthy.

Reactive Ion -Beam Etching was devel-oped to allow the separation of variablesin a dry etch tool. The energies of theions created in plasma etchers and reactiveion etchers are not easily controlled inde-pendently of other parameters, such aspressure and power. In a RIBE tool, theseenergies are electrically controlled byaccelerating grids. In this way, an extradegree of control is possible, at least pro-viding the opportunity for modifying thechemistry at the surface without changingthe bombardment processes due to theions. This technique will certainly be underintense development in the near term.

Magnetically -enhanced Ion Etching(MIE) is another technique under inves-tigation by RCA Laboratories. It uses amagnetic field to increase the efficiency ofthe discharge, obtaining high directionalityat low dc self -bias voltages. (See thesidebar.)

The longer -term development of thistechnology will likely proceed along twofronts. The first deals directly with the

basic requirements of future devices. Thecritical dimension of advanced circuits(such as the Very -High -Speed IntegratedCircuit program of the Department ofDefense) is now below 1 micrometer. Ageneral rule of thumb is that dimensionalcontrol must be 10 percent of the designedsize. If the smallest feature is 7500 Ang-stroms (0.75 micrometers), the toleranceis plus or minus 750 Angstroms. Thisdegree of control was absolutely unthink-able a few years ago, and it is not clear ifit is obtainable at all without major innova-tion. There will probably be design inven-tions that relax some of the size constraints,but the trend is still to smaller and smallerdevices. Improvements in resist technologyand the development of x-ray and electronbeam lithographies must occur, or even a"perfect" dry etch process will be severelylimited.

The second area will be in the applica-tion of these techniques to commercialetching tools. Dry etchers will becomemore intelligent by combining sophisticatedsensors with computer processing that goesbeyond simple process monitoring. Adap-tive control of the etch process will becomeimportant so that, for example, if the etchcomputer determined that the etch wasnot proceeding uniformly, it would adjustthe machine settings to obtain a betterresult.

An expert system will interrogate boththe circuit designers and the processingengineers about the desired results. It will


determine the processing parameters andrun the etchers. This implies a distinctlack of people in the processing line, andthat is yet another area of applicationbeing addressed. The level of automationwill most certainly increase to the pointwhere no human interaction will be neces-sary. Wafers will be transported to andthrough the etch tool automatically. Processmonitoring will be interfaced to centralizedmanufacturing control systems to recordthe detailed history of each wafer, as wellas to alert engineers to problems.

SummaryDry etching is an extremely useful tech-nique in integrated circuit manufacture,and is essential for advanced devices. Theinteractions between the design goals andthe processing steps are complex, andrequire great attention to detail. Dry etchtools are complicated systems that demandsophisticated monitoring and control. Whilethe understanding of the processes occur-ring in a plasma is increasing, the field isstill relatively young, and many problemsremain unsolved.

Further readingThe best sources of information are thetechnical journals. The Journal of theElectrochemical Society devotes a majorityof its pages to solid state science andtechnology. The Society has two divisionsthat address semiconductor issues, and itorganizes symposia in plasma processingand other related areas. The IEEE Trans-actions on Electronic Devices is anothergood source of information, as is theJournal of Vacuum Science and Tech-nology. There are many Japanese journalsthat contain useful information. Your RCAlibrary has access to the computerized data-bases and a few keywords will obtainhundreds of references in this area in ashort time.

One of the most useful ways to explore

the literature is by using the Citation Index.This source lists those papers that refer toa particular article, rather than the articleitself. This procedure works forward intime rather than backward. If you havefound an article by Smith that interestsyou, by searching Citation Index you canfind those papers that cite Smith as areference. Thus, by knowing just a few ofthe names mentioned in this article, youcan find most of the important papers inthe field.

There are few books in this field, sinceit is both moving too rapidly, and thetechnology is often considered proprietary.One source is Glow Discharge Processesby Brian Chapman (Wiley-Interscience,1980), which contains many useful refer-ences and a wealth of basic informationabout discharges.

Bibliography1. G.C. Schwartz and P.M. Schaible, J. Vac. Sci.

Technol, 16(2), p. 410 (1979). This is a short,simple discussion of reactive ion etching of silicon inchlorine -based plasmas. It illustrates many of thephenomena obtained in low-pressure discharges.

2. L.M. Ephrath, IEEE Trani Electron Dew.., ED -28(l I),p. 1315 (1981). This is a short, review -type paper onthe uses of dry processing for VLSI. It concentrateson reactive ion etching, with examples of equipmentand etching of typical device structures.

3. Brian Chapman, Glow Discharge Processes, Wiley-Interscience, New York (1980). This book, whilesomewhat short on practical dry processing informa-tion, does contain a useful list of references and areasonably complete treatment of the physics ofdischarges.

4. M.F. Leahy, Proc. 3rd Microelectronics MeasurementTechnology Seminar, Benwill Publishing Corp., Boston(1981). Examples of optical endpoint detection for anumber of important plasma processes are given.Silicon, silicon nitride, and aluminum etching dis-charges are presented.

5. H.W. Lehmann and R. Widmer, J. Vac Sci Technol,15, p. 319 (1978). This is an early silicon dioxideetching study showing profile control of sidewallsand sub -micron resolution of grating patterns.

6. D.L. Flamm and V.M. Donnelly, Plasma Chemistryand Plasma Processing, 1(4), p. 317 (1981). Thisreview article gives 143 references up to about 1981.Concise treatments of a number of important areasare provided.

Dr. Michael F. Leahy is a Member of theTechnical Staff at RCA Laboratories,Princeton, N.J. He received a BS inChemistry from Long Island University,and an MS and PhD in Physical Chemistryfrom Rutgers University for thesis work inMossbauer and Raman spectroscopy oforganotin compounds. He spent two yearsat Princeton University as a VisitingResearch Fellow studying the spectra ofvisual pigments, as well as directing thecomputerization of the Raman spectroscopyfacilities. After joining RCA in 1978, hewas involved in various aspects ofintegrated circuit fabrication, includingwafer cleaning and contamination control.He is responsible for the development ofdry processing technologies for advancedCMOS, CCD, and high -power MOS devices.He also chairs a Technology Committeeof the Solid State Division, which isresponsible for all technical evaluations ofdry processing equipment. He received anRCA Laboratories Outstanding AchievementAward in 1983 (with M. Duffy and J.S.Maa) for the development of dry etchingtechniques. Dr. Leahy is the author of 9papers and holds one U.S. Patent. He is amember of the Electrochemical Society.Contact him at:RCA LaboratoriesPrinceton, N.J.Tacnet: 226-3037


James Warner John Slitz

How to evaluate and shop forcomputer graphics software

Jim Warner is President of Precision Visuals, Inc., a vendor ofdevice -independent computer graphics software, and John Slitzis Vice President of Marketing. Here they provide a "shopper'sguide" to graphics software, offered from the perspective of thesupplier.

For many years, scientists and engineersconsidered computer graphics as an applica-tion in and of itself, to be taken off theshelf and used when they needed to presentinformation, such as summaries of researchresults. In the mid -1980s, graphics evolvedinto a necessary, integral component ofmany computer applications, as systemusers and developers discovered the effective-ness with which graphics can communicatecomplex technical data.

Parallel growth has taken place in theuse of graphics at the interfaces betweensystems and their operators. Now, systemdesigners are using graphics more often toincrease human performance, as well ashuman understanding. In this context,graphic displays are becoming common-place in control and simulation applica-tions, while graphics -based systems are

Abstract: In any computer applicationwith a major graphics component, thekeystone is the graphics software, whichcan greatly affect the cost and schedule ofprogram development, as well as theproductivity of the end user. We willpresent a methodology that can helpensure the selection of the optimalgraphics software development strategy,both for a specific application and for thelong-term needs of the organization.

©1985 RCA Corporation.Final manuscript received Sept. 18, 1985Reprint RE -30-5-11

beginning to realize their extraordinarypotential to increase productivity in designand manufacturing.

In a hardware/software system thatemploys graphics, the software is oftenthe ingredient that has the most impacton system development time and cost,while being the most difficult item forwhich to plan. Over the life of the system,software will have a major impact on theproductivity of the user. Also, the main-tenance and upgrading of software mayrepresent a large share of the system'soperating cost.

In considering the software elements ofsystems incorporating graphics, it is usefulto divide them into three separate levels:application software, tools packages, anddevice drivers.

For application software, developershave basically four choices: buy it as partof a turnkey system along with hardware,buy it ready to run, write their own codefrom scratch, or build each neededapplication using a software tools package.

All four approaches present tradeoffsbetween performance and flexibility. Turn-key systems and special-purpose applicationsoftware both tend to be fast and efficient,because they are optimized for an applica-tion, specific hardware, or both. However,they are usually restricted to a small familyof devices, are seldom portable to a differ-ent hardware environment, and are onlydesigned to solve a specific kind of problem.

If a turnkey solution or off -the -shelf

application software can be found thatvery closely matches all of your require-ments, then we would by all means recom-mend that you acquire it. Almost certainly,though, either of these approaches willentail compromises, because both are gen-eralized for a class of users, not optimizedfor your specific situation.

With enough time and programmingresources you might be able to developcompact, efficient application code withoutusing a tools papkage by focusing only onthe immediate needs of your end usersand a small number of devices. On theother hand, the resulting application willnot readily port to new devices in thefuture unless a considerable effort is madeto design it to be device -independent. With-out special graphics programming expertiseand a strong need for graphics functionalitythat is not commercially available, it isdifficult to justify writing original graphicsapplication software without using a provengraphics tools package (see Fig. 1).

A tools package can be used to createmany different applications, even on anad hoc basis, and most of today's toolsare based on standards that ensure a highdegree of machine and device indepen-dence. Most tools vendors also offer exten-sion packages that can get you part of theway toward a complete application, suchas in mapping or presentation graphics.

Compared with turnkey systems andspecial-purpose application software, thetools approach offers much more flexibility.


{Ready to run

Programmingrequired

Turnkey Systems (hardware & software)

Application Systems (software only)

Graphics Extension Packages (application -dependent)

Tools Systems (application -independent)

Vendor -supplied Software (device drivers)

Fig 1. Generally speaking, the more specific the application, the easier apackage is to use.

