Anti-Aircraft Fire Control andthe Development of IntegratedSvstems at Sperry, 19251940

d

David A. Mindell

Thedawn of the electrical age broughtnewtypes of control systems. Able totransmit data between distributed compo-nents and effect action at a distance, thesesystems employed feedback devices aswell as human beings to close controlloops at every level. By the time theoriesof feedback and stability began to becomepractical for engineers in the 1930s a tra-dition of remote and automatic controlengineering had developed that built dis-tributed control systems with centralizedinformation processors [I]. These twostrands of technology, control theory andcontrol systems, came together to producethe large-scale integrated systems typicalof World War II and after.

Elmer Ambrose Sperry (I 860-1930)and the company he founded, the SperryGyroscope Company, led the engineeringof control systems between 1910 and1940. Sperry and his engineers built dis-tributed data transmission systems thatlaid the foundations of todays commandand control systems. Sperrys fire controlsystems included more than governors orstabilizers; they consisted of distributedsensors, data transmitters, central proces-sors, and outputs that drove machinery.

This article tells the story of Sperrysinvolvement in anti-aircraft fire controlbetween the world wars and shows howan industrial firm conceived of controlsystems before the common use of controltheory. In the 1930s the task of fire controlbecame progressively more automated, asSperry engineers gradually replaced hu-man operators with automatic devices.Feedback, human interface, and systemintegration posed challenging problemsfor fire control engineers during this pe-

108

riod. By the end of the decade these prob-lems would become critical as the countrystruggled to build up its technology tomeet the demands of an impending war.

Anti-Aircraft Artillery FireControl

Before World War I, developments inship design, guns, and armor drove theneed for improved fire control on Navyships [2]. By 1920, similar forces were atwork in the air: wartime experiences andpostwar developments in aerial bombingcreated the need for sophisticated fire con-trol for anti-aircraft artillery. Shooting anairplane out of the sky is essentially aproblem of leading the target. As air-craft developed rapidly in the twenties,their increased speed and altitude rapidlypushed the task of computing the lead outof the range of human reaction and calcu-lation. Fire control equipment for anti-air-craft guns was a means of technologicallyaiding human operators to accomplish atask beyond their natural capabilities

During the first world war, anti-aircraftfire control had undergone some prelimi-nary development. Elmer Sperry, as chair-man of the Aviation Committee of theNaval Consulting Board, developed twoinstruments for this problem: a goniome-ter, a range-finder, and a pretelemeter, afire director or calculator. Neither, how-ever, was widely used in the field [3].

When the war ended in I 9 18 the Armyundertook virtually no new developmentin anti-aircraft fire control for five to sevenyears. In the mid-1920s however, theArmy began to develop individual compo-nents for anti-aircraft equipment includ-ing stereoscopic height-f inders,

searchlights, and sound location equip-ment. The Sperry Company was involvedin the latter two efforts. About this timeMaj. Thomas Wilson, at the FrankfordArsenal in Philadelphia, began develop-ing a central computer for fire control data,loosely based on the system of directorfiring that had developed in naval gun-nery. Wilsons device resembled earlierfire control calculators, accepting data asinput from sensing components, perform-ing calculations to predict the future loca-tion of the target, and producing directioninformation to the guns.

Integration and DataTransmission

Still, the components of an anti-aircraftbattery remained independent, tied to-gether only by telephone. As Preston R.Bassett, chief engineer and later presidentof the Sperry Company, recalled, nosooner, however, did the components getto the point of functioning satisfactorilywithin themselves, than the problem ofproperly transmitting the informationfrom one to the other came to be of primeimportance. [4] Tactical and terrain con-siderations often required that differentfire control elements be separated by up toseveral hundred feet. Observers tele-phoned their data to an officer, who manu-ally entered it into the central computer,read off the results, and telephoned themto the gun installations. This communica-tion system introduced both a time delayand the opportunity for error. The compo-nents needed tighter integration, and sucha system required automatic data commu-nications.

IEEE Control Systems

In the 1920s the Sperry GyroscopeCompany led the field in data communi-cations. Its experience came from ElmerSperrys most successful invention, a true-north-seeking gyro for ships. A significantfeature of the Sperry Gyrocompass was itsability to transmit heading data from asingle central gyro to repeaters located ata number of locations around the ship. Therepeaters, essentially follow-up servos,connected to another follow-up, whichtracked the motion of the gyro withoutinterference. These data transmitters hadattracted the interest of the Navy, whichneeded a stable heading reference and asystem of data communication for its ownfire control problems. In 1916, Sperrybuilt a fire control system for the Navywhich, although it placed minimal empha-sis on automatic computing, was a sophis-ticated distributed data system. By 1920Sperry had installed these systems on anumber of US. battleships [5].

Because of the Sperry Companys ex-perience with fire control in the Navy, aswell as Elmer Sperrys earlier work withthe goniometer and the pretelemeter, theArmy approached the company for helpwith data transmission for anti-aircraft firecontrol. To Elmer Sperry, it looked like aneasy problem: the calculations resembledthose in a naval application, but the physi-cal platform, unlike a ship at sea, anchoredto the ground. Sperry engineers visitedWilson at the Frankford Arsenal in 1925,and Elmer Sperry followed up with a letterexpressing his interest in working on theproblem. He stressed his companys expe-rience with naval problems, as well as itsrecent developments in bombsights,work from the other end of the proposi-tion. Bombsights had to incorporate nu-merous parameters of wind, groundspeed,airspeed, and ballistics, so an anti-aircraftgun director was in some ways a recipro-cal bombsight [6]. In fact, part of the rea-son anti-aircraft fire control equipmentworked at all was that it assumed attackingbombers had to fly straight and level toline up their bombsights. Elmer Sperrysinterests were warmly received, and inI925 and 1926 the Sperry Company builttwo data transmission systems for theArmys gun directors.

The original director built at Frankfordwas designated T- 1, or the Wilson Direc-tor. The Army had purchased a Vickersdirector manufactured in England, but en-couraged Wilson to design one that couldbe manufactured in this country [7].Sperrys two data transmission projects

April 1995

Fig. 1. Simplified system layout and data flow diagrum for Sperry T-6 anti-aircraft gundirector computer.

were to add automatic communicationsbetween the elements of both the Wilsonand the Vickers systems (Vickers wouldeventually incorporate the Sperry systeminto its product). Wilson died in 1927, andthe Sperry Company took over the entiredirector development from the FrankfordArsenal with a contract to build and de-liver a director incorporating the best fea-tures of both the Wilson and Vickerssystems.

From 1927 to 193.5, Sperry undertooka small but intensive development pro-gram in anti-aircraft systems. The com-pany financed its engineering internally,selling directors in small quantities to theArmy, mostly for evaluation, for only theactual cost of production [S]. Of the nearly10 models Sperry developed during thisperiod, it never sold more than 12 of anymodel; the average order was five. TheSperry Company offset some develop-ment costs by sales to foreign govem-ments, especially Russia, with the Armysapproval 191.

The T-6 DirectorSperrys modified version of Wilsons

director was designated T-4 in develop-ment. This model incorporated correc-tions for air density, super-elevation (theneed to aim a bit high to compensate forthe droop of the trajectory due to gravity),and wind. Assembled and tested at Frank-ford in the fall of 1928, it had problemswith backlash and reliability in its predict-ing mechanisms. Still, the Army found theT-4 promising and after testing returned itto Sperry for modification [lo]. The com-

pany changed the design for simplermanufacture, eliminated two operators,and improved reliability. In 1930 Sperryreturned with the T-6, which tested suc-cessfully. By the end of 193 1, the Armyhad ordered 12 of the units. The T-6 wasstandardized by the Army (i.e. accepted asoperational) as the M-2 director [I 11.

Since the T-6 was the first anti-air-craft director to be put into production, aswell as the first one the Army formallyprocured, it is instructive to examine itsoperation in detail. A technical memoran-dum dated 1930 explained the theory be-hind the T-6 calculations and how theequations were solved by the system. Al-though this publication lists no author, itprobably was written by Earl W. Chafee,Sperrys director of fire control engineer-ing [ 121. The director was a complex me-chanical analog computer that connectedfour three-inch anti-aircraft guns and analtitude finder into an integrated system(see Fig. 1). Just as with Sperrys navalfire control system, the primary means ofconnection were data transmitters, simi-lar to those that connected gyrocompassesto repeaters aboard ship.

The director takes three primary in-puts. Target altitude comes from a stereo-scopic range finder. This device has twotelescopes separated by a baseline of 12feet; a single operator adjusts the anglebetween them to bring the two images intocoincidence. Slant range, or the raw targetdistance, is then corrected to derive itsaltitude component. Two additional op-erators, each with a separate telescope,track the target, one for azimuth and one

109

for elevation (these telescopes are physi-cally mounted on the director). Each sight-ing device has a data transmitter thatmeasures angle or range and sends it to thecomputer. The computer receives thesedata and incorporates manual adjustmentsfor wind velocity, wind direction, muzzlevelocity, air density, and other factors.The computer calculates three variables:azimuth, elevation, and a setting for thefuze. The latter, manually set before load-ing, determines the time after firing atwhich the shell will explode (correspond-ing to slant range of the predicted positionof the target). Shells are not intended to hitthe target plane directly but rather to ex-plode near it, scattering fragments to de-stroy it.

The director performs two major cal-culations. First,pvediction models the mo-tion of the target and extrapolates itsposition to some time in the future, basedon an assumption of constant course,speed, and altitude. Prediction corre-sponds to leading the target. Second, theballistic calculation figures how to makethe shell arrive at the desired point in spaceat the future time and explode, solving forthe azimuth and elevation of the gun andthe setting on the fuze. This calculationcorresponds to the traditional artillerymans task of looking up data in a precal-culated firing table and setting gunparameters accordingly. Ballistic calcula-tion is simpler than prediction, so we willexamine it first.

The T-6 director solves the ballisticproblem by directly mechanizing the tra-ditional method, employing a mechani-cal firing table. Traditional firing tablesprinted on paper show solutions for agiven angular height of the target, for agiven horizontal range, and a number ofother variables. The T-6 replaces the firingtable with a Sperry ballistic cam. Athree-dimensionally machined cone-shaped device, the ballistic cam or pinfollower solves a pre-determined func-tion. Two independent variables are inputby the angular rotation of the cam and thelongitudinal position of a pin that rests ontop of the cam. As the pin moves up anddown the length of the cam, and as the camrotates, the height of the pin traces a func-tion of two variables: the solution to theballistics problem (or part of it). The T-6director incorporates eight ballistic cams,each solving for a different component ofthe computation including superelevation,time of flight, wind correction, muzzlevelocity. air density correction. Ballistic

cams represented, in essence, the storeddata of the mechanical computer. Laterdirectors could be adapted to differentguns simply by replacing the ballisticcams with a new set, machined accordingto different firing tables [ 131. The ballisticcams comprised a central component ofSperrys mechanical computing technol-ogy. The difficulty of their manufacturewould prove a major limitation on theusefulness of Sperry directors.

The T-6 director performed its othercomputational function, prediction, in aninnovative way as well. Though the targetcame into the system in polar coordinates(azimuth, elevation, and range), targetsusually flew a constant trajectory (it wasassumed) in rectangular coordinates-i.e.straight and level. Thus, it was simpler toextrapolate to the future in rectangularcoordinates than in the polar system. Sothe Sperry director projected the move-ment of the target onto a horizontal plane,derived the velocity from changes in po-sition, added a fixed time multiplied by thevelocity to determine a future position,and then converted the solution back intopolar coordinates. This method becameknown as the plan prediction methodbecause of the representation of the dataon a flat plan as viewed from above; itwas commonly used through World WarII. In the plan prediction method, theactual movement of the target is mechani-cally reproduced on a small scale withinthe Computer and the desired angles orspeeds can be measured directly from themovements of these elements. [ 141

Together, the ballistic and predictioncalculations form a feedback loop. Opera-tors enter an estimated time of flight forthe shell when they first begin tracking.The predictor uses this estimate to performits initial calculation, which feeds into theballistic stage. The output of the ballisticscalculation then feeds back an updatedtime-of-flight estimate, which the predic-tor uses to refine the initial estimate. Thusa cumulative cycle of correction and re-correction brings the predicted futureposition of the target up to the point indi-cated by the actual future time of flight.1151

A square box about four feet on eachside (see Fig. 2) the T-6 director wasmounted on a pedestal on which it couldrotate. Three crew would sit on seats andone or two would stand on a step mountedto the machine, revolving with the unit asthe azimuth tracker followed the target.The remainder of the crew stood on a fixed

platform; they would have had to shufflearound as the unit rotated. This was prob-ably not a problem, as the rotation angleswere small for any given engagement. Thedirectors pedestal mounted on a trailer,on which data transmission cables and therange finder could be packed for transpor-tation.

We have seen that the T-6 computertook only three inputs, elevation, azimuth,and altitude (range), and yet it requirednine operators. These nine did not includethe operation of the range finder, whichwas considered a separate instrument, orthe men tending the guns themselves, butonly those operating the director itself.What did these nine men do?

Human ServomechanismsTo the designers of the director, the

operators functioned as manual servo-mechanisms. One specification for themachine required minimum dependenceon human element. The Sperry Com-pany explained, All operations must bemade as mechanical and foolproof as pos-sible; training requirements must visual-ize the conditions existent under rapidmobilization; . .. The lessons of WorldWar I ring in this statement; even at theheight of isolationism, with the countrysliding into depression, design engineersunderstood the difficulty of raising largenumbers of trained personnel in a nationalemergency. The designers not onlythought the system should account forminimal training and high personnel turn-over, they also considered the ability ofoperators to perform their duties under thestress of battle. Thus, nearly all the workfor the crew was in a follow-the-pointermode: each man concentrated on an in-strument with two indicating dials, one theactual and one the desired value for aparticular parameter. With a hand crank,he adjusted the parameter to match the twodials.

Still, it seems curious that the T-6 di-rector required so many men to performthis follow-the-pointer input. When theexternal rangefinder transmitted its data tothe computer, it appeared on a dial and anoperator had to follow the pointer to actu-ally input the data into the computingmechanism. The machine did not explic-itly calculate velocities. Rather, two op-erators (one for X and one for Y) adjustedvariable-speed drives until their rate dialsmatched that of a constant-speed motor(the adjustment on the drive then equaledvelocity). When the prediction computa-

110 IEEE Control Systems

XFig. 2. The Sperry T-6 director: A. Spotting scope. B. North-south rate dial and handwheel.C. Future horizontal range dial. D. Super-elevation dial and handwheel. E. Azimuth trackingtelescope. F. Future horizontal range handwheel. G. Traversing handwheel (azimuthtracking). H. Fire control ojlicers platform. .I. Azimuth tracking operators seat. K. Timeofflight dial and handwheel. L. Present altitude dial and handwheel. M. Present horizontalrange dial and handwheel. N. Elevation tracking handwheel and operators seat. 0.Orienting clamp. (Courtesy Hagley Museum and Library)

tion was complete, an operator had to feedthe result into the ballistic calculationmechanism. Finally, when the entire cal-culation cycle was completed, another op-erator had to follow the pointer to transmitazimuth to the gun crew, who in turn hadto match the train and elevation of the gunto the pointer indications.

Fig. 3 shows the crew arrayed aroundthe T-6 director, in an arrangement thattoday seems almost comical. Strange asthese operations seem, they reveal Sperryengineers conception of what the humanrole in the operation of an automated sys-tem ought to be. The numerous follow-the-pointer operations were clearlypreferable to data transmission by tele-phone; in that sense the system was auto-mated. Operators literally supplied thefeedback that made the system work, al-though Sperrys idea of feedback was

April 1995

rather different from the one prevalenttoday:

In many cases where results are ob-tained by individual elements in the cycleof computation it is necessary to feed theseresults back into the mechanism or totransmit them.

The Sperry document acknowledgesthe possibility of doing these operationsautomatically, but does not find it the pref-erable option:

When mechanical methods are em-ployed, it is necessary to use some form ofservo-motor, and electrical servo-mo-tors are used to a limited degree for feed-ing back data into the computer.

It has been found in many cases to bemuch easier to rely on a group of opera-tors who fulfill no other function than toact as servo-motors... This operation canbe mechanically performed by the opera-

tor under rigorous active service condi-tions. [ 161

Human operators were the means ofconnecting individual elements into anintegrated system. In one sense the menwere impedance amplifiers, and hencequite similar to servomechanisms in othermechanical calculators of the time, espe-cially Vannevar Bushs differential ana-lyzer [17].

The term manual servomechanismitself is an oxymoron: by the conventionaldefinition, all servomechanisms are auto-matic. The very use of the term acknow-ledges the existence of an automatictechnology that will eventually replacethe manual method. With the T-6, thisprocess was already underway. Thoughthe director required nine operators, it hadalready eliminated two from the previousgeneration T-4. Servos replaced the op-erator who fed back superelevation dataand the one who transmitted the fuze set-ting. Furthermore, in this early machineone man corresponded to one variable,and the machines requirement for opera-tors corresponded directly to the data flowof its computation. Thus the crew thatoperated the T-6 director was an exactreflection of the algorithm inside it.

Why, then, were only two of the vari-ables automated? Where the Sperry litera-t u r e p r o u d l y t r u m p e t s h u m a nfollow-the-pointer operations, it barelyacknowledges the automatic servos, andeven then provides the option of manualfollow-ups if the electrical gear is notused. This partial, almost hesitating auto-mation indicates there was more to thehuman servo-motors than Sperry wantedto acknowledge. As much as the companytouted their duties are purely mechanicaland little skill or judgment is required onthe part of the operators, men were stillrequired to exercise some judgment, evenif unconsciously. The data were noisy,and even an unskilled human eye couldeliminate complications due to erroneousor corrupted data. Noisy data did morethan corrupt firing solutions. The mecha-nisms themselves were rather delicate anderroneous input data, especially if it indi-cated conditions that were not physicallypossible, could lock up or damage themechanisms [ 181. The operators per-formed as integrators in both senses of theterm: they integrated different elementsinto a system, and they integrated mathe-matically, acting as low-pass filters to re-duce noise.

I l l

Fig. 3. The Sperry T-6 Director mounted on a trailer with operators. Note power supply atleft and cables to other system elements. (Courtesy Hagley Museum and Library)

Later Sperry DirectorsWhen Elmer Sperry died in 1930, his

engineers were at work on a newer gen-eration director, the T-8. This machinewas intended to be lighter and more port-able than earlier models, as well as lessexpensive and procurable in quantities incase of emergency. [19] The companystill emphasized the need for unskilledmen to operate the system in wartime, andtheir role as system integrators. The opera-tors were mechanical links in the appara-tus, thereby making it possible to avoidmechanical complication which would beinvolved by the use of electrical or me-chanical servo motors. Still, army fieldexperience with the T-6 had shown thatservo-motors were a viable way to reducethe number of operators and improve reli-ability, so the requirements for the T-8specified that wherever possible electri-cal follow-up motors shall be used to re-duce the number of operators to aminimum. [20] Thus the T-8 continuedthe process of automating fire control, andreduced the number of operators to four.Two men followed the target with tele-scopes, and only two were required forfollow-the-pointer functions (for the tworate follow-ups). The other follow-the-pointers had been replaced by follow-upservos fitted with magnetic brakes toeliminate hunting (the inclusion of thesebrakes suggests that the hesitating use ofservos in earlier models may have beendue to concerns about their stability). Sev-eral experimental versions of the T-8 were

112

built, and it was standardized by the Armyas the M3 in 1934.

