"Heath Row." I probably shouldn't dwell on it as much as I do, but there's something ineffably weird about Britain's inability to decide what to call London's main airport in the first generation of its existence. As the statistics show, it was also a very foggy place in the late 1940s and 1950s, due to all of that low-quality coal being burnt in power generation facilities which really ought to have been retired, but weren't, as electrical demand was growing so quickly. If only those old-timers had grasped just how easy it would turn out to be to stifle demand and stop economic growth in its tracks!
Ah, well, we have to let bygones be bygone, and focus on the important part, which is the development of automatic landing capabilities to the point where, even if the modern Vancouverite can't afford a house, they can afford to fly to Mexico or the Caribbean and back at Christmas, and never for a second think that they might land into the runways at YVR, instead of on them.
|From Sir John Charnley, "The RAE Contribution to All-Weather Landing." [msword document]|
Along the way, we'll learn a bit more about the coming of the transistor era; and, specifically, why it happened in America. Well, okay, we don't need to learn that. "Military Industrial Complex" and all of that.
The research question of the day was, "What happened to the Smith's Electric Pilot?" It turns out that the SEP1 [Seldom Ever Performs Once] was the RAF Mark IX Autopilot in civilian dress, latest in a series of rapidly iterated electro-mechanical autopilots going back before 1925, when the RAF was fiddling with an "unmanned aerial torpedo," as they said back in the day.
To call this "obscure" is not to understate the case. The very brief discussion at Wikipedia seems devoted to the figure of Archibald Montgomery Low, a colourful member of our modern breed of Super-Geniuses,
who was certainly involved in the Larynx, but wouldn't have understood the mathematics of automatic control if they bit him in his self-promoting ass. In reality, the Larynx tested the latest edition of the RAE Pilot's Assister, a pneumatic device with a takeoff from a pitot tube and a gyroscope for directional control, along with an assortment of emergency stops for when the system inevitably began hunting. Here's The Aeroplane for 15 January 1930, complete with C. G. Grey blaming his rag's sorry journalism on the Secret List. From the sound of the aircraft fitted, it was a fairly bulky equipment, although not so much so as an actual pilot. That said, the idea of putting a high explosive warhead on the Larynx and launching it into the wild blue yonder seems . . . futuristic.
With that bit of detective work out of the way, I can get back to F. W. Meredith, an RAE "junior" then involved in the quite separate task of getting aircraft on the ground when it was too foggy to see the ground. Early experiments (pre-Meredith) involved a lead dangling from the tail skid, with a lever and dashpot at the top to adjust the tailplanes, an arrangement that seems to have sufficed when landing an SE5 on a typical airfield of the day. The introduction of the dangerous and limiting concept of the "runway," however, dashed this promising approach, as also aircraft with approach velocities much above walking speed. Meredith, a maths-talking lad then employed by the mysterious boffins of the "aircraft stability branch," contributed an arrangement that basically involved depending on a Vickers Virginia's dependable, slow, porpoising to put it on the ground automatically at a measured interval after the ground-detecting line detected the ground.
Extrapolating wildly from Charnley's anecdotal, oral history of Old Days at Farnborough, it seems as though Meredith had found his niche, and might have been thrown in with the crowd working on the Pilot's Assister, a crowd that included, I now discover, several other, even more obscure scientists, beavering away on the theory of automatic control within RAE, and the practical problems of applying it to autopilots. Honestly, I didn't know half this stuff, in spite of reading, like, all three of the monograph histories of automatic control I could find, a decade or more ago.
This is probably as good a point as any to introduce a theoretical interlude, which is going to look like shit, because I've never learned to use Latex, and at my age, don't propose to do so. The basic principle of automatic control is amazingly simple: the thing is deviating from the direction it is supposed to be going in, so you push it back with a force proportional to the amount of deviation; but in mathematical terms you get -d2/dt(x)=x, which is one of those dreaded partial differential equations that get thrown at you in third year. Fortunately, this one has a solution, which I typed with the Equations editor of Microsoft Word, unlike the dog's breakfast above, on account of I am going to work a night shift in an hour (bleeah) and can't be arsed, which is just very well, because Blogger won't let me cut and paste that stuff: F(t)=Asin(wT +B) seems like a passable way of writing it. Anyway, it's a trig function, and a harmonic. To visualise it, just rotate your thumb in a flat circle in front of your eyes, and compare it (in your imagination, if necessary) to a mass oscillating on a spring. They're both ways of modelling this equation, and equivalent to the control problem. It's one of those freaky, illuminating moments when mathematics really does seem to open a door into a higher level understanding of the universe, and it's got all of the good stuff, too: Music theory, celestial spheres, Pythagoras, even Mind (if we accept that that is what automation is about.)
|I don't, particularly|
Reality is rarely so kind as to supply us with a simple case of Simple Harmonic Motion, or music would be no fun at all. The simple case of a force proportional to, and inverse of, displacement, may be complicated by a residual velocity or velocity-dependent friction term (a "displacement component" and "damping force" in the simple, happy, pre-vector-everything days of the mid-Twentieth-Century papers I am frequently summarising here). Even a delay in the system, so that the restoring force is delivered at a predictably wrong time, a "lag," renders the equation unsolvable by analytical methods and forces the insertion of what Meredith would call, in that 1949 paper, a "mechanical differentiator or integrator," Meredith preferring the latter on score of design virtues. Nowadays, of course, we just use our magic computers. From here we can move on to all the lovely features of beat and harmonics that make music theory so interesting in the eyes of those who do it.
