Tuesday, March 15, 2011

Fall of France, 9: Manpower, Part 6, 1: Computers. No. Seriously.

It is dawn on 21 May, 1940. 

We War of the Spanish Succession historians know this geography

The Dyle Line on which the Allies planned to halt the Germans if the Belgian Meuse/Maaß fortresses fell has been compromised, and Lord Gort has fallen back on a line of waterways. At Ghent in Flanders, the river Escaut/Scheldt joins the Lys and turns sharply westward towards Antwerp, below which it falls into the commingled estuary of the Meuse/Maaß and Escaut/Scheldt, a now-managed estuarine terrain prone to fall back into its prelapsarian chaos when heavy artillery or human ingenuity interferes with the drains, and, appropriately enough, called Zeeland. This isn't a sound defensive line, and between Ghent and the sea the Belgians have instead lined up on the 19th Century Ghent-Ternuizen Canal. South of Ghent, the line follows the river Escaut/Scheldt, with the army boundary between the Belgians and the BEF somewhere around Oudenarde. At Maulde on the frontier, the line abandons the Escaut/Scheldt and follows the Scarpe to Amiens, where the BEF has its GHQ, protected by a battle group, Petreforce. Between Maulde and Arras, however, the line is held by French troops, including a North African division in which it appears that the Commander-in-Chief, Lord Gort, has a misplaced lack of confidence. East of Arras is Prioulx's refitting Cavalry Corps, now up to its full strength of 3 armoured divisions.

 South of Arras along the Canal du Nord, Gort has deployed three whole divisions of the BEF that were brought over to France as "labour" units to quicken the work of building the BEF's infrastructure, a task made much more arduous by the decision to extend its lines of communication all the way back to the Norman and Breton coasts rather than to the Channel Ports, as in the last war. In practice these units are reasonably well trained infantry by the standards of the Battle of France, when ever army, in Karl-Heinz Friesen's phrase, was an iron spearhead on a wooden shaft. Unfortunately, they almost entirely lack artillery (and by this I mean actual gunnery-type persons and maybe tractors, since the one thing not lacking in any country's arsenals except Germany's is leftover WWI gun tubes). This makes them tit-useless in modern warfare, except possibly in urban terrain. Too bad that doctrine calls for them to deploy on the outskirts of towns, on the assumption that the massed siege artillery legions of WWI will promptly level every structure in sight! To add to the perversity of it, one of the labour divisions, the 23rd, is a "doubled" motorised division, and even has its organic mechanised reconnaissance battalion. I'm not clear if it has its attached motor companies, but since a large portion of it was overrun while making a mounted march, it may have. 

By the 21st, the labour divisions have disappeared, the Canal du Nord has been passed, and the Germans are extending that hook shown on the map all the way to the coast. Cue crazy talk about Hitler "letting" the BEF escape, and even more dubious talk about how Gort just sensed that the Belgians were about to crack, or had given telepathic permission for the BEF to bug out. Because that's what's about to happen. At Dunkirk to the north, there is a vast area of inundatable land behind the coastal dunes, with a Seventeenth Century channel cut through them to allow Dunkirk to function as a protected port and agricultural market town. Behind those inundations, the drained fields around Dunkirk, there is enough room for the BEF to stand around and await evacuation. The alternative, since the BEF has been cut off from ammunition resupply, is one last offensive southwards from Arras, and Gort has lost faith in that idea. 

But back up to the sad fate of the artillery-less labour divisions. In a country that has 1500 AA guns pointing skyward, how exactly does that come to happen? I've asked that question any number of times with the aim of putting a cynic's spin on it: cui bono

Up for examination this time round? The computer industry.

Looking at the mass of material that I have to go through, I'm thinking that this might take a while. So here's my summary theory: the "modern," which is to say, High Sixties computing industry coalesced from a wide range of fields. One of the largest was the automation/amplification/control industry that had many applications in 1933, but metastatised during the rearmament period and during the war due to the pressures of air defence. Those who were in the industry in 1933, including many defence contractors, particularly those supplying existing naval fire control, jumped on the bus with enthusiasm. So enthusiastically that I'm seeing a feedback loop of self-interest independent of the natural concerns about strategic bombing. In short, if the government is going to pay your staff to muck around with computers (that it buys from you!), that's like free training, and who is going to turn that down? Indeed, if the way is open, won't you lobby for more?

So there's my conspiracy theory that explains how the AA Command got so big, so quickly, even as, and not unrelated to, the way that the Territorial Army field artillery was failing to keep pace --in terms of personnel, not equipment.  

But, hey, theory, schmeory. The interesting stuff is in the details!

Here's Geoffrey Shakespeare, First Lord of the Admiralty, introducing the Navy Estimates 1939–40and explaining why ships are being held up by the novel problem of fire control delivery delays.   

 "The cost of the fire control gear on the 1912 King George V battleships was £11,000; that for the King George V of 1939 was £213,000. Taking inflation into account, the total cost has tripled.
(650) Development of fire control has been very extensive, the government’s demand has virtually created a new precision light engineering industry. Before rearmament began there were only 3 firms capable of building complete naval fire control gear; now there are 28. (345 House of Common Debates (647-50):

There's my bill of attainder. Britain is building a whole new industry devoted to shooting German planes down when they fly over battleships. Oh, and factories, too. Against German tanks, though, the French are on their own.

So this whole industry --I think that we're all agreed that fire control was one of the original computer applications. It's the focus of Norbert Wiener's Cybernetics, and the whole vast, diffuse project of integrating social and industrial planning with national defence and the work of individual machines starts with the intoxicating experience of the Air Defence of Great Britain Headquarters, and the notion of controlling everything from the track a train takes from Liverpool to Southampton down to the direction an individual AA gun is pointing and making it all, like, efficient or something. Of course, in recent years, thanks to Stephen Roskill,* Jon Sumida, Correlli Barnett and David Harmer, we've had it that all that British work was uniquely futile, because the Americans got it right with their Mark 37 Naval Fire Control, and the whole British scientific establishment was stupidheads for not just buying lots of Mark 37s instead of working on all their stupid British crap. 

Given the failure of the first generation of the British computing industry, one might be tempted to agree. But, on the other hand, where are the Seven Dwarfs now? We might have moved into an era of corporate stability in the computing field now that Apple has resisted its doomsayers. But aren't we shorting Microsoft now? Anyway, there's a much more substantive way of addressing this issue. Actual research, the kind that David Mindell did. (By the way, those interested might want to check out Jon Agar and David Noble. I found both these books to be misfires, but not in uninteresting ways.) 

So, anyway: Lund on David Mindell on the history of the magical Mark 37:

 The original (USN) computing fire control was the Ford Rangekeeper, first seen at sea in July 1916. Ford, a company started by Hannibal Ford of Sperry, who was apparently encouraged to depart the parent company, in marked contrast with Carl Norden, whose separation was acrimonious. Ford established his own firm working exclusively with the USN along with the Arma Engineering Company and GE. ( 16–24). Ford and Sperry began to be exposed to the ideas of Pollen (Mindell segues through Sumida, 28–9 andff). Sperry “relied on its technical skills and its relationship with the navy. It also had something the navy did not have: access to the technical details of British fire control. In 1913, Reginald Gillmor and Tom Morgan went to London, where they founded Sperry Gyroscope’s British subsidiary, the Sperry Gyroscope Company Ltd. From England, Gillmor corresponded regularly with Elmer Sperry. The inventor frequently asked Gillmor’s opinions, and he responded with detailed reports that Sperry then distributed to his naval contacts in the United States. Gillmor conducted a technical survey of European fire control systems and sent it to Sperry. Gillmor also sent Sperry details of Pollen’s work and recommended that Sperry Gyroscope license and manufacture Pollen’s system. (It never did, and Pollen believed that Sperry Gyroscope stole his invention.) In 1916, Gillmor reported about Jutland and the role of the Sperry gyrocompass in the engagement. Gillmor proved so knowledgeable in these matters and had developed so many contacts within the Admiralty that in April 1917 he returned to the U. S. navy as flag secretary on the staff of Admiral Simms . . . . A twenty-four page memo from Gillmor dated 1 August 1916 compared the tactical and technical issues of British fire control with those of American fire control with a clarity unsurpassed in any BuOrd documents of the time. Sperry passed the memo on to the bureau. Through what Elmer Sperry called “channels which insured their freedom from censorship” Gillmor transferred fire control technology from the Royal Navy to Sperry Gyroscope, and only then to the U. S. Navy.”(citations in general to the Sperry papers and Sperry’s letters to J. Strauss, head of BuOrd. Quotation is from “Sperry to Lieutenant Commander F. C. Martin, 31 August 1916, EAS Papers, box 32, Lt. Logan Cresap Fire Control Correspondence folder.) Ford’s Rangekeeper was soon supplemented by his Battle Tracer. It used an unsatisfactory Sperry electrical transmission system. Ford and Sperry consistently worked together, licensing each others’ patents and designing systems that would work together. (Mindell, 34) Sperry took on the gyro work, Ford the rangekeeper. The Mark 1 Ford was succeeded by the lighter and simpler Mark 2, which was ordered for virtually every ship from gunboat up. ((Mindell, 40) Ford’s secrecy and intimate arrangements with the navy served it well; for while Sperry and Pollen agreed that the Ford violated Pollen patents, higher ups in the US government connived to protect them. Franklin Roosevelt guaranteed “to hold and save you harmless against any and all suits” brought for infringing patents on British fire control.” The navy virtually stole technology from abroad and protected the companies that produced it from legal action.” (Mindell, 42; Roosevelt’s private guarantee is in a letter to the Ford Instrument Company, 30 April 1918, RG 74, E-30, box 747, subject file 30309, folder 201–250.) While Ford flourished in connection with the navy, Sperry did not. Even its electrical transmitters proved problematic, because they did not self-synchronise, and by 1918 the navy had lost faith that Sperry was genuinely willing to solve the problem. At last, Sperry called Gillmor back from the navy to become CEO and solve the company’s organisational problems. Gillmor, in a letter to BuOrd, was soon able to report success in reorganisation and attribute previous problems to “corrupt management.” (46) That was not going to solve the synchronsation problem, which was a technical, not a 'corruption' issue. Sperry lost fire control contracts after 1920 to GE.
At the same time, a new head of fire control at BuOrd was selected; Lieutenant William R. Furlong. Furlong, a 1905 Naval Academy graduate, had taken a master’s in electrical and radio engineering at Columbia, and served with the Grand Fleet in 1918, evaluating British fire control systems. He would rise to direct BuOrd in 1937–41 before passing on to oversee salvage at Pearl Harbor. Furlong shows up in David Zimmerman, Top Secret Exchange, as the man who obdurately opposed drawing BuOrd within the tent of the Tizard exchange. When Admiral Stark, with Tizard in his office, phoned Furlong and asked him to explain his actions, Stark came to the conclusion that technical sharing would reveal that the Ford Predictor was believed to infringe British patents now held by Paul and Isherwood(?) that were embroiled in ongoing patent suit. This brought the question of protection in the case of transferred WWII-era technology to a head and led to the British free-sharing policy (Zimmerman, 113–14.) .
As for GE, it won its place originally with a self-synchronising electrical 'relay:' the Selsyn, this time a German technology (48–9). The Selsyn was not, however, a terribly satisfactory answer, as Mindell will discuss, and  the thyraton and amplidyne were all considered in the 1930s, although whether BuOrd had the money to support these engineering efforts during the Depression is another matter. Meanwhile, Arma Engineering, founded by David Mahood, formerly of Sperry, and Arthur P. Davis, developed the German Anschütz gyrocompass into the navy’s stable element. Again, the use of a semi-bootleg civilian technology as perhaps the most critical 'computing' element points towards trouble ahead, even as the The Ford Rangekeeper and Arma stable element are linked in a whole series of naval fire control technologies, notably the famous 'fruit machine,' the Torpedo Data Computer. (56)
In 1929, Elmer Sperry sold his stake in Sperry Gyroscope.  North American Aviation, awash in aviation boom cash, bought Sperry, and in 1933 GM bought North American Aviation. With the Fishers in charge, Sperry had a much more powerful patron, and its relationship with the armed forces was on a new foot both politically and in terms of capitalisation. A new holding company was formed, the Sperry Corporation, which also held Sperry Gyroscopes (UK). Gillmor remained at the head of Sperry Gyroscopes Division. The Fishers also bought Ford and brought it into Sperry,   retaining Hannibal Ford as head of the division, and soon added the navy’s hydraulic systems manufacturer, Waterbury Tool Corporation, and the main aviation hydraulics supplier, Vickers of Detroit (Mindell, 60, 80–2). The navy, while not best pleased by Sperry’s past, was stuck with its new partner. Meanwhile, back in n 1925 Ford built an antiaircraft gun director, the Mark 19, which was a single purpose HA director and pretty much the contemporary British HACS style system, evidently heavy enough that it was really only suitable for larger ships. By 7 December 1941, “42 of these devices had been installed in the fleet” (italics mine, 61).

 “Prewar naval antiaircraft fire control culminated with the Mark 37 director, introduced in 1939 and sate of the art in 1940.” (Mindell, 61) It directed the new DP 5"/38 against both air and surface targets (High Angle/Low Angle in Admiralty parlance.) It was self-contained on its own barbette, had a 15ft stereoscopic rangefinder, and could track targets with a 400kt horizontal and 250kt vertical speed, the latter to counter dive bombers. This last was the feature that impressed Stephen Roskill's generation, and there might have been large sales had utopian production projections held up. As it was, 30 were produced by 1940, and 800 in all, not enough for the USN, never mind RN, given that destroyers typically carried 1, cruisers 2, battleships 4, and destroyer escorts, implicitly, none. (62) The good old Ford Mark I survived, albeit now promoted to a computer. The Mark 37 director followed the target, and the computer picked up its data input by following the motorised tracking, this being called the automatic rate control feature. Operators need only keep the targets centred in the director sights until the  computer converged on a solution. 

Unfortunately, this is where the piper is paid for the fairly cavalier approach to system errors implied in the hodgepodge development of the previous twenty years. The feedback loop introduced stability problems. Low stiffness is an intrinsic problem with the Selsyn. This finally became evident to the BuOrd after problems with the new 6" mounting emerged in 1937. It is not clear to me that the causes of the low damping --Sperry carelessness and corner-cutting development-- was equally well understood at the time. In any case, cumulatively these meant poor response to driving. The Mark 37 oscillated under optical input and took impractically long to converge on solutions against dive-bombing targets. But it was when radar inputs were added that the system crashed, as intrinsic noise caused complete stability breakdown, although by the time this was evident, the equipment was already in production. (66) As Friedman indicates, the single-cam mechanical predictor for the Bofors 40mm, the Mk 51, was often preferred in action. Major Kerrison's design might not give an accurate solution, but at least it generated one! The army was able to fix their Sperry AA fire controls of the same design by retrofitting smoothing circuits drawn from telephone engineering. This worked pretty well versus the buzz bombs, but was far too complex a jury rig to go to sea. Faced with the inadequacies of the Mark 37, MIT's Radiation Lab was contacted to replace the Mark 37 with the Mark 56. (Since my trashing of the Selsyn is based in part on the "Components Handbook" in the Radiation Laboratory's Radar Handbook series, take it with a small grain of salt.) This is where Wiener comes in. The Mark 56 used a mechanical computer, a GE amplidyne servo, and fired live ammo first in December 1944 (269) Radiation Lab was doing its best to get Mk 56 with its “full” automation to sea on a crash schedule. 5 prototypes were on order when Japan surrendered, 2 near completion. Production models reached the fleet in 1947, and developments remained standard in the US Navy into the 1970s. (Mindell, 270–3). 

So there we go: a revisionist junking the Mark 37. This isn't to say that the various British High Angle Control Systems worked either. In fact, they mainly didn't work, because the engineers kept saying that they couldn't deliver what the services wanted. The point here is only to say that they shouldn't be trashed for not delivering what Sperry was disingenuously claiming to provide. We need to look at them in their own terms, investigate the problems they set out to solve, and the technology that they developed in the process, and see it all converging on 21 May, 1940. 

*I see that Chris Bell has written a replacement. And I've had a book recalled! I think I'll head out to UBC and see if it's actually on the shelves. It could happen!
**,E. J. H. Douch, "The Use of Servos in the Army during the Past War,” Jour. Elec. Eng. 94, IIA (1947) (Convention on Automatic Regulation and Servo Mechanisms): 177–89] 

1 comment: