Mechanische Analog-Computer für schwere Flab-Kanonen, ca. 1930–1945: Technische Studien ab historischen Quellen — Funktionsweise des mechanischen Analog-Rechners SPERRY (USA)

Complete English translation of the original German-language document (26 pages).

Mechanical Analog Computers for Heavy Anti-Aircraft Artillery

ca. 1930–1945 — Technical Studies from Historical Sources

André Masson, Langenthal

[Caption: Acquired by Switzerland from 1937 onward and built under license by HASLER: the Hungarian fire-control computer GAMMA-JUHASZ. On the left, the two telescopes for tracking the aircraft.]

[Caption: 12 of the 14-man crew operate the trajectory computer: the German Kommandogerät 36 with range-finder mounted on top. In the center of the image is the round “flight-direction table.”]

[Caption: In 1935 the Swiss Army purchased the American SPERRY computer on a trial basis — six years before the world’s first functional computer (Zuse Z3).]

Archeo-Informatics — Military Gear-Drive Computers (ca. 1930 to 1945):

Operating Principle of the Mechanical Analog Computer SPERRY (USA)

Large firing distances — and correspondingly long shell flight times — in heavy anti-aircraft defense make computational assistance indispensable. Optical sighting of the aircraft provides data to the computer, which uses that data to determine the point at which the aircraft and the shells will meet. The computer continuously supplies three changing quantities to the guns: azimuth angle, elevation angle, and fuze setting (i.e., the time delay of the projectile’s time fuze). In the 1930s, the calculations in most of these “fire-control computers” (Kommandogeräte) were only capable of handling straight and level flight.

In 1935, Switzerland acquired two computers on a trial basis from the American firm SPERRY (information from B. Benz, Museum Dübendorf). One unit has been lost; the second is today on display at the Flieger-Flab Museum in Dübendorf. Experience with this device and the associated 7.5 cm VICKERS guns apparently did not prove satisfactory. Trials were therefore conducted with the Hungarian GAMMA-JUHASZ device, which was purchased in quantity, produced under license by HASLER, and further developed. From 1944 onward (Model 1943), the HASLER device was fitted with new extrapolation capabilities for turning-flight. The associated guns were 7.5 cm SCHNEIDER-CREUSOT cannons.

Photographs of the surviving SPERRY device at the Flieger-Flab Museum, Dübendorf:

[Caption: Left side: The large, round display shows the flight altitude (supplied by the range-finder). To its left is the display of the aircraft’s elevation angle, with coarse and fine scales. Under the lid with the small chain: switch from aerial to ground or naval targets. The computer stands on a large tripod.]

[Caption: Right side: Display for velocities in the North/South and East/West directions. Lower left: handwheel for lateral slewing of the device. Upper left, under covers: parallax adjustments (the difference between the position of the guns and the position of the computer).]

[Caption: Rear side: Here various manual corrections are applied to improve the accuracy of the shot. At the very top, mounted on a rotating axle, were originally three telescopes for tracking the target (the telescopes are no longer present today).]

Written documentation (technical manuals, operating instructions) for the SPERRY device has unfortunately not survived at the Flieger-Flab Museum. However, the book “Flak-Kommandogeräte” (Chief Engineer Alfred Kuhlenkamp, 1943, VDI — Verein Deutscher Ingenieure) (cf. Ref. 1) contains a schematic of the computing gear trains and more detailed information on the mathematical operating principle of the mechanical SPERRY computer. Alfred Kuhlenkamp (1901–1973) worked from 1931 until the end of the war in the Army Ordnance Office on the development of anti-aircraft fire-control computers and anti-aircraft sighting devices. The photographs in his book show that the SPERRY type known to him is fortunately almost identical to the surviving specimen at the Flieger-Flab Museum Dübendorf. A later SPERRY type (described in Ref. 1, p. 144 as “modern”) looks quite different in its construction; both its operating method and its computing approach were substantially redesigned.

The same book by A. Kuhlenkamp also contains more detailed information on the German Kommandogerät 36, but not a single word about the already-existing KdoGt 40 (of which 746 units had been produced by 1942!). It also explains the British VICKERS device. Only summary information is given for the Hungarian GAMMA-JUHASZ, the French AUFIÈRE and OPL, and for the

Czech KODA (which the Russians copied as the PUAZO) and the Dutch HAZEMEYER. These were in general devices captured and examined during the war. The devices described in detail — SPERRY, KdoGt 36, and VICKERS — differ considerably in their mathematical approach and in their practical construction. At the Flieger-Flab Museum Dübendorf there are also numerous documents relating to the GAMMA-JUHASZ (later production by HASLER).

The German Kommandogerät 40 is described, including its functional diagram, in Ref. 3, pp. 60–63.

A human lifetime before our own time, computer architecture still looks very different from today!

SPERRY Computer: A Look Inside

What is the optimal operating system? How many gigabytes, how many MHz? These questions were not yet being asked in 1935.

What one sees is nothing but gear wheels, levers, gear ratios, transmissions, springs, and even a taut thread.

There is as yet no memory. For every new task, a new computer must be built.

To understand how mechanical computation works, it is essential first to know the principle of the “follow pointer” (Folgezeiger):

Operating the Computer — Almost Exclusively via the Follow Pointer

Early computers required up to 12 men, who had to continuously set current values (Kommando-Hilfsgerät 35, Kommandogerät 36). Later, the personnel requirement was reduced. All personnel at the computer had to continuously update “their” variable (distance, angle, etc.): using a handwheel, a rotating indicator needle had to be kept in constant agreement with a reference pointer that was being driven by the machine. The follow pointers had, on the outermost ring, a fixed numerical scale for reading off the value; then a middle, rotatable ring with a mark; and an inner, rotatable center disc, also with a mark. Of the two movable parts (ring / center), the center is driven by the computer, and the operator must use a handwheel to continuously keep the ring so adjusted that the two marks are always exactly aligned.

At first one wonders why the crew had to keep entering via handwheels variables that the device already knows. There is no other activity at the SPERRY than

always keeping “mark on mark” — setting aside the telescopes, which must also always keep the aircraft (in azimuth and in elevation) centered in the crosshairs.

[Caption: Altitude (right) and elevation angle (left) must be continuously “tracked” by hand at the computer. Lower right: the handwheel for manual tracking of the altitude, which is already known to the device (input from the range-finder).]

[Caption: At the gun: The gunner sets the elevation of the barrel according to the computer’s instructions — the mark on the central disc must align with the mark on the middle ring. Left: coarse range; right: fine range. Value set on the ring: 474 mils.]

Understanding and interpreting these “follow-pointer trackings”:

It is a simple activity that can probably be performed even under stress. The required movement goes directly from eye to hand; no “thinking” is required.
The individual variables are processed by the computing mechanism through numerous mechanical gear stages arranged in series. Mechanical force must repeatedly be fed into the computing process “by hand.” The electrical synchronous transmission from the range-finder to the fire-control computer, and from the latter to the guns, transmits only the signal, but not the mechanical force needed to actually carry out the subsequent processes. Precision, slip (in the integrators), friction, and required force are all interrelated.
It is probable that the human operator, in performing “tracking,” is able to automatically compensate for and smooth out all manner of oscillations and jerky movements. It is possible that only these “empty,” apparently unnecessary manual inputs make a more precise computation possible (this could only be assessed if one were to observe a fully connected and operating device).
When a given quantity is “tracked” at a handwheel, it is not always the case that the movement of the handwheel changes the quantity labeled there — through tracking, a completely different, previously unknown quantity may also be found at the correct value. A mathematical operation can thus be embedded between the handwheel and the follow pointer: the (unknown) input is then correct when the (known) output is correctly set. Through follow-pointer tracking, the operator can therefore find the correct value of a previously unknown quantity. Example in the SPERRY computer (see the description below): finding the unknown map-distance to the aircraft by correctly tracking the already-known elevation angle (vertical angle to the aircraft). The operator has the feeling that he is tracking the elevation angle, but in reality he is adjusting the map-distance. The computer ensures purely mechanically that the equation [tan Γ = altitude / horizontal distance] is satisfied at all times, and the operator must — by manually varying the horizontal distance — arrive at a value of Γ that matches the optically measured value in the telescope. At that moment, all three variables correspond to the current state. Of course, all quantities must be continuously tracked to follow the aircraft.

In later years, “tracking” was increasingly automated through electrical solutions, which saved on operating personnel. In the SPERRY type discussed here, only

a single quantity (in two components) is tracked automatically. Electrical tracking appears to be a delicate matter, because oscillations can arise, because tracking should be precise yet also fast, and because (according to Ref. 1, p. 116) when the power is off, manual manipulations must not damage the sensitive electrical contacts (with gaps of tenths of a millimeter). The SPERRY computer, however, has a quite different type of tracking; there are no tenths-of-a-millimeter contacts, as drawn in the diagrams of Ref. 1, pp. 113–115. In one of the earlier SPERRY types, the T6 (the Swiss device is perhaps the T8 — uncertain), seven men at the computer alone each operated one follow pointer (English: “follow the pointer”) and set a quantity via handwheel that the computer already knew. In addition there were three men for azimuth, elevation, and distance measurement, plus — for each gun — three men setting azimuth, elevation, and fuze, all using follow pointers (illustration found on the internet). The internal feedback of the projectile flight time is clearly visible here: the hit point can only be found once the flight time of the shells is known — and to find the flight time, the hit point is needed.

http://web.mit.edu/STS.035/www/PDFs/sperry.pdf

Functional Description of the SPERRY Device (possibly T8, acquired on a trial basis by the Swiss Army in 1935):

Every “line” in the following schematic represents a rotating shaft that corresponds to a variable. The variable can be represented either by the angle of rotation of the shaft or by its rotational speed. At both ends there are often gear wheels to transmit the information around corners to the next stages.

Rectangular coordinates (East/West, North/South, altitude) are marked in green.
Polar coordinates (angle, distance) are marked in red.
Times are marked in blue.
A small square with a diagonal cross indicates addition of the quantities fed into it.

Abbreviations:

HR = Handrad (handwheel)
FZ = Folgezeiger (follow pointer)
Variables with subscript M in the schematic always denote “…for the measurement point,” i.e., the aircraft.
Variables with subscript T in the schematic always denote “…for the hit point,” i.e., the predicted intercept point ahead of the aircraft.

The functional schematic of the SPERRY computer is taken from Ref. 1, p. 139.

A) Inputs / Outputs — The Computer’s Connections to Its Environment

Data Input:

HR 4 (fine, two speeds) and HR 5 (coarse): Setting the azimuth angle. The entire (heavy!) device is rotated laterally toward the aircraft via worm gear 6, then fine-adjusted to follow the aircraft’s motion. The observer stands at telescope 1 and follows the target in azimuth. The line of sight into the telescope is always horizontal, even when the aircraft is at high elevation.

HR 8: Setting the elevation angle. The observer stands at telescope 2 and follows the aircraft vertically. The movement of HR 8 is brought on one hand to display FZ 10 (coarse and fine), and on the other hand, the handwheel simultaneously rotates shaft 9 vertically, on which all three telescopes are mounted. Telescope 3 in the center serves the fire-control officer for observing the shots; he will issue manual corrections as needed.

The range-finder (telemeter) is a stereoscopic optical distance-measuring device. In Switzerland, tubes with a 3-meter baseline were standard; in Germany, larger ones up to 4 m and 6 m were also used, with Zeiss even producing 8 m and 12 m instruments. In the SPERRY, remarkably, it is not the distance but the flight altitude that is transmitted to the computer and displayed prominently at FZ 14. The range-finder itself has azimuth angle, elevation angle, and distance, and can therefore calculate the aircraft’s altitude without difficulty (from 1942 onward; previously, distance was called out verbally). — In Switzerland, the range-finder was traditionally mounted separately on its own tripod. In Germany and France, the range-finders were permanently attached to the computing device — the long tubes make the computer heavier and more unwieldy during rotation, but of course there was as yet no talk of fast jet aircraft.

With the GAMMA-JUHASZ, the British VICKERS, and the KODA, the range-finders were always set up separately on their own tripod. There are photographs (see below, p. 14) showing that in Switzerland trials were also conducted with the GAMMA-JUHASZ-derived GAMMA-HASLER device mounted directly with the range-finder, everything placed on a heavy trailer. Such configurations were never used with operational troops. Direct mounting on the computer means that the measured distance to the aircraft and the set angles are automatically picked up by the computer (no cables are needed, fewer operating personnel required).

Data Output:

From transmitter 52, the azimuth angle goes to the guns; from transmitter 58, the elevation angle together with the necessary ballistic barrel elevation; and from transmitter 61, the fuze setting (time delay for the projectile’s fuze). Technical implementation of the data transmission: see photograph below, p. 13.

The device has numerous possibilities for manual correction of the shot solution. These are not discussed here, as they are on one hand not important for understanding the basic principle of the mechanical computation, and on the other hand are easily understood (with the exception of the wind correction). In the order they appear in the schematic, the manual corrections are:

Control	Function
54	Azimuth correction
39	Parallax N/S
62	Fuze time correction
30	Wind N/S
42	Parallax E/W
63	Parallax altitude
31	Wind E/W
57	Elevation correction, hit point
16	Altitude correction, aircraft

Parallax: The computing device does not stand in the middle of the guns but rather several hundred meters to one side, possibly also higher or lower. The altitude parallax correction capability has not yet been found on the museum device in Dübendorf.

B) Determining the Aircraft’s Position: Polar → Rectangular Coordinates (N/S and E/W)

The azimuth angle to the aircraft, unchanged as measured by the telescope, rotates via a worm gear one disc carrying the azimuth slot of the coordinate converter 25–27 (cf. photograph, p. 11). The disc beneath it, engraved with the distance spiral, is adjusted via the map-distance e_km (projection onto the terrain plane), which is found using HR 18 and HR 22. In the coordinate converter, the spiral provides a linear relationship between map-distance (angle of rotation of the disc) and the local offset of the engraved groove from the center. A pin that passes through everything forces the two slides of the rectangular coordinates into their correct positions, so that the East/West and North/South coordinates can be read off and used further. The linear displacement

of the two arms 27 is immediately reconverted via rack-and-pinion into rotational motion, which is passed on to further gear stages.

To determine the map-distance, a clever and ingenious solution is employed. If m (Gamma) is the elevation angle to the aircraft, e_km is the map-distance to the aircraft, and h is the flight altitude, then the following always holds: tan m = h / e_km. This relationship is used to determine e_km. The flight altitude h is known from the range-finder and is re-entered using HR 13 (FZ 14). Lever 17 is a switchover between aerial and naval targets (not discussed here). The altitude h set at HR 13 rotates the cam body 11, which is displaced horizontally by the (still unknown) map-distance e_km. Cam body 11 now solves the entire equation tan m = h / e_km, so that the sensing stylus 12 gives, as its distance from the rotation axis, the elevation angle m (cam body: see photographs below, p. 12). The displacement of sensing stylus 12 is converted via rack-and-pinion into rotational motion. The elevation angle m that has been found is matched via HR 18/22 at FZ 10 to the elevation angle m measured by the telescope — and with that, the map-distance e_km has been correctly found! Cam body 11 combines a trigonometric function and a division (a three-variable function table). The rectangular coordinates found are fed on one hand into the tachometer generators 28 and 29, where the rectangular velocity components of the aircraft are determined. The tachometer generators are a type of clockwork mechanism: over a fixed time interval, the increment of position is determined, which is proportional to velocity. On the other hand, the aircraft coordinates are also needed in the additions (40, 43) to add the respective lead distance (E/W, N/S), which yields the coordinates

of the hit point. These are then converted back into polar coordinates by the second coordinate converter 48. HR 18 adjusts the map or horizontal distance. So as to avoid continuous adjustment during flight, HR 22 sets the rate of change of distance (approach or recession — not the aircraft’s speed), which is ultimately integrated to the horizontal distance by a disc integrator, as follows: Motor 20 drives a round disc at constant speed (seen here from the side). Pressed against it are two balls or a rotatable measuring disc, which is set more or less far from the center of rotation via spiral thread using HR 22. The balls or the measuring disc transmit their rotation to roller 19, which rotates faster when the balls are further out. Either an increase or decrease in distance is possible, since the balls rotate in the opposite sense when the motor’s axis of rotation is crossed. The operator need only track the slowly changing rate of radial approach/recession. Whether the required adjustment of the elevation angle using two handwheels is simple in practice could only be determined by experience. HR 18 corresponds to an integration constant.

C) Determining the Lead Distance and Hit Point: How Far Does the Aircraft Fly During the Shell’s Flight Time? — Rectangular Coordinates

Velocity multiplied by time gives distance: in multipliers 32 through 36, the aircraft velocity (entered manually via follow pointer) is multiplied component-by-component by the shell flight time — this yields the additional distance traveled by the aircraft up to the hit point (assuming straight-and-level flight). The shell flight time is read from the lower portion of cam body 37, which is displaced along its axis by the flight altitude, and rotated by the map-distance e_kt

to the hit point. Cam body 37 is rotated through only half the angle compared to cam body 55 when the map-distance changes. For each pair of altitude / map-distance, the shell flight time must be known (presumably from theoretical calculation?) in order to fabricate the cam body. The output of the two multipliers corresponds to the distance traveled by the aircraft during the shell’s flight time. This distance is added component-by-component in adders 40 and 43 to the current position of the aircraft — this addition produces the coordinates of the predicted hit point. The guns are aimed at this point.

D) Converting the Hit Point: Rectangular → Polar Coordinates, Output to the Guns; Fuze Information to the Guns

The guns, however, require angles and not rectangular coordinates. In azimuth, the computed lead cannot simply be added. Therefore, the polar coordinates must be freshly computed from the Cartesian components. Unfortunately, the coordinate converters do not work in the direction from Cartesian to polar form — this necessitates an indirect approach. In the second coordinate converter 48, (initially provisional) values for azimuth and distance to the hit point are entered. The resulting Cartesian coordinates E/W and N/S are compared at 44 and 45 with the already-known correct values of the hit point: the inputs to the coordinate converter are changed by two motors 46 and 47 until the Cartesian values at the output are correct. At that moment, the input (polar coordinates to the hit point) is also correct. The comparison of actual value and set-point value is carried out by an electrical follow-up system (shown dotted in the schematic). The two motors replace two operators who would be doing the same thing at two follow pointers.

However, the follow-up system here is a delicate matter, because both values to be found (distance and azimuth to the hit point) each affect both Cartesian coordinates E/W and N/S! The two operators at the follow pointers would interfere with each other. The electrical follow-up system is implemented in a considerably more complex manner than is shown in A. Kuhlenkamp’s schematic, where motor 47 for azimuth tracking is controlled solely from the E/W component, and motor 46 for distance solely from the N/S component. The two motors, which at the output each change only one quantity, in reality take their input — depending on the position of the hit point — now from one source (comparison of the E/W component) and now from the other (comparison of the N/S component). This is apparent from US Patent No. 2,065,303 of 1933 / 1936 (filing / publication dates?). The electrical control of the follow-up motors as published in that patent is shown further below in the Appendix, pp. 14/15: numerous relay contacts influence the flow of information in the electrical follow-up system! There is even a hysteresis, so that the motors do not switch too frequently between one source and the other in special positions of the aircraft.

The azimuth angle found is transmitted directly to the guns via transmitter 52. The distance to the hit point goes to cam bodies 55 and 37, which (together with the flight altitude) compute the barrel elevation including ballistic elevation correction (the “fall” of the shells during flight time must be compensated — body 55), as well as the shell flight time for lead computation (body 37, lower portion), and the fuze-setting information for setting the fuze (body 37, upper portion).

Photographs of the SPERRY Fire-Control Computer:

[Caption: One of the two coordinate converters sits at the very top of the device (lower left is the photographer’s foot, for scale). Polar coordinates (angle, distance) are converted here into rectangular coordinates. At the very bottom is a distance disc with a spirally milled groove (distance proportional to angle of rotation). Above it is the azimuth disc with a slot corresponding to the compass bearing to the aircraft. A sliding pin passes through everything and carries along the two arms that slide at right angles: they indicate the pure East/West and North/South displacement of the target. Rack-and-pinion gearing converts this to rotational motion that is passed on. At right: synchro for the altitude display, value arriving from the range-finder.]

[Caption: A cam body (center of image, behind the perforated metal plate) is a precisely milled three-dimensional table in steel. It can rotate (first variable); the ball bearings for longitudinal displacement (second variable) are also clearly visible. The “result” is picked off horizontally with a fixed stylus pressing on the surface, shown more clearly in the next photograph: distance from the rotation axis = third variable. Above it: the addition gearbox of the photograph after next, where one of the inner gear wheels is visible. Diagonal feed of one of the summands. According to A. Kuhlenkamp’s schematic, a total of 13 addition gearboxes are needed throughout the entire computer. What is being added here could not be determined.]

[Caption: The surface of the cam body, seen from a slight angle. In the background, slightly out of focus, the stylus that reads off the appropriate third variable for the two variables of rotation and longitudinal displacement. If this computer corresponds to the schematic published by A. Kuhlenkamp in 1943, the cam body solves the equation: flight altitude (rotation) divided by horizontal distance (displacement) equals tangent of elevation angle; the latter is picked off by the stylus and set by hand to match the measured telescope elevation angle — thereby the horizontal distance is found. As H increases: rotation counter-clockwise. As distance increases: cam body approaches the observer. Below the cam body (see above) is a heater; in front of it, the spirit level.]

[Caption: An addition gearbox that is simpler to manufacture than the classical differential gear. Inside the bright, round, self-rotating housing are two gear wheels (not visible here), each meshing with one of the small peripheral gears. The two summands are fed in axially — visible at front, barely visible at rear — the sum consists in the rotation of the round housing.]

[Caption: Below, on a movable wheel: electrical contacts by which the two follow-up motors select for themselves the appropriate input variable (North/South or East/West coordinates) by which to regulate the outputs (distance, azimuth). Cf. the schematic from the patent specification, further below in the Appendix, p. 15. These must be the overlapping pickup contacts, approximately in the middle of the drawing on p. 15. This is a more efficient solution than that drawn in A. Kuhlenkamp’s schematic (which, however, is limited to the mechanical information path only).]

[Caption: Setting the parallax, i.e., the horizontal distance between the guns and the fire-control computer. The guns must aim in a different direction than the one computed. The East/West and North/South components are each adjustable separately. The setting for the difference in altitude has (not yet) been found. Diameter of the handgrips: just under one centimeter. These settings are concealed under a screw-off cover; they need to be reset only after a change of position. The parallax values are set using correction knobs 39 and 42 in the schematic — these are they. Upper left: the worm gear adjusts the angle of a large wheel.]

[Caption: Electrical synchronous transmission of rotational values to the guns, or from the range-finder to the fire-control computer. (Image from the Dübendorf documents on the GAMMA-JUHASZ, adapted for Swiss languages.) The 3 rotor coils (actively set on the transmitter side) remain parallel on both sides. In the SPERRY the arrangement will be similar (likewise 5 cables per channel). Only the signal is transmitted — no torque!! In Ref. 3, p. 41, a MIWIKO-G fire-control computer is depicted for which the outgoing cable is labeled: “torsion cable for data transmission” — a mechanical solution! (Plus electrical lines for lighting and telephone). MIWIKO-G was designed by K. Papello, possibly 1937 or earlier. Possibly an abbreviation for Militärisches Winkel-Kommando-Gerät (Military Angle Fire-Control Device).]

[Caption: Experimental GAMMA-HASLER 43 on trailer. This configuration with the range-finder mounted directly on top was never used with operational troops. Azimuth and elevation angles to the aircraft, as well as distance (or possibly altitude), are coupled more directly to the computer than with a separately standing range-finder.]

[Caption: A later version, the GAMMA-HASLER 43/50 RS, received radar data (in place of the range-finder) from the British radar set “Mark VII.” This was the predecessor of the electronic analog computers “Fledermaus” and “Superfledermaus” by Siemens-Albis (from approximately 1963 onward).]

Appendix — Special Topics:

Electrical follow-up system for coordinate conversion in the SPERRY device
Unclear matters regarding ammunition and initial velocity adjustment in the SPERRY
How are the guns loaded? Setting the exact time delay
Turning flight
Comparison of fire-control computers from different manufacturers

A) Units produced — documented types B) Mechanical elements and building blocks used — comparison C) Different mathematical models for finding the hit point D) A look inside the GAMMA-JUHASZ E) Gear-train diagram of the German KdoGt 36 and the British VICKERS Predictor

The computer in the complete anti-aircraft battery (for laymen)
Bibliography

Electrical Follow-up System for the Conversion of Cartesian to Polar Coordinates

The patent specification at http://www.google.com/patents/US2065303 contains brief explanations of a computer that closely resembles the SPERRY device in Dübendorf (housing and functional schematic). The numbers of the individual elements in the patent do not correspond to the numbers in A. Kuhlenkamp’s schematic, since Kuhlenkamp studied captured devices ten years later in Germany (and probably never saw these patent specifications). Reading the patent explanations requires some flexibility, as the automatic scanning has misread numerous characters (i instead of 1, a zero read as eight, etc.).

The contacts labeled 83 and 86 at the very top of Fig. 8 are the two comparators 44 and 45 in Kuhlenkamp’s schematic — one lies in the path of the East/West component, the other in the path of the North/South component. The various relay contacts determine which of the two follow-up

motors 96 and 98 (Kuhlenkamp’s 46 and 47) uses which comparator as its input. The follow-up motors regulate the two output quantities, azimuth and distance. Fig. 9 shows that, for example, for hit point T1 it is more efficient to work with the East/West component in order to quickly achieve changes in distance (and North/South works better for changes in azimuth). For hit point 1, therefore, the follow-up motor for azimuth must be controlled via the North/South component; for hit point 2, it would be exactly the opposite. With this circuit, the two follow-up motors select for themselves the most efficient input for control! The sector switchers L, L1, L2, L3 in the center of the schematic are shown in one of the photographs above.

Unresolved Questions Regarding Initial Velocity Adjustment

In the lower portion of cam body 37, the shell flight time is determined in order to derive from it the lead distance and thus the hit point. It is not entirely clear how, when different ammunition is used (or even only with strongly differing temperatures), the initial velocity v₀ is set, because to swap out cam body 37 (and presumably cam body 55 as well) for different bodies whenever v₀ changes would be very laborious.

At the 7.5 cm gun in the Dübendorf Museum, two marks are painted in red for the recoil travel of the barrel at different initial velocities (for range-practice use with the towed target sleeve versus for war-issue ammunition). How this recoil is determined in the midst of heavy powder smoke is a separate problem — but in any case, initial velocity is important, since a change in it feeds very directly into the hit position and flight time. In the SPERRY’s functional schematic, no v₀-dependent adjustment of flight time or fuze time is visible (it would have to be located between cam body 37 and the multipliers, because after the component-wise decomposition, two corrections would be required). Manual correction 63 corrects the flight altitude that enters the computation (i.e., range-finder uncertainty and/or parallax altitude). Correction knob 63 has not yet been found on the Dübendorf device — but that does not mean it does not exist.

Since the situation is not entirely clear, a passage from Kuhlenkamp (Ref. 1, p. 143) is quoted verbatim:

“The cam body can be additionally displaced in the height direction by an amount settable via operating knob 63. Segment 64 additionally shifts spindle 56, so that an addition of this value to the measured value is effected. The value settable in this manner is the vertical positional difference, which can lie between −400 and +450 m.” (An altitude difference of 400 m between guns and computer is inconceivable given cable lengths.) “Some of the interior- and exterior-ballistic influences are accounted for approximately. The influence of rifling twist is covered by a correction to the azimuth; air density by a correction to the initial velocity; and the change in initial velocity by a correction to the measured altitude.”

In Ref. 1, p. 58, a trajectory chart is reproduced for a standard trajectory with v₀ = 607 m/s and for a v₀ reduced by 5.5%, i.e., 573 m/s. At an altitude of 3,000 m, horizontal distance of 3,000 m, barrel elevation 50°, and flight time of approximately 14 seconds, a 5.5% reduction in initial velocity results in a distance error of approximately 120 m by which the aircraft is missed — which is far too much. The two red marks on the museum cannon in Dübendorf (v₀ = 550 m/s for range practice with towed sleeve; v₀ = 805 m/s for war-issue ammunition) represent a mutual deviation of more than 30%. The ballistic cam bodies of the GAMMA-JUHASZ device being used therefore had to be exchanged when a different type of ammunition was fired. The initial computation and manufacture of the ballistic cam bodies must have been an extremely laborious undertaking.

If it was indeed impossible to adjust the initial velocity on the trial device purchased by the Swiss Army, and if the Swiss were using different ammunition from the Americans, with a cam body not matched to that ammunition — then it would be understandable why the results did not prove satisfactory to the Swiss.

At any rate, “matching” VICKERS guns were apparently purchased together with the SPERRY computer (which were later replaced by SCHNEIDER-CREUSOT guns when the switch was made to the GAMMA-JUHASZ fire-control computer).

“Die Flakartillerie in Dessau” (www.militaermuseum-anhalt.de/Regionale Militärgeschichte) outlines how the initial velocity of the shells was measured (Germany, Second World War):

“The gun barrels of the anti-aircraft batteries were under constant care and inspection. To maintain their precision, the V-zero values had to be measured at regular intervals — that is, the velocity of the shell on leaving the barrel. While permanent stations were set up for this purpose in the home territory, at the front a motorized team moved from one anti-aircraft regiment to the next to attend to each gun. Magnetized shells were fired through two coils set up at a precisely measured distance in front of the gun barrel. The time required for the shell to travel from the first coil to the second was determined by the so-called Boulanger apparatus [Le Boulengé — not Boulanger!], of which — to increase measurement accuracy — two were housed together in a special trailer. As the shell flew through the coils, the shell’s magnetic field induced a current impulse that, via a relay, triggered two falling weights in the Boulangé apparatus in succession, the second (smaller) of which actuated a cutting blade that struck a notch into the first. The fall height was measured with a gauge, yielding the new V-zero value for each gun. This made it possible for the battery’s shells to burst at precisely the intended position.”

From Ref. 3, pp. 62–63, it is not apparent how initial velocity was set in the German KdoGt 40 (though at least a “service grade” is mentioned).

From Ref. 1, pp. 127–136, it is also not apparent how v₀ was set in the German KdoGt 36. At least numerous “ballistic drums” are present, on which several correction sheets or films were wound, and which were easier to exchange than cam bodies (among other things, p. 131 on the “service grade” of the barrels: after many shots, the initial velocity changes). For photographs of these drums, see also Ref. 3, pp. 27, 29.

How Are the Guns Loaded? Precise Timing Consistency Is Essential

The shells are inserted into the fuze-setter machine at the gun, to which the computer continuously transmits the correct time delay. When the fire bell sounds, the shell is fuzed; the heavy cartridge must then be inserted by hand into the barrel, the breech closes, and when the red lamp extinguishes, the trigger lever is released — the shot fires.

[Caption: At the Flieger-Flab Museum Dübendorf: The 7.5 cm cartridge sits in the fuze-setter machine; the fuze at the tip of the projectile is correctly set. The loader then pulls the cartridge out, turns it (in this case by 90°, even more for a steeper barrel angle), and inserts it from the rear into the gun. Because this is done by hand, this operation does not always take the same amount of time. The anticipated loading delay is entered into the computer in advance. The weight of a 7.5 cm cartridge was approximately 11 kg (Bofors, Schneider).]

From the moment the cartridge leaves the fuze-setter until the shot is fired, a certain time elapses that is communicated to the computer in advance (loading delay). The time delay is adapted to the physical condition of the loaders involved. The crew was known personally and their individual work speeds were accounted for. All guns were supposed to fire as simultaneously as possible. From the moment the red light at the gun is extinguished (Kügel, p. 68) (possibly three flashes, P. Blumer), the trigger lever must be released with as little delay as possible. The red light is controlled by an adjustable clock at the gun, independent of the computer.

This account is based on the recollections of then-Kanonen-Korporal Ulrich Wegmann (Flab basic training 1962 on the 7.5 cm gun) and applies to the GAMMA-JUHASZ computer and the 7.5 cm SCHNEIDER-CREUSOT gun as operated in Switzerland. The SPERRY was never used operationally in Switzerland, but the procedure is likely to have been similar in all cases.

In the operating manual (Museum Dübendorf) for the Hungarian GAMMA-JUHASZ device, a loose slip of paper was found on which someone had listed at what timing error a target miss of 25 meters* would not be exceeded, this being still considered acceptable given the fragmentation effect. Title of the list: “Permissible fuze-setting deviation in 1/100-second intervals, at v₀ = 805 m/s.” For a flight time of 1 to 2 seconds, a maximum time error of 0.03 seconds is permitted; for a flight time of 10 seconds, approximately 0.06 seconds; for a flight time of 20 seconds, approximately 0.08 seconds — thus consistently below one-tenth of a second. (* calculated solely from the shell’s path — the aircraft’s own motion adds to this!)

Ref. 3, p. 24, on the German Kommandogerät 36: “The time required for the command and loading delay could be set on the device and thus taken into account (typically 13 seconds).” That seems almost impossible!? Elsewhere, much shorter values are found: three to six seconds between shots.

Turning Flight

The GAMMA-JUHASZ device purchased by the Swiss Army was built under license by HASLER AG and later further developed. Originally it was designed for pure straight-and-level flight; by 1941 (or earlier), climbing or diving flight was already being accounted for. From 1944 onward, it was extended by HASLER to correctly predict turning flight as well. Who would like to verify this? Who understands this circuit for the turning-flight extrapolation? Attempt to read and understand unfamiliar mechanical computing gear trains! Shown here is the newly installed gear train with the turning-flight extension from 1944, with a switch between “Fixed Course” and “Turning Flight.” This switch also existed in the “Super-Fledermaus.”

Symbols:

D1, D2, D3 — additions or subtractions (the direction is clarified at D1)
φ̇ — rate of change of azimuth angle (tracking the aircraft from the computer, i.e., how fast the measured azimuth angle is changing)
α̇ — rate of change of flight angle. “Flight angle”: the angle between the aircraft’s course (its compass bearing) and the line of sight from the computer to the aircraft, always referenced to the map. Even in straight-and-level flight, the flight angle is continuously changing. The flight angle as seen from the pilot: the angle between the aircraft’s nose and the direction toward the computing device (possibly ± 180°).
ψ̇ — rate of change of the aircraft’s course, i.e., how many degrees per second the aircraft’s course (compass reading) is changing.
δ — swivel angle of the chord of the flight path: the angle between the current aircraft course (tangent) and the chord to the future hit point on the curved flight path.

(These are now non-rectangular coordinates: much more difficult, unfamiliar. Cf. further below, page 21: Different Mathematical Models.)

[page 18: figure only]

Comparison of Fire-Control Computers of Different Origins

The fire-control computers of different countries differ greatly, both in their technical implementation and in the underlying conceptual approach — that is, the mathematical methods employed and the coordinate system used to determine the future point of impact. Only a brief discussion of some differences is possible here, since a deeper excursion into an additional mathematical framework would represent a substantial undertaking.

A) Produced Devices — Better-Documented Types

The operating principle of the SPERRY device from the USA is examined here (possibly the T8 model, purchased by Switzerland in 1935 and tested in the summer of 1936). Also known from the literature (Ref. 1 through Ref. 3) are the functions of the Kommandogerät 36 and the Kommandogerät 40 (both German), as well as the British VICKERS. The museum in Dübendorf holds extensive documentation on the Hungarian GAMMA-JUHÁSZ and on the later versions GAMMA-HASLER 40 and 43, which subsequently (from 1950 until approximately 1962) also processed radar data in addition to telemeter range data.

Knowledge remains fragmentary (a few schematics, individual photographs) regarding various auxiliary fire-control computers from Germany (WIKOG, MIWIKOG, Kommando-Hilfsgerät 35, Zeiss Kommandorechner C2 (1927–37), Pschorr 27 (from the firm NEDINSCO in Holland, where Zeiss manufactured due to restrictions imposed by the Treaty of Versailles), Schönian 1918). Regarding the term “auxiliary device”: fire-control computers were exceedingly expensive to develop and manufacture, and they were always available in far too small numbers. Simpler auxiliary devices were therefore also attempted.

From other countries there are isolated references or photographs of fire-control computers from France (designs AUFIÈRE or OPL), Holland (HAZEMEYER, built under license from Siemens-Halske), Czechoslovakia (ŠKODA), the latter copied by Russia (PUAZO). A final PUAZO type 6-60 (from 1953 or 1956) apparently weighed 7 tonnes and required 2.5 kW of electrical power for its motors, etc. Mentioned only very briefly in Ref. 1, p. 10 and p. 123 is a fire-control computer GOERTZ from Vienna: “built in a single unit and delivered to Russia.” Finland is said to have worked from 1929 onward with a Goertz that did not function correctly (?). The “KERRISON PREDICTOR” (GB) was apparently optimized for low-flying aircraft with high angular velocities.

Developments are observable everywhere — earlier types and later copies and improvements. A curious passage on SPERRY devices was found on the internet (was America scarcely threatened by foreign aircraft??):

http://web.mit.edu/STS.035/www/PDFs/sperry.pdf

Of the nearly 10 models Sperry developed during this period (1927–35), it never sold more than 12 of any model; the average order was five. The Sperry Company offset some development costs by sales to foreign governments, especially Russia, with the Army’s approval [9].

Germany produced, between 1936 and 1945 at Zeiss alone, a total of 2,053 Kommandogerät 40 units as well as a further 2,376 fire-control computers of other types (Ref. 3, p. 43).

B) Construction of the Devices: Mechanical Elements and Components Used

What follows are fragmentary partial details of the better-known devices — they nevertheless show the widely differing approaches by which the ever-identical task (determining the point of impact ahead of the aircraft) could be tackled using mechanical gear-train computation. Due to the continuous further development of all types, these figures and years must not be taken too literally!

	Vickers Predictor	SPERRY	Kdo Gt 36	GAMMA-JUHÁSZ	Kdo Gt 40
Year	1928	1935	1936	(1934)/37	40
Origin	GB (US)	US	D	Hun/Sui	D
Additions	9	13	9	12 ¹⁾	20
Integrators	3 ²⁾	1	3	0	7
Form bodies	1 ³⁾	4 ⁴⁾	0	19	14 ⁶⁾
Cam drums	2	0	3 ⁵⁾	0	3
Motors	0	2	3 ⁷⁾	18	16
Electrical follow-ups	0	2	0	14	10

¹⁾ 12 differentials, plus 2 lever-type multiple-addition mechanisms
²⁾ One of them with two friction wheels
³⁾ Plus three cam discs (one dimension fewer)
⁴⁾ 4 curves on three bodies
⁵⁾ Plus 3 tables, plus 2 plotters
⁶⁾ Plus 5 cam discs
⁷⁾ Of which one is a spring motor

Integrators

Friction-wheel gears (or, in Germany, also spherical-segment gears) can be used very versatilely: as multipliers, as integrators, and — with a different choice of variables — for division and differentiation, and for blending different quantities.

Form Bodies

“Three-dimensional tables”: Carefully ground, three-dimensional shapes that rotate and translate in front of a pickup stylus. The distance between the axis of rotation and the pickup stylus constitutes the third variable.

Function Drums

Rotating cylinders or drums were frequently used; paper or foil sheets carrying entire families of curves could be mounted on them. The operating personnel selected (depending on flight altitude, speed, etc.) the appropriate curve and had to continuously keep a pointer exactly on the curve using a handwheel.

Electrical Follow-Up Systems

An electric motor drives a variable in such a way that the variable itself influences the motor contacts (forward, reverse). This allows an already-known variable to be “copied” and its setting transmitted with greater force than previously to the next gear train. Operating personnel are thereby saved.

Special Experiments

The MIWIKOG fire-control computer by Karl Papello (1890–1958) is said to have employed flywheels on the manual drives for azimuth and elevation angle and for range, so that the device could continue computing unchanged — even if the aircraft was temporarily no longer visible due to natural or artificial fog (Ref. 3, p. 41). Similar in concept are the new handwheels on the GAMMA-HASLER from 1943 onward, electrically actuated.

C) Different Mathematical Models for Finding the Point of Impact

The SPERRY device computes in an intuitive manner using rectangular coordinates: the current aircraft position plus the additional flight distance traveled during the projectile’s flight time gives the point of impact. This is clear, logical, and easy to visualize. The “addition operation” succeeds simply only when the additional flight distance is expressed in meters (East/West and North/South separately) rather than in angles.
There are other possibilities for the advance computation: instead of imagining the motion of the aircraft, one stays with the directly measured values: azimuth angle, elevation angle, slant range. All three quantities each have their own rate of change — the point of impact is computed in advance using these quantities, not using the aircraft’s velocity. This gives entirely different calculations than the rectangular method. Having spent weeks with the rectangular method, it is not entirely easy to suddenly slip into a new way of thinking. Angles add differently than kilometers, and while at least the aircraft’s speed remains constant during straight-and-level flight at constant altitude, this no longer holds for the azimuth and elevation angular rates or for the closing rate.

This “angular-velocity method” is considered less accurate, since, for the predicted curve to conform to the actual flight path of the aircraft, the accelerations of all three quantities should also be taken into account (analogous to a power-series expansion). This is, however, hardly feasible, since the determination of acceleration is burdened with large uncertainties (quotation from Kuhlenkamp, see below).
The Kommandogerät 36 and the GAMMA-JUHÁSZ compute in cylindrical coordinates: azimuth angle, horizontal range, flight altitude. This again constitutes an entirely new way of thinking: the flight path is reproduced purely geometrically — with a “target-path plate” (in the case of the Kdo Gt 36, a so-called “flight-direction table”); there are movable carriages, elevation slides, and azimuth rulers. Motors and lead screws trace out distances, flight directions, and lead distances to scale.
The French fire-control computers, the HAZEMEYER, and the KODA use rectangular coordinates — but not geographically fixed directions (N/S, E/W); instead, directions tied to the aircraft’s heading: the azimuth angle from the computing device to the aircraft, the perpendicular to it, and the flight altitude.
The British VICKERS and the German WIKOG (Zeiss) compute algebraically, in that rather complex equations of spherical geometry are first solved on paper in advance and independently of the specific aircraft course. The computing device must solve — simultaneously — two mutually coupled equations derived from the measured angles to the aircraft (azimuth, elevation) and the computed values for the point of impact; these equations contain a total of 12 trigonometric functions, as well as the time of flight and the observed angular rates in elevation and azimuth.

The elevation lead angle can only be determined once the azimuth lead angle is already known. Computing the azimuth and elevation lead requires knowing the point of impact — that is, requires the two sought lead angles to already be known. The time of flight and the two lead-angle values are likewise mutually interdependent.

Complex lever gears and (in part double) friction-wheel gears take over the evaluation. The computation within the gear trains can just barely be followed — but visualizing where the aircraft is located, etc., is no longer possible. The gear-train schematic of the VICKERS is shown further below on page 24. More detail can be found in Ref. 1, pp. 122–124, or Ref. 2, pp. 43–46. See also Ref. 5 and Ref. 6 — all of it is difficult to understand.

Accelerations: With every time derivative (i.e., obtaining velocity from observed position or angle, or computing accelerations from velocities), unavoidable errors, uncertainties, and fluctuations become increasingly disruptive. On this point, a clear quotation from A. Kuhlenkamp, state of knowledge 1943 (Ref. 1, p. 50):

“Despite all measures taken, it has to this day not been possible, in practical field use, to determine pointing values with such uniformity and accuracy that a derivative higher than velocity can be measured. There has been no lack of attempts to measure at least the angular accelerations, nor of gear-train-based measurement possibilities for acceleration. However, these attempts have repeatedly foundered on the fact that the achievable pointing accuracy — particularly with difficult target-track types — is still insufficient, and that excessively large errors in acceleration therefore result.”

D) A Look Inside the GAMMA-JUHÁSZ: The Philosophy of the Chosen Coordinate Systems

The philosophy of the chosen coordinate systems naturally has implications for the mechanical gear trains that take over the mathematical functions. Two excerpts from the overall drawings of the GAMMA-JUHÁSZ — reproduced in Switzerland — follow.

The first drawing is dated 12.2.1941, Kdo Flieger- und Flabtruppen (Museum inventory 045925 old / 12271 new). The second drawing is dated 12.10.1943, K+W Thun (inventory 039469 old / 5123 new). Drawings in the Dübendorf Museum.

Target-path plate of the GAMMA-JUHÁSZ: A large circle in which the principal elements of the flight are reproduced purely geometrically in plan view and elevation view. Included: two flight-direction arms in the large circle at right — two smaller, rotatable arms at the same angle, mounted on a carriage (wheels indicated) moved by a lead screw and tension cable. Hence the long toothed rollers 42, 45, 68.

Azimuth-lead ruler: Long needle 47, nearly horizontal, slightly above center, determines the azimuth-angle lead.

Beneath it, a tension cable for measuring the range to the point of impact 14 (= range to aircraft 26 plus lead distance).

Elevation slide 53, lower left, executes the up/down movement of the elevation ruler 52, shown horizontally in the figure.

Range ruler 51: Pivot at lower left, rising at approximately 40° toward upper right. The aircraft altitude 13 is labeled h_m, and the slant range e_m is kept matched to the telemeter value via the handwheel at upper left.

Eight motors are visible in the excerpt, two handwheels (slant range and azimuth angle), three tension cables. At right in the small circle, a display on the outer casing (“target-path recorder”), for visual monitoring of range and flight direction.

[page 23: figure only — cascade of form bodies from the GAMMA-JUHÁSZ gear-train drawing]

A whole cascade of form bodies that are directly interdependent in a complex manner.

In rotation, the seven form bodies bearing a letter inside are all controlled by the same shaft (one quarter of the image width from the left edge, presumably the projectile flight time). In translation (vertical displacement in the image), the form bodies likewise all depend on the same quantity (lower left corner of the image, second shaft from the bottom — elevation of the barrels; a major change occurred here between 1941 and 1943!). The pickup stylus, which on the labeled form bodies presses laterally against the surface, controls in each case the translation of the seven flat form bodies that are controlled pairwise synchronously in rotation (i.e., by a total of four variables). The output of the flatter form bodies passes through a lever cascade to three “micrometers,” each connected to a handwheel (manual corrections for fuze time, azimuth angle, elevation) and carrying three electrical contacts at the opposite end: a center contact can make connection with “upper” or “lower” and controls the three nearby motors, whose output in turn influences the movement of the micrometer housings. This therefore constitutes an electrical follow-up of the variable — i.e., a re-injection of force. The electrical wires are not drawn here. The motors and handwheels are labeled “fuze correction,” “azimuth-angle correction,” and “elevation correction,” with the motors bearing a clear summation symbol. The lever linkages at the output of the form bodies can thus blend several quantities together, all contributing to the corresponding correction. The “flat” form bodies (each on the left) appear to determine the relative weighting during blending; the “tall” form bodies provide the corrections (from left to right) for air density, tailwind/headwind, muzzle velocity, crosswind — always as a function of projectile flight time and barrel elevation. The crosswind (upper far right) has an effect only on the azimuth angle. The handwheels allow final corrections if the observed rounds land incorrectly in elevation or azimuth, or if the shells detonate too early or too late.

Taken together, all these form bodies effect final fine corrections to the shot placement: all necessary angles and fuze times for the guns have already been determined elsewhere from the aircraft’s motion. Mixed in at the end are additional influences such as wind or a changed muzzle velocity. At the very top is the decomposition of the wind into a longitudinal and a lateral component. To the right of that (on the two uppermost shafts) are two “differentials”: these are addition mechanisms, analogous to the differential gear in an automobile — shaft speed of drive shaft = ½ × (speed of left wheel + speed of right wheel). With these, two independent quantities can be added, i.e., blended together.

E) Gear-Train Diagram of the German Kommandogerät 36 and the British VICKERS PREDICTOR

[page 24: figures only — uncommented reproduction from Ref. 1, pp. 127 and 154]

The Computer in the Complete Anti-Aircraft Battery:

Display in the Dübendorf Museum, shown here using the GAMMA-JUHÁSZ computer as an example. Far left: telemeter for range determination; then the computer; a central power unit (with gasoline engine and battery); longer cable to the guns (labeled: up to 6 × 100 m); distributor; gun cable 50 m. Each gun requires 8 men; the computer requires 9; the telemeter requires 5 — a total of 53 men for the battery (plus relief crews).

References Cited

Book “Flak-Kommandogeräte,” Senior Engineer Alfred Kuhlenkamp, 1943, VDI-Verlag GmbH (Association of German Engineers). 176 pages. The expertise of the Army Ordnance Office specialist is evident. Many geometric sketches. Individual gear trains, integrators, electrical follow-up systems. Good but very dense explanations of the overall computer architecture for three devices. Nothing on operation, personnel, battery organization, guns, or effectiveness on target. The book is held in the ETH Zurich special library HDB under call number R 1977 / 752, location: Kammer 7. A further copy is held at the Fachhochschule Jena (Ernst Abbe University of Applied Sciences Jena) under call number A05107, location: Magazin 2. The “Bibliothek am Guisanplatz” in Bern also holds a copy. Reproduction of the historical diagrams is made with the permission of the (successor) VDI-Verlag.
Special issue “Flugabwehr,” A. Kuhlenkamp, VDI-Verlag GmbH, 3rd edition 1940 (1st, 2nd editions 1938, 39). Guns, anti-aircraft sights, searchlights, sound-ranging devices, fuzes, etc. are also discussed. Twenty-six pages cover computing gear trains and fire-control computers, including detailed treatment of friction-wheel gears (integrators). Among the fire-control computers, only the VICKERS is described in detail.
Jenaer Jahrbuch zur Technik- und Industriegeschichte, vol. 11, 2008, Verein Technikgeschichte in Jena e.V. Contribution by Klaus-Dieter Gattnar (employed at ZEISS, Jena, 1956–1991): “Fire-control computers for anti-aircraft defense — developed and produced at the Zeiss works from 1915 to 1945.” Gattnar contribution: 91 pages. Products other than ZEISS are barely discussed. Thematically extremely broad in scope, with many photographs and sketches, construction details, manufacturing. References to the development of radio-measuring devices FuMG (= radar). The book is available from booksellers or (address as of 2014) at [email protected].
Original Pathé film, 1939, on the British “Gun Predictor VICKERS,” shown in (symbolic, i.e., promotional) operation: http://www.britishpathe.com/video/predictions-while-you-wait
Description of the VICKERS predictor, from pp. 11–13 (also other British devices): http://sydney.edu.au/engineering/it/research/tr/tr223.pdf (If the address is no longer active: British mechanical gunnery computers in World War II, Allan G. Bromley)
Larger work “Computing before computers,” Chapter 5: Analog Computing Devices, Allan Bromley: broad collection, including tide prediction by Lord Kelvin. Anti-aircraft gun predictor VICKERS on pp. 186–190. http://ed-thelen.org/comp-hist/CBC.html

Image Sources

Title page, top: “Fliegerabwehr,” Herrmann Schild, p. 19, Dübendorf 1982, new edition 2005. With kind permission of the Flieger- und Flab-Museum Dübendorf.

Title page, bottom: From Ref. 3, p. 25. With kind permission of the Jenaer Jahrbuch zur Technikgeschichte and the Podzun-Pallas-Verlag (formerly Nebel-Verlag).

All other photographs were taken by the author at the Flieger- und Flab-Museum Dübendorf.

The technology excavator: André Masson, CH-4900 Langenthal — Autumn 2014