Or, instead, we can move on to the second aspect of the theoretical problem. Control theory begins gets complicated because we are talking about an intrinsically stable aircraft flying along, and getting upset in some manner of linear displacement or oscillation about any or all of three axes of freedom, leading the autopilot to apply a correctional force to the various rudders, flaps and tailplanes involved, resulting in the lags and dampening effects and displacements mentioned above. But taking a step back from this, the method of detecting those rogue motions is a challenge.
Consider a pencil, which you balance on its end on a table. Very carefully not thinking about Heath Ledger, you flick it on the tip, and it falls over. Now take the same pencil, and balance it on its end, and then give it a sharp twist with your palms, so that it spins rapidly in place. Congratulations, you have great manual dexterity. Now, flick the tip. The pencil does not fall over. It tips a bit, and continues to spin, even though it is, to all appearances, defying gravity to do so. In reality, the defying bit is covered off by a wobble of the tip of the pencil, which will probably be slowly rotating around the vertical axis it formerly conformed to. The reason for this is that, as Newton tells us, momentum is conserved, and "spin" is a form of momentum. The pencil will tip over on its side, but only if you flick it hard enough to remove the component of angular momentum that makes it point upright to begin with.
It's all a bit mysterious, and it is more usually demonstrated with a child's toy, a top.
It gets a lot more mysterious when you have to explain the wobble, or precession, which has driven great physicists like Ernest Mach MAD. It's even more arcane when applied to aircraft control theory, because you have to throw out the first principle, and accept that your gyroscope is going to be power-driven, and that the important fact isn't its remarkable stability, but the force with which the tip of the pencil pushes back when you push it with your finger. The power-driving keeps the gyroscope pointing in the right direction, so that it does indicate the right direction, and the "pickoff" --I hope I'm using the word right-- uses the force to measure the amount of deflection, and tell the motors of the autopilot how much to push rudders, planes and flaps.
MAGIC! Also, the whole to-do with lag, dampening and displacements comes into effect here, in all three axes and in periodic and non-periodic motion, which means that you need a corrective on the gyroscope, which, as I understand these things, is usually one or more additional gyroscopes. F. W. Meredith makes a lot of to-do about the correctional gyroscopes, since the method eventually developed for the family of British service autopilots employs the rate, or second-order derivative of the extraneous displacement, (or first order derivative of the rate of displacement --vectors really do make this simpler, damn them) to correct the directional gyroscope, whereas proprietary Sperry technology uses the displacement component. Displacement-rate versus rate-rate, American versus British, mathematically jejune versus mathematically awesome. In the RAE formulation. By the way, don't ask me why Honeywell has dropped out of the conversation in 1948. Question for another day, I guess.
Speaking of theory in the prewar era, the discussion here comes back to where I started in on control engineering, the 1937 unsigned The Engineer series that ran through most of the spring. I'd lilnk or something, but you know where it is (Grace's Guide), and you have to pay for pdf downloads. I think the series may have begun in February, but it definitely ran through March and the first half of April, and the Leading article discussing it appears, in typical The Engineer ass-backwards way, in mid-April. As I say, I'd give you a more precise date, but it would involve using up one of my paid downloads. Or two, if I decided to fart around with The Engineer's surprisingly useless index.
The extremely sparse historiography of control engineering assigns authorship to Nicholas Minorsky, but it is very hard to believe that The Engineer would deny authorial credit to a Russian-American consulting engineer, or that Minorsky would put up with it. In fact, it would be a pretty unprecedented violation of publishing norms, compared with the alternative possibility that it was written by a British civil servant, in which case an unsigned (non-leading) article is precisely in line with publishing norms.
I probably shouldn't make such a big deal of it. As far as I can tell, the attribution derives from one of Minorsky's executors discovering a copy of the entry (actually, entries) in this series that describes Minorsky's 1923 autopilot, installed in the turboelectric battleship New Mexico, in Minorsky's files. Practically unsuccessful but tremendously influential, Minorsky's device gets two installments in a twelve-part series that begins with temperature control but mostly concerns itself with various autopilot designs. The only really important issue is that the very thin historiography appears unaware of the RAE working group, and this is a good example of the way in which history of science is written by the best publicists. It also edits out an important contribution made by state-funded research in national security-adjacent laboratories. I keep arguing that the national security state played a more important role in the development of modern technological society than is generally acknowledged. Well, here's an example.
So what do autopilots have to do with blind landing? Essentially, the problem of blind landing was solved by automating landing entirely. This seems strange, because it adds a step. By itself, the blind landing problem seems obvious enough: You point a beam at the plane, the plane follows the beam in. At most, short of a perfect "glide path," the plane needs to know its altitude, and the magic of the radio altimeter, if not radar, can tell you that. Or, if not, a ground controller at a radar can tell you what you need to know. Instrument Landing System or Ground Controlled Approach, either or both.
In practice, the problem of combining automation with "following the beams" goes back well before 1948. One of the pleasures of my current very slow read-through of James Goodchild's revisionist account of R.V. Jones' "Most Secret War" has been a very lucid discussion of X-Gerat, the Lorenz blind-landing system-based bombing aid that, rather ambitiously, including an automatic bomb-dropping mechanism activated when the bomber flew over the target beam. I'm not really seeing the gain in accuracy this provides, but no-one has expressed any interest in my criticisms of excessive faith in automation yet, and people from 85 years ago are unlikely to be the ones that start.
This gives us a window into the "blind approach" community's thinking, way back in the 1930s when the Luftwaffe was adapting the Lorenz system into a blind bombing aid, but Meredith puts it more clearly in his 1949 Royal Aeronautical Society talk, summarised here, in Flