Anti- Fire Control and the Development of Integrated Svstems at Sperry, 19251940 d David A. Mindell

he dawn of the electrical age brought riod. By the end of the decade these prob- searchlights, and sound location equip- T new types of control systems. Able to lems would become critical as the country ment. The Sperry Company was involved transmit data between distributed compo- struggled to build up its technology to in the latter two efforts. About this time nents and effect action at a distance, these meet the demands of an impending war. Maj. Thomas Wilson, at the Frankford systems employed feedback devices as Arsenal in Philadelphia, began develop- well as human beings to close control Anti-Aircraft Artillery Fire ing a central for fire control data, loops at every level. By the time theories Control loosely based on the system of “director of feedback and stability began to become Before , developments in firing” that had developed in naval gun- practical for engineers in the 1930s a tra- ship design, guns, and armor drove the nery. Wilson’s device resembled earlier dition of remote and automatic control need for improved fire control on Navy fire control calculators, accepting data as had developed that built dis- ships [2]. By 1920, similar forces were at input from sensing components, perform- tributed control systems with centralized work in the air: wartime experiences and ing calculations to predict the future loca- information processors [I]. These two postwar developments in aerial bombing tion of the target, and producing direction strands of technology, and created the need for sophisticated fire con- information to the guns. control systems, came together to produce trol for anti-aircraft artillery. Shooting an the large-scale integrated systems typical airplane out of the sky is essentially a of World War II and after. problem of “leading” the target. As air- Integration and Data Elmer Ambrose Sperry (I 860-1930) craft developed rapidly in the twenties, Transmission and the company he founded, the Sperry their increased speed and altitude rapidly Still, the components of an anti-aircraft Gyroscope Company, led the engineering pushed the task of the lead out battery remained independent, tied to- of control systems between 1910 and of the range of human reaction and calcu- gether only by telephone. As Preston R. 1940. Sperry and his engineers built dis- lation. Fire control equipment for anti-air- Bassett, chief engineer and later president tributed data transmission systems that craft guns was a means of technologically of the Sperry Company, recalled, “no laid the foundations of today’s command aiding human operators to accomplish a sooner, however, did the components get and control systems. Sperry’s fire control task beyond their natural capabilities to the point of functioning satisfactorily systems included more than governors or During the first world war, anti-aircraft within themselves, than the problem of stabilizers; they consisted of distributed fire control had undergone some prelimi- properly transmitting the information sensors, data transmitters, central proces- nary development. Elmer Sperry, as chair- from one to the other came to be of prime sors, and outputs that drove machinery. man of the Aviation Committee of the importance.” [4] Tactical and terrain con- This article tells the story of Sperry’s Naval Consulting Board, developed two siderations often required that different involvement in anti-aircraft fire control instruments for this problem: a goniome- fire control elements be separated by up to between the world wars and shows how ter, a range-finder, and a pretelemeter, a several hundred feet. Observers tele- an industrial firm conceived of control fire director or calculator. Neither, how- phoned their data to an officer, who manu- systems before the common use of control ever, was widely used in the field [3]. ally entered it into the central computer, theory. In the 1930s the task of fire control When the war ended in I 9 18 the Army read off the results, and telephoned them became progressively more automated, as undertook virtually no new development to the gun installations. This communica- Sperry engineers gradually replaced hu- in anti-aircraft fire control for five to seven tion system introduced both a time delay man operators with automatic devices. years. In the mid-1920s however, the and the opportunity for error. The compo- Feedback, human interface, and system Army began to develop individual compo- nents needed tighter integration, and such integration posed challenging problems nents for anti-aircraft equipment includ- a system required automatic data commu- for fire control engineers during this pe- ing stereoscopic height-finders, nications.

108 IEEE Control Systems In the 1920s the Sperry Gyroscope Company led the field in data communi- cations. Its experience came from Elmer Sperry’s most successful invention, a true- north-seeking gyro for ships. A significant feature of the Sperry Gyrocompass was its ability to transmit heading data from a single central gyro to repeaters located at a number of locations around the ship. The repeaters, essentially follow-up servos, connected to another follow-up, which tracked the motion of the gyro without interference. These data transmitters had attracted the interest of the Navy, which needed a stable heading reference and a system of data communication for its own fire control problems. In 1916, Sperry built a fire control system for the Navy which, although it placed minimal empha- Fig. 1. Simplified system layout and data flow diagrum for Sperry T-6 anti-aircraft gun sis on automatic computing, was a sophis- director computer. ticated distributed data system. By 1920 Sperry had installed these systems on a were to add automatic communications pany changed the design for simpler number of US. battleships [5]. between the elements of both the Wilson manufacture, eliminated two operators, Because of the Sperry Company’s ex- and the Vickers systems (Vickers would and improved reliability. In 1930 Sperry perience with fire control in the Navy, as eventually incorporate the Sperry system returned with the T-6, which tested suc- well as Elmer Sperry’s earlier work with into its product). Wilson died in 1927, and cessfully. By the end of 193 1, the Army the goniometer and the pretelemeter, the the Sperry Company took over the entire had ordered 12 of the units. The T-6 was Army approached the company for help director development from the Frankford standardized by the Army (i.e. accepted as with data transmission for anti-aircraft fire Arsenal with a contract to build and de- operational) as the M-2 director [I 11. control. To Elmer Sperry, it looked like an liver a director incorporating the best fea- Since the T-6 was the first anti-air- easy problem: the calculations resembled tures of both the Wilson and Vickers craft director to be put into production, as those in a naval application, but the physi- systems. well as the first one the Army formally cal platform, unlike a ship at sea, anchored From 1927 to 193.5, Sperry undertook procured, it is instructive to examine its to the ground. Sperry engineers visited a small but intensive development pro- operation in detail. A technical memoran- Wilson at the Frankford Arsenal in 1925, gram in anti-aircraft systems. The com- and Elmer Sperry followed up with a letter dum dated 1930 explained the theory be- pany financed its engineering internally, expressing his interest in working on the hind the T-6 calculations and how the selling directors in small quantities to the equations were solved by the system. Al- problem. He stressed his company’s expe- Army, mostly for evaluation, for only the rience with naval problems, as well as its though this publication lists no author, it actual cost of production [S]. Of the nearly recent developments in , probably was written by Earl W. Chafee, 10 models Sperry developed during this “work from the other end of the proposi- Sperry’s director of fire control engineer- period, it never sold more than 12 of any tion.” Bombsights had to incorporate nu- ing [ 121. The director was a complex me- model; the average order was five. The merous parameters of wind, groundspeed, chanical analog computer that connected Sperry Company offset some develop- airspeed, and ballistics, so an anti-aircraft four three-inch anti-aircraft guns and an ment costs by sales to foreign govem- gun director was in some ways a recipro- altitude finder into an integrated system ments, especially Russia, with the Army’s cal [6]. In fact, part of the rea- (see Fig. 1). Just as with Sperry’s naval approval 191. son anti-aircraft fire control equipment fire control system, the primary means of worked at all was that it assumed attacking connection were “data transmitters,” simi- had to fly straight and level to The T-6 Director lar to those that connected gyrocompasses line up their bombsights. Elmer Sperry’s Sperry’s modified version of Wilson’s to repeaters aboard ship. interests were warmly received, and in director was designated T-4 in develop- The director takes three primary in- I925 and 1926 the Sperry Company built ment. This model incorporated correc- puts. Target altitude comes from a stereo- two data transmission systems for the tions for air density, super-elevation (the scopic range finder. This device has two Army’s gun directors. need to aim a bit high to compensate for telescopes separated by a baseline of 12 The original director built at Frankford the droop of the trajectory due to gravity), feet; a single operator adjusts the angle was designated T- 1, or the “Wilson Direc- and wind. Assembled and tested at Frank- between them to bring the two images into tor.” The Army had purchased a Vickers ford in the fall of 1928, it had problems coincidence. Slant range, or the raw target director manufactured in England, but en- with backlash and reliability in its predict- distance, is then corrected to derive its couraged Wilson to design one that could ing mechanisms. Still, the Army found the altitude component. Two additional op- be manufactured in this country [7]. T-4 promising and after testing returned it erators, each with a separate telescope, Sperry’s two data transmission projects to Sperry for modification [lo]. The com- track the target, one for azimuth and one

April 1995 109 for elevation (these telescopes are physi- cams represented, in essence, the stored platform; they would have had to shuffle cally mounted on the director). Each sight- data of the . Later around as the unit rotated. This was prob- ing device has a data transmitter that directors could be adapted to different ably not a problem, as the rotation angles measures angle or range and sends it to the guns simply by replacing the ballistic were small for any given engagement. The computer. The computer receives these cams with a new set, machined according director’s pedestal mounted on a trailer, data and incorporates manual adjustments to different firing tables [ 131. The ballistic on which data transmission cables and the for wind velocity, wind direction, muzzle cams comprised a central component of range finder could be packed for transpor- velocity, air density, and other factors. Sperry’s mechanical computing technol- tation. The computer calculates three variables: ogy. The difficulty of their manufacture We have seen that the T-6 computer azimuth, elevation, and a setting for the would prove a major limitation on the took only three inputs, elevation, azimuth, fuze. The latter, manually set before load- usefulness of Sperry directors. and altitude (range), and yet it required ing, determines the time after firing at The T-6 director performed its other nine operators. These nine did not include which the shell will explode (correspond- computational , prediction, in an the operation of the range finder, which ing to slant range of the predicted position innovative way as well. Though the target was considered a separate instrument, or of the target). Shells are not intended to hit came into the system in polar coordinates the men tending the guns themselves, but the target plane directly but rather to ex- (azimuth, elevation, and range), targets only those operating the director itself. plode near it, scattering fragments to de- usually flew a constant trajectory (it was What did these nine men do? stroy it. assumed) in rectangular coordinates-i.e. The director performs two major cal- straight and level. Thus, it was simpler to Human Servomechanisms culations. First,pvediction models the mo- extrapolate to the future in rectangular To the designers of the director, the tion of the target and extrapolates its coordinates than in the polar system. So operators functioned as “manual servo- position to some time in the future, based the Sperry director projected the move- mechanisms.” One specification for the on an assumption of constant course, ment of the target onto a horizontal plane, required “minimum dependence speed, and altitude. Prediction corre- derived the velocity from changes in po- on ‘human element.“’ The Sperry Com- sponds to “leading” the target. Second, the sition, added a fixed time multiplied by the pany explained, “All operations must be ballistic calculation figures how to make velocity to determine a future position, made as mechanical and foolproof as pos- the shell arrive at the desired point in space and then converted the solution back into sible; training requirements must visual- at the future time and explode, solving for polar coordinates. This method became ize the conditions existent under rapid the azimuth and elevation of the gun and known as the “plan prediction method” mobilization; . ..” The lessons of World the setting on the fuze. This calculation because of the representation of the data War I ring in this statement; even at the corresponds to the traditional artillery on a flat “plan” as viewed from above; it height of isolationism, with the country man’s task of looking up data in a precal- was commonly used through World War sliding into depression, design engineers culated “firing table” and setting gun II. In the plan prediction method, “the understood the difficulty of raising large parameters accordingly. Ballistic calcula- actual movement of the target is mechani- numbers of trained personnel in a national tion is simpler than prediction, so we will cally reproduced on a small scale within emergency. The designers not only examine it first. the Computer and the desired angles or thought the system should account for The T-6 director solves the ballistic speeds can be measured directly from the minimal training and high personnel turn- problem by directly mechanizing the tra- movements of these elements.” [ 141 over, they also considered the ability of ditional method, employing a “mechani- Together, the ballistic and prediction operators to perform their duties under the cal firing table.” Traditional firing tables calculations form a feedback loop. Opera- stress of battle. Thus, nearly all the work printed on paper show solutions for a tors enter an estimated “time of flight” for for the crew was in a “follow-the-pointer” given angular height of the target, for a the shell when they first begin tracking. mode: each man concentrated on an in- given horizontal range, and a number of The predictor uses this estimate to perform strument with two indicating dials, one the other variables. The T-6 replaces the firing its initial calculation, which feeds into the actual and one the desired value for a table with a “Sperry ballistic cam.” A ballistic stage. The output of the ballistics particular parameter. With a hand crank, three-dimensionally machined cone- calculation then feeds back an updated he adjusted the parameter to match the two shaped device, the ballistic cam or “pin time-of-flight estimate, which the predic- dials. follower” solves a pre-determined func- tor uses to refine the initial estimate. Thus Still, it seems curious that the T-6 di- tion. Two independent variables are input “a cumulative cycle of correction and re- rector required so many men to perform by the angular rotation of the cam and the correction brings the predicted future this follow-the-pointer input. When the longitudinal position of a pin that rests on position of the target up to the point indi- external rangefinder transmitted its data to top of the cam. As the pin moves up and cated by the actual future time of flight.” the computer, it appeared on a dial and an down the length of the cam, and as the cam 1151 operator had to follow the pointer to actu- rotates, the height of the pin traces a func- A square box about four feet on each ally input the data into the computing tion of two variables: the solution to the side (see Fig. 2) the T-6 director was mechanism. The machine did not explic- ballistics problem (or part of it). The T-6 mounted on a pedestal on which it could itly calculate velocities. Rather, two op- director incorporates eight ballistic cams, rotate. Three crew would sit on seats and erators (one for X and one for Y) adjusted each solving for a different component of one or two would stand on a step mounted variable-speed drives until their rate dials the computation including superelevation, to the machine, revolving with the unit as matched that of a constant-speed motor time of flight, wind correction, muzzle the azimuth tracker followed the target. (the adjustment on the drive then equaled velocity. air density correction. Ballistic The remainder of the crew stood on a fixed velocity). When the prediction computa-

110 IEEE Control Systems tor under rigorous active service condi-

X tions. ” [ 161 Human operators were the means of connecting “individual elements” into an integrated system. In one sense the men were impedance , and hence quite similar to servomechanisms in other mechanical calculators of the time, espe- cially ’s differential ana- lyzer [17]. The term “manual servomechanism” itself is an oxymoron: by the conventional definition, all servomechanisms are auto- matic. The very use of the term acknow- ledges the existence of an automatic technology that will eventually replace the manual method. With the T-6, this process was already underway. Though the director required nine operators, it had already eliminated two from the previous generation T-4. Servos replaced the op- erator who fed back superelevation data and the one who transmitted the fuze set- ting. Furthermore, in this early machine one man corresponded to one variable, and the machine’s requirement for opera- tors corresponded directly to the data flow of its computation. Thus the crew that operated the T-6 director was an exact reflection of the inside it. Fig. 2. The Sperry T-6 director: A. Spotting scope. B. North-south rate dial and handwheel. Why, then, were only two of the vari- C. Future horizontal range dial. D. Super-elevation dial and handwheel. E. Azimuth tracking ables automated? Where the Sperry litera- telescope. F. Future horizontal range handwheel. G. Traversing handwheel (azimuth tracking). H. Fire control ojlicer’s platform. .I. Azimuth tracking operator’s seat. K. Time ture proudly trumpets human offlight dial and handwheel. L. Present altitude dial and handwheel. M. Present horizontal follow-the-pointer operations, it barely range dial and handwheel. N. Elevation tracking handwheel and operator’s seat. 0. acknowledges the automatic servos, and Orienting clamp. (Courtesy Hagley Museum and Library) even then provides the option of manual follow-ups “if the electrical is not used.” This partial, almost hesitating auto- mation indicates there was more to the tion was complete, an operator had to feed rather different from the one prevalent human servo-motors than Sperry wanted the result into the ballistic calculation today: to acknowledge. As much as the company mechanism. Finally, when the entire cal- “In many cases where results are ob- touted “their duties are purely mechanical culation cycle was completed, another op- tained by individual elements in the cycle and little skill or judgment is required on erator had to follow the pointer to transmit of computation it is necessary to feed these the part of the operators,” men were still results back into the mechanism or to azimuth to the gun crew, who in turn had required to exercise some judgment, even transmit them. ” to match the train and elevation of the gun if unconsciously. The data were noisy, to the pointer indications. The Sperry document acknowledges and even an unskilled human eye could Fig. 3 shows the crew arrayed around the possibility of doing these operations eliminate complications due to erroneous the T-6 director, in an arrangement that automatically, but does not find it the pref- or corrupted data. Noisy data did more today seems almost comical. Strange as erable option: than corrupt firing solutions. The mecha- these operations seem, they reveal Sperry “When mechanical methods are em- nisms themselves were rather delicate and engineers’ conception of what the human ployed, it is necessary to use some form of erroneous input data, especially if it indi- role in the operation of an automated sys- ‘servo-motor, ’ and electrical servo-mo- cated conditions that were not physically tem ought to be. The numerous follow- tors are used to a limited degree for ‘feed- possible, could lock up or damage the the-pointer operations were clearly ing back’ data into the computer. mechanisms [ 181. The operators per- preferable to data transmission by tele- It has been found in many cases to be formed as integrators in both senses of the phone; in that sense the system was auto- much easier to rely on a group of opera- term: they integrated different elements mated. Operators literally supplied the tors who fulfill no other function than to into a system, and they integrated mathe- feedback that made the system work, al- act as servo-motors... This operation can matically, acting as low-pass filters to re- though Sperry’s idea of feedback was be mechanically performed by the opera- duce .

April 1995 Ill the strength of the Sperry Company: manufacturing of precision mechanisms, especially ballistic cams. By the time the U.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 Ford Motor Company as subcontractor was a “real choke” and could not keep up with production of the 90mm guns, well into 1942 [23]. The army had also adopted an English system, known as the “Kerrison Director” or M5, which was less accurate than the M7 but easier to manufacture. Sperry redesigned the M5 for high-vol- ume production in 1940, but passed on manufacturing responsibility to the Singer Sewing Machine and Delco companies in 1941 [24]. By 1943, an electronic comput- ing director developed at Bell Labs would Fig. 3. The Sperry T-6 Director mounted on a trailer with operators. Note power supply at supersede the M7, and the M7 ceased left and cables to other system elements. (Courtesy Hagley Museum and Library) production (the Western Electric/Bell Labs gun director will be the subject of Later Sperry Directors built, and it was standardized by the Army another article in this series). When Elmer Sperry died in 1930, his as the M3 in 1934. engineers were at work on a newer gen- Throughout the remainder of the ‘30s Conclusion: Human Beings as eration director, the T-8. This machine Sperry and the army fine-tuned the direc- System Integrators was intended to be lighter and more port- tor system as embodied in the M3. Suc- The Sperry directors we have exam- able than earlier models, as well as less ceeding M3 models automated further, ined here were transitional, experimental expensive and “procurable in quantities in replacing the follow-the-pointers for tar- systems. Exactly for that reason, however, case of emergency.” [19] The company get velocity with a velocity follow-up they allow us to peer inside the process of still emphasized the need for unskilled which employed a ball-and-disc integrator automation, to examine the displacement men to operate the system in wartime, and [2 11. The M4 series, standardized in 1939, of human operators by servomechanisms their role as system integrators. The opera- was similar to the M3 but abandoned the while the process was still underway. tors were “mechanical links in the appara- constant altitude assumption and added an Skilled as the Sperry Company was at data tus, thereby making it possible to avoid altitude predictor for gliding targets. The transmission, it only gradually became M7, standardized in 194 1, was essentially mechanical complication which would be comfortable with the automatic communi- similar to the M4 but added full power involved by the use of electrical or me- cation of data between subsystems. Sperry control to the guns for automatic pointing chanical servo motors.” Still, army field could brag (perhaps protesting too much) in elevation and azimuth [22]. These later experience with the T-6 had shown that about the low skill levels required of the systems had eliminated errors to the point servo-motors were a viable way to reduce operators of the machine, but in 1930 it where the greatest uncertainty was the the number of operators and improve reli- was unwilling to remove them completely varying time it took different crews to ability, so the requirements for the T-8 from the process. Men were the glue that manually set the fuze and load the shell held integrated systems together. specified that wherever possible “electri- into the gun. Automatic setters and loaders As products, the Sperry Company’s cal follow-up motors shall be used to re- did not improve the situation because of anti-aircraft gun directors were only par- duce the number of operators to a reliability problems. The M7 model also tially successful. A decade and a half of minimum.” [20] Thus the T-8 continued added provision for entering azimuth ob- development produced that the process of automating fire control, and servation from radio locator equipment, could not negotiate the fine line between reduced the number of operators to four. prefiguring the of radar for target performance and production imposed by Two men followed the target with tele- observations. At the start of World War II, national emergency. Still, we should scopes, and only two were required for the M7 was the primary anti-aircraft direc- judge a technological development pro- follow-the-pointer functions (for the two tor available to the army. gram not only by the machines it produces rate follow-ups). The other follow-the- Following 15 years of work at but also by the knowledge it creates, and pointers had been replaced by follow-up Sperry, the M7 was a highly developed by how that knowledge contributes to fu- servos fitted with magnetic brakes to and integrated system, optimized for reli- ture advances. Sperry’s anti-aircraft direc- eliminate hunting (the inclusion of these ability and ease of operation and mainte- tors of the 1930s were early examples of brakes suggests that the hesitating use of nance. As a mechanical computer, it was distributed control systems, technology servos in earlier models may have been an elegant, if intricate, device, weighing that would assume critical importance in due to concerns about their stability). Sev- 850 pounds and including about 11,000 the following decades with the develop- eral experimental versions of the T-8 were parts. The design of the M7 capitalized on ment of radar and digital .

112 IEEE Control Systems When building the more complex systems Sperry Gyroscope Company Records, Box [2 l] For an explanation of the Sperry veloc- of later years, engineers at Bell Labs, MIT, 3. A document clearly intended for foreign ity servo, see Allan G. Bromley, “Analog Computing Devices” in William Aspray, and elsewhere would incorporate and governments allowing Sperry to customize their directors to different types of guns. ed., Computing Before Computers, Ames, build on the Sperry Company’s experi- [lo] Note 7 above, pp. 12-14. Iowa, 1990, p. 190. ence, grappling with the engineering dif- [l l] Note 7 above, pp. 9-16. [22] Automating the pointing of the gun was ficulties of feedback, control, and the [12] Sperry Gyroscope Company, “Anti- a more difficult problem than data transmis- augmentation of human capabilities by Aircraft Gun Control,” Publication No. 20- sion because it involved significant power technological systems. 1640, Brooklyn, New York: Sperry amplification. The Sperry Company had Gyroscope Company Inc., 1930. Sperry Gy- purchased the rights to the Nieman torque Notes and References roscope Company Papers. This document system from Bethlehem Steel in (All documents referred to in Elmer Sperry does not list an author, but its language and 1926. During the next several years, Sperry Papers and Sperry Company Papers are lo- explanations are quite similar to those in an applied power controls to a number of indi- cated in the archival collection of the Hagley article published by Chafee, “A Miss is as vidual guns. None was incorporated into Museum and Library, Wilmington, DE.) Good as a Mile,” in Sperryscope, the official actual systems, due to Army concerns about [l] Another important element of control Sperry Company organ, in April 1932. reliability, and perhaps also to problems of technology in this period was process con- [ 131 Robert Lea, “The Ballistic Cam in Dean stability. No further work was done on trol. For a detailed exploration of control Hollister’s Lamp,” Sperry Company Papers. power controls for guns for almost 10 years. technologies in the 1930s see Stuart Ben- [14] Note 12 above, p. 21. It was not until 1939 that Sperry began a nett, A History of Control Engineering: [ 151 Note 12 above, p. 32. contract to develop an electro-hydraulic re- 1930-1955, London, 1993, IEE Press, Chap- [ 161 Note 12 above, pp. 24-25. mote control system for the Army’s new ter 1. [17] Henry M. Paynter, “The Differential 90mm A.A. gun. This work was well under- [2] John Testuro Sumida, In Defence of Na- Analyzer as an Active Mechanical Instm- way when reports from the British in combat val Supremacy: Finance, Technology, and ment,” Keynote speech to the 1989 Ameri- with Germany indicated that automatic British Naval Policy 1889-1914, London: can Control Conference, IEEE Control power controls for gun pointing “was not Routledge 1989. Systems Magazine, December 1989, pp.3-7. only desirable but was absolutely neces- [3] Elmer Sperry to T. Wilson, Frankford [ 181 Anti-Aircraft Defense, Harrisburg, PA, sary.” Sperry Company Report, “Power Arsenal, July 10, 1925. Elmer Sperry Pa- 1940, reprints the manuals for the Sperry Controls,” Feb. 7, 1944. Sperry Company pers, Box 33. M-2 director and discusses the mechanical Records, Box 40. [4] Sperry Company memorandum, prob- problems that can be caused in this genera- [23] Harry C. Thompson and Lida Mayo, ably Preston R. Bassett, “Development of tion of Sperry directors by contradictory in- The United States Army in World War II: Fire Control for Major Calibre Anti-Aircraft put data. The Ordnance Department, Volume 2: Pro- Gun Battery,” p. 2. Sperry Gyroscope Com- [19] Note 12 above, p. 18 curement and Supply, Washington, DC, pany Records, Box 33. [20] “Universal Director and Data Trans- 1960, p. 86. [5] Thomas P. Hughes, Elmer Sperry: Inven- mission System,” Sperry Company Publica- [24] Sperry Company memo on M-5 and tor and Engineer (Baltimore, 197 l), p. 233. tion no. 14-8051, Aug. 1, 1932, p. 6. Sperry M-6 directors. Sperry Company Records, [6] Elmer A. Sperry to T. Wilson, Frankford Company Records, Box 2. This document is Box 33. Also see Bromley (note 21 above) Arsenal, July 10, 1925. Elmer Sperry Pa- essentially a specification for the T-8. pp. 186-191 pers, Box 33. [7] United States Army, Ordnance Depart- ment, “History of Anti-Aircraft Director De- velopment,” no date, probably prepared in David A. Mindell is an electrical engineer and a doctoral student in the the fall of 1935. Sperry Gyroscope Company at the Massachusetts Institute of Technology. He is writing Records, Box 4. a dissertation on the history of control systems from 19 16- 1945. Before coming [8] Note 4 above. to MIT he was a staff engineer at the Deep Submergence Laboratory of the Woods [9] Note 4 above. See also Sperry Company Hole Oceanographic Institution, where he developed the control system for Form #1607, “Sperry Universal Director: JASON, a remotely operated vehicle for deep-ocean exploration. Information to be Furnished by Customer.”

April 1995 113