Partial DeviceIndependence

DeviceDependent

Device IndependenceWith Escapes

Pure Device Independence

Fig 2. Machine independence is usually a matter of degree.

The application can run on many differentmachines and graphics devices, and canbe tailored to your own particular needs.This flexibility, though, will be gained atthe expense of some performance, becausetools packages have to be large and sophis-ticated enough to address many machines,devices, and applications. Nevertheless,using standards -based tools may result inimmediate, net performance gains, becausethey put you in control of selecting thehardware that meets your needs as yousee them. Also, they will give you accessto future advances in graphics devices andprocessing power and keep your hardwareoptions open. As long as present trendscontinue, you should be able to move theapplication gradually to new devices withbetter performance and lower cost.

The third level of graphics software in

our scheme consists of device drivers, theprograms you need to use a particulargraphics device with a device -independenttools package. You will not have decisionsto make about device drivers if you buyyour hardware and software from the samesource (e.g., Tektronix or a turnkey ven-dor), because the vendor will integratethem for you. In the more common caseof mixing many different devices withdevice -independent software, the driverswill influence the overall performance ofthe system nearly as much as the applica-tion software and the design of the deviceitself.

The number and quality of device driversoffered by software tools vendors vary.The quality of a device driver is measuredmainly in terms of the intelligence withwhich it can make use of hardware fea-

tures. For relatively simple devices youmay be able to write a driver in-house,and in many instances hardware and soft-ware suppliers provide skeleton drivers tomake this job easier. For high-performancedevices, writing a driver is seldom feasiblefor the user, and the availability of a high -quality driver is crucial to the optimal useof the device.

General criteria for evaluatinggraphics softwareBefore comparing graphics software, it isimportant to define your problem suc-cinctly and understand your needs com-pletely. The importance of a patient,thorough approach to defining needscannot be overemphasized.

This process requires gathering knowl-edge of your environment. You will haveto develop a profile of the end users oroperators who will use the application.Also, examine the blend of programmersand trainers who are available to supportthe application, because their skills deter-mine the level of usability that you willrequire in the software you select. Forexample, a target user who is an engineeror electronics technician with solid com-puter skills will need less training, andwill be able to deal with a more complexsystem interface (requiring less programmertime to create), than a casual computeruser who needs a friendlier interface andmore instruction.

Analyze the constraints and resourcesof your existing and planned computersystems, graphics devices, and any otherapplications that are planned or alreadyin place. Graphics may place heavy de-mands on CPU time and communicationlinks. Machine independence is usually amatter of degree-in any case, you shoulddemand assurance that any software youobtain will install successfully and runefficiently on the available CPUs. Evenon standalone workstations, some graphicssoftware products will run much betterthan others (see Fig. 2).

Borrowed, shared, or inherited hardwaremay narrow your range of software choices.The best -quality devices tend to be thebest-selling models and consequently theones with the best available software sup-port, but be skeptical of the quality ofgraphics software available for castoffhardware from little-known manufacturers.A small saving in hardware cost can vanishquickly if several weeks of programmingsupport are needed to integrate a deviceinto the system.


A shopping list for thegraphics tools buyer

Device independence: Devicedrivers Most hardware features of the

device supported by the driver Many devices supported-

vendor actually has thedevices in-house

Vendor willing to developquality drivers for new devices

System features:

Does a given feature actuallyrepresent a solution to apotential problem, or is itmerely excess baggage?

Debugging and error handling:

Speaks English.Presents "snapshot" showingthe subroutine called andparameters passed.Allows programmers to

specify the severity of an errorthat halts execution.

System architectureDevice -independent routinesand device drivers.Separate pipelines for two -and three-dimensionalviewing.

View them as guidelines, notgospel. Usability is moreimportant than absolutefidelity.

Documentation Ask to see the manual before

you purchase.In reference section, are allpackage capabilitiesillustrated and specialcontingencies covered?

In tutorial section-does itcontain: An overview of textbook

graphics concepts and howthose concepts areimplemented?

A definition of technicalterms?Frequent illustrations?Real -world examples withgood annotation?

Sample modules of frequentlyencountered graphics tasks?Device driver documentation?

Training

A range of courses offered.Presented at the vendor'ssite-at your site-on aregional basis.User references: use them, butdon't abuse them.

Your commitment

Do you have sufficient humanresources to:Design, code, document, andmaintain your own customapplication?Install new device drivers?Incorporate new features intothe system as the needarises?

Defining needs

List your needs in your own terms, not inthe language of graphics. For example,say "I need to generate pictures in the laband display them later at several otherfacilities," not "I've got to have a metafilecapability." Or, "I have to move objectsaround on the screen," not "I need asegmented data structure with pick inputand image transformations."

Arrange your list of needs into a detailedhierarchy of "musts," "wants," and "don'tcares" by assessing the task at hand andthe environment you will be working in.Then, challenge the vendors to match theirproducts' features to each item on yourlist. In this way, you will be concentratingon finding a software answer to each need,rather than wasting time trying to under-stand features that may not be relevant toyour situation.

Conceptual model of a graphicssystem

Choosing a graphics software system foryour application consists of more thanfinding software features that correspondto your discrete functional requirements.Tools packages-even those based on thesame standards-are also differentiated bytheir architectures, which can influence

how well they run on a specific computersystem. A modular design, for instance,allows device drivers to run on a work-station in a distributed processing environ-ment. The structure of the tools systemmight also greatly affect the efficiency withwhich it accesses the special, advanced(that is, non-standard) features of high-performance hardware.

Tools packages are libraries of subrou-tines that can be called by an applicationprogram to perform basic graphics func-tions. Internally, the subroutines addressan idealized "virtual" device that representsthe combined functionality of many typesof devices, such as plotters, CRT displays,joysticks, and image recorders. In actualoperation, the subroutines must interfaceat some point to the individual personalityof each device used. This takes place atthe virtual device interface (VDI). Foreach device, a driver passes graphics inputacross the VDI to the device -independentlayer of the tools package. Output com-mands from the device -independent layerpass through the VDI and are translatedby the device driver into device -dependentcommands that can be understood by theindividual device. If an output commandis inappropriate for the particular device,or if the driver is incapable of interpretingit, the system should ignore the command

or attempt to simulate the function insoftware. An intelligent device driver per-forms every possible function on the device,using software to simulate only what cannotbe done by the device. Through simulation,a particular color on a raster display canappear as a crosshatching pattern whenthe image is drawn on a plotter.

A tools system might be designed toperform every graphics function (movinga plotter pen, drawing a line, formingpolygons or markers, altering a viewingtransformation, changing a color, etc.),always in the device -independent layer.At the other extreme, the package couldbe designed to handle only simple graphicsfunctions at that level, letting the devicedriver, through user -selected parameters,decide whether more complex functions(such as drawing text characters or calcu-lating a modeling transformation) will becarried out slowly in software or morerapidly on intelligent devices that havethese capabilities built into hardware orfirmware.

For high-performance applications usingadvanced devices, the system that worksbest will be the one with the most deviceintelligence, doing most of the work whereit can be done most efficiently-in thehardware or firmware of the device. Thisarchitecture uses device drivers that are

Warner/Slitz: How to evaluate and shop for computer graphics software 85

Evaluation checklist forsoftware vendors

d Products that solve yourproblem.

it Reliable service and timelyanswers.

EI Updates and enhancements.E: Device drivers for newest

graphics devices.it Support for viable ANSI/ISO

standards. Productivity enhancements

beyond the standards.

E Documentation, training, andtelephone support.

E Installation and on -sitetechnical assistance.

Market recognition andacceptance.Business staying power.

harder to design, but that give a device -independent system the ability to use devicefeatures to maximum advantage. Althoughsome device drivers are bundled with toolspackages or even offered free, you willultimately get what you pay for, becausehighly capable drivers require a consider-able engineering effort by the supplier. Apackage that supports sophisticated devicesat the driver level will allow you to buyonly what you need for each device, aswell as providing better overall systemperformance.

Thus, in summary, the overall design ofthe tools system and the sophistication ofits device drivers should be evaluatedcarefully in matching software tools toapplication requirements.

If a graphics application will be usedon a large scale to amass a large graphicaldatabase (such as mechanical engineeringdrawings) over a long period, careful plan-ning must be devoted to how the datawill be stored and retrieved. A graphicssoftware system that is dedicated to aspecific application can manage graphicaldata in a simpler fashion than a systemthat has to be flexible enough to deal withmany different kinds of graphical data.The application -independent system, forexample, manipulates data in a virtualdevice coordinate system. The application -specific system can store information interms of "world" coordinates that areappropriate to the application.

For each application database, program-

ming resources must be allocated to mainte-nance, because changes made at the virtuallevel must be preserved in the coordinatesthat are meaningful to the application.Software can perform the translations, butthe programmer must ensure the integrityof the application database so if the hard-ware goes down, all pictures can be regen-erated from that database. The graphicssoftware system can thus remain free ofany one application so it can be used in astandard way for many applications.

As graphics applications have grown,users are recognizing the need to handletheir graphical data as valuable resources,using a metafile system for archival andactive storage of pictures. A graphicsmetafile capability allows pictures to begenerated at one time and place, storedsomewhere else, and displayed at anothertime and place, all without re -executingthe original application program thatcreated them.

Metafiles can also be used to combinemultiple pictures into one image, so theyare useful for applications where someparts of a picture remain constant whileother parts are updated. If a metafile capa-bility is needed for an application, youcan expect to have a significant investmentin graphical data, so the issues to considerhave to do with compatibility betweensoftware systems (see Fig. 3).

The role and reality of graphicsstandards

One of the criteria for comparing graphicssoftware is adherance to either de facto orformally sanctioned standards. Standardscomformance has different meaning andvalue to different users, but its implicationscan be profound for a buyer who is con-sidering long-term investments in graphicssystems.

In 1979, the first major attempt wasmade to standardize the use of graphics atthe application level when the Associationfor Computing Machinery's Special InterestGroup on Graphics proposed the Corestandard to the American National Stan-dards Institute (ANSI). Core achieved verywidespread acceptance among softwaresuppliers, hardware manufacturers, andthousands of users, but never became aformal standard. Its concepts, however,served as a foundation for the GraphicalKernel System (GKS) that only recentlybecame the first true international graphicsstandard when it was formally adoptedby the International Standards Organization(ISO) and ANSI.

Presently, Core and GKS are the onlytwo standards that have any immediateimpact on the buyer's software strategydecisions. Basically, the choices in toolspackages are between those based on thewell -proven Core model and the newer,but as -yet incomplete, GKS. Some userswill have both, and will choose betweenthem for each application the way theywould choose between programming lan-guages, because each system has its prosand cons for any one application. Core's3-D capability gives it the edge in manysituations, but GKS, which currently isonly a 2-D system, is slowly gaining accep-tance and is likely to be the basis forseveral standards that are now beingconsidered.

One of these is PHIGS, the Program-mer's Hierarchical Interactive GraphicsSystem. PHIGS is a proposed standardfor high-performance 3-D graphics. TheComputer Graphics Interface (CGI) willattempt to establish standard ways fordevice -independent software subroutinepackages to talk and listen directly andefficiently to advanced devices such asinteractive workstations, in addition to theconventional hardware that presentlyrequires device drivers.

The Computer Graphics Metafile(CGM) and Initial Graphics ExchangeSpecification (IGES) attempt to establishstandards for storing, transmitting, andretrieving graphical data. If surrmful, thesestandards will simplify the exchange ofpictures that were created or recorded bydifferent sources.

Adherance to the standards has manybenefits to any organization that has manyusers and applications. By standardizingon a few common graphics software tools,the organization can cut training costs anddevelop a pool of programmers withgraphics experience who can help newgraphics users get their applications intoproduction faster. Application code basedon a standard is much easier to maintainthan highly customized programs.

But following the graphics standardsoffers no assurance of being able to takeadvantage of the best available graphicstechnology. Because of the process bywhich the standards are established, theyevolve slowly through an arduous processof compromise and negotiation, laggingseveral years behind both hardware tech-nology and market expectations. As currentpractice races ahead of the standards,software developers are facing increasingpressure to extend the functionality of theirtools offerings beyond that mandated by


the standards. Some packages are marketedwith claims of compliance to standardsthat have not yet been officially sanctioned.The risk to the buyer is that applicationsoftware and graphical databases built withthese products might not be compatiblewith the final versions of the standards.

Within the boundaries of the standards,there is significant room for tools devel-opers to use their software engineeringskills to make their packages easier to usethan others. Some designers are better thanothers at making their packages easy todebug, test, and maintain. In the case ofGKS, some of the first packages that havereached the market are only partial imple-mentations of the standard, and may nothave been designed for optimal growthinto the higher levels, or into the 3-Dextensions of GKS that are now beingformulated by the standards -making bodies.

In the final analysis, consider the stan-dards as guidelines, not gospel. If goodgraphics software offers you conformanceto one or more of the official standards,so much the better, but remember that itis more important for the software to beeffective and easy to use in your environment.

Key functionality to look for ingraphics software toolsAbove all, a tools package must solveyour problem-it has to work, completelyand reliably, in your application. The toolspackage must have proven compatibilitywith your CPUs, operating systems, andlanguages. It must also provide some wayfor you to acquire or produce device driversoftware for every device that you intendto use.

The process starts with honest evaluationof your needs. Then, look to a softwarevendor with a quality reputation, goodhistory, and sound financial situation. Thesalespeople should know their softwareand be able to work with you on yourapplication needs. Above all, be sure youget the names and phone numbers of peo-ple in your industry currently using theproduct you are evaluating. Call them.Check the documentation. Call the helpline. Ask about training. The quicker youare up and running, the quicker you putgraphics to work for you.

Documentation

Scrutinize the documentation as carefullyas the package itself. While the quality ofsoftware manuals has steadily improved,vendors often presuppose too much knowl-

Application Program

Device IndependentGraphics Tools

Package

Device Driver

Application Database(real -world data)

Graphics Database(virtual data)

Virtual Device Interface

GraphicsDevice

Device -LevelDisplay List

Fig. 3. System configuration with metafile capability.

edge on the part of the reader, particularlywhen he or she is a programmer. There-fore, demand a copy of the manual beforeyou sign the contract, and look it overcarefully.

You shouldn't have to be a "graphicsprogrammer" to begin using a graphicspackage. The documentation should pro-vide an overview of standard textbookgraphics concepts. Just as important, themanual should show how those conceptsare implemented by the software. Technicalterms should be defined concisely, andwhenever possible, abstract topics shouldbe explained using real -world analogies.For example, the concept of "virtualcamera" helps clarify three-dimensionaltransformations, while a metafile can bethought of as a picture library.

Documentation should include both areference and a tutorial section. Referencedocumentation should not only explainall the capabilities of the package, butshould account for special contingencies.What happens, for example, if the virtualcamera is positioned coincident with thepoint it is viewing? If ten picture segmentsare stored, in what order are they displayedon the screen? If you are looking straightdown on an object, which way is up? Afew sentences describing the not -everydayoccurrences can save hours of writing testprograms.

How to evaluate graphicssoftware documentation

E Overview of textbook graphicconcepts.

E How concepts areimplemented by the software.

Reference and tutorialsections.

Illustrations-the more thebetter.

E Programming examples-themore the better.

E Answers to "what if" and"how to" questions.

Detailed documentation oneach device driver.

Evaluate the documentation ascarefully as you evaluate thesoftware.

Tutorial documentation should walk theprogrammer through several programs,narrating each step of the way. For graphicsmaterial, of course, an abundance of illus-trations is a must. Also highly useful areprogram modules that perform commonlyused functions: rotating an object, menuselection using graphics input, and clearing

Warner/Slitz: How to evaluate and shop for computer graphics software 87

part of the screen. Such examples willsave the programmer from having to createcode that is already perfected.

Device driver documentation will alsoincrease programmer productivity. High-performance devices do much of the workin firmware, and they all do it differently.Good driver documentation shows the pro-grammer how the device -independent capa-bilities of a tools system relate to thespecific features of a given device.

Training and user groupsVendors take different approaches tocustomer training. Optimally, a variety ofcourses are offered-from basic ones de-signed for managers to advanced sessionstailored for experienced programmers.Inquire as to where the courses areheld-at the vendor's headquarters, at yoursite, or either. Sometimes, classes are heldon a regional level as well.

User groups are an excellent way tobecome familiar with the product, bothbefore and after purchase, because theyallow you to rub shoulders with systemdesigner involved with a cross-section ofapplications.

Can you make the commitment?The final question on your list should notbe directed at your vendor, but atyourself-can you commit enough humanresources to make a graphics tools packagework? Like most worthwhile endeavors,graphics application programming takes

time and commitment-often a fewmonths, sometimes in excess of a year.

Maintenance is also a factor. You shouldhave the in-house expertise to install newdevice drivers when additional equipmentis purchased, and to incorporate new fea-tures as the need arises.

James Warner is President and founder ofPrecision Visuals, a supplier of productsthat aid programmer and end -userproductivity in scientific and engineeringapplications. He has been active incomputer graphics as an educator,developer, vendor, and user for 14 years.Prior to forming Precision Visuals in 1980,Mr. Warner held positions at theUniversities of Michigan and Colorado.

Mr. Warner is a regular contributor tocomputer graphics publications, writingarticles and tutorials on graphicsstandardization, hardware implementationstrategies, and the uses of graphicssoftware tools. He is a frequent speaker atmajor computer conferences here andabroad.

John Slitz is Vice President of Marketingat Precision Visuals. Prior to that he heldseveral positions at IBM. He has a BA inEconomics from the State University ofNew York at Cortland, an MALS inPsychology/Sociology from the GraduateFaculty of the New School for SocialResearch, and an Executive MBA inManagement from Fairleigh-DickensonUniversity.Contact them at:Precision Visuals6260 Lookout RoadBoulder, Colorado 80301(303)530-9000


©ETD U© ©Woff the job

S. Kreitzberg J. Krupa

Engineering an electrically -powered automobile

Engineers are inveterate tinkerers, and as a result there is aChevrolet Vega in Moorestown, N.J. that will never be thesame . . .

After sitting in gas lines in the 1970s, wedecided to answer an advertisement in ado-it-yourself magazine, headlined "ConvertYour Car To Electric-Send for Our Plan."It looked like the perfect project forelectrical engineers who loved to tinkerwith cars.

The plans were quickly obtained andjust as quickly discarded. They wereunclear and geared toward converting oneparticular model of car. We had to designour own conversion. To begin, we posteda notice on the RCA Moorestown bulletinboard to enlist the aid of some co-workers.To our surprise, we received 43 positiveresponses, and an RCA electric car clubwas born.

Members ranged from people who hadnever opened the hood of a car to a manwho had designed and built an electrictractor. Employee Relations, through theMoorestown Recreation Association, agreedto support us as a hobby club, and wewere on our way. Plans started to takeshape in regular meetings.

The car and its componentsFinding a suitable car to convert, thenumber one problem, was solved when

Abstract: Dozens of RCA Engineershave tried their hand at building anelectric car. Two, driven to succeed tellhow they actually completed a roadworthycar and generated many new ideas in theprocess.

©1985 RCA Corporation.Final manuscript received April 26, 1985Reprint RE -30-5-12

the number one Kreitzberg son departedfor college, leaving an old Chevrolet Vegabehind. The club had its car. It wasrelatively light in weight, with a four -speed transmission and ample space inthe engine compartment. The gasolineengine was removed and discarded (Fig.1), but the four -speed transmission andclutch were left intact according to ourplanned approach.

Next, we had to locate a dc electricmotor capable of moving a 2300 -poundvehicle at a reasonable road speed. Investi-gation revealed that others were using oldsurplus aircraft starter/generator unitswhich were made for continuous operationwith forced -air cooling. After several callsto aircraft junkyards on the west coast,we found a 20 -horsepower motor. This75 -pound monster came equipped withroller bearings on both ends of thearmature, a series/parallel (compound)field, and the capability to accommodatea dc input ranging from 24 to 96 volts.Numerous other parts such as relays, volt-age regulators, and motors were alsoobtained from aircraft sources. The batteriesselected were heavy-duty marine SearsDieHards which were designed for repeateddeep discharge.

Once the project got into full swing, itoccupied every evening after work, week-ends, and holidays. Family life became athing of the past. We were redesigning acar! Considering that the original designcost millions of dollars and took thousandsof man-hours, one can appreciate what ittakes to redesign about a third of thevehicle. In addition to replacing everythingin the engine compartment, the followinghad to be designed: a completely new

dashboard with voltage and current meters,a heating system for winter use (no morehot engine coolant available), an electricalpedal (formerly called the gas pedal) andassociated motor controller, an on -boardcharging system, battery holders with pro-vision for removing fumes emitted whilecharging, a 12 -volt accessory system withits own on -board charger, and an adaptorto mate the clutch to the electric motor.

The effort was quite rewarding, however,since our hard work contributed to whatwill undoubtedly be the way of the future.Our club motto became "The future isnow."

Coincidentally, the group learned of asimilar student project at CinnaminsonHigh School. After speaking to the teacherin charge of the project, we invited themto join our club. From that day, our meet-ings were filled with students eager todiscuss and learn about electric cars fromRCA personnel. We helped them a greatdeal, but no less than their enthusiasmhelped us to continue with this project.Many of these students went on to engi-neering colleges after graduation. We hopeto see some of them back at RCA againas engineers.

Ready for the proving groundSomehow, the big day finally arrived. Thecar, shown in Fig. 2, was finished andready for testing. It had operated stationary(transmission in neutral) many times, butthis was the ultimate test. Batteries werefully charged, and specific gravities werecarefully checked. Tires were pumped to36 pounds to minimize rolling friction.We all held our breath. Shift into first


Fig. 1. With the gasoline engine removed, the Vega'sengine compartment yields ample space for conversion.

gear, step on the electric pedal, release theclutch, and away we go-it worked! Thecar rode beautifully. It had enough torqueto get rolling up to about 30 miles perhour, but we soon ran out of power, so itwas back to the drawing board. Thisproblem was solved simply by going froma 48 -volt to a 72 -volt battery system.

Our top speed was then over 40 mph,a reasonable level for local driving. Webegan to use the car for daily commutingto and from work as part of the debuggingprocess. The sound emitted by this vehicleis somewhat like a loud elevator whir, soneedless to say, it generated a lot ofattention wherever it went.

Our trip to the N.J. State InspectionStation was an especially memorable occa-sion. The first electric car under theirinspection, a curiosity, was quickly sur-rounded by the entire inspection staff pluseveryone who had been waiting in line.The car passed with flying colors on thefirst try.

Into the record booksIt didn't take long for two newspapersand WPVI Action News to contact theclub, via RCA. We also received a requestfrom RCA Moorestown to exhibit the carfor Open House. For this occasion, weupgraded the car (Fig. 3) by transferringthe entire electric drive system to anotherVega chassis of the same year but in bettercondition, with a newly painted body.

Moving from the outside upgrade to

Fig. 2. The first model produced by RCA's electric carclub is ready for the road.

Fig. 3. An improved car on display for the RCA Moorestown Open House. Theentire electric drive system is emplaced in a better chassis.

the inside "nuts and bolts," the followingparagraphs present the more detailed tech-nical aspects of our design and discussfeatures planned for future models.

The technical detailsWhen converting a conventional gasolineengine car to an electric drive, the standardtransmission and differential are usuallyleft intact. This greatly simplifies the conver-

sion process. Unfortunately, the speed -torque characteristics of the gasoline engineare not similar to those of the electricmotor. Therefore, we had to match theelectric motor characteristics to the existingtransmission requirement. Calculationsshowed that a 3:1 speed reduction of theelectric motor through a planetary gearsystem would be satisfactory. Accordingly,we obtained a 3:1 planetary gear reductionfrom a Ford automatic transmission, mod -


Fig. 4. A planetary gear system provides a fixed 3:1 speed reduction of theelectric motor to match the transmission requirement. (Since this reduction istoo high for normal highway use, a variable speed ratio system is now beingdeveloped.)

ified it, and connected it between theelectric motor and the standard four -speedtransmission. The electric motor is con-nected to the sun gear, the ring gear isheld stationary, and the 3:1 speed reductionfrom the carrier is connected to the trans-mission. This is shown in Fig. 4. Theplanetary gears run in oil and provide aquiet, efficient speed reduction.

The car performs very well with thisplanetary gear. This speed reduction is, how-ever, too high for normal highway use.We soon realized that a variable speedratio would be preferable to a fixed ratio.To provide a variable speed, the planetarygear reduction is currently being replacedwith a belt drive. The electric motor isnow to be mounted on the fender wall,and the belt will drive a pulley mountedon the four -speed transmission. There aretwo sprockets on the electric motor shaftand two on the transmission box. Thusany one of four reduction ratios may beselected by simply lifting the belt tensionerarm and moving the belt to the desiredsprocket.

This speed reduction system furnishesother advantages. The problem of aligningthe electric motor to the four -speed trans-mission is eliminated, and the electric motoris now readily accessible for maintenance.

Ideally, the transmission for an electricdrive automobile should consist of aninfinitely variable speed ratio-the dcmotor operates most efficiently when runwithin its nominal rating. The efficiencyof the dc motor drops rapidly when it isoverloaded, and this usually happens duringacceleration. Unfortunately, such a trans-mission is not currently available; accord -ingly, several club members have under-taken the task of designing and buildingsuch a transmission. The system consists

of an integral package containing theelectric motors and planetary gear trains.

An alternative to totalconversion-the hybrid carA small auxiliary gasoline engine mayalso be connected to the system, creatingthe popular hybrid electric car. With thisarrangement, the car is powered on levelroads by a small gasoline or diesel enginewhich drives a dc generator. For uphillclimbs or acceleration the 72 -volt batterieswill automatically provide the extra power.

Many hybrid cars are in operation atthe present time. The owners in manycases register more than 70 miles per gallonof fuel. With the integral package con-taining the dc motors and the matchedtransmission, conversion will be mucheasier and less expensive. The driving rangeof the hybrid car is almost limitless because,whenever the car does not require the fulloutput from the engine, its power willautomatically charge the batteries.

Related development projectsThe following paragraphs describe someother development projects undertaken: animproved mechanical speed control, a newtype of primary battery, and an entirelynew approach to converting electricalenergy into mechanical energy.

Speed control

We designed and built several differentsolid-state speed controllers using silicon -controlled rectifiers (SCRs) or high -currentDarlington transistors. These units canswitch in excess of 350 amps. They arereliable and provide very smooth speed

Fig. 5. Front view of a mechanicalspeed controller. The five -inch rotatingdisc provides a variable voltage controlon only 15 watts of power.

control for the electric automobile. Onedrawback with the solid-state speed controlis that the power dissipated by the switch-ing elements can approach 200 watts whencontrolling 350 amps. A mechanical speedcontrol was designed to provide a verysimple, inexpensive, and efficient speedcontrol using only 15 watts of power. Theunit consists of a 5 -inch copper disc withfour isolating segments as shown in Fig.5. The disc is rotated at 1000 rpm by asmall permanent -magnet dc motor. Acopper -composition brush is held againstthe disc. As the disc rotates, it will makeor break the current depending on whetherthe brush is on the copper segment or theisolating segment. By moving the brushradially along the disc, the ON to OFFratio is varied, providing a variable voltagecontrol. Unlike other mechanical speedcontrols, the disc and brush are run entirelysubmerged in light oil. This eliminatesarcing, reduces the brush wear, and pro-vides cooling.

Primary battery

The greatest drawback of electric auto-mobiles is the heavy, bulky, and relativelylow capacity storage battery. With presentstorage battery systems the driving rangebetween the charges is approximately 40miles. The ideal battery would be onethat did not require recharging but wouldsimply be replaced, at a nominal cost,after perhaps 500 miles of service. Thiswould approximate filling up the fuel tankon a gasoline -driven automobile.

A primary battery is very simple, con-sisting of plates of two dissimilar alloys inan electrolyte. The potential differencesbetween the plates depends on the alloysused. If the plates are connected to an

S. Kreitzberg/J. Krupa: Engineering an electrically -powered automobile 91

external circuit, an electrical current isprovided. This type of battery cannot berecharged, but it will supply current untilthe plates are dissipated. The current canbe controlled by simply varying the depthof immersion in the electrolyte. The"current per pound" of alloy depends onthe alloys that are used. One such alloy,which was originally manufactured as analuminum solder, will provide approxi-mately 3 X 96,000 coulomb per 30 grams,or about 1210 ampere -hours per pound.The voltage of this cell is 0.8 volt. There-fore, one pound will provide about onekilowatt hour. A typical electric automobileconsumes about 6 kilowatts (100 amps at60 volts) when traveling at 40 mph. There-fore, 6 pounds of alloy would be requiredto drive the car 40 miles. This type ofbattery has one serious drawback. As cur-rent is drawn from the plates they becomecoated with a chemical film that decreasesthe current density, eventually reducing itto zero. To prevent the chemical filmfrom forming, an ultrasonic scrubber hasbeen tried. The preliminary results werevery encouraging. Presently, a method ofdirecting the ultrasonic waves at all theplates from a single generator is beinginvestigated.

New type of electric motor

The new alloy Nitinol, which has unusualcharacteristics, makes it possible to convertheat directly into mechanical energy. Apiece of wire about 1/16 inch in diametermade from this alloy is very pliable andcan be easily bent into a semicircle. If thissemicircle is immersed in hot water (about160°F), it will snap back into its originalstraight form with unusual force.The company that manufactures this alloyhas designed a small wheel -motor thatwill rotate as long as part of the wheel isimmersed in hot water.

The potential use of this alloy for ahigh -efficiency propulsion device for electricautomobiles has been investigated. Insteadof using hot water, the wire was heatedby passing a pulse of current through it.The results were the same-the wire revertsto its original shape. To increase themechanical force, thicker wires were used.However, the thicker wire required moreforce to bend and did not revert to thesame original straight shape. To rectifythis, many thin straps of the alloy wereconnected together at the ends, makingthe composite strap very pliable and easyto bend. A piece of this strap (approxi-mately 1/4 -inch square) exerts a force ofover 10 pounds. Additionally, when the

Sid Kreitzberg joined RCA in 1962 andhas worked in the Program ManagementOrganization on the TALOS, BMEWS, andAEGIS programs. In 1979 he becameinterested in electric cars, and subsequentlyfounded the RCA Electric Auto Association,of which he is President. The club hassince become the Alternate Energy Cluband is currently experimenting with suchdiverse items as fuel cells and super -high -mileage carburetors as well aselectric cars. Mr. Kreitzberg, a SeniorMember of the Engineering Staff, is currentlyengaged in project management activityon the AEGIS Operational Readiness TestSystem (ORTS) and the CG53 UnderwaterShock Trails.Contact him at:RCA Missile and Surface Radar DivisionMoorestown, NJTACNET: 224-2842

John E. Krupa joined RCA Montreal in1953 and transferred to RCA Moorestownin 1959. He has been associated withmost of Moorestown's projects as a SignalProcessor Design Engineer. Currently, heis with the System Integration and Testactivity. Mr. Krupa became interested inelectric drives while serving with the RoyalAir Force in World War II. More recently,he designed the dc motors and controlsfor a 140 -hp diesel electric tractor. Nowhe is designing an integrated electric drivefor a 300 -hp electric tractor. His hobby-toy electric trains.Contact him at:RCA Missile and Surface Radar DivisionMoorestown, NJTACNET: 224-3024

strap was inserted in a strong magneticfield, the response to the same currentpulse was much faster, allowing morecycles per second. Development work withthis unusual alloy is continuing. The goalis to develop a method of converting theseimpulses into smooth rotary motion.

The future electric automobileThe popularity of the electric car hasfluctuated since its inception just beforethe turn of the century. Several years ago,during the gasoline shortage, there werethousands of these vehicles on the road.Many large manufacturers were preparingfor full-scale electric automobile produc-tion. With the sudden abundance of gaso-line and declining prices, many plans forelectric car production have been put inlimbo. However, strong interest andresearch in electric automobiles iscontinuing.

The future electric car will featureelectric motors designed specifically forautomobile use. The motor will be anintegral part of the wheel, and independentfour-wheel drives will be used. This prin-ciple has been successfully demonstratedon a large four-wheel drive tractor. Theprime source of power will almost certainlybe the low -temperature, hydrogen fusiongenerator. In general, nuclear physicistsdo not consider this generator a workablesource of power because of its low outputand efficiency. Nonetheless, in light of thefact that one gallon of sea or lake watercontains enough deuterium to drive a carfor 20 hours at 40 mph, this low -efficiencyfusion generator looks very attractive. Withon -board power no longer a problem, thesize of the electric automobile will increase,and acceleration will be compatible withthe present gasoline -driven car . . . and,oh yes, the electric models will be availablein any color-except black.


Engineering News and Highlights

Hospodor named President of Stockton is new Division VP Webster named to new postAmericom and GM at ASD

The election of Andrew T. Hospodor asPresident and Chief Executive Officer ofRCA American Communications, Inc., hasbeen announced by Eugene F. Murphy,RCA Executive Vice President of Communi-cations and Electronic Services. Mr. Hos-podor succeeds Dr. James J. Tietjen, whohas become Vice President, RCA Labora-tories. Since 1981, Mr. Hospodor has servedas Division Vice President and GeneralManager of Automated Systems Divisionin Burlington, Massachusetts, where hehad overall responsibility for the designand manufacturing of automatic test equip-ment, vehicle test equipment, and commandand control systems.

Mr. Hospodor received his BS in Mech-anical Engineering from Cornell Universityin 1960, and his MS in Mechanical Engi-neering and MBA from Lehigh University.In 1976, he completed the Harvard Univer-sity Program for Management Development.

He is a member of the Association of theUnited States Army, the Armed Forces Com-munications and Electronics Association,the Air Force Association, the Associationof Old Crows, the Army Aviation Associationof America, and the Navy League of theUnited States. He is also a member of theAmerican Defense Preparedness Associa-tion, the American Helicopter Association,Inc., the American Institute of Aeronauticsand Astronautics, and the Security AffairsSupport Association.

w.

John D. Rittenhouse, RCA Executive VicePresident, Aerospace and Defense, hasappointed Eugene M. Stockton as DivisionVice President and General Manager ofRCA Automated Systems Division. Auto-mated Systems Division designs and manu-factures a wide range of automatic testequipment, vehicle test equipment and com-mand and control systems. It was elevatedto Division status earlier this year by RCAPresident and Chief Executive Officer RobertR. Frederick.

Prior to his appointment, Mr. Stocktonwas Division Vice President, Engineering,at ASD. He was responsible for engineeringservices and for C3I programs including theRemotely Monitored Battlefield Sensor Sys-tem (REMBASS). Mr. Stockton's engineeringorganization also incorporated artificialintelligence techniques into command anal-ysis functions. Since joining RCA AutomatedSystems in 1955 as an Associate Engineer,Mr. Stockton has served in a number ofengineering management positions, includ-ing nine years as the Division's ChiefEngineer.

Mr. Stockton holds a BS in Electrical Engi-neering from Michigan State University. Hereceived an MSEE in electrical engineeringfrom Drexel University in 1959, and an MBAfrom Western New England College in 1981.He also completed the Harvard BusinessSchool Advanced Management Program.

The appointment of Dr. William M. Websteras Vice President and Senior TechnicalAdvisor has been announced by Roy H.Pollack, Executive Vice President, ElectronicProducts and Technology. Dr. Webster, whohas served as Vice President, RCA Labora-tories, since 1968, will now have responsi-bility for providing technical advice and coun-sel to all RCA activities.

He has been a leader in the field of solidstate physics, and joined RCA Laboratoriesin 1946, making numerous contributions totube and transistor developments. From1954 to 1959, he was Manager of AdvancedDevelopment for the RCA Semiconductorand Materials Division. He returned to RCALaboratories as Director of the ElectronicResearch Laboratory in 1959. He was ap-pointed Staff Vice President, Materials andDevice Research in 1966, and assumedresponsibility for the management of RCALaboratories in 1968. He was elected aCorporate Vice President in 1969.

Dr. Webster holds patents in such diversefields as television, vacuum tubes, gastubes, circuitry, and semiconductor devices.A member of the National Academy ofEngineering, Dr. Webster was awarded theFrederik Philips Award by the Institute ofElectrical and Electronics Engineers in 1980"for sustained leadership in the manage-ment of research and development."

He received a BS in Physics from UnionCollege in 1945, and a PhD in ElectricalEngineering from Princeton University in1954.


Tietjen to Head RCALaboratories

Dr. James J. Tietjen, who has headedRCA American Communications, Inc., since1983, has been appointed Vice President,RCA Laboratories by Roy H. Pollack, Execu-tive Vice President, Electronic Productsand Technology. He will be responsible forall of the research activities of RCA Labora-tories, and will report to Mr. Pollack.

Dr. Tietjen is widely known for his workin materials and components research. Anative of New York City, he received hisBS, cum laude, in Chemistry from lonaCollege in 1956. He was awarded his MSand PhD degrees in Physical Chemistry in1958 and 1963 from Pennsylvania StateUniversity.

After joining RCA Laboratories in 1963,Dr. Tietjen worked primarily on preparationof a broad range of semiconductors bychemical vapor transport reactions. Thiswork led to the development of a series ofnew and improved electronic devices. Hehas published more than 30 technicalpapers in the area of materials and com-ponents research.

He was appointed a Research GroupHead in 1969, and was named Director ofthe Materials Research Laboratory in 1970.In 1977 he was appointed Staff Vice Presi-dent, Materials and Components Research,the position he held until being namedPresident and Chief Operating Officer ofRCA American Communications, Inc., in1983. He became President and ChiefExecutive Officer in 1984.

Dr. Tietjen was twice a recipient of RCA'shighest technical honor, the David SarnoffAward for Outstanding Technical Achieve-ment, in 1967 and 1970. He was alsoawarded RCA Laboratories OutstandingAchievement Awards in 1965 and 1969. InJune of this year he was a recipient of thePennsylvania State University DistinguishedAlumnus Award, the university's highesthonor.

Pate named SSD Ed Rep Galton is named ASD TPA

Virginia Pate joined the Standards Groupof RCA Solid State in 1978 as a TechnicalWriter, with responsibility for maintenanceof Engineering Specifications. She later trans-ferred to Engineering Publications, and waspromoted to Technical Writer/Editor. Herinitial duties involved the preparation of datasheets, catalogs, and Databooks coveringSolid State products. Most recently she hasbeen named Project Editor for publicationsconcerning semicustom IC products, andhas been given additional responsibilitiesin the preparation of promotional news-letters and brochures as well as applicationnotes and technical manuals.

Ms. Pate received a BS in Chemistry(Honors), with a minor in Mathematics, fromMorgan State University in 1963. She subse-quently worked in the fields of analytical,research, and quality -control chemistry withcompanies such as Reed and Carnrickand CI BA Pharmaceutical, as well as withJohn Hopkins University.

Eugene B. Galton, Manager, EngineeringStaff, has been appointed Technical Publi-cations Administrator at RCA AutomatedSystems Division, Burlington, Mass. Mr.Galton joined RCA in 1956 as an Engineer-ing Group Leader in Camden, New Jersey,and has held positions as Engineering Man-ager, Plant Manager, Marketing Manager,and PMO Manager with responsibilities inall of the product areas at RCA Burlington.Prior to joining RCA, Mr. Galton worked forBell Laboratories and the Hazeltine Elec-tronics Corporation as a design and devel-opment engineer on telephone headsets,IFF equipment, and airborne and shipboardelectronics.

He is a graduate of Cornell Universitywith an AB in Chemistry, and a BEE (withdistinction). He also received an MEE fromNYU and an MBA from Temple University.He is a member of Eta Kappa Nu, Tau BetaPi, Phi Kappa Phi, Association of Old Crows,and a Senior Member of the IEEE.

Staff announcements

Aerospace and Defense

John D. Rittenhouse, Executive Vice Pres-ident, Aerospace and Defense, announceshis organization as follows: William V.Goodwin, Division Vice President and Gen-eral Manager, Missile and Surface RadarDivision; Eugene M. Stockton, Division VicePresident and General Manager, AutomatedSystems Division; Joseph Pane, DivisionVice President and General Manager, Gov-ernment Volume Production; Lawrence J.Schipper, Division Vice President and Gen-eral Manager, Communication and Infor-mation Systems Division; Charles A. Schmidt,

Division Vice President and General Man-ager, Astro-Electronics Division; Joseph C.Volpe, Division Vice President and GeneralManager, Broadcast Systems Division;James B. Feller, Staff Vice President, Tech-nology; James R. Foran, Staff Vice President,Business Planning; Leibard V. Fox, StaffVice President, Finance; Donald L. Gilles,Staff Vice President, Employee Relations;Joseph B. Howe, Staff Vice President, Sys-tems and Manufacturing; and Francis H.Stelter, Jr., Staff Vice President, Marketing.

James B. Feller, Staff Vice President, Tech-nology, announces his organization as fol-

94 RCA Engineer 30-5 Sept. Oct. 1985

Kriesman is ASD Ed Rep

Linda Kriesman, Publications Engineer atRCA Automated Systems Division, Burling-ton Massachusetts, has been named Edi-torial Representative to the RCA Engineer.Ms. Kriesman joined RCA in 1984, and isresponsible for the publication of propos-als, technical reports, and conferencepapers. She is also strongly involved inwriting and directing ASD video productions.Prior to joining RCA, Ms. Kriesman workedas a technical writer for a training firm inWashington state, where she wrote over35 user manuals on such diverse subjectsas chemical preparation, paper machinetroubleshooting, and automatic control sys-tems.

Ms. Kriesman served as an IntelligenceOperations Specialist and Czech linguistfor four years with the U.S. Army. She nowserves as an Intelligence Specialist withthe Naval Intelligence Reserve at SouthWeymouth Naval Air Station.

Ms. Kriesman graduated summa cumlaude from Boston University with a BS inBroadcasting and Film. She is currentlypursuing a Master's Degree in Technicaland Professional Writing at NortheasternUniversity.

lows: Ronald A. Andrews, Staff Vice Pres-ident, Advanced Technology Laboratories;John M. Herman, Staff Vice President, SolidState Technology Center; Thomas A. Martin,Director, Technical Planning; and Paul J.Nicholas, Director, Special Programs.

Joseph B. Howe, Staff Vice President, Sys-tems and Manufacturing, announces hisorganization as follows: Joseph B. Howe,Acting, Manufacturing Planning; and DennisJ. Woywood, Staff Vice President, Systemsand Advanced Programs.

Dennis J. Woywood, Staff Vice President,Systems and Advanced Programs, an-nounces his organization as follows: PrestonN. Shamer, Manager, Advanced ProgramDevelopment; and Arthur R. Wentz, Tech-nical Director, Strategic Defense Initiative.

Americom

Eugene F. Murphy, RCA Executive VicePresident of Communications and ElectronicServices, announces the election of AndrewT. Hospodor as President and Chief Exec-utive Officer.

A. L. Edwards, Manager, Program Manage-ment, announces the appointment of R. A.Siddiqui as Manager, Major Programs.

Murray Fruchter, Director, Terrestrial Sys-tems, Technical Operations, announces theappointment of G. D. Abernathy as Manager,Equipment Engineering.

Muhammad Ashraf, Manager, Terrestrial Fac-ilities Engineering, announces his organi-zation as follows: Frank W. De Mille,Manager, Digital Systems; Henry E. Foon,Manager, Commercial Systems; andLeonard Derkach, Manager, Facilities andConstruction.

Muhammad Ashraf, Manager, Terrestrial Fac-ilities Engineering, announces the appoint-ment of Frank W. De Mille as Manager,Digital Systems.

Astro-Electronics DivisionCharles A. Schmidt, Division Vice Presidentand General Manager, announces theappointment of Frank A. Boyer as Manager,Astro Quality Process Improvement andPlanning.

Communication andInformation Systems Division

Lawrence J. Schipper, Division Vice Pres-ident and General Manager, announces theappointment of Alfred C. Thompson as Div-ision Vice President, Business Development.

Lawrence J. Schipper, Division Vice Pres-ident and General Manager, announces theappointment of Bernard S. Sacks as Direc-tor, Product Assurance.

John F. Serafin, Division Vice President,Program Operations, announces the appoint-ment of Edy J. Mozzi as Director, FutureSecure Voice Systems.

H. Andre Carron, Director, ManufacturingOperations, announces the appointment ofVincent J. Mazzaglia as Manager, PlantEngineering.

Consumer ElectronicsOperations

D. Joseph Donahue, Vice President, Con-sumer Electronics Operations, announceshis organization as follows: Robert C.Arnett, Division Vice President, Manufac-turing; Bennie L. Borman, Division Vice Pres-ident, Manufacturing Engineering and Tech-nology; James E. Carnes, Division Vice Pres-ident, Engineering; Keith U. Clary, DivisionVice President, Consumer Electronics Emp-loyee Relations; Larry A. Cochran, Director,Monitor Operations; David E. Daly, DivisionVice President, Consumer Electronics Pro-duct Planning; Michael B. Evans, DivisionVice President, Consumer Electronics Fin-ance; John P. Keating, Division Vice Pres-ident, Purchasing; Harry Anderson, DivisionVice President, Program Management; andJames R. Smith, Division Vice President,Product Assurance.

Robert C. Arnett, Division Vice President,Manufacturing, announces his organizationas follows: Terry J. Burns, Director, Con-sumer Electronics -Mexican Operations andGeneral Manager, RCA Componentes-S.A.de C.V.; Robert E. Fein, General Manager,Productos Electronicos de La Laguna-S.A.de C.V.; G. Bruce Dilling, Plant Manager,Bloomington Plant; Dennis L. Dwyer, Direc-tor, Production Planning, Transportation andDistribution; David D. Eden, Director, CrownWood Products Operations; Kenneth D.Lawson, Director, Facilities Management;James D. MacKay, Plant Manager, RCA,Inc. (Canada); J. B. Thomas, Plant Manager,Indianapolis Plant; and Kenneth S. Williams,President and General Manager, RCA Tai-wan Limited.

Kenneth S. Williams, President and GeneralManager, RCA Taiwan Limited, announcesthe appointment of Glen E. Ernst as Direc-tor, Manufacturing Operations, RCA TaiwanLimited.

Kenneth S. Williams, President and GeneralManager, RCA Taiwan Limited, announcesthe appointment of Thomas A. Egold asManager, Materials, RCA Taiwan Limited.

Terry J. Burns, General Manager, RCAComponentes S.A. de C.V., announces theappointment of David R. Crawford as Man-ager, ACI Technology for RCA CornponentesS.A. de C.V.

James E. Carnes, Division Vice President,Engineering, announces his organizationas follows: James A. McDonald, Director,New Products Laboratory; Eugene Lemke,Staff Technical Coordinator; George C.Waybright, Director, Display Systems Engi-neering; Robert P. Parker, Director, SignalSystems; Richard A. Sunshine, Director,Mechanical Design Engineering; and WillardN. Workman, Director, Product Engineering.


Harry Anderson, Division Vice President,Program Management, announces theappointment of Alfred L. Baker as Manager,CTC140.

George C. Waybright, Director, Display Sys-tems Engineering, announces his organi-zation as follows: Basab B. Dasgupta, Man-ager, Magnetics Engineering; William V.Fitzgerald, Jr., Manager, Deflection Subsys-tems; Peter R. Knight, Manager, Deflectionand Power Supply; and David E. Laux, Man-ager, Advanced Yoke Development.

Stephen H. Morrall, Director, TelevisionProduct Planning and Development,announces the appointment of Edward E.Thompson as Manager, Television ProductLine Development.

J. B. Thomas, Plant Manager, IndianapolisPlant, announces his organization as fol-lows: James R. Arvin, Manager, Operations -Components; Elliott N. Fuldauer, Manager,Materials; James P. Gallagher, Manager,Plant Financial Operations; Peter C. Hill,Manager, Ferrite Operations; Randall R.Mitchell, Jr., Manager, Plant Quality Control;Richard E. Molyneux, Manager, Operations -Plastics; E. Rene' Parks, Manager, EmployeeRelations, Consumer Electronics -Indianap-olis; and Walter E. Todd, Manager, FacilitiesServices -Consumer Electronics -Indianapolis.

Larry A. Olson, Manager, ManufacturingTechnology Center, announces the appoint-ment of Robert A. Monat as Manager, Com-puter Integrated Manufacturing Systems.

Corporate Staff

Robert R. Frederick, President and ChiefExecutive Officer, announces the appoint-ment of George D. Prestwich to the newposition of Corporate Quality Executive.

Roy H. Pollack, Executive Vice President,Electronic Products and Technology, an-nounces the appointment of William M.Webster as Vice President and SeniorTechnical Advisor.

Distributor and SpecialProducts Division

E.A. Boschetti, Division Vice President andGeneral Manager, announces the appoint-ment of Richard J. Klein as Director, Engi-neering Quality Assurance.

Globcom

Anthony Longo, Manager, Central OfficeEngineering, announces that the EngineeringRecords activity is transferred from Construc-tion and Installation to Central Office Engineer-

ing. Anthony J. Fradella will continue asLeader, Engineering Records.

John P. Shields, Director, Network Engi-neering, announces the appointment of JohnA. Kruk as Manager, Project Engineering.

William A. Klatt, Director, Network Opera-tions, announces the appointment of JosephCantelmo as Manager, Gateway Operations.

Sharyn M. Yensko, Manager, Product Plan-ning and Development, announces theappointment of John P. Begley as ProgramManager.

Government VolumeProduction

Joseph Pane, Division Vice President andGeneral Manager, Government Volume Pro-duction, announces his organization asfollows: William J. Bodo, Director, Finance;Joseph B. Christopher, Director, Manufactur-ing Operations; Robert M. Lisowski. Man-ager, Technical Support; Paul R. Morrison,Manager, Contracts; Fred Picus, Manager,Special Programs; William L. Reid, Manager,Materials; Nicholas B. Sher, Manager, Prod-uct Assurance; Bennie E. Tyree, Director,SATCOM Programs; James D. Wilson, Man-ager, FSVS Programs; and James C. Woolley,Director, COMSEC Programs.

Robert M. Lisowski, Manager, TechnicalSupport, announces the appointment ofWilliam H. Turner as Manager, LogisticsSupport.

Fred Picus, Manager, Special Programs,announces the appointment of NormanHalem as Deputy Manager, Programs.

NBC

Michael Sherlock, Executive Vice President,Operations and Technical Services, an-nounces his organization as follows: ArthurTimpani, Director, Program Operation Ser-vices; S. Merrill Weiss, Director, BroadcastSystems Engineering; Paul N. Beck, Man-ager, Construction Contracts; Kim L. Caster,On -Air Technical Manager, Satellite Opera-tions; Ruediger Reinke, Facilities Manager;Doug Erb, Manager, Election Communica-tions; Ferdinand Galli, Technical FacilitiesPlanning Manager; Robert Martinez, Man-ager, Electronic Maintenance; MichaelMathews, Technical Manager, NetworkNews; Robert E. Bartnik, Technical Manager;Lee Goldman, Manager, Video Tape Opera-tions; Wayne Grennier, Manager, SportsTechnical Services; Alexis Janaro, Manager,Electronic Maintenance; and Richard Post,Technical Manager, Network News.

New Product Division

Erich Burlefinger, Division Vice Presidentand General Manager, New Products Divi-sion, announces his organization as follows:Don R. Carter, Division Vice President, TubeOperations; John H. Cook, Director, Em-ployee Relations; Mark L. Frankel, DivisionVice President, Business Development;Harold R. Krall, Division Vice President,New Product Development; Reginald R.Pattey, Director, Strategic Planning and Ser-vices; Ronald G. Power, Vice President,Emitters and Detectors, RCA Inc. (Canada);Carlton L. Rintz, Division Vice President,Closed Circuit Video Equipment; RandolphC. Rose, Director, Finance; and Eugene D.Savoye, Director, CCD and Silicon TargetTechnology.

RCA Laboratories

Roy H. Pollack, Executive Vice President,Electronic Products and Technology, an-nounces the appointment of James J. Tietjenas Vice President, RCA Laboratories.

Jon K. Clemens, Staff Vice President, Con-sumer Electronics Research, announces hisorganization as follows: Curtis R. Carlson,Director, Information Systems Research Lab-oratory; Jack S. Fuhrer, Director, TelevisionResearch Laboratory; David D. Holmes, Dir-ector, Advanced Television TechnologyResearch; Arthur Kalman, Director, DigitalProducts Research Laboratory; and Arch C.Luther, Senior Staff Scientist.

Bernard J. Lechner, Staff Vice President,Advanced Video Systems Research Labora-tory, announces the appointment of WalterH. Demmer as Head, Digital Video Research.

RCA Service Company

Donald M. Cook, President, RCA ServiceCompany, announces the appointment ofLeonard J. Schneider to the newly createdposition of Division Vice President, Con-sumer and Business Services and Systems.

Donald M. Cook, President, RCA ServiceCompany, announces the appointment ofJohn C. Phillips as Director, BusinessDevelopment and Services.

Leonard J. Schneider, Division Vice Pres-ident, Consumer and Business Services andSystems, announces his organization asfollows: A. Gene Chumley, Director,Operations-Business Communications Ser-vices; Richard L Layton, Director, Engineer-ing and Product Development; Earle A. MalmII, Division Vice President, Marketing andSales; J. William McGee, Division Vice Presi-dent, Field Operations-Consumer and Com-mercial Services; and Leonard J. Schneider,Acting, Planning and Administration.


1.1=M P Pp r

J. William McGee, Division Vice President,Field Operations, announces the appoint-ment of Roy A. McFadden as Director,Industrial Electronic Services and OperationsSupport.

Francis J. Hanson, Division Vice President,Contracting and Operations Analysis,announces his organization as follows:James H. Stryker, Manager, OperationsAnalysis; William A. Corner, Manager, Valueand Quality Engineering; and Philippe H.Melroy, Manager, Contracting.

Richard L. Layton, Director, Engineeringand Product Development, announces hisorganization as follows: John W. Bakas,Manager, Product Evaluation and RegulatoryAffairs; Paul R. Dicke!, Manager, Regulatoryand Inter -Company Liaison; John I. Matheus,Manager, Future Secure Voice Systems;George A. Drechen, Manager, Field Engi-neering Services; Joseph D. Kilpatrick,Manager, Communications Systems Plan-ning; Montgomery C. Teague, Adminis-trator, Product Development-Healthcare;and Walter F. Turley, Manager, ProductDevelopment.

Solid State Division

Robert P. Jones, Division Vice President,Power Products and International Manu-facturing, announces the following titlechanges: Donald E. Burke, Director,Engineering-Power Products, and GeorgeW. lanson, Director, Product Marketing-Power Products.

John R. Steiner, Director, Environmentaland Plant Engineering, announces theappointment of F. Douglas Rue as Manager,Plant Engineering-Somerville.

Professional activities

Parker receives Eta Kappa Nu award

Inventor Recognition Program

The RCA Corporate Inventor RecognitionProgram was established in 1973, and eachyear recognizes those RCA inventors whoare granted U.S. patents based on theirRCA work. Milestone awards of jewelry,plaques, desk accessories, or crystal glass-

ware are made to inventors when theyreceive their 1st, 5th, 10th, 20th, 30th and50th patents. The program provides for theinformal presentation of a plaque in theform of a replica of an inventor's first U.S.patent upon the granting of that patent.

Zurich inventors: Dr. Walter J. Merz, John R. Sandercock(5), Peter E. Haferl (30), and Ronald A. Mills.

The most recent RCA employee to receivean Eta Kappa Nu award is Robert P.Parker, Director of Signal Systems, Con-sumer Electronics Operations. He was oneof three individuals selected for Honora-ble Mention in 1984. Mr. Parker was citedfor "his contributions to the fields of colortelevision technology and engineering man-agement, and for his involvement in com-munity activities." He is shown here receiv-ing the award at a ceremony held in NewYork City. Also shown are (left to right)Cecelia Jankowski, Grumman AerospaceCorporation (Honorable Mention), DonaldChristiansen, Publisher of IEEE Spectrum,(Master of Ceremonies), and Earl Steele,President of Eta Kappa Nu.

Awards for the other milestones are pres-ented at award dinners held each year inPrinceton, Lancaster, Indianapolis, CherryHill, Somerville, Burlington, and Zurich.

In 1984 there were 437 U.S patents grantedto 322 different RCA inventors. Of thisnumber, 73 received their first patent, 33their fifth, 21 their tenth, 5 their twentieth,and 4 their thirtieth. Those who received anaward at one of the award dinners this yearare shown in one of the group photographs.The numeral following each person's nameindicates the milestone award received for1984.

Those who received awards for five or morepatents but are absent from the photographsinclude: Allonse A. Acampora (5), John F.Corboy, Jr. (5), Thomas J. Faith, Jr. (5),Lubomir L. Jastrzebski (5), Ronald W. Kipp(5), George L. Schnable (10), Minoru Toda(20), Leon J. Vieland (5), and George J.Whitley (5).


Cherry Hill inventors: Top Row-George S. Zorbalas (5); CharlesW. Reno (5); Kevin J. Phillips (5); John D. Rittenhouse, ExecutiveVice President, RCA Aerospace and Defense; Charles E. Profera(5). Bottom Row-Alfred Schwarzmann (10); Kenneth C. Hudson(10); Leslie A. Torrington (20); Willard T. Patton (5).

Lancaster inventors: Top Row-Gilbert N. Butterwick (10);Larry M. Hughes (10); F. Carl Farmer, Jr. (5); Donald W. Bartch(10); Erich Burlefinger, Division Vice President and GeneralManager, New Products Division; Martin K. Brown (5). BottomRow-Robert E. McHose (5); Charles M. Tomasetti (5); StanleyA. Harper (10); Paul P. Webb (5); Myron H. Wardell, Jr. (10).

Somerville inventors: Top Row-William C. Hittinger, ExecutiveVice President, Robert E. Wilson (10); Carl R. Turner, DivisionVice President and General Manager, Solid State Division.Bottom Row-Joel R. Oberman (5); Robert H. lsham (5);Bhupendra P. Patel (5); Robert C. Shambelan (5).

Indianapolis inventors: Top Row-William V. Fitzgerald, Jr.,(10); Thomas F. Kirschner (5); Billy W. Beyers, Jr. (10); Clyde F.Coleman (10); D. Joseph Donahue, Vice President, ConsumerElectronics Operations. Bottom Row-Larry A. Cochran (10);Jack S. Fuhrer (10); Robert P. Parker (10); Robert L. Shanley, II(30); William A. Lagoni (10).

Princeton inventors: Seated-Edward A. James (5); Roger G. Stewart (30); Eugene S.Poliniak (20); Mahesh Kumar (5); Markus Nowogrodzki (5); Leonard Schiff (5); Frank W.Wendt (5). Standing-Steven A. Lipp (5); Howard G. Scheible (5); Peter A. Levine (20);Philip M. Heyman (10); Donald J. Sauer (10); Kenneth W. Hang (20); Terrence R. Smith(5); Ashok N. Prabhu (10); Allen L. Limberg (30); Brain W. Faughnan (5); Werner Kern(10); Christopher H. Srolle (5); William M. Webster, Vice President, RCA Laboratories;Thomas V. Bolger (10); William C. Hittinger, Executive Vice President.


Technical excellence

Second-quarter MSRD TEC Awards

Best

Lampe

Osterman

DiCurcio

Mehling

Pugh

William J. BestFor outstanding contributions to the integ-ration and testing of the EDM-4 SignalProcessor. Mr. Best came to the assign-ment as a junior team member with no

previous integration experience, and hasbecome one of the team's most valuedmembers after only 18 months. Of particularnote are his expertise in microprocessorfirmware analysis and his ability to trouble-shoot all modes of the system.

John A. DiCiurcioFor extraordinary analytical and problem -solving accomplishments during the integ-ration and testing phase of the EDM-4Signal Processor development. His specificcontributions ranged from resolution ofhigh false -alarm rates and severe transientproblems to optimization of range biasvalues for special functions. His analyticalskills and innovative equipment modifica-tions have been key factors in establishingthe Signal Processor as a practical oper-ating subsystem.

Ross W. LampeFor major contributions to the IFF phasedarray antenna design for ADAR-M. He con-figured the antenna system, determinedelement spacing and array excitation, andverified the designs through simulation.Especially notable was his design of aprinted circuit folded dipole radiating ele-ment. Dr. Lampe also coordinated relateddesign activities to assure a workable sys-tem, and developed and maintained ex-cellent working relationships with customerrepresentatives.

John A. MehlingFor outstanding leadership in the softwaredevelopment for the PMTC AN/FPS-16radar system upgrade. In MSRD's firstapplication of a VAX 11/780 for real-time

operations, Mr. Mehling led the team thatovercame a variable and unpredictableinterrupt response problem, developing thesystem without modifying the vendor'soperating system. Their top-level systemarchitecture and design has resulted in ahighly modular, structured, and maintain-able software system.

Dale M. OstermanFor technical leadership and applicationof his systems expertise to the EDM-4Signal Processor integration and test effort.His intimate knowledge of the SignalProcessor, down to the frame and modulelevels, contributed heavily to his extra-ordinary skill in troubleshooting. This abil-ity, in combination with his personal dedic-ation, has made him the key contributor tothe integration and test of the EDM-4 SignalProcessor.

Mikel PughFor conceiving, designing, and implemen-ting a unique redundancy and reconfig-uration control system for the AN/SPY-1 BWaveform Generator. His innovative ap-proach enables the off-line generator tobe isolated from the on-line generator forrepair and fault isolation testing. His creat-ive approach to the timing control systemmade maximum use of PROMs to providea highly flexible timing control system withmajor advantages over previous designs.

The award winners will be honored at aspecial function to be held early next year,and each will receive a commemorativeplaque and a current text or referencebook.

Mountaintop TEC winners

Neilson VanHorn

John Neilson and Clyde VanHorn wererecipients of the RCA Solid State DivisionTechnical Excellence Annual Award. Theywere selected from nominees throughoutthe Solid State Division.

John Neilson is a 27 -year RCA employee,and has a BS in Physics from AlbrightCollege, and an MS in Physics from theUniversity of Pennsylvania. He is a SeniorMember, Technical Staff in the Design andProduct Engineering group. John presentlyholds thirteen patents and is a past recipientof the David Sarnoff Award and the Mountain-

top Technical Excellence Award. He wasinstrumental in the design of the high voltagesilicon rectifier, stud rectifier, and high volt-age stack rectifier devices. His design andimplementation of the first triac gave RCA atwo-year lead on the industry. Currently,John is active in the design and developmentof Power MOSFETS.

Clyde VanHorn, Senior Systems Analyst,has been the main architect for writing thesoftware that allows the interconnecting oftest equipment to the required inputs from


worldwide Marketing, Applications, Quality,and Sales. He has made numerous trips toMalaysia to assist in incorporating a com-puter database link to the U.S. A recipient ofMountaintop's Technical Excellence Award,Clyde is a 19 -year RCA employee.

Palm Beach Gardens

Frank McCarty and Chris Blauvelt of RCASolid State Division, Palm Beach Gardens,were awarded second-quarter individual

MEP meeting at Skytop

Anthony J. Btanculli, left, corporate coor-dinator for the MEP program, presentsHans K. Jenny with an award in recogni-tion of his long service to the program.Mr. Jenny is one of the five "pioneers"who conceived or implemented the MEPprogram ten years ago. The other fourpioneers are William J. Underwood, FrankMcClure, George R. Field, and Fred Dixon.

The tenth annual meeting of the leaders ofthe RCA Minorities in Engineering Program(MEP) was held at Skytop Lodge,Pennsylvania on August 13-15. This year'sspeakers included Roy H. Pollack, ExecutiveVice President Electronic Products and Tech-nology, William C. Hittinger, Executive VicePresident, and Edward L. Scanlon, SeniorVice President, Employee Relations.

MEP is a program to encourage minorityhigh school students to consider technicalcareers. The program operates at 14 RCAlocations and has, over the past decade,reached more than 2000 students. Technicalcoordinators at participating RCA locationsare responsible for student engineeringorientations, and an Employee Relationscoordinator is responsible for liaison be-tween RCA and the community and schools.

Technical Excellence Awards. Frank wasselected for specifying and implementinga real-time Engineering Analysis Compu-ter System utilizing HP3000, HP1000, andHP150 personal computers. This systemreduced engineering analysis time fromdays to hours. Chris was selected for thedevelopment of a pseudo -automatic test -program generator for the Fairchild testsystems. This developement resulted inthe Semicustom product line achieving a24 -hour turnaround time for design verifi-cation samples.

Americom TEC awards

Cosmo Loguidice, Member, EngineeringStaff, has been awarded the Technical Ex-cellence Award for outstanding efforts inthe complete reconfiguration of the multiplemultiplex system at the Pt. Reyes California

Marion TEC Award

Richard L. Gore received the TechnicianAward for the second quarter of 1985. Dick'sefforts were directed toward solving particle -caused problems in the factory. He initiatedand continues to use a mount cleanlinesstest, which has demonstrated a differencein cleanliness of mounts among mount ven-dors. A source of microspheres, whichcaused blocked apertures, was identifiedby this test.

Earth Station. This reconfiguration, withno interruption to our traffic, resulted inspace savings capable of accommodating28 additional racks in a location where nofloor space was available for expansion.

Want article reprints?Call Louise Carr. Tacnet: 226-3092

Want to write an article?Call Tom King, Tacnet: 226-3090


Editorial Representativesand Technical Publications Administrators

Contact your Editorial Representative at the Tacnetnumbers listed below to plan your RCA Engineer articleand to announce your professional activities.

Advanced Technology Laboratories*Merle PietzEd Master

Moorestown, New JerseyMoorestown, New Jersey

Aerospace & Defense Staff*Theresa Law

(Approvals only)

American Communications*Anita Labib

(Approvals only)Carolyn Powell

Astro-Electronics Division*Frank Yannotti

Carol Coleman

Cherry Hill, New Jersey

* Technical Publications Administrators, responsible forreview and approval of papers and presentations, areindicated here with asterisks before their names.

Tacnet National Broadcasting Company253-6429 *Bob Mausler New York, New York 324-4869

New Products Division*Art Sweet

Bob McIntyre Ste Anne de Bellevue 514-457-9000

253-6436

222-5833

Princeton, New Jersey 258-4346


Princeton, New Jersey 229-2544Princeton, New Jersey 229-2919

Automated Systems Division*Gene Galton Burlington, Massachusetts

Linda Kriesman Burlington, Massachusetts

Broadcast Systems Division*Harry Green Gibbsboro, New Jersey 266-3971

Consumer Electronics Operations*Eugene JansonJohn HopkinsLarry Olson

Indianapolis, IndianaIndianapolis, IndianaIndianapolis, Indiana

326-3775326-3001

422-5208422-5217422-5117

Corporate Information Systems & Services*Sue Handman Cherry Hill, New Jersey 222-6242

Global Communications*Dorothy Unger Piscataway, New Jersey 335-4358

Communication and Information Systems Division*Dan Tannenbaum Camden, New Jersey 222-3081

Bruce White Camden, New Jersey 222-3908

Solid State Technology Center*Susan Suchy Somerville, New Jersey 325-7492

Missile and Surface Radar Division*Don HiggsGraham BooseJack Friedman

Nov. 14, 1985

Moorestown, New JerseyMoorestown, New JerseyMoorestown, New Jersey

224-2836253-6062224-2112

Network Services*Bill Brown

Patent OperationsGeorge Haas

RCA CylixHank Cocozza

RCA Laboratories*Julie Dann (Approvals only)

Eva Dukes

RCA Records Division*Greg Bogantz

RCA Service Company*Murray Kaminsky

Dick DombroskyRay MacWilliams

Tacnet

Lancaster, Pennsylvania 227-6878



Memphis, Tenn. 739-1177

Princeton, New Jersey 226-2061Princeton, New Jersey 226-2882

Indianapolis, Indiana 424-6141

Cherry Hill, New JerseyCherry Hill, New JerseyCherry Hill, New Jersey

222-6247222-4414222-5986

RCA Technical Excellence Center*Tony Bianculli Princeton, New Jersey 226-2111

(for Corporate Approvals)

Solid State Division*John Schoen

Sy SilversteinVirginia PateHarold Ronan

Dick MoreyJohn Young

Somerville, New Jersey 325-6467

Somerville, New Jersey 325-6168Somerville, New Jersey 325-6704

Mountaintop, Pennsylvania 327-1473or 327-1264

Palm Beach Gardens, Florida 722-1262Findlay, Ohio 425-1307

Video Component and Display Division*Ed Madenford

Lou DiMattioNick MeenaJ.R. Reece

Lancaster, PennsylvaniaScranton, Pennsylvania

Circleville, OhioMarion, Indiana

227-6444329-1435432-1228427-5566

Fractals and forgeries Page 4

Computer graphics is generally concerned withthe construction of images from mathematicalmodels. However digital methods can also beused to create realistic forgeries of real worldscenes by merging pieces taken from actualphotographs. This photograph was constructedfrom two source images, one of a head andshoulders, the second of an apple in front of thehead and supported by a pole. Pyramidtechniques were used to place a reduced versionof the face on the apple and to remove thesupport. The result is a natural looking compositewithout seams or artifacts.

-Peter BurtRCA Labs

MCEZ/ EngineerA technical journal published bythe RCA Technical Excellence Center"by and for the RCA engineer"

13 Roszel Road, P.O. Box 432, Princeton, N.J. 08540-0432

Address correction requested

7EJ NGSSEN203 ELKINS RDCHERRY HILLNJ 08034

Printed in the U.S.A.Form No. Re -30-5

A cloud can be thought of as a superposition ofmany circles of different sizes. Here, using apyramid technique, six component images weremade, each containing randomly positionedcircles of a particular size. Summing thecomponents gives a self -similar, fractal-likeimage. The "raw" fractal is then thresholded andpseudo -colored to give a cloud. If eachcomponent is moved at a different rate andsummation is done repeatedly, realtimeanimation of random fractal scenes is possible.

-Joan OgdenRCA Labs

1GSDGC

BULK RATEUS Postage

PAID

Phila., Pa.Permit No. 2906

Documents

mcoa · The astute reader might point out that a single package could do the job by building the MEVDP model from scratch, ... are published in a document titled "Manufacturing Design