Throughout the remainder of the 30sSperry and the army fine-tuned the direc-tor system as embodied in the M3. Suc-ceeding M3 models automated further,replacing the follow-the-pointers for tar-get velocity with a velocity follow-upwhich employed a ball-and-disc integrator[2 11. The M4 series, standardized in 1939,was similar to the M3 but abandoned theconstant altitude assumption and added analtitude predictor for gliding targets. TheM7, standardized in 194 1, was essentiallysimilar to the M4 but added full powercontrol to the guns for automatic pointingin elevation and azimuth [22]. These latersystems had eliminated errors to the pointwhere the greatest uncertainty was thevarying time it took different crews tomanually set the fuze and load the shellinto the gun. Automatic setters and loadersdid not improve the situation because ofreliability problems. The M7 model alsoadded provision for entering azimuth ob-servation from radio locator equipment,prefiguring the addition of radar for targetobservations. At the start of World War II,the M7 was the primary anti-aircraft direc-tor available to the army.

Following 15 years of work atSperry, the M7 was a highly developedand integrated system, optimized for reli-ability and ease of operation and mainte-nance. As a mechanical computer, it wasan elegant, if intricate, device, weighing850 pounds and including about 11,000parts. The design of the M7 capitalized on

the strength of the Sperry Company:manufacturing of precision mechanisms,especially ballistic cams. By the time theU.S. entered the second world war, how-ever, these capabilities were a scarce re-source, especially for high volumes.Production of the M7 by Sperry and FordMotor Company as subcontractor was areal choke and could not keep up withproduction of the 90mm guns, well into1942 [23]. The army had also adopted anEnglish system, known as the KerrisonDirector or M5, which was less accuratethan the M7 but easier to manufacture.Sperry redesigned the M5 for high-vol-ume production in 1940, but passed onmanufacturing responsibility to the SingerSewing Machine and Delco companies in1941 [24]. By 1943, an electronic comput-ing director developed at Bell Labs wouldsupersede the M7, and the M7 ceasedproduction (the Western Electric/BellLabs gun director will be the subject ofanother article in this series).

Conclusion: Human Beings asSystem Integrators

The Sperry directors we have exam-ined here were transitional, experimentalsystems. Exactly for that reason, however,they allow us to peer inside the process ofautomation, to examine the displacementof human operators by servomechanismswhile the process was still underway.Skilled as the Sperry Company was at datatransmission, it only gradually becamecomfortable with the automatic communi-cation of data between subsystems. Sperrycould brag (perhaps protesting too much)about the low skill levels required of theoperators of the machine, but in 1930 itwas unwilling to remove them completelyfrom the process. Men were the glue thatheld integrated systems together.

As products, the Sperry Companysanti-aircraft gun directors were only par-tially successful. A decade and a half ofdevelopment produced machines thatcould not negotiate the fine line betweenperformance and production imposed bynational emergency. Still, we shouldjudge a technological development pro-gram not only by the machines it producesbut also by the knowledge it creates, andby how that knowledge contributes to fu-ture advances. Sperrys anti-aircraft direc-tors of the 1930s were early examples ofdistributed control systems, technologythat would assume critical importance inthe following decades with the develop-ment of radar and digital computers.

IEEE Control Systems

When building the more complex systemsof later years, engineers at Bell Labs, MIT,and elsewhere would incorporate andbuild on the Sperry Companys experi-ence, grappling with the engineering dif-ficulties of feedback, control, and theaugmentation of human capabilities bytechnological systems.

Notes and References(All documents referred to in Elmer SperryPapers and Sperry Company Papers are lo-cated in the archival collection of the HagleyMuseum and Library, Wilmington, DE.)[l] Another important element of controltechnology in this period was process con-trol. For a detailed exploration of controltechnologies in the 1930s see Stuart Ben-nett, A History of Control Engineering:1930-1955, London, 1993, IEE Press, Chap-ter 1.[2] John Testuro Sumida, In Defence of Na-val Supremacy: Finance, Technology, andBritish Naval Policy 1889-1914, London:Routledge 1989.[3] Elmer Sperry to T. Wilson, FrankfordArsenal, July 10, 1925. Elmer Sperry Pa-pers, Box 33.[4] Sperry Company memorandum, prob-ably Preston R. Bassett, Development ofFire Control for Major Calibre Anti-AircraftGun Battery, p. 2. Sperry Gyroscope Com-pany Records, Box 33.[5] Thomas P. Hughes, Elmer Sperry: Inven-tor and Engineer (Baltimore, 197 l), p. 233.[6] Elmer A. Sperry to T. Wilson, FrankfordArsenal, July 10, 1925. Elmer Sperry Pa-pers, Box 33.[7] United States Army, Ordnance Depart-ment, History of Anti-Aircraft Director De-velopment, no date, probably prepared inthe fall of 1935. Sperry Gyroscope CompanyRecords, Box 4.[8] Note 4 above.[9] Note 4 above. See also Sperry CompanyForm #1607, Sperry Universal Director:Information to be Furnished by Customer.

