West Coast Contributions to the Development of the General...

The genesis of this history of NorthropAircraft’s pioneering work in digital computersduring the 1946 Snark missile program was a13 May 1996 meeting of myself [DonaldEckdahl], Irving Reed, and Harold Sarkissian at

Sarkissian’s Newport Beach, California, home.I subsequently undertook research that includ-ed numerous visits to the National CashRegister (NCR) Corporation archives in Dayton,Ohio, and a review of the extensive oral histo-ries of many key early participants in the com-puter industry collected in the 1970s as part ofa Smithsonian Institute–sponsored program. Ihave drawn particularly on the histories ofFloyd George Steele, Richard Sprague,Sarkissian, Reed, and myself.

Briefly, Northrop’s “computer group” devel-oped a digital differential analyzer (a machinecapable of analyzing and solving a system ofordinary differential equations), which it calledthe Maddida (Magnetic Drum Digital Differen-tial Analyzer), and later founded ComputerResearch Corporation (CRC). Purchased by NCRin 1953, CRC became the technical and organi-zational underpinning of that company’s elec-tronics and computer structure. CRC personnelplayed a major role in NCR’s transformation,during the 1970s and into the 1980s, from amechanical-based office equipment companyinto a highly profitable electronics and com-puter systems company.

Most histories place almost all of the pio-

West Coast Contributions to theDevelopment of the General-Purpose ComputerBuilding Maddida and the Founding of Computer Research Corporation

Donald E. Eckdahl(Deceased 23 July 2001)

Irving S. ReedUniversity of Southern California

Hrant H. (Harold) Sarkissian(Deceased 15 February 2001)

The story told here is of the members of a computer group—established by Northrop Aircraft in the mid-1940s to develop aguidance system for a US Air Force missile—who subsequentlyfounded a company that designed and built a general-purpose,digital computer. Successful, but in need of funds, the company waseventually absorbed by a larger enterprise.

4 IEEE Annals of the History of Computing Published by the IEEE Computer Society 1058-6180/03/$17.00 © 2003 IEEE

Editor’s NoteIn addition to collaborating with his coauthors, Donald Eckdahl

received, during 1996 and 1997, editorial assistance and guidancefrom Eric Weiss, then biography editor of the IEEE Annals of theHistory of Computing. In mid-2000, Eckdahl presented a more orless finished manuscript to another editor, John Simon, who sug-gested some further refinements and restructuring and subse-quently edited the manuscript.

Coauthor Harold Sarkissian passed away in February 2001 whilethis work was in process. The work was essentially completed andthe manuscript had been accepted by Annals when Eckdahl passedaway in July 2001. Simon subsequently worked with survivingcoauthor Irving Reed to incorporate the helpful suggestions andaddress the insightful observations of the Annals reviewers, forwhich both are grateful.

Given the circumstances under which the article was draftedand Simon’s unfamiliarity with the subject matter, there is clearlyroom for errors of omission and misinterpretation. Readers’ addedrecollections and corrections are welcome.

neering work in the development and manu-facture of the digital electronic computers fromwhich the current major computer industryevolved in the eastern half of the US.1,2 Thishistory documents how a group of people inthe western US developed, between 1946 and1953, a digital computer and digital computerdesign concepts that resulted in

• the invention and implementation of thedigital differential analyzer;

• the development of logic and circuit designthat used Boolean algebra comprehensivelyto implement germanium diode logic;

• the development of digital magnetic record-ing concepts and an airborne magnetic tapeunit and early magnetic drum that was usedas the main memory device of the digitaldifferential analyzer;

• a series of digital computers that includeddigital differential analyzers and general-

purpose computers the size of a small refrig-erator, not the small room size that was stillthe norm in 1952;

• the first transcontinental shipment of a dig-ital differential analyzer and its successfuloperation within 24 hours of arriving at itsdestination (to enable John von Neumann,then at the Institute for Advanced Study inPrinceton, New Jersey, to operate themachine and evaluate its implications);

• the formation of a commercial computercompany that would later be sold to NCR andbecome the basis of NCR’s position in thecomputer industry in the 1960s and 1970s;

• the development of the piezo-electric quartzcrystal oscillator clock;

• the definition of logic propositions as highor low voltage rather than the presence orabsence of a pulse; and

• the spawning of about 14 other computercompanies (see Figure 1).

January–March 2003 5

Figure 1. Genealogy of West Coast aircraft and computer industries. (Courtesy of the Charles BabbageInstitute, University of Minnesota, Minneapolis.)

Northrop and Project MX-775Aeronautical engineer Jack Northrop found-

ed Northrop Aircraft in 1939. Between enteringthe aircraft industry in 1918 and being namedLockheed Aircraft’s chief engineer in 1972, hepioneered, among other novel design concepts,the monocoque or stressed-skin constructionof the highly successful Lockheed Vega. DuringWorld War II, Northrop produced the P51 BlackWidow, the first night interceptor aircraft, someof which were still test-flying prototypes ofguidance equipment modules in 1946 whenthe company landed a US Air Force contract fora new missile that was to use a unique, auto-matic, extremely accurate guidance system forlong-range missions.

The contract apparently did not specify arocket drive, and the company chose the BlackWidow, with a range and speed slightly greaterthan that of the latest WWII manned bombers,using navigation by the stars as was commonbefore loran (long-range navigation, a tech-nique that employed radio to plot the positionof a ship at sea). The required range was greaterthan 5,000 miles with a required accuracy ofone-tenth of a nautical mile. The original idea,to take star sights automatically every so oftenand then rapidly and accurately compute thetrue position, clearly required a digital compu-tation system.

Further study led Northrop and the AirForce to call the system inertial navigationbecause using a gyroscope to maintain a stableplatform in space was the same as locking a tel-escope onto different stars. Both conceptsrequired continuous knowledge of the horizonor true vertical, and because the force ofmotion in a fast-moving aircraft or missile can-not be distinguished from the force of gravity,the computation continuously required thesolution of a set of differential equations fromthe known initial conditions of the craft.Because navigating as done by ships at sea—byusing the horizon to make star sights—was notpossible in a rapidly moving aircraft, inertialguidance was the only way to go.

The development organization In the spring of 1946 I had just been dis-

charged from the US Navy, having served mostrecently as a naval officer on a supply ship in thePacific. I had graduated from the University ofSouthern California (USC) in 1944 while still inthe Navy, with a BS in electrical engineering.Now I wanted and needed a job, preferably as anengineer, as I’d had a long-term love affair withelectronics and had held an amateur radiolicense since the age of 13. Having a child on the

way helped me decide to take the job rather thana graduate fellowship at the California Instituteof Technology (Caltech), although throughevening studies at USC I was able to earn an MSin electrical engineering in 1949.

Engineering jobs were scarce in 1946. Ifound a few opportunities to work as an elec-tronics technician, but none as a developmentengineer. After much searching, I obtained aninterview with Eric Ackerlind at NorthropAircraft in Hawthorne, California. Northropmanagement thought that a computer groupwas needed to help with the development ofthe guidance system. Ackerlind had been hiredto create that group on the basis of his manyyears of experience in radio and communica-tions development.

I was familiar with the mechanical and elec-tronic analog computers used in the Navy, butnot with the new digital computers. Evidentlysatisfied, however, by my excellent grades fromthe university and electronics experience (byway of my longtime ham radio hobby),Ackerlind hired me. Many times since August1946 I have thought, How could a 22-year-oldhave been so lucky to find employment in sucha difficult time and be dropped right into thebeginning of the digital computer revolution?

Early on I learned that, although he visual-ized the use of digital computer techniques tohelp solve the guidance problem, Ackerlind didnot presume an actual digital computer to bepart of the in-flight equipment. He was awarethat all of the digital computers built or pro-posed up to that time were “rooms full ofequipment.”

My new direct boss, Floyd George Steele, hadbeen hired by Ackerlind in March 1946. Aged28, with a BA in physics, a BS in mechanicalengineering, and an MS in aeronautical engi-neering, he had worked at Douglas Aircraft inoperations research from 1941 to 1944 and as aradar technician in the US Navy from late 1944to early 1946. Ackerlind added Richard Spragueto the computer group in July 1946. Aged 24,Sprague held a degree in electrical engineeringfrom Purdue University where he had donecourse work under a professor named Siskind inwhat was then the secret area of radar.

“He called it something else,” Spraguerecalled,

but I took one of the first courses in radar atPurdue in 1941, as did many of us majoring inelectronics. Upon graduation, I joined theGeneral Electric Company in their test-engineer-ing program at Schenectady [New York]. I stayedin that program for a year and a half and sort of

6 IEEE Annals of the History of Computing

Building the Maddida

managed to talk myself into radar projects all thetime I was with them.3

In 1944 the Navy was looking for electricalengineering graduates to become radar officers;Sprague had enlisted and become a naval offi-cer at Fort Schuyler, New York. Following a yearof training, he was assigned to radar designwork for the Naval Research Laboratory inWashington, D.C. “My view,” he reflected, “isthat the technologies that went into radardevelopments were just as important as almostanything else you could name in the begin-nings of the electronic digital computer field.”

The senior manager of Project MX-775, El Weaver, was an experienced Northrop projectengineer from the airframe side of the business.He organized the MX-775 missile program witha comprehensive technical and administrativemanagement team that represented the manydisciplines needed to develop such a complexsystem, bringing in scientists, mathematicians,and designers in optics, celestial mathematics,gyros, communications, computers, and gener-al airborne electronic equipment systems.

Knowing that the analog computers used asin-flight equipment in aircraft control systemscould not deliver the required accuracy for aguided missile system, and that the size of pro-posed digital computers rendered them unsuit-able for field use, never mind in-flight use,Weaver’s team of technical and systems expertsproposed to build or buy a “function genera-tor” (or perhaps digital computer) for use onthe ground. The generator was expected todevelop a data string that would be recorded ontape by the new “magnetic” recording methoddemonstrated by the Germans during WWII.

In the months that followed my arrival, thecomputer group studied and learned. We readeverything we could about digital computers,including the ENIAC design papers by J. PresperEckert and John Mauchly at the University ofPennsylvania’s Moore School of ElectricalEngineering, reports from the Ballistic ResearchLaboratory at Aberdeen Proving Grounds, theENIAC’s purchaser and user, and the ElectronicDiscrete Variable Automatic Computer(EDVAC) proposal, written, I believe, by Eckertand Mauchly while still at the Moore School.The EDVAC proposal described an internallyprogrammable machine that could store itsown instructions and then modify its paththrough those instructions. It also proposedand conceptually elaborated the future “gener-al-purpose” computer. We subsequently builtand tested electronic circuits, taught ourselveshow to build digital “gates” and “mixers” using

vacuum tubes and germanium diodes, devisedways to build special-purpose computers andfunction generators, and considered storingdigital numbers on magnetic wire and taperecorders.

Concurrently, the computer group began toexpand. Will Dobbins joined in November1946, Bernie Wilson in February 1947. In June1947 two Caltech graduate students signed onas summer consultants. Herman Kahn stayedonly for the summer, leaving for the RandCorporation and later authoring the bookThermonuclear War (we encountered him again,for a short time, as an employee of CRC circa1951). Irving Reed was later among thefounders of CRC. “One of my jobs,” herecalled, “was to try to find a guidance criteri-on using star trackers for guiding the missile.”

Reed had earned a PhD in mathematics witha minor in physics from Caltech before joiningNorthrop as a full-time employee in June 1949.While a graduate student at Caltech, he hadtaken a course in symbolic logic from E.T. Bell.The former Navy electronics technician hadasked Bell about the possible application ofsymbolic logic to the physics of relay switches.Bell had directed him to the seminal paper thatClaude Shannon had authored in 1938 aboutthe use of Boolean algebra in logic design forrelay devices. Reed recognized this potentialmore clearly when he later became involvedwith digital computers at Northrop, where heintroduced the concept and Shannon’s paperto Floyd Steele. Steele later explained to othersthe mathematical concept of Boolean algebraand how it might be used in the design of elec-tronics as well as relay logic.

In September 1947, Ackerlind and Steelehired Hrant (Harold) Sarkissian to be the com-puter group expert on magnetic recording.Upon graduating from the University ofCalifornia at Berkeley with a degree in electri-cal engineering in 1944, Sarkissian had enteredthe Signal Corps. He spent two and a half yearsworking with microwave equipment, eventu-ally in Germany. “The reason I got the job [atNorthrop],” Sarkissian recalled,

at least I believe it is the correct reason, asidefrom the fact that they were beginning to hireelectronic engineers again, was that I had seenone of the first tape recorders that the Germanshad made called the Magneto-phone. That wasas close to tape recording as anybody atNorthrop had been, apparently, so that qualifiedme as a magnetics expert. In any case they putme into the computer group … under … Dr. EricAckerlind.


… My immediate supervisor was Don Eckdahl.The group numbered 15 or so, but I do notremember exactly because they were hiring morerapidly at that time to get into the guidance andcontrol of the Snark missile system.

From theories to devicesSteele, about five years older than the rest of

us and the computer group’s genius andspokesperson, was our conceptual leader. Helived with his family in Manhattan Beach, asmall coastal town south of Los Angeles. I metSteele often for dinner and walks along thebeach during which we discussed subjects bothtechnical and intellectual, ranging from musicto books by Dostoevsky and Oswald Spenglerto the revolution that we believed was going tobe precipitated by the effect of the new digitalcomputers on business, entertainment, militarysystems, and so forth. He was our teacher; we,his disciples.

Dependent on the rest of us to develop andimplement his concepts, theories, and ideas,Steele pushed us hard to apply the Booleanalgebra concept to the implementation of logicin electronic digital circuits. He expected us toreadily develop the methods for moving fromBoolean equations to the design of a digitalelectronic logic system that represented thoseequations. He also promoted the notion that acomputer was a set of stable states that couldchange in sequence in accordance with a logi-cal system. He proposed that a logical proposi-tion be defined as being equivalent not to thepresence or absence of an electronic pulse, aswas then common practice, but to a high volt-age (true) or a low voltage (false).

DIDASteele, who had read Lord Kelvin’s paper on

the ball and disk integrator and about the oper-ation and use of the large mechanical differen-tial analyzers proposed and built by VannevarBush, had by late 1946 conceived the conceptand general structure of a digital differential ana-lyzer, later nicknamed DIDA. The device was tohave two electronic registers for each integrator.Transferring a binary number from one registerinto the other as directed by a pulse stream, theinput rate generated an output each time thesecond register overflowed; the sum of the out-put stream of pulses was the integral of the num-ber in the first register times the input rate.

Asked by Steele to employ this concept todesign and construct a digital differential analyz-er, Sprague and Wilson designed the integratorsand a pair of registers. Each stage comprised oneflip-flop, which usually employed two vacuum

tubes. A register that could hold an 18-bit binarynumber would thus require 36 vacuum tubes. Apair of registers became an integrator. The build-ing of multiple copies of these flip-flop circuitsand electronic registers was subcontracted to thethen-fledging Hewlett-Packard Company, at thetime a small but successful producer of qualitytest instruments for the electronic equipmentindustry (primarily radio receivers and audioamplifiers). Because each integrator existed as apair of physical registers, a 10-integrator DIDAthus had 10 pairs of electronic registers. Integra-tors were interconnected according to the differ-ential equation to be solved, the wires from theoutput of one integrator being connected to oneof the two inputs of one or more of the otherintegrators. Sprague and Wilson built several inte-grators, eventually reaching the stage of waitingfor parts needed from Hewlett-Packard to con-duct a system test of a small DIDA.

Quartz clock For a missile guidance system to continually

solve differential equations required extremelyaccurate initial conditions at the outset of com-putation. Among these conditions were launchsite and time. Whereas the launch site could bespecified accurately using longitude and latitude,the state of the art in timekeeping technologywas accuracy only to about one-tenth of a sec-ond. The government broadcasted from Wash-ington, D.C., on a radio station with call lettersWWV, a signal by which a clock could be set toan accuracy of about one ten-thousandth of asecond, which was adequate for missile guidanceif a clock could be built to keep time to the sameor better accuracy. We concluded that a digitalsystem could be devised to turn temperature-controlled quartz (piezo-electric) into a “clock.”Steele assigned this development task to me.

Digital counters were designed and con-structed that would start with the crystal oscilla-tor’s pulse rate, which turned out to be 100,000cycles per second. We subcontracted the build-ing of the constant-temperature oven needed tokeep the crystal stable and prevent its frequencyfrom drifting. We also developed the circuitry toconvert the oscillator output from sine waves topulses and a digital counter to reduce the rate tospeeds that could be displayed as hours, minutes,seconds, and tenths and hundredths of seconds.The digital counters thus had to divide by (orcount to) specific numbers such as 60. Thedetailed design for this project taught us a greatdeal about digital circuit design with vacuumtube flip-flops and germanium diodes.

The successful development of this clockdevice was the occasion for many evenings and



late nights spent with other guidance systempeople flight-testing early modules. Our testbedwas a P51 Black Widow aircraft. The quartz,piezo-electric, crystal oscillator clock was a pre-cursor of today’s quartz digital watch, whichsubstitutes integrated transistor circuits for vac-uum tubes to minimize size.

Magnetic digital recording and playback Concurrently, Sarkissian was leading anoth-

er small group engaged in the development ofan airborne magnetic tape unit that was torecord and play back digital signals unattendedover an eight-hour period. Not only the circuit-ry, but the methods for recording and playingback digital numbers rather than audio voice ormusic, had to be developed. Recalled Sarkissian:

There were wire recorders, so one of the firstthings I was supposed to do was pick out digitalsignals from a wire recorder. The specs were verydifficult to meet for that time. The wire recorderwas supposed to record digital information andthen play it back in a precise manner. … Uponasking for background information I suddenlydiscovered that there was not any on taperecorders, either at Northrop or even in the liter-ature, except some very basic things about theearly history of wire recording or magneticrecording. This launched me, more or less, intoresearch of what was then the state of the art andhow we could make use of it.

Sarkissian’s group exceeded the extremely tightreliability specifications established by the sys-tem designers for the digital magnetic recordingendeavor. Sarkissian and I often drove togetherto Northrop Field in the evenings to help theteam carry out the interesting, and educational,flight tests of the airborne magnetic tape unit.

BinacIn December 1947, an R. Rawlins, then assis-

tant project engineer for guidance, let a con-tract to Philadelphia-based Eckert-MauchlyComputer Corporation to design an airbornedigital computer to solve the in-flight guidanceproblem. This was a curious step, given that therest of the technical and administrative man-agement of Project MX-775 did not accept thepracticability of an airborne computer solution.(Our computer group did not begin to acceptthe possibility of an airborne computer untilthe Maddida [pronounced “mad eyeda”] wasclose to a reality.) Eckert and Mauchly calledthe machine they delivered 18 months laterthe Binary Automatic Computer (Binac). TheBinac, although small compared to other early

computers, at 10 to 20 times the size of theMaddida was not even close to being a proto-type for an airborne version of a computer (seeFigure 2).

“It was sort of a preposterous request to bean airborne computer,” reflected Sarkissian.

It was a best-effort thing. Perhaps they had otherreasons for asking for this; it may have been away for the government to sponsor develop-ment, which apparently was what was going on.Ike Auerbach, an engineer at Eckert-Mauchly,was on the other end of this at that time and hecould probably see more of what was happening… than we could. I knew a little more about ayear or so later when Northrop management wasthinking about sending someone back to Eckertand Mauchly for training. Until that time, theEckert-Mauchly program was hardly visible onour end. I think our computer group was essen-tially only advisory, with respect to the projecthead, on this particular program.

Genesis of the Maddida In January 1948, after nearly completing

work on the electronic digital clock, I was askedby Steele to develop a magnetic drum memoryversion of the digital differential analyzer (thus,Maddida). Steele believed Sarkissian to be farenough along in his study of magnetic record-ing to lead the development effort. I was to beproject engineer and team leader. Jack Donanand Carl Isborn joined the computer group inthe spring of 1948. Al Wolfe and Dobbins wereamong the other team members. Sprague con-tinued to work on DIDA until January 1949,then, leaving its completion to Wilson, becamean active participant in the Maddida develop-ment effort.


Figure 2. The Binac computer delivered infulfillment of the contract to design an airbornedigital computer. (Courtesy of the ComputerHistory Museum, copyright 2003.)

ConceptThe integration concept for the Maddida

involved adding and subtracting the contentsof two registers, the number y in the Y registerto or from the R register. This addition–subtraction update of the integrator wasaccomplished at the occurrence of an input dif-ferential dx (±1 divided by a scale factor, somepower of 2). (The type of integrator used inMaddida is shown in Figure 3.) The R registerhad the same length as the Y register andwould overflow or underflow depending on thesize of the number y multiplied by the sign (±1)of the differential dx needed to form the outputdifferential dz = ydx of the integral z = ∫ydx,where, again, the differential dz is ±1 dividedby the appropriate power-of-2 scale factor. Themethod of integration in Maddida had the

advantage over previous techniques of numer-ical integration of having minimal round-off-error growth. Reed successfully proved thevalidity of the Maddida technique of numeri-cal or digital integration.4

In the overall Maddida system, a sequence ofpairs of numbers, namely, the Y and R registers,were recorded along two tracks of a magneticdrum. These pairs, which represented a finite setof integrators (originally 22 integrators), wereoperated on sequentially every turn of the drumas they passed by the read and write heads. (Theflow of information through the Maddida isshown in Figure 4.) These operations caused (ordid not cause) two possible things to occur:increment the number y in the Y register by thedifferential dy, and add the number y, multi-plied by dx, to the R register to form the outputdifferential dz (±1 divided by the appropriateoutput scale factor of the integral).

Technical trainingEarly in 1948 we learned that Alvin Sugar

was to offer at USC, during the winter semes-ter, a graduate course in electrical engineeringcalled “Programming a General PurposeComputer.” Steele, Sprague, Sarkissian, andfour or five others subsequently took advantageof this extraordinary opportunity. Sugar, whohad recently been engaged as a consultant byEckert and Mauchly, had been granted permis-sion to use the planned instruction set for theirUnivac computer, which was then under devel-opment. This timely course, which ran a fullsemester, involved writing a program to per-form matrix inversion. The Univac machinelanguage’s straightforward, one-address struc-ture mirrored the computer’s logical design.The concept of assembly and high-level pro-gramming languages had not yet been con-ceived, nor did the word software yet exist; allprogramming in that period was done inmachine language.

Following development of the Binac in thesummer of 1948, Mauchly began visitingNorthrop every two to three months. In addi-tion to whatever business he might have hadwith Northrop management, he routinely lec-tured members of the computer group aboutdigital computer concepts and design, includ-ing logic, circuits, and memories. Being a pro-fessor of physics, Mauchly was an excellentteacher. He was also friendly and open. Overthe course of more than a year, we benefitedfrom about five half-day lectures.

Eckert’s and Mauchly’s design conceptsinvolved the use of vacuum tubes (no germa-nium diodes) for both flip-flops and logic.



Figure 3. Diagram of the type of integrator used in the Maddida. (FromMaddida Preliminary Report, Project MX-775, Report No. GM-545,Northrop Aircraft Inc., 26 May 1950; personal files of D. Eckdahl.)

Figure 4. The flow of information through the Maddida. (FromMaddida Preliminary Report, Project MX-775, Report No. GM-545,Northrop Aircraft Inc., 26 May 1950; personal files of D. Eckdahl.)

Moreover, they defined logical propositions asthe presence or absence of a pulse. Finally,Eckert and Mauchly did not use Boolean alge-bra in logic or system design.

Logic and circuit designHaving been acquainted by Steele with the

background and mathematical concepts ofBoolean algebra and how germanium diodenetworks could be formed to represent anyBoolean algebra expression, Sprague and I,while developing the Maddida’s logical design,found a way to apply Boolean algebra to a totalsystem of logic and circuits. Our method was tocreate and write the logic equations thatdefined the input to each input side of everyflip-flop in the system. Each input equation wasbased on the states of all the other flip-flops.The result was the set of Boolean equationsboth necessary and sufficient to define themachine.

The mathematical logic (logical proposi-tions) was represented electronically as a highvoltage for true and low voltage for false. (Inthe ENIAC, Binac, and Univac, logical states orpropositions were represented by the presenceor absence of an electronic pulse.) The basiclogic premises, electronic circuits, and total sys-tem design using Boolean algebra were thus alldifferent and new in the Maddida and themachines descended from it.

Memory drum and NRZ magnetic recordingSarkissian and Isborn developed the

mechanical structure and system of adding themagnetic recording that was to become theMaddida’s memory. The drum was about eightinches in diameter and an inch and a halfthick. The magnetic surface was made by firstdissolving the coating off magnetic tape andthen spraying it on with a small airbrush-sizepaint sprayer (we did not yet know enoughabout magnetic coating to formulate our own).

Sarkissian related an instance of the give-and-take that culminated in the developmentof the digital magnetic recording system andcircuits. “Don Eckdahl and I,” he recalled,

sat in a hotel room in New York City and solvedthis problem, and then … proved that it could bedone. … [W]e devised a system that did notchange the character of the logical signals, as wasthe case with the methods that other people used.… We felt people would be more satisfied with asolution that did not require a completely differ-ent technique to transform from one form andthen retransform … back. Such things were occur-ring in mercury memories and in electrostatic

memories and a variety of other schemes. …In lay … terms: … to record 0s and 1s you have

to be able to detect … a zero or a one. The prob-lem is how do you detect a zero. Many differentschemes were tried to make a one and a zero lookdifferent on a drum. You can record it by sayingif the magnetic field is north, that is a one; if it issouth it is a zero; and so on. … The method weused, which is now old hat, is called the NRZ(non-return-to-zero) system. If the logical signal,say, is a one, we change it so that the recording isnorth. We stay there, we do not change anything,and the head continues to record north, north,north, north. … Then, when it goes from a oneto a zero … we go from all north to all south. Allwe are worried about is that change. … [I]f thatchange occurs then we change the flip-flops. …

Misunderstanding andmiscommunication

In April 1949, Ackerlind, apparently worriedthat he lacked authority to develop an airbornecomputer and that other jobs seemed to havemuch higher priority at Northrop, began toreassign members of the computer group toother activities. In May, in a memo to Ackerlindcopied to Frank Bell, George Fenn, and a P.H.Taylor, Steele stated that the Maddida “alreadyhas” all of the timing and intelligent featuresto control star trackers and so forth and evenestimated the total number of vacuum tubesrequired. If asked, as project engineer I wouldnot have supported this highly optimistic sug-gestion that the Maddida might be adapted tothe requirements of a full guidance system.

In July and August, some members of thesomewhat free-running computer group weredispatched to Philadelphia to learn about theBinac. Donan, Dobbins, Isborn, and BobPrathers visited Eckert-Mauchly ComputerCorporation in July; Steele, Sprague, Reed, JerryMendelson, and I visited in August. Each groupstayed for about a month. Probably arranged bymanagement to keep us out of their hair, the vis-its also served to familiarize some engineers withthe Binac before Northrop took delivery of it.

“The people we met with,” recalledSprague,

included Ike Auerbach, Al Auerbach, Bob Shaw,and Grace Hopper. … Shaw was the logical design-er of Binac and an incredible guy. He was respon-sible almost single-handedly for the logical designof all of the early Eckert and Mauchly machines.… He was almost blind. … He also had a spasticcondition. … He had to walk on crutches. … Inother words, he was pretty severely handicapped.But he was a brilliant guy and did a lot of the


design work in his head .…When you are dealing

with … pulse circuits, gates,etc. you really had to drawthese circuits on … circuitdiagrams. … Shaw had help… drawing … circuit dia-grams, but he had no help …reading them. … In order toread one of his own dia-grams, he had to put it aboutthree to four inches from hiseyes and he could only see alittle bit of it at any givenpoint. … The thing that wasvery impressive was hewould have … a piece ofdrawing paper four feet longand three feet high com-pletely covered with Binaccircuitry. And he would pickit up, look at one point on it,and immediately shift hisway around on that diagramby moving it in front of hiseyes to some other point. …It was obvious he had thewhole diagram stored in hishead. He could not havedone it otherwise. … It wasincredible. … Al Auerbachwas also a terrific circuitdesigner on the Binac.

Sprague recalled IkeAuerbach being astute inthe marketing area andHopper being a remarkableprogrammer. Of Eckert andMauchly, he recalled how

we would walk out into the laboratory andEckert would be out there talking about the lat-est developments … in very esoteric terms, usingmathematical expressions while he was dis-cussing whatever it happened to be. And all thetime he would be talking with somebody, he’dbe standing there twirling his key chain in a widecircle. … Carefree, would you say? … Mauchly[was] at the other extreme, very quiet.

Our group’s political problems at Northropworsened during the Maddida’s initial debug-ging. Looking back, I see that we caused mostof the problems ourselves. The project man-agement must have had trouble keeping trackof our group. Moreover, although accepted asthe leader and spokesman of those in the com-

puter group, Steele was never able to convinceothers, either technically or politically, of thepromise or practicality of our work. The resultwas a sad and somewhat humorous successionof misunderstandings and conflicts betweenmembers of the computer group and theNorthrop management structure. The Maddidawas nearly a bootlegged or skunk works-typeproject; management tried many times to splitit up, disband it, and reassign key personnel toother projects. Ultimately, management lookedto other managers such as Lee (Leo) Ohlingerto try to control us.

We trusted Steele to represent our viewpointsto management; we believed that he really triedto demonstrate the importance of our develop-ments in digital computers and the impact theymight have on the problem of inertial guidanceof the Snark missile. We also presumed that hehad emphasized the probable business oppor-tunities that a pioneering position in the then-new commercial computer sector might affordNorthrop Aircraft. But we later found that Steelehad, in fact, instead communicated to both theproject management and to Jack Northrop thefar more radical ideas that the new techniquesassociated with digital computers were so pow-erful that the company should scrap its entireguided missile program and begin anew with anall-digital system for all elements of the prod-uct, save the airframe. He almost surely did notpropose a specific plan or offer to lead this totalredesign. He must have believed that we haddone and proven so much that it should beobvious to management that Northrop’s fullresources should be applied to this proposedchange in direction.

The Maddida on tour By the time I returned from my visit to

Eckert and Mauchly, Sarkissian and Wolfe hadmade considerable progress. Although I had bythen been directed to leave the project, I nev-ertheless returned to the project area in orderto continue to debug the Maddida (manage-ment rescinded its stay-away order about aweek later). Debugging progressed throughmany stages, including intercommunicationbetween integrators. By November 1949 wehad a working, 22-integrator digital differentialanalyzer, as Figure 5 shows.

Von Neumann’s appraisalThe pressure on Northrop management to

explain the apparent success and importanceof this “bootlegged” machine was severe. Inlate February 1950, Steele and Reed convincedmanagement to let us transport the Maddida to



Figure 5. This is the Maddida designed,built, and tested at Northrop. It used avery different construction, even forthat time. A large diode board, ringedwith vacuum tube flip-flops, was hinge-mounted on the surface of a “teacart”provided with rollers for mobility. Thedrum memory was also mounted onthe surface; situating the vacuum tubecircuitry directly on the drum structureminimized the length of the leadsbetween the magnetic heads andamplifiers. The power supply for thedrum and logic unit was installed underthe cart. The original Maddida is part ofthe permanent collection of theComputer History Museum. (Courtesyof the Computer History Museum,copyright 2003.)

the Princeton Institute for Advanced Study forreview and evaluation by John von Neumann.Steele, Reed, and I left for Princeton after pack-ing and air shipping the Maddida. Also makingthe trip as a representative of Northrop man-agement was Ohlinger, our new boss.

Our arrival at Princeton and subsequentinteraction with von Neumann is chronicledby Reed:

The first day we arrived in Princeton, themachine was put in a fourth-floor hotel room ofthe Princeton Inn. … We realized that we did nothave any way to get the necessary power. We hada three-phase power supply. … I think it was Don[Eckdahl] who noticed that there was an electriccompany across the street, so before we went tobed we decided … to see if we could get a cablestrung across the street. That … was done. A cablewas run from the power company to the windowon the fourth floor of the Princeton Inn. …

Von Neumann told us that he had done someprior work investigating the machine and … hadread what few reports there were. He told us thathe had ideas similar to this before and that oneof the reasons Whirlwind had come into exis-tence was because he and some others had theidea that to build aircraft of the future theywould need giant simulators. One of the purpos-es of the Whirlwind computer was aircraft simu-lation, initially the simulation of controlsystems. He said that he had another idea alongthese lines to build a differential analyzermachine. I think somebody at that time hadbuilt an extensive analog differential analyzer,but I do not remember whom.

The big problems at that time were related tothe fact that we were just getting into the jet ageand people were worried about high-speed air-craft and the new control systems. … No longercould people steer airplanes directly with cables;they had to go through power units and servosystems and … we would have to build simula-tors to simulate the performance of suchmachines to make sure they worked. VonNeumann thought one possible application of aMaddida-type machine could be this simulator.I think he said as much in the document that hewrote to Jack Northrop. …

[Von Neumann] asked to see a diagram of themachine. One of us, probably Don, pulled out along list of equations and told him that this wasour best description. He was very impressed andsaid that he always thought one could design amachine this way. I think he reminisced to someextent back to Turing and … EDVAC and ENIAC.

[Von Neumann] asked about programming themachine and … I got up to the blackboard and

showed him how we did it. All my student lifeand my professional life … I had known aboutvon Neumann as a great mathematician and Ithought he was almost omniscient in his capa-bility. I said that we were going to program theBessel equation and he asked me to write it downbecause he could not remember it. He was verynice in the sense that he played down the factthat he may have known the equation. So I wroteit on the board very quickly. … He raised my egoby having me do that. After this I showed himhow we did the programming. I think that Dongave a short lecture on the logic and finally wecame out with the logic diagrams. This all tookplace at his office before he ever saw the machine.

I think Don explained a logic design formulato von Neumann and showed how we used it togo about the design of diode “and” and “or”logic circuits. Von Neumann was very impressed.I think to some extent the structure was also dis-cussed. … There were only four channels on thedrum. One was a clock track, which was only aread track. There was a circulating register trackand … a pair of channels, which were both readand write, which served for the integrators andthe code.

I think Don … handled most of the explana-tion because Steele was not very good at explain-ing the details of how something worked. He wasmore of a philosophist, whereas Don was veryexact, crisp, and to the point. …

As Jerry Mendelson told it to [Robina]Mapstone [oral history interviewer for theSmithsonian’s National Museum of AmericanHistory], Steele would conceptualize on such andsuch and within a couple of days or [a] week …Eckdahl would have it manifested.

Von Neumann was very impressed by the factthat we got this machine on an airplane and flewit to Princeton.

It seems to me that during that afternoon wewent to another location in Princeton and got acomplete rundown on the [Institute forAdvanced Study] IAS machine. It was under con-struction and quite complete, but not yet opera-tional. We met both [Julian] Bigelow and[Herman] Goldstine. I do not think Bigelow par-ticularly reacted, so I do not know if he wasimpressed or not. I have seen him since then, butI do not think he recalls that we were there. Wedid not make much of an impression on him.

Before von Neumann came over to see theMaddida in our hotel room, we were invited overto his house where we had one of his … very drymartinis. He was famous for his martinis, whichwere probably pure gin. Bigelow was with him,and so was his wife. She was a scientist or amathematician and I think she actually had


something to do with computing machines. Sheunderstood what we were doing more or less.Bigelow came over and I remember him sittingnext to me on the couch. I still remember thefact that [von Neumann’s] furniture was as wornas mine. He was not a very ostentatious man andhe had what I would call rather plain furnish-ings, probably second-hand. This made me thinkhe was really a scientist, although he lookedmore like a businessman. He seemingly enjoyedliving and … was a humorist.

I was very impressed because he was every-thing I thought a mathematician and scientistshould be. He was a hero to me, certainly, par-ticularly since when in school I had studiedgame theory and had read part of his book,which is not easy to read.5

The next morning I programmed the machinewith a particular case of the Bessel differentialequation. It took quite a while … because thedata had to be put in with push buttons. Donwas the expert on loading it. He was really theonly person who knew how to operate it. If ithad not been for Don, I think it would have beena disaster. … Floyd and I were good at guessingwhere the trouble might be, whereas Don hadeverything under control and knew exactly whatwas going on within the machine.

To make a long story short, we had the machineprogrammed to do J of order one-half (J1/2(x)),which is a simple program and easy to put intothe machine, and Don was checking the programto make sure it went through the right value at acertain time, when von Neumann arrived unex-pectedly. He … asked what we were computingand I told him we were computing the Besselfunction. … He sat down and with a pencil andpaper computed where the machine should be atthe next checkpoint. The machine would calcu-late up to T = 1 and then it would stop so wecould read … then … go on to T = 2, stop, and soon. There were three or four minutes betweeneach of the values. We were working to a veryhigh accuracy so it took many, many turns of thedrum before reaching the value of T = 1. VonNeumann computed what it should be the nexttime it stopped and I remember being quiteimpressed by that because he did not have anytables. The value of J of order one-half is thecosine of X divided by the square root of X. Ithink I made the boundary conditions slightly dif-ferent, so it wasn’t exactly that; it … was a [linear]combination of J of order one-half and J of orderminus one-half. He computed that. He did it thefirst time. He got better at it the second time. Hewas really good. I watched him as he did it, and asI recall, he had a very quick method of calculatingthe cosine of X and was very rapid at making esti-

mates. … After the demonstration, he leftextremely impressed. He was a humble man andhe told us that it was a great privilege to meet usand he wished us all success. I think that was thelast time I ever saw him.

Von Neumann subsequently communicat-ed to Jack Northrop, in a letter reproduced inits entirety in the sidebar “Letter from vonNeumann to Jack Northrop,” his account ofour visit and opinions of the Maddida.

Evaluation by the Air ForceOn the way to Princeton, Steele had visited

with a Colonel Bubb at Wright-Patterson AirForce Base and obtained approval for us todemonstrate the Maddida there on our returntrip. We consequently shipped the Maddida byair to Dayton, Ohio, where we were joined bySarkissian and Wolfe. Northrop’s Dayton salesoffice made arrangements for transportationand hotels, and Bubb organized the meetingsand demonstrations.

The Air Force wanted us to run a particularproblem on the Maddida using documentationfrom a previous solution worked out on themechanical Bush Differential Analyzer early inWorld War II. That solution had been workedin conjunction with a study of an electromag-netic cannon that the Germans were purport-edly developing. As a double check, theproblem had been rerun to four decimal placeson Harvard University’s Mark I. We were askedto demonstrate the Maddida in such a way asto make direct comparisons with the problemsolutions from the tables computed by theMark I and curves plotted by the BushDifferential Analyzer at the MassachusettsInstitute of Technology (MIT).

Because the problem normally would haverequired more integrators than even theMaddida had, Steele and Reed spent most ofone night reworking the problem so that theMaddida could be programmed to run it. Themachine was demonstrated initially with somesimpler problems. Many of the reviewers,among them Air Force computing experts andspecial consultants such as Harry Goode, pro-fessor of electrical engineering at the Universityof Michigan, were initially skeptical that theMaddida really was a differential analyzer. Thepurpose of the electromagnetic cannon prob-lem was to put the Maddida and all of us fromNorthrop to a real test of credibility.

The next morning, after Steele and Reedprogrammed the problem, which involvedsolving a complicated scaling problem, theMaddida successfully duplicated the Mark I




Ed. note: The following letter was reproduced in itsentirety in the Annals of the History of Computing,vol. 9, no. 3/4, 1988, pp. 364-365.

Letter from von Neumannto Jack Northrop

The Institute for Advanced StudySchool of MathematicsPrinceton, New JerseyMarch 14, 1950

Dear Mr. Northrop,Mr. L.A. Ohlinger, as well as Messrs. F. Steele, I. Reed, and

D. Eckdahl, have been in Princeton. We had several detaileddiscussions regarding the basic principles, the technical exe-cution, the operational characteristics, and various possibleand probable uses of the magnetic digital differential ana-lyzer which this group has developed, as well as regardingother computing and control devices which could bedesigned on the basis of the same principles. They alsodemonstrated the differential analyzer very successfully. Ourdiscussions took place on March 8th and parts of March 9thand March 10th. The demonstration was held in theevening of March 9th and the morning of March 10th.

I have given Mr. Ohlinger my conclusions from the dis-cussions and the demonstration, and also discussed someother technical points with the other members of thegroup after Mr. Ohlinger’s departure. The following are asummary of these conclusions.

I think that your magnetic digital differential analyzer is amost remarkable and promising instrument. The principlesinvolved in its design and its engineering embodiment seemto me very sound. I would view the machine as it now standsas an example of what can be done, rather than as the actu-al solution for any particular purpose, but I consider that youhave established the principles of a whole family of very newand most useful instruments. The present instrument has var-ious limitations that your group will certainly be able toremove, as soon as they go into construction for specificapplications. However, even as the instrument now stands,it is very remarkable and far ahead of other equipment forsimilar and similarly varied purposes. To be specific, I thinkthat it is obsoleting the analogy type differential analyzer inits mechanical, as well as its electrical form. The propositionthat instruments of your type can surpass analogy differen-tial analyzers in precision, logical capacity, compactness, flex-ibility, and probably equal the best of them in speed, seemsto me clearly established. It seems to me clear, too, that theyare inherently cheaper in man-hours as well as in money.

Further developments on the basis of the principles thathave been established by your digital differential analyzerare obviously called for. Your equipment can probablyserve as a basis for worthwhile all-purpose computers ofthe intermediate type, but I would not consider this as themost important implication. It seems to me more interest-ing as a basis for a family of special purpose machines todeal with matrix problems. Among these I would mention

the following: Performing linear transformations in n vari-ables, solving n linear equations in n variables, determin-ing the proper values and proper vectors of matrices oforder n, solving the problems of game-strategy and of lin-ear programming of order n, all of this for values of n ofthe order of 100 and even higher. The importance of theproblems of the two first categories requires no comment.(To mention just a few obvious examples: The first cate-gory includes the Fourier-transformations that are neededin interpreting x-ray-crystallographic data. The second cat-egory occurs in the solution of flutter problems, also in thesolution of various problems of chemical analysis of com-plicated mixtures of organic compounds.) The third classwill have applications in connection with quantum-theo-retical chemistry, in particular in determining molecularand atomic wave functions although here considerable fur-ther mathematical explorations will be required. Solutionof problems of the last class will be of great importance,and may well be decisive, in certain phases, for enterpriseslike Project SCOOP, of the Air Materiel Command, andProject RAND, of the Army Air Forces. I shall be glad to giveyou further details on these subjects if you so desire.

I have started to discuss these questions with the mem-bers of your group, but I have thought about them moresince they left, and I see now in more specific detail howcertain variants of your equipment suit these purposes.

Please let me know whether you want at this time aspecifically technical discussion of any phase of the pres-ent machine, and of possible variants, or whether it ispreferable to reserve this for later, oral discussions.

In conclusion, I would like to say that the fact that yourmachine could be transported by airplane and by truckfrom Los Angeles to Princeton and be satisfactorily runningwithin 24 hours after its delivery is one of the most impres-sive engineering feats I have ever observed in this field. Onehas to be familiar with the great difficulties of runningequipment of this type even under the most ideal labora-tory conditions in order to appreciate the exceptional tourde force of your group, who were able to effect a brilliantand convincing demonstration under what amountedalmost to “field conditions.” They have certainly demon-strated the practicality of your equipment in a way whichis outstanding for computing machines of this degree ofcomplication.

It was a great pleasure to have had this occasion to dis-cuss these problems with the members of your group andto see the demonstration.

I am, Sincerely yours,

/s/ John von NeumannJohn von NeumannJVN:lo

Mr. John K. NorthropNorthrop Aircraft, Inc.Northrop FieldHawthorne, California

results to better than three decimal places. TheAir Force people, particularly Bubb, were quiteimpressed.

This is not to suggest that our achievementwas not hard-won. “One of my (battle) scars,”recalled Sarkissian, who had the assignment ofreading and recording much of the data outputduring these demonstrations,

came from a conversion I was asked to do. Ourreadout was on an oscilloscope; the pulses werelaid out in a scattered way and I had to interpretthem into a binary and then a decimal number. Imade a mistake, a whole day’s worth, and Ialmost got shot in the process.

Show-stealer at Rutgers University While the Maddida was being introduced to

von Neumann in Princeton, Sarkissian andSprague were attempting to convince WalterCerny, a key member of the Northrop manage-ment structure, to permit the machine to bepresented at the first national meeting of theAssociation for Computing Machinery to beheld at Rutgers University on 28–29 March1950. Cerny’s initial response, reflecting thelack of a market study and the fact that theMaddida was classified secret, was an emphat-ic “No.” But in the face of the ensuing uproar

among the computergroup, Northrop accededto further discussions and,ultimately, to the Rutgersvisit. Ohlinger was onceagain to represent thecompany and initiate amarket survey, and Steeleand I were to be permittedto give papers. Sarkissian,Sprague, and Reed accom-panied the Maddida,which Northrop decidedcould be declassified (itwas later discovered thatno secret classification hadever existed) to the ACMconference.

The Maddida was dis-played in an expansive hallwith equipment from othercompanies. The papersdelivered by Steele and mewere entitled “Maddida

General Theory” and “Mad-dida—Design Features,”respectively.6 The abstractssubmitted in fulfillment ofACM requirements are

reproduced in the sidebar, “Abstracts of PapersPresented to the ACM Conference at RutgersUniversity.’’ Figure 6 shows a copy of the trans-parency that I showed at that presentation.

“There was tremendous publicity,”Sarkissian recalled of the Rutgers visit,

because it was just what the [news]papers wantedto hear—automatic factories and all the usualfuturistic science fiction stuff. That was quite aninteresting thing. The dean of engineering ofRutgers was ecstatic over this. I was with thecomputer in the booth. Next to me was JanRajchman from RCA and he had this newSelectron tube. He was showing that and nobodywas around him; they were all around theMaddida, about fourteen deep trying to see thisthing. We literally stole the show.

Reed recalled

that the Maddida was the only full-scale, work-ing, electronic computing machine demonstrat-ed at the meeting and one of the fewtransportable machines in existence at that time.The IBM punched card machines were portable,but basically were electromechanical machineswith relays and so forth, although they may haveused some vacuum tubes.



Figure 6. A copy of the transparency prepared and presented by Donald Eckdahl at theRutgers conference, showing examples of a portion of the Boolean algebra equations thatdefined the Maddida. (From the personal files of D. Eckdahl.)


Ed. note: The following abstracts are reprinted as theywere originally written.

Abstract by the Author of Paper Given Before The Association for Computing Machinery

Conference at Rutgers University, New Brunswick, New Jersey

March 28-29, 1950“Maddida” — General Theory

By Floyd G. Steele

Two devices are normally used to simulate the mathe-matical process of integration—the analogy integrator andthe numerical accumulator.

It is possible by employing a numerical technique toproduce a device having the logical characteristics of ananalogy integrator and the accuracy and repeatabilitycharacteristics of a numerical device.

A set of digital integrators may be inter-coupled in themanner natural to the ordinary Differential Analyzer tosolve ordinary differential equations or sets thereof, eitherlinear or non-linear.

Further, numerical integrators may be stored in a mem-ory in a manner which yields simplicity of utilization andcommunication.

A magnetic drum digital differential analyzer calledMaddida has been built and put into operation. It wasdesigned to have 44 digital integrators, each having anaccuracy of 1 part in a million. Fifty-six tubes wererequired, not including power supply and readout.

It is apparent that a machine having about 132 numer-ical integrators can be built with a total of about 100 tubes.

The digital differential analyzer appears to have theadvantage of fewer tubes, economy, speed, accuracy, andpossibly greater reliability over the normal analoguedevices which solve ordinary differential equations.

In computation, it has the advantage of easy coding,accuracy, and considerable economy.

In control, it will be found to intervene between senso-ry devices and effectors more directly than the ordinarynumerical machine and to require fewer components thanstandard analogy devices. It should prove of use in thosecontrol problems which require either extensive facility orhigh accuracy.

It is hoped that this type of machine can serve both toenlarge the scope of automatic control and to provide indi-vidual computation.

Acknowledgment is made to Donald Eckdahl and DickSprague for general logical contributions on the digital dif-ferential analyzer and to Donald Eckdahl, William Collison,Harold Sarkissian, and Dick Sprague for specific logical con-tributions on the machine Maddida. Valuable mathematicassistance has been rendered by Dr. Irving Reed.

Abstract by the Author of Paper Given Before The Association for Computing Machinery

Conference at Rutgers University, New Brunswick, New Jersey

March 28-29, 1950“Maddida” — Design Features

By Donald E. Eckdahl

Maddida—Magnetic Drum Digital DifferentialAnalyzer—is an electronic digital computer which inte-grates differential equations directly, as do the analogueDifferential Analyzers. Integration is accomplished by oper-ating on pairs of numbers by an additive transfer process.A register n digits in length containing a number “y” isadded into a register “R” of equal length upon each occur-rence of an input pulse “dx.” All carries from register “R”are used as incremental outputs “dz.” The number in “y”can be increased or decreased by an input “dy.” The resultis that dz = ydx/Bn where “B” is the base of the numbersystem used. A paper by the originator of this type ofmachine, Floyd G. Steele, at this conference, covers fullythe theory of such integrators and a number of differentmethods of possible realization. One machine, which willbe discussed by Richard E. Sprague at some later date, uti-lizes complete flip-flop storage.1

Maddida stores the registers mentioned on a magneticdrum and operates on each pair of numbers serially. A pairof numbers is called an integrator. A total of 22 such inte-grators are stored on the drum. The control and arithmeticcenter sees only one digit of one integrator at a time. Thenumber system is binary and the additive transfer processis accomplished by serial binary addition. A one-word reg-ister on the drum stores and routes the outputs of the inte-grators to inputs of other integrators. Each integratorcontains a code section which defines the manner of con-nection with other integrators in the machine.

The memory stores 22 pairs of 48 digit words on twochannels. A one-word register is provided on a third chan-nel by closer spacing of the heads. A clock or timing sourceis permanently recorded on the drum for synchronizationand constant delay adjustment of the memory. The drumis used as a delay memory in constant re-circulation. Anon-return to zero system is used, the memory outputcontrolling a flip-flop on two grids. At some later date, apaper by Harold H. Sarkissian will cover fully the details ofthis memory.1

The machine contains 60 vacuum tubes including thoseused in the memory. Additional tubes are present in thefour regulated power supplies. One thousand germaniumdiodes are used for logical operations and for clampingand coupling on the flip-flops.

Abstracts of Papers Presented to the ACM Conference at Rutgers University

continued on p. 18

Figure 7 shows the Maddida at the RutgersUniversity presentation.

After Rutgers, the Maddida was shipped toWashington, D.C., for another demonstrationto the Air Force and others at the Pentagon.“What we were trying to do,” explainedSprague,

was to show the technology, both the digital dif-ferential analyzer technology and the computerdesign technologies that went into it, to a numberof important people in the Air Force who spon-sored and were the contracting agency for the MX-775, the Snark missile. … Some of the people whoattended the meeting … were … involved with theSnark. Others were invited to attend a demonstra-tion and … discussion of the machine by noneother than Charles Lindbergh [who] was in a sortof evaluation status at that time as an Air Forcegeneral. Part of his responsibility was to visit aero-space firms around the country, evaluating thingsthat were on the frontier from a technology sense.When he came to Northrop Aircraft in 1948, wedemonstrated Maddida to him … and the assistantsecretary of the Air Force. …

[The meeting] was held in … a sort of audito-rium with a slanting floor. I will never forget theMaddida on a podium, a raised stage, in a spot-light with red cloth draped over it, really donewell, with all these Air Force generals and offi-cers out there. … Floyd … gave the introducto-ry presentation on the digital differentialanalyzer concept. …

It was a classified presentation, by the way, sowe talked about the missile and the guidanceproblems and what Maddida’s original objectivehad been. … Then we demonstrated themachine solving some differential equations.

The reaction first was skepticism. … Then …just tremendous enthusiasm. Not only for themachine but what its implications were. …[S]ure, it was not yet airborne, which was theultimate objective, but it was not very far fromit. The size was getting down to the point whereit looked like it could be made to fly. Reliabilityand withstanding shock were something elseagain, but it was well received.

Figure 8 shows the control panel of theMaddida.



Design of the machine is based on the complete appli-cation of symbolic logic to parallel the model of a com-puter which is stated as a group of stable state devices thatsuccessively change themselves to a new configuration. Alogical algebra using the “+” and “x” operators to definethose operations directly accomplished by diodes, namely,the “inclusive or” and the “and,” is utilized to write com-plete equations about each grid of every flip-flop in themachine. The memory record functions are also written inthis algebra.

We have been using this algebra to design digital elec-tronic circuitry for about two years, but this is our firstmachine in which it is applied in its complete form. The set oflogical equations completely replaces the block diagram forthe machine. All operations, including counting, are accom-plished in this one general manner. The result is a machinewhich, in internal operation, is completely parallel, all flip-flops changing at the same time, i.e., at each clock interval.