Sperry Gyroscope Company Records, Box

essentially a specification for the T-8.

3. A document clearly intended for foreigngovernments allowing Sperry to customizetheir directors to different types of guns.[lo] Note 7 above, pp. 12-14.[l l] Note 7 above, pp. 9-16.[12] Sperry Gyroscope Company, Anti-Aircraft Gun Control, Publication No. 20-1640, Brooklyn, New York: SperryGyroscope Company Inc., 1930. Sperry Gy-roscope Company Papers. This documentdoes not list an author, but its language andexplanations are quite similar to those in anarticle published by Chafee, A Miss is asGood as a Mile, in Sperryscope, the officialSperry Company organ, in April 1932.[ 131 Robert Lea, The Ballistic Cam in DeanHollisters Lamp, Sperry Company Papers.[14] Note 12 above, p. 21.[ 151 Note 12 above, p. 32.[ 161 Note 12 above, pp. 24-25.[17] Henry M. Paynter, The DifferentialAnalyzer as an Active Mechanical Instm-ment, Keynote speech to the 1989 Ameri-can Control Conference, IEEE ControlSystems Magazine, December 1989, pp.3-7.[ 181 Anti-Aircraft Defense, Harrisburg, PA,1940, reprints the manuals for the SperryM-2 director and discusses the mechanicalproblems that can be caused in this genera-tion of Sperry directors by contradictory in-put data.[19] Note 12 above, p. 18[20] Universal Director and Data Trans-

mission System, Sperry Company Publica-tion no. 14-8051, Aug. 1, 1932, p. 6. SperryCompany Records, Box 2. This document is

[2 l] For an explanation of the Sperry veloc-ity servo, see Allan G. Bromley, AnalogComputing Devices in William Aspray,ed., Computing Before Computers, Ames,Iowa, 1990, p. 190.[22] Automating the pointing of the gun wasa more difficult problem than data transmis-sion because it involved significant poweramplification. The Sperry Company hadpurchased the rights to the Nieman torqueamplifier system from Bethlehem Steel in1926. During the next several years, Sperryapplied power controls to a number of indi-vidual guns. None was incorporated intoactual systems, due to Army concerns aboutreliability, and perhaps also to problems ofstability. No further work was done onpower controls for guns for almost 10 years.It was not until 1939 that Sperry began acontract to develop an electro-hydraulic re-mote control system for the Armys new90mm A.A. gun. This work was well under-way when reports from the British in combatwith Germany indicated that automaticpower controls for gun pointing was notonly desirable but was absolutely neces-sary. Sperry Company Report, PowerControls, Feb. 7, 1944. Sperry CompanyRecords, Box 40.[23] Harry C. Thompson and Lida Mayo,The United States Army in World War II:

pp. 186-191

The Ordnance Department, Volume 2: Pro-curement and Supply, Washington, DC,1960, p. 86.[24] Sperry Company memo on M-5 andM-6 directors. Sperry Company Records,Box 33. Also see Bromley (note 21 above)

David A. Mindell is an electrical engineer and a doctoral student in theHistory of Technology at the Massachusetts Institute of Technology. He is writinga dissertation on the history of control systems from 19 16- 1945. Before comingto MIT he was a staff engineer at the Deep Submergence Laboratory of the WoodsHole Oceanographic Institution, where he developed the control system forJASON, a remotely operated vehicle for deep-ocean exploration.

April 1995 113