The equations written define when a pulse shouldappear on the grid of a particular flip-flop in terms of theconditions of other flip-flops in the machine. The flip-flopscontrolled by the memory enter into these equations inthe same way as all others. Each independent equation issimplified by the use of the theorems of the algebra. Anattempt to find the most economical arrangement ofdiodes for groups of equations is then made by represent-ing an arrangement in symbolic form and counting thediodes needed for each arrangement tried.

The resulting diode combinations are designed by a

technique which results in applying two simple formulasrepetitively starting at the highest level and working downto the basic flip-flop drivers. Often, some conditions in anet of diodes are impossible due to known relationshipsamong the driving propositions, and these known rela-tionships are taken into account during the design. Theresult is a selection of all resistors in the net and a tabula-tion of maximum load on each flip-flop plate. In only onecase, in the present machine, was it necessary to parallel aflip-flop with another tube in order to handle the load. Themaximum load allowed on a flip-flop is held well belowmaximum rating in order to allow for aging of the tubes.

After construction of the machine, about two monthswere required for initial check out. Very high reliability hasbeen experienced. Fifty hours of operating time wererecorded from the end of the initial check out to the firstfailure, which was found to be a critical overload causedby design error. Modifications to include a slightly moreconvenient readout were made at this time, and furtherreliability checks have not yet been undertaken.

Acknowledgment is made to Harold Sarkissian and Carl Isborn for their developmental work on the magneticmemory, and to Harold Sarkissian, Al Wolfe, and Dick Sprague for contributions to the logical and circuitdesign of the machine. Willis Dobbins and Jack Donan ren-dered valuable assistance during check out of the machineand in design of the read-out system.

Reference and note1. Such a paper was, to our knowledge, never written.

continued from p. 17

Disenchantment and disaffectionBy the time we returned to Northrop, we

were in a high state of enthusiasm and lookingforward to an opportunity to tell the story ofthe various adventures of the Maddida toProject MX-775 management and personnel.Some of us carried “requests for bids” from var-ious organizations that had seen and becomeexcited about the Maddida. Our frame of mindwas such that we definitely expected someform of commendation or praise for our workand results.

However, shortly after returning, we weresummoned by Weaver to a meeting, held inCerny’s office, that included Cerny, Ohlinger,Taylor, and another manager, named Silliman.Weaver produced a paper that he said wasapproved by Jack Northrop that containedstrongly worded orders to all of us in the com-puter group to “return to your jobs on theProject” and “bring up anything you have inmind” with Taylor “by going through chan-nels.” Neither Weaver nor Cerny nor any of theothers asked to hear about what had gone onduring the weeks we had spent back East. AsCerny left the meeting, I asked him if they weregoing to discuss anything further. He replied“No.” We were crushed. Not only had the com-mendation and praise we had anticipated notbeen lavished on us, but we had clearly beenreprimanded. MX-775 project management,entirely frustrated with this “wild group andtheir bootlegged program,” was in no mood tohear glowing tales of outside interest generat-ed by our trip.

With respect to whether Northrop manage-ment had rejected the Maddida “because theythought it couldn’t handle the problem athand or … because they did not want to gocommercial,” Reed was “convinced that it wasa rejection of the people. We had become toohard to manage and control, and bright as wewere, they must have decided that it was betterto let us go.”

More than a year earlier, Herbert Metcalf, theresident patent counsel for Northrop, hadassigned John Matlago, a young patent engineerwho had been with the company for about twoyears, to work with, and follow the develop-ment and possible patent activities of, the com-puter group. Following our disastrous meetingwith the project management, Matlago hadpleaded with Metcalf to encourage JackNorthrop to set up a department or company todevelop and manufacture computers. Metcalfagreed, but asked Matlago to find out, in theabsence of any oral or written report on itsactivities, what had transpired during the com-


Figure 7. A copy of the picture and story from the 28 March 1950edition of the Newark Evening News indicates that the other exhibitorsat Rutgers were Alan B. Dumont Co. of Clifton, New Jersey; ReevesInstrument Corp. of New York; IBM Corp. of New York; RCALaboratories of Princeton; Raytheon Manufacturing Co. of Waltham,Massachusetts; and the Zator Co. of Boston. This picture was madefrom a partial photograph of the Steele-only part of the picture fromthe Newark Evening News archives combined with the microfilm copy ofthe original news article and picture.

Figure 8. Photo of the front control panel of acommercial Maddida. Northrop brochureadvertised the Maddida as a “deskside”computer. (Photo courtesy of the National Airand Space Museum, Smithsonian Institution.)

puter group’s travels with the Maddida. Initially we were defensive, submitting to

Matlago only outlines of our work and therequests for bid that we had received. But aftermaking clear that we had fully expected to givea verbal report of, and be questioned about, ourtrip, and that any lack of disclosure was not dueto our lack of interest or readiness to discloseour activities, we agreed, somewhat reluctant-ly, to prepare a report, which was delivered toMetcalf within a few days. Metcalf, after review-ing the report, told Matlago to return one copyto the computer group and stated that it shouldbe submitted to the group’s immediate super-visor. Still, Northrop management would notinaugurate any exchange or discussion thatmight lead to resolution of the divisive schismthat had developed between the computergroup and Project MX-775 management.

A meeting held by some of the group withTaylor served only to reveal that Taylor pos-sessed neither an understanding of the digitaldifferential analyzer concept nor any interest inhelping us find a use for the Maddida. In anoth-er meeting with Taylor, we (Reed, Sarkissian,and I) were advised by a Dr. Barthay, an AirForce advisor to Project MX-775 who had beenevaluating the missile’s guidance system, thatthe Northrop company and missile project werecommitted to a definitive prototype productionschedule using the present system and that theexisting contract would not allow the develop-ment of an all-digital system of the type beingproposed in conjunction with the Maddida.Barthay added that, in his opinion, an all-digitalsystem would definitely have been superior, butwould have had to have been proposed anddeveloped while the contract was in theresearch and development phase.

Sarkissian and I concluded that because themilitary system using the Maddida had been“blocked in the project,” it was too late tosecure a new development contract to pursuean alternate approach. In retrospect, it is alsonow clear that misunderstanding and lack ofcommunication extended beyond Project MX-775 management to the entire Project MX-775team and the rest of the scientific developmentorganization.

Founding of Computer ResearchCorporation

“We had been talking about forming a com-pany for months,” Sarkissian recalled.

Floyd would vacillate: one day yes and one dayno. The situation kept getting worse. … After themanagement meeting, I remember Don and I

talking about the lack of any good future atNorthrop. I suggested we get a temporary jobsomewhere until Floyd and the others decidedthey had had enough, and then we could gettogether and do something. I actually arrangedfor a temporary job for both of us [at ElectronicEngineering Company of California]. …

There were five of us involved: Don Eckdahl,Dick Sprague, Floyd Steele, Irving Reed, andmyself, Harold Sarkissian. … The five of usformed the company. We sat down in ManhattanBeach and formulated the company more or lesson broad principles. Don and I started the processby quitting our jobs and leaving Northrop. … Therest of them quit over the next three weeks. Inthe meantime, Don and I were running aroundgetting things set up. We rented … space on thesecond floor of a knit shop in Manhattan Beach;we could walk to work. On May 8 [1950] the fiveof us formed a formal partnership.

Then we started looking for contracts. Weneeded money. We did not take any salary for aslong as we could afford not to. We simply builtup a backlog of debts. As I remember it, this com-pany was really supported by widows andorphans; it was really a family affair. … Therewere no large sums; there were many little ones. Ido not think our financing was any great shakes.As I remember, it was something like $30,000.

Reed and I recall the total being about$70,000, including a large contribution from aReed cousin. During this period we struggled tofind customers that might contract for thedevelopment of some type of computer. Steeleevery so often would make a plea to “pull outall the telephones and go back to design.” Butthe rest of us, intent on securing contracts,were preoccupied with sales-related activities.

On 16 July 1950, in the Longmont, Col-orado, office of attorney Lyman P. Weld, weformed Computer Research Corporation. Sixdays later the then-new CRC board voted tooffer to buy (with shares of CRC stock) the orig-inal partnership. The five partners accepted theoffer. Figure 9 shows a press release thatannounced the founding of CRC.

Dobbins, Isborn, Matlago, and Donan sub-sequently joined the firm, together with AlbertWolfe and Bernie Wilson. Senior electronicstechnicians employed as development engi-neers, Wolfe and Wilson were heavily involvedin the construction and testing of the digitaldifferential analyzer CRC later built for NorthAmerican Aviation. Dobbins and Donan, bothelectrical engineers, worked on circuit and logicdesign for many CRC products. Isborn, an elec-trical engineer and physicist, worked with



Sarkissian on magnetic drum and tape memo-ries. Matlago, with degrees in both engineeringand law, wrote all of CRC’s patents.

First development product contractOn 22 September 1950, North American

Aviation signed a $45,750 contract with CRCfor the development and manufacture of a dig-ital differential analyzer with the followingspecifications:

• 50 integrators,• 150 pounds and 7 cubic feet maximum,• powered by 400 cycle ac (airborne) conven-

tional aircraft (propeller or jet), and• operational up to 25,000 feet pressure alti-

tudes and ambient temperatures rangingfrom 0 to 35 degrees Centigrade.

Steele and Reed probably were instrumentalin cultivating the interest of the correspondinggroup at North American Aviation before leav-ing Northrop. They had met with that group,which was developing a similar missile systemcalled the Navaho, to discuss common prob-lems associated with guidance systems. WhenNorth American engineers Jerry Weidermanand Lester Kilpatrick subsequently visited CRCto discuss how the Maddida might be used inthe missile guidance problem, Reed and I seizedthe opportunity to negotiate and close the con-tract with them.

Just prior to formally landing the NorthAmerican contract (see Table 1, next page), wehad moved CRC to somewhat larger quarters inTorrance, California. We converted the secondfloor over a bakery, which housed six apart-ments—each containing a kitchen, bath, andliving area—into development labs, parts stor-age, and a prototype construction area. It wasin this building the following spring that wecompleted construction of the digital differen-tial analyzer for North American Aviation.Outside facilities were used to run the environ-mental tests stipulated by the contract. Themachine passed all tests with minimum trou-ble and required no significant modifications.The product was delivered and accepted by thecustomer on 15 May 1951.

The sustaining AFCRL contractIn late summer 1950, we were visited by John

Marquette, head of the US Air Force’s CambridgeResearch Lab (AFCRL), George Valley, and twoor three Air Force officers who inquired if wewere the people who had developed theMaddida and written proposals about the detec-tion and observation of aircraft using a network

of radars. The answer, of course, was yes.“Since we had so many ideas along these

lines,” Reed recalled,

they decided to see how well we could do andthey paid our way … to Cambridge … ResearchLabs. We talked to the people there about whatthey had in mind for detecting and observing air-craft, what their thinking was concerning theSAGE [semi-automatic ground environment] typething of the future, and the types of radar theywould probably use. We were then taken to visitthe Instrumentation Laboratory and met JayForrester and all the people on [the] Whirlwind I


Figure 9. This press release in the Redondo Beach,California, Daily Breeze tells the story of the founding of CRC.

computer. When we saw this gigantic machine,we were convinced even more that we were real-ly on the right track by building reasonably smallprocessors rather than these gigantic machines,which, we thought, would bankrupt the country.

We outlined what we might do under contractand … Marquette was very impressed with us. Hewas very interested in CW [continuous wave]radar, and we saw a demonstration of the con-tinuous wave radar that had been built. Part ofour contract was to study how to detect andtrack aircraft with a set of CW radars. We wereawarded, probably about December 1950, a two-part contract: one was to study the problem ofdetecting and tracking aircraft with CW radarand the other … was to design and build a gen-eral-purpose computer that could be put at CWradar sites. As I recall, the contract was exactly$150,000, split $75,000 for the study and$75,000 to build a full-scale, general-purposecomputer. We performed on both halves reason-ably well. … We managed to do the whole job onless than $150,000; we built a full-scale general-

purpose computer calledCadac [Cambridge Air ForceDigital AutomaticComputer], which was theprototype of what was latercalled the CRC102A.8

My primary job was direc-tor of theoretical research; Ithink that was the title.Basically, I had … responsi-bility for the study part ofthat contract. I hired severalmathematicians: ChesterStone, Leonard S. Abrahams,… Hal W. Banbrook. …

Sprague described theWhirlwind as

a great, big thing stretchingin one direction as far as theeye can see—an aisle goingdown the middle and thenracks running off to twosides going out maybe 25 to30 feet in either direction onboth sides. … [Y]ou had tohave a road map becauseyou go down five aislespaces and then over to theright six racks, let’s say, toget to a part of the machine.I always had this great ambi-tion of taking the CRC102,which had a 1,000 word

memory—the Whirlwind had 500 words in themain memory, or the core …—and wheeling theCRC102 through the doors, down one of theaisles, and over to one of the corners.

Marquette also arranged for us to visit anexisting, manually structured aircraft schedul-ing and control section at La Guardia Airportin the New York area. There we learned a greatdeal about current systems and the significantscope of the problem of tracking aircraft andmissiles on an automatic system.

The AFCRL contract was defining for CRC,generating both sufficient funds and pressureto drive the design and manufacture of a gen-eral-purpose computer. As Reed and his groupembarked on a study phase, Steele, Sprague,Sarkissian, and I concentrated on design anddevelopment.

A decision was made to go forward with thefollowing set of preliminary design specifica-tions. Drum memory was to be about 1,000words, each word a 33-bit binary number



Table 1. Excerpt from table reproduced in J.E. Sammet, “Answers to Self-StudyQuestions.”7

Year Month Name Place

1951 Jan.–Feb. Harvard Mark III Cambridge, Mass.February Ferranti Mark I Manchester University

Burroughs Lab. Comp. Detroit, Mich.April LEO Lyons & Co., London

Whirlwind Cambridge, Mass.June Univac, #1 Philadelphia, Pa.

Maddida, #1 Northrop, Los AngelesMaddida, #2 North American Aviation, Los Angeles

Summer IAS Princeton, N.J.Late Pilot ACE NPL, Teddington, England

1952 January Cadac USAF Cambridge Research CenterFebruary Hughes Airborne Wright-Patterson AFB, DaytonMarch Maniac Los Alamos

Ordvac Aberdeen, Md.April EDVAC Philadelphia, then Aberdeen, Md.

Harvard Mark IV Cambridge, Mass., then Dahlgren, Va.August Illiac Urbana, Ill.Summer Teleregister SpeedH, #1 Griffis AFB, Rome, N.Y.

Univac, #2 The Pentagon, USAFTeleregister SpeedH, #2 Griffis AFB, Rome, N.Y.

August Magnetronic Reservisor New York (LaGuardia Airport?)November Elecom 100, #1 Aberdeen, Md.

Univac, #3 US Army Map ServiceMaddida, #3 Univ. of Utah, App. Phys. LabMaddida, #4 Univ. of Utah, App. Phys. LabMaddida, #5 Univ. of Utah, App. Phys. LabMaddida, #6 Arnold Engineering

December Narec Naval Research Lab, Washington, D.C.IBM 701, #1 IBM WHQ, New York City

recorded in binary serial on the drum. Thedrum was to contain a clock channel that, atnormal speed, would generate a square wave ofpulses 10 microseconds wide and 10 microsec-onds apart, yielding a clock frequency of50,000 cycles (that is, Hertz) per second. Weplanned to use a three-address command struc-ture, the first two octal digits being the com-mand and the next three the address of thefirst operand; the second three being theaddress of the second operand; and the thirdthree being either the address at which theanswer was to be stored or the address desig-nated by a compare or jump command. I recallSteele being a party to these parameters andfrequencies, which were similar to thoseemployed for the Maddida, and having con-ceived the concept of a three-address com-mand structure. Sprague and I subsequentlycommenced work on the computer’s logic andcircuit design, Sarkissian and Isborn on thedevelopment of a drum memory.

The Maddida patent application issueIn late 1950, CRC resolved to try to secure a

license from Northrop to avoid later argumentsover patent rights to our various planned prod-ucts. We had all signed Northrop patent agree-ments that clearly stated that all inventorswould assign their rights to Northrop Aircraft.We tried to promote the rather weak argumentthat the Maddida had been developed “in spiteof Northrop Aircraft” and, therefore, was real-ly partly owned by the inventors. But in a keymeeting, Steele, Matlago, and Richard Dabneywere unable to convince Jack Northrop of thevalidity of this position. We then developedanother questionable strategy of withholdingthe signatures of the inventors as a way to forceNorthrop to permit the patent to slip into thepublic domain as a result of late filing. On 20 March 1950, the five investors still at CRC(Steele and William Collison had left) offeredto sign the patent application provided thatcertain “conditions” were met. Northrop atthat point decided to locate and file, with Steeleand Collison only, a potentially illegal or, atbest, weak patent rather than put up with fur-ther antics on the part of the CRC people.Consequently, the filed patent application byNorthrop Aircraft omitted Sarkissian, Sprague,Isborn, Wolfe, and me.

In later years, there was an “interference”fight at the patent office level betweenCRC/NCR and Northrop Aircraft, but there wasnever any move by Northrop to legally interactwith CRC or NCR on the patent rights to theMaddida.

Inventing the flow diagram system oflogic design

Applying the logic equation concepts of theMaddida to the general-purpose structureinvolved a series of command sequences thatevery so often interrelated the logic with mem-ory operations. Progress was slow until one all-night session in Washington, D.C., duringwhich the flow diagram and program counterconcepts, and a major part of the design, wereconceived.

Sprague described that episode in detail in“A Western View of Computer History” pub-lished in the Communications of the ACM.9

In the winter of 1950 … we had reached a pointin the logical design of the CRC102 that calledfor some clear thinking uninterrupted by the restof the problems of a small, growing company.We were to go to Washington, D.C.; we decidedto take all of the 102 logic diagrams and Booleanequations with us and hole up in our hotel untilwe solved the problem.

We had reservations at the Statler Hilton, buton arrival discovered they had overbooked andwe were spilled over to the Lee House aboutthree blocks away. The Lee House Hotel was veryold and cold and appeared to have been con-structed at the time of General Lee’s surrender.Between 6:00 p.m. and 10 p.m. we spread dia-grams and equations all over the beds, the floor,the walls, and two card tables we were able toborrow. Each bellhop who arrived bringing morecoffee looked around the room incredulously.

Around midnight we were in our pajamas andjust beginning to see a glimmer of light on ourproblem when the hotel fire alarm sounded. Weran out of the room and looked for the fire exit,calculating that the elevators might be filled withsmoke. The fire door led down an inside circularstairwell and as we entered it we wondered whyno one else appeared in the hallway with us.

We ran as fast as we could down eight flights ofstairs, burst into the lobby, and found completeserenity and no smoke. We dashed up to the deskin our pajamas and bare feet and asked where thefire was. The night manager looked up over thetop of his glasses and said: ‘‘Oh, that alarm is onthe fritz. It goes off by itself every so often.”

Laughing, we took the elevator and went backto our design task without thinking that a realfire would have destroyed the CRC102.

By 4 a.m. we had evolved the flow diagrammethod, although we had not realized it yet. Wewere on our fourth pot of coffee when we hearda loud crashing sound outside our door. Wefound an elderly, gray-haired man lying in thehallway. … His head was bleeding.


While Don helped him I rushed back to thephone and called the front desk. When I told theclerk what had happened, he asked me what theman looked like. I described him and he said:“Oh yes, that’s the Colonel. He’s always doingsomething like that.” Fortunately for theColonel, he was not as hurt as the blood led usto believe. The hotel doctor didn’t arrive forabout half an hour and only then were we ableto return to the design of the 102.

By morning we were finished. The flow dia-gram technique was perfect and the CRC102 log-ical design entered its final phase. I suppose theLee House should be given some sort of credit forkeeping us awake to complete our invention.

One important objective in our first gener-al-purpose computer design was to significant-ly reduce the number of flip-flops. At that time,each flip-flop required two vacuum tubes,which were costly, generated heat, and tendedto be somewhat unreliable. We thought wecould share flip-flops between various func-tions, the same flip-flops used during com-mand setup, for example, being used also fortemporary storage for carry operations associ-ated with arithmetic commands. Other flip-flops that did not have to be dedicated, such asthose needed to time functions of mechanicaldevices such as printers, could also be used forcomputation functions.

At the time Sprague and I went to the LeeHouse, we had begun to use individual sheetsof large accounting paper to describe whatwent on at a particular time or in a particularstep (for example, which side of which flip-flopwas being used for a particular equation). Witheach large accounting sheet representing thelogic and connections for a particular time orsequence slot, it followed naturally that multi-

ple operations could be described by stackingthese sheets. It occurred to us that, each ofthese sheets being like a block in a flow dia-gram, a program counter could be made to con-trol the sheets of sequences.

The result was a method of defining themachine in the time domain by associating theprogram counter with the flow diagram steps.Previously, machines had been defined in a spa-tial sense, that is, by identifying in drawingsthe accumulator register, adder, commanddecoder, and so forth. In the case of spatialdescriptions, the time sequence was implied orexisted only in the designer’s mind or memo-ry. For example, complete sets of spatial draw-ings existed for the Binac and Univac, but therespective time domains and sequences existedonly in Bob Shaw’s head. The physical drawingof the CRC102 was just a list of flip-flops byname. A flow diagram described the machine’stime sequence and a set of Boolean equationsdescribed the logic between the programcounter, flow diagram, and all the flip-flops andstable state devices in the machine. Figure 10shows a reduced copy of the CRC102/Cadacflow diagram.

Progress on the CRC102, which we renamedCadac (for Cambridge Air Force DigitalAutomatic Computer), was significant throughthe rest of November 1950, with Sprague and Imaking strides in the logic, Sarkissian andIsborn on the drum memory, and Dobbins,Wolfe, Donan, and Wilson on the circuit andmechanical designs. By May, the logic wascomplete.

Cash and management woesRichard Dabney was recruited, at Steele’s

instigation, in October 1950 as CRC’s businessmanager. He brought many key skills fromplanning and manufacturing management atNorthrop Aircraft. Hired in mid-November thatsame year as finance manager, Jack Warshauer(a friend of Dabney’s) was previously a profes-sor of finance and accounting at the Universityof Alaska in Fairbanks. He arrived just as wewere trying to secure our first progress pay-ments from the Air Force. We were paid in earlyDecember following a progress payment auditon 30 November 1950.

Under the business and financial leadershipof Dabney and Warshauer, we completed thecomputer hardware design and preparedreports on the system and technical memoryresearch. These actions were required to con-tinue to receive progress payments from the AirForce and thereby stay alive financially. We alsocompleted the debugging and testing of the air-



Figure 10. This is a much-reduced copy of the Cadac/CRC102 flowdiagram dated 8 December 1950 and signed by Donald Eckdahl,Richard Sprague, and Will Dobbins (the original is 9 feet long by 3 feethigh). Each small block on the diagram is a step in the flow andoriginally was one of the large sheets of accounting paper mentionedby Sprague in the main text.

borne digital differential analyzer for NorthAmerican Aviation.

Angels from Los AngelesAlthough the progress payment system asso-

ciated with the Air Force contract contributeda great deal, CRC nevertheless found itself cashstrapped as 1951 wore on. About March,Dabney, through some previous contacts, man-aged to enlist the help of two “angels.” GeorgeFuller and Gordon Turnbull, two wealthy LosAngeles businessmen, agreed to loan us moneyagainst future contracts and progress payments,solving CRC’s financial problems for the rest of1951 and into early 1952.

It also helped that on 15 May 1951, CRCdelivered the digital differential analyzer (DDA)pictured in Figure 11 to North AmericanAviation for payment, after it had passed pres-sure, shock, and vibration tests for airborne use.

Diverging philosophiesSteele was often gone from the operation

during this period, pondering and inventingnew ways to think about digital systems andcomputers. He was interested in the logic designconcepts we had developed, but never got closeenough to see, for example, how a complete setof Boolean equations could be written and tiedtogether by a flow diagram. Believing, never-theless, that he totally grasped the flow diagramconcept, he moved directly to try to conceivethe next step in what he called the “tabula rasa”(that is, blank tablet) machine. I believe he actu-ally began to envision how future machinesmight be built using the technique that hascome to be known as microprogramming.Steele would literally disappear from the officefor weeks. The rest of us would muse that “hehad been out in the desert, sitting on a sharprock and thinking.” We recognized him to be agenius and believed that eventually he wouldland on something vital, but his attitude andactions were threatening the nature and con-tinued existence of our new company. Evenbefore we left Manhattan Beach, Steele hadwanted us to stop work on Cadac and start overby finding a way to design his novel concept ofthe tabula rasa-like machine.

In March 1951, about the same time CRC’sangels came through, the schism betweenSteele and those of us who had been his disci-ples came to a head. Our commitment to moveforward as a businesslike organization to meetour schedules and contract terms with AFCRLwas not shared by Steele. He was seldom pres-ent and not prepared to add his thinking to theproblems of money, contract terms, and other

business matters. Our design activities also suf-fered as Steele withheld his creativity. Theteacher–disciple relationship frayed and ulti-mately snapped. Only Reed (whom I believenever fell as far as the rest of us into blind-faithmode) seemed able to maintain communica-tion with Steele.

We finally decided to hold a board meetingto make the formal motions and secure thevotes to remove Steele and install Dabney aspresident. Steele’s brother, Arnold, the attorneywho had helped us create CRC and subse-quently advised us, both provided guidance inhow to conduct the meeting and representedand protected his brother. The meeting was dif-ficult and strained. A few weeks later, Steeleresigned from CRC. It was a difficult and emo-tional time for all of us.

Cadac meets the pressDebugging of the Cadac, which continued

throughout the summer and fall of 1951, was a


Figure 11. (a) Airborne Digital Differential Analyzerdelivered to North American Aviation. (b) Logicsection. (c) Drum memory.

(a)

(b)

(c)

24-hour process. Sprague and I took some shifts;Dobbins and Wolfe, others. Present during mostshifts to provide drum memory support waseither Sarkissian or Isborn. We recorded onmagnetic tape events, problems, and solutionsas they occurred so that sometimes brief verbalsummaries by tired people could be supple-mented with detailed recorded recountings.Because the germanium diodes were mountedin clips on large diode boards, we could executeuntested commands one at a time in an order-ly process through step-by-step insertion of thediodes that would activate them. In the earlyyears of digital computer development thedebugging process was sometimes referred to as“commissioning,” a probably more descriptiveterm that we believe originated with UKengineers.

By mid-November, the Cadacsystem was fully operational, andwe showed and demonstrated it toeveryone, including our originalinvestor family and Turnbull andFuller, as well as people from theAir Force. A press conference anddemonstration held at CRC,which drew several members ofthe local press, generated someunexpected excitement as Spraguelater recounted in the ACM article“A Western View of ComputerHistory.”10

Sprague, as vice president of mar-keting, held a press conference atwhich a reporter asked whether themachine being demonstrated could“think.” Sprague explained, usingthe analogy of chess playing, thatcomputers of the type being demon-strated, appropriately programmedby experts and much faster, mightplay and win often against humanchess masters. The reporter’s papersubsequently ran a story about an“electronic brain” that could playchess against a human being and winevery time. Even as Sprague tried tocalm the other, rather upset, CRCofficers, the UP wire service pickedup the story and it spread nationally.Things got worse as a rival general-purpose computer maker challengedthe Cadac to a chess match, radioprograms solicited interviews, andtheir producers requested that themachine appear on Edward R.Murrow’s “See It Now” program andon the “Dean Martin–Jerry Lewis

Show” to play chess with (and lose to) the lat-ter. The Air Force, to Sprague’s considerable sur-prise, was pleased with the publicity and thematter was laid to rest by explaining that thecomputer had to be shipped directly to theCambridge Research Lab to perform importantdefense-related work.

The Cadac was shipped to AFCRL on 13 December 1951. In January 1952 it wasmoved to and made operational at MIT’sLincoln Laboratory in Massachusetts. Figures12 through 15 show photos of the Cadac, itsmemory, and some of its peripherals.

CRC maturesCRC’s success in acquiring contracts to build

larger computers was accompanied by changesin facilities and personnel.



Figure 12. Like the Maddida, the Cadacused large boards to mount the manygermanium diodes used for logicoperations. The vacuum tubes,primarily flip-flops, were mounted ontop. The power supply was under-neath the diode board structure.

Figure 13. The Cadac drum memoryincluded the clock channel and datachannels that could store 1,024 32-bitwords, or about 4,000 bytes.

Figure 14. The Cadac and peripherals.Richard Sprague, left; Donald Eckdahl,right.

Figure 15. Cadac just before delivery toCambridge Air Force ResearchLaboratory. Richard Dabney, left;Donald Eckdahl, right.

Contracts for larger computers During the late spring and early summer of

1953, we courted a possible contract to devel-op a much larger general-purpose computerthat would satisfy all of the US Navy Bureau ofAeronautics’ business applications—to includepayroll computation, spare parts management,and many other nontechnical applications. Weproposed an ambitious system involving a10,000-word (versus Cadac’s 1,000-word) drummemory, one track of which was to containapproximately 1,000 words with 10 heads pertrack to speed memory access.

The system also involved contracting for thedevelopment of a high-speed mechanical print-er and developing a magnetic tape storage unitto supplement drum storage. Although IBMpunched card input-output units were to be akey part of the system, we also committed todevelop a special input preparation device(called TEPU, for tape editing and printingunit) that would accommodate keyboarding tomagnetic tape. The machine would store andcalculate with decimal numbers (coded in bina-ry) instead of direct binary storage of arithmeticoperations as was the case with Cadac.

On 10 November 1953 we signed, with theNavy Bureau of Aeronautics, a contract to buildwhat we called the CRC107. Shortly thereafter,we secured from the White Sands ProvingGround a contract to build a similar machineto perform missile data reduction and other sci-entific activities for the White Sands MissileRange, New Mexico (see Figure 16).

Changes in plant and peopleBefore the Cadac was shipped in early 1952,

CRC, needing more space for both engineeringand manufacturing, moved from over the bak-ery in Torrance to a 21,600-square-foot indus-trial building in Hawthorne, California, closeto Northrop Aircraft. About the same time,Reed, anguished over the falling-out among thefounders that had preceded Steele’s departure,left CRC. A good friend of Steele, he neverthe-less appreciated the viewpoints of the otherCRC founders. Reed subsequently joinedGeorge Valley at the prestigious LincolnLaboratory. He later became a professor of elec-trical engineering and computer science at USCand, ultimately, an eminent mathematician.

When businessman/angel Turnbull died inearly 1952, Fuller became concerned aboutgoing forward without his partner. He had alsobecome concerned about our current methodof financing. The cash advances on anticipatedprogress payments were still somewhat satis-factory, but we had begun to design a commer-

cial model of the Cadac that we called theCRC102A, and he doubted that he could pro-vide the capital to finance inventories, produc-tion facilities, and sales activities. Dabney, whohad been our principal contact with the part-ners, advised us that Fuller’s position was thatwe had plenty of time, but should begin to lookfor other methods and sources of financing.

We still were unsuccessful in interesting thetraditional investment community, and theventure capital industry was at the time almostnonexistent. One Boston-area new-ventureinvestor, American Research and Development,declined a solicitation by Dabney andWarshauer.12 Our other option was to sell CRCto a company already in the electronics oroffice equipment field.

In search of a buyerThe first prospective buyer we approached

was Pasadena, California-based ConsolidatedElectrodynamics, a smallish but well-estab-lished company that was one of the producersof the mass spectrometer, a large scientificinstrument that identified samples of unknownmaterials through spectral analysis. Philip Foggwas the company’s president; Jim Bradburn, its


Figure 16. Data reduction at McDonnell Douglas,as at “every other aircraft company, typically usedwhat can only be described as ‘stockyards’ ofhuman ‘computers’ who sat at desks and usedmechanical calculators like the popular Friden orMarchant models of the day. The scene was rightout of Dickens: Rows of crewcut young men as faras the eye could see in shirtsleeves and skinny tiesfilling in calculation sheets month after month,year after year.”11 (Photo courtesy of P. Ceruzzi,Beyond the Limits: Flight Enters the Computer Age,MIT Press, 1989.)

vice president of engineering. My recollectionis that Fuller made the contact and was instru-mental in establishing a framework for thenegotiations. Fuller was looking to the compa-ny to make loans against CRC contracts as heand Turnbull had done and buy him out for anegotiated price. This apparently was the onlyarrangement, no price or value being estab-lished for the rest of CRC or its other share-holders. When a New York banker representingConsolidated Electrodynamics made a defini-tive offer of price and conditions, Fuller was sodisappointed and angry that he broke off nego-tiations immediately and left the meeting.13

Sprague, largely on his own, made overturesto a brother of Walt Disney named Larry, whorepresented himself to be in charge of “DisneyEnterprises,” only to discover that the latterhad neither funds nor any authority or respon-sibility for his brother’s money.

In July 1952, Warshauer and I traveled tothe Midwest in search of a buyer. Warshauerhad arranged a meeting with some of the topmanagement of Admiral Radio in Chicago, buttwo hours of courteous, exploratory discussionfailed to develop into any serious interest.

An adding machine salesman from NCR,with whom Sarkissian, at the behest of hisbrother-in-law, had agreed to meet, had askedif it would be all right to tell his regional man-ager about CRC. He subsequently did, and wesoon received a visit from Jim Boyle, theaccounting machine sales manager out of thecompany’s Los Angeles office. Boyle, extreme-ly enthusiastic about CRC, arranged for us tovisit NCR headquarters in Dayton on 29 July.

NCR gave us a red-carpet welcome. We hadlunch in the Horseshoe Room, which at thetime seated about 200. Afterward, like all spe-cial guests of the company, we were presentedwith a picture that had been taken duringlunch (see Figure 17). In attendance were all ofthe top management and supervisory manage-ment at the headquarters facilities, whichincluded the large Dayton factory.

After lunch, we toured the headquarters andfactory complex and a nearby residential areaand visited the last home of Orville and WilburWright, which was now owned by NCR andvariously used as a guest house and for enter-tainment and meetings. We sat on the openfront porch and were joined by our lunch hosts,Charles Keenoy, director of product planning,and Joseph Desch, senior director of specialengineering. Williams, the engineering vicepresident [first name not known], did not joinus. We were served cocktails in this afternoonsetting and a little later were joined by execu-tive vice president Robert Oleman. Warshauerand I told the CRC story, and the NCR team lis-tened intently. Desch later made a definitiverecommendation that his management proceedaggressively to forge a relationship with CRC.

The CRC/NCR relationshipCRC was subsequently visited by Desch and

some of his subordinates: Keenoy and othersfrom the product planning organization, exec-utive vice president Oleman, and, toward theend, by NCR president Stanley C. Allyn. BySeptember 1952 we had signed an agreementletter outlining general terms. CRC was toremain a separate corporation, under its ownname. NCR was to buy 80 percent of CRC’sstock, including that of the original investors,and pay off the long-term loan owed to Fuller.The founders and other employees were toretain 20 percent ownership of the company.The total value of this transaction was approx-imately $1 million. Following the negotiations,Oleman hosted a fancy dinner party for thefounders and their spouses at Chasen’s restau-rant in Hollywood. All of us were familiar withthis prestigious, upscale Hollywood restaurant,but had never imagined dining there. Positiveattitudes prevailed on the part of both CRC andNCR, which made for a fine evening.

Owing to a Justice Department consentdecree dating to its early years, NCR wasrequired to testify in federal court on its plansto acquire a controlling interest in CRC.Sarkissian and I participated in the two to threemonths of preparation for what turned out tobe a two-day hearing in federal court in



Figure 17. At lunch on 29 July 1952 in NCR’sHorseshoe Room were (left to right): Engineeringvice president Williams, Donald Eckdahl, CharlesKeenoy, Jack Warshauer, and Joseph Desch.(Courtesy of the NCR Archive at the MontgomeryCounty Historical Society.)

Cincinnati, Ohio. Court approval for the trans-action was received in December 1952. ByFebruary 1953 the NCR transaction was com-plete. Figure 18 shows the announcement.

Culture clashAfter an initial honeymoon period, prob-

lems began to develop in CRC’s relationshipwith NCR. Although it had bought CRCbecause of its general-purpose computer prod-uct, NCR was not a systems company; few ofthe sales, product planning, and engineeringpersonnel at Dayton had even a remote under-standing of how or why IBM, which was a sys-tems company, had been developing andselling applications for its punched card systemproducts. NCR had difficulty seeing beyondfixed applications that extended its freestand-ing machines. Nor did it understand the sys-tems concept of programming the capability toexecute radically new business applications andsystems into a general-purpose computer.

The idea that customers might conceive anddevelop the applications of the future (as cus-tomers of punched card systems had beendoing for years) fit neither NCR’s view nor theoffice equipment business philosophy. CRCwas subsequently beset by questions from NCR:How can this “binary” computer ever be animportant tool in business? Is it only a “scien-tific” computer? Why do you use punchedcards and punch card readers and punches,products that aid and abet IBM, for input andoutput? (CRC had been using a punched cardsystem to meet labor and other cost distribu-tion reporting requirements for its many gov-ernment contracts.)

NCR directed us to replace punched cardswith paper tape as the input/output mediumfor CRC computers. NCR neither used IBMequipment directly nor contracted with pro-cessing companies that employed it, andbecause NCR equipment could not performcost distribution and similar functions, thecompany’s manufacturing and accountingsimply did without. Because NCR machinestended to work on the fringes of the operationsof typical industrial corporations, even such anobvious application of a general-purpose com-puter as payroll distribution was not reallyunderstood in its philosophical context.

That Sprague and his small sales organiza-tion had by spring of 1953 accumulated animpressive list of customers for the CRC102A(a fully production-engineered version of theCadac), some of which planned to use the newmachine for commercial (nonscientific) appli-cations, did not alter the view of NCR senior

management, which increasingly verbalizedand wrote about CRC’s “scientific computer.”In June, CRC delivered its first real productioncomputer, a CRC105 (a digital differential ana-lyzer closely resembling the CRC102A in Figure19) to Lockheed Aircraft in Burbank, California.

Meanwhile, the relationship betweenDabney and Allyn had become strained.Dabney viewed Allyn as an equal, a perceptionAllyn did not share. Around midsummer, whenhe tried to reach him by telephone, Allyn foundDabney in the middle of a business trip relax-ing for a few days at a first-class resort in theCaribbean. That Dabney’s attitude had evident-


Figure 18. The NCR press release describing theCRC deal. (Courtesy of the NCR Archive at theMontgomery County Historical Society.)

Figure 19. The CRC102A and CRC105 productionfloor in Hawthorne, California, about 1953.14

(Courtesy of the NCR Archive at the MontgomeryCounty Historical Society.)

ly been anything but humble when the NCRpresident finally reached him incensed Allyn,who ordered that Dabney be fired. The formalfiring fell to Oleman and neither CRC nor otherNCR management had the power to undo it.

ReorientationIn July, NCR dispatched Robert Pierson from

Dayton to be president of CRC. Pierson was amember of the product planning organizationand, like Keenoy, had previously been highlysuccessful in sales and sales management.Pierson decided to run marketing directly, withSprague as his vice president of sales. He askedWarshauer to be vice president of finance and meto be vice president of operations, which includ-ed engineering and manufacturing. Sarkissiancontinued as vice president of engineering. Ourvice president of manufacturing, Harry Swanson,had considerable manufacturing systems experi-ence and had been involved in a number of tech-nology manufacturing plans at Northrop.

Swanson and the people he hired andbrought with him had a significant impact onCRC’s manufacturing philosophies and con-cepts. He was particularly knowledgeable aboutthe learning curve theory developed byNorthrop and other modern airframe compa-nies of that period. He also introduced thenotion that manufacturing assembly workerscould and did contribute more by helping toidentify superior manufacturing methods thanby working harder or faster. These concepts notonly permeated CRC, but also became thefoundation of NCR’s manufacturing philoso-phy in the electronics and computer era.

In October, in his “keeping up with the busi-ness” column in NCR’s Dayton house organ,Allyn represented CRC as the development,manufacturing, and sales arm of NCR’s com-

puter business. By the end of 1953, CRC hadmade considerable progress thanks to a newpresident with a strong sales background andintimate knowledge of NCR politics and people.

The end of CRC NCR’s announcement in late January 1954

that CRC was no longer to be a freestandingcorporation, but instead a division of NCR,required that all remaining founders’ stock besold back to NCR, a severe disappointment tothe CRC founders. This was accomplished on 5 February 1954. We were again not offeredstock options, so had ended up with no finan-cial interest in either CRC or NCR. (During theinitial NCR negotiations we had not even askedfor stock options, in fact, did not know to ask,none of our senior advisors or attorneys havingexplained the concept to us.)

Pierson was subsequently named generalmanager of NCR’s Electronics Division, and henamed me assistant division manager.

First CRC102A delivered Delivery of the first CRC102A to Holloman

Air Force Base (New Mexico) was an importantmilestone for the new division. With the 102Ain full production on our manufacturing floor,subsequent deliveries went smoothly; by Octo-ber, 16 of these computers had been success-fully delivered. Figure 20 shows the CRC102A.

Many more CRC102As and CRC105s weredelivered during the remainder of 1954. NCRmanagement paid the division many visits dur-ing this period to review the progress of bothnew product deliveries and research and devel-opment activities. Figure 21 shows the NCRand CRC management group in 1954.

In mid-1954, the division initiated produc-tion of the CRC102D, a version of the CRC102designed to allay NCR’s concern about the useof binary numbers for internal computation(arithmetic operations were done in binary-coded decimal numbers and output was in dec-imal notation). But continued dissatisfactionon the part of NCR management, particularlyAllyn, with the division’s inability to be a prof-it producer rather than an expense, togetherwith other signals of NCR’s displeasure, ledSprague to carefully document customer satis-faction with the CRC102 computer. Figure 22shows a memo that summarizes the level ofcomputer reliability in the CRC102.

CRC107 deliveredOn 18 August 1953, the division delivered

to the Navy Bureau of Aeronautics inWashington, D.C., the CRC126 magnetic tape



Figure 20. CRC102A computer and output printer.(Courtesy of the NCR Archive at the MontgomeryCounty Historical Society.)

unit, the first piece of the large CRC107 com-puter system that we were contracted to deliv-er. In January 1954, the CRC127 TEPU andCRC128 high-speed printer were delivered. Byfall, the two large units that comprised the cen-tral processor and 10,000-word memory drumhad been installed. Subsequent problems withthe bearings in the memory drum wereresolved through extensive rework and testing.Rendered operational in late 1955, the Navyused the computer for many years thereafterfor payroll, accounting, and parts system pro-cessing. Figure 23 is an artist’s rendition of thesystem delivered to the Navy Bureau ofAeronautics.

Changing of the mission and the guardIn late November 1954, Owen Gardner, head

of accounting machine sales for NCR, assumedsales responsibility for the Electronics Division.In February 1955, Sprague left NCR to pursue analternate career. In March, NCR returnedPierson to his previous position in Dayton andnamed me manager of the Electronics Division,charged with terminating all manufacturingactivities and converting the division charterexclusively to research and development.Significant budget cuts were made throughoutthe organization. Sarkissian left in September1955. A decade later, he founded Major DataCorporation and, frequently ahead of his time,developed electronic voting equipment for

which the US is only now ready.I remained with the Electronics Division,

which, five years later, was renamed the DataProcessing Division, with responsibility for allNCR computer systems’ research, development,and manufacturing. I subsequently was namedvice president in charge of that division and


Figure 21. NCR and CRC management group atCRC facilities in early 1954. From left to right:Robert Chollar, NCR vice president of research;Bob Pierson, CRC president (from NCR); JackWarshauer, CRC treasurer; Stanley C. Allyn, NCRpresident; Dick Sprague, CRC vice president ofsales; Bob Oleman, NCR executive vice president;Donald Eckdahl, CRC vice president ofoperations; and Harold Sarkissian, CRC vicepresident of engineering. (Courtesy of the NCRArchive at the Montgomery County HistoricalSociety.)

Figure 23. Artist’s rendition based on design datafor the CRC107 installed at the Navy Bureau ofAeronautics facility in Washington, D.C.Photographs could not be found of the CRC106or CRC107. (Courtesy of the NCR Archive at theMontgomery County Historical Society.)

Figure 22. A summary memorandum of reliabilityand general customer satisfaction relative to theCRC102A computers installed in November 1954.

later transferred to NCR headquarters inDayton to head, as senior vice president of theengineering group, all NCR engineering andmanufacturing worldwide.

The legacy of the Maddida“Maddida,” the Project MX-775 report

author summarized,

is unique among large scale computers because ofits compactness, simplicity of construction, andaccuracy. A prototype model of Maddida requiredless than 600 man hours for its construction. Itoccupies only 7 1/2 square feet of floor space, yetit contains 22 integrators capable of six decimalplace accuracy. To date, this model has operatedfor a total of more than 270 hours and for asmuch as 60 hours without a failure. Transportedextensively by air, rail, and truck, it has been con-sistently placed in operation within 24 hoursafter delivery. The performance of this simplemodel has been remarkable; the potentialities offuture models appear unlimited.15

Other chroniclers expounded on thosepotentialities. “Maddida represents,” wrote DagSpicer,

a transitional period between two key technolo-gies. While remaining faithful to its roots in theanalog analyzers with which its inventors werecomfortable, Maddida took a bold, bright stepforward into the then-new and computational-ly-driven world of jet aircraft, missiles, and rock-ets. It was such advanced computation, providedeconomically and reliably by machines likeMaddida, that enabled both the computing andaerospace industries to move forward.11

Spicer further observed that “the technologicaladvances of so many computing projects areoften equaled or surpassed by the formation ofcomputer experts trained by the projects them-selves, who then go on to propagate into anddefine the industry.”11

“When Maddida appeared to be successful,”recalled Fred Gruenberger,

a group of men at Northrop … impatient at theefforts to promote such equipment … spun offinto a new company, Computer ResearchCorporation. CRC was ultimately absorbed byNational Cash Register. Northrop meanwhilewent ahead with Maddida and later sold therights to Bendix. Glenn Hagen, who was incharge of the Maddida project, left Northrop tohelp form Logistics Research, Inc., which laterproduced the ALWAC. And Floyd Steele, also of

the Maddida group, formed Digital Controls, acompany later absorbed by Litton Industries.Thus, the digital computer group at Northrop …had a finger in many computing pies, one way oranother.16

AcknowledgmentsAll figures for which no other source is givenare courtesy of the Montgomery CountyHistorical Society, Dayton, Ohio, which gener-ously provided the authors access to its exten-sive NCR archives.

References and notes1. T.C. Bartee, I.L. Lebow, and I.S. Reed, Theory and

Design of Digital Machines, McGraw-Hill, 1962.2. T.J. Bergin, ed., 50 Years of Army Computing: From

Eniac to MSRC, Army Research Laboratories andU.S. Army Ordnance Center and School, 2000.

3. Unless noted otherwise, all quoted material isfrom the Computer Oral History Collection of theNational Museum of American History.

4. T.C. Bartee, I.L. Lebow, and I.S. Reed, Theory andDesign of Digital Machines, McGraw-Hill, 1962,sections 10-8, 10-9, and 10-10.

5. O. Morgenstern and J. von Neumann, Theory ofGames and Economic Behavior, Princeton Univ.Press, 1944.

6. Copies of the papers apparently do not exist; Ibelieve they never existed as I remember givingmy paper from notes, save for a projected trans-parency that contained examples of a portion ofthe Boolean algebra equations that defined theMaddida.—D. Eckdahl.

7. J.E. Sammet, “Answers to Self-Study Questions,”Annals of the History of Computing, vol. 11, no. 1,1989, p. 43. Sammet writes: “As can be clearlyseen in the list … the most common computer in1951–1952 was the Maddida, not the Univac.”

8. Cadac was the name given to the CRC102 com-puter developed under contract for the Air Forceand the CRC102/Cadac was the prototype for asubsequent commercial version of the computercalled the CRC102A.

9. R. Sprague, “A Western View of Computer Histo-ry,” Comm. ACM, vol. 15, no. 7, July 1972, pp.691-692.

10. R. Sprague, “A Western View of Computer Histo-ry,” Comm. ACM, vol. 15, no. 7, July 1972, pp.690-691.

11. D. Spicer, “Maddida: Bridge Between Worlds,”CORE: A Publication of the Computer Museum His-tory Center, vol. 1.3, Sept. 2000, pp. 2-5.

12. But several years later, when the investment com-munity began to understand that there might besome demand for these “giant brains” or “digitalcomputers,” American Research and



Development financed the start-up computercompany Digital Equipment Corporation.

13. Consolidated Electrodynamics subsequently cre-ated its own computer company, Electro-Data,which later became the basis of Burroughs Cor-poration’s entry into digital computers.

14. Tibor Fabian recalled a visit to the CRC facilitymade with UCLA Management Sciences ResearchProject “cubby-mate” Dick Canning in late 1952or early 1953. “After security arrangements, wewere led into a large industrial area, a relativelyempty structure. On one side stood metalframes, similar to industrial metal shelves, in along row, one behind the other. Those in thefront were fairly well filled with electronic equip-ment and a great deal of wiring; those towardthe middle were gradually emptier, while thosetoward the end were almost empty. It was clearthat this was an assembly process, in which theassemblers moved gradually from the first to thelast of the metal frames and added the electron-ics and wiring in the step-wise fashion of anassembly line. As Dick explained, the racks werecomputers-to-be.” From E. Weiss, “Roger SissonBiography,” IEEE Annals of the History of Comput-ing, vol. 18, no. 2, Summer 1996, pp. 67-76.

15. Northrop Aircraft, Maddida (Preliminary Report),Project MX-775, report no. GM-545, 26 May1950.

16. F.J. Gruenberger, “A Short History of Digital Com-puting in Southern California,” Annals of the His-tory of Computing, vol. 2, no. 3, July 1980, p. 248.

Donald E. Eckdahl was born29 April 1924 in Los Angelesand died 23 July 2001 in Tuc-son, Arizona. Eckdahl joinedNorthrop Aircraft in 1946,where he worked as a projectengineer in helping to developcomputer and guided-missile

systems, including the Maddida. In 1950, he was acofounder of Computer Research Corporation, whicheffectively launched a 30-year career with NCR,where he served as vice president and general man-ager of operations of the Electronics Division and,ultimately, as senior vice president of NCR’s engi-neering and manufacturing group. In later years,Eckdahl was a partner in an executive search firm(1980–1985), joined Branford, Connecticut-basedMultiflow Computer as president and CEO(1985–1990), and retired from ESA, a Connecticutconsulting and executive search firm, in 1995. Eck-dahl earned a BS and an MS in electrical engineeringat the University of Southern California. He was amember of both the IEEE and the ACM.

Irving S. Reed is the CharlesLee Powell Professor of Electri-cal Engineering and ComputerScience at the University ofSouthern California in LosAngeles, which he joined in1963. Reed’s career began atNorthrop Corporation in 1949,

where he helped develop the Maddida computer. In1951, he joined MIT’s Lincoln Laboratory and, in1960, Rand Corporation. Reed, the principal origina-tor of the Reed-Muller and Reed-Solomon error-correcting codes, has made many researchcontributions, most recently in adaptive arrays, digi-tal signal processing, detection theory, and VLSIdesign of coders and decoders. Reed received a BS anda PhD from the California Institute of Technology.He served in the US Navy and is an IEEE fellow and amember of the National Academy of Engineering.Reed’s many awards include the IEEE’s Richard W.Hamming Medal in 1989.

Hrant H. (Harold) Sarkissianwas born in 1922 in Fresno,California, and died 15 Febru-ary 2001 in Newport Beach,California, as this article wasbeing submitted to the Annals.Following his years at NorthropAircraft and CRC with his co-

authors, Sarkissian maintained a consulting practiceand founded, in the mid-1960s, Major Data Corpo-ration. His passion, according to wife Martha Sarkiss-ian, was being an electronics engineer, whether onhis knees earning a handsome consulting fee forremoving a piece of chewing gum that had stoppeda room-size dinosaur of a computer in its tracks or, sooften ahead of his time, developing decades ago elec-tronic voting equipment for which the US is onlynow ready. A 1944 graduate of the University of Cal-ifornia at Berkeley, he held a BS in electronic engi-neering. Sarkissian was a member of the Signal Corpsin WWII and was a Life Member of the IEEE.

Readers may contact Irving Reed at Univ. of South-ern California, Dept. of Electrical Engineering, EEBRoom 500, 3740 McClintock Ave., Los Angeles, CA90089-2565; (213) 740-7335.

For further information on this or any other com-puting topic, please visit our Digital Library athttp://computer.org/publications/dlib.


West Coast Contributions to the Development of the General...

Documents

Transcript of West Coast Contributions to the Development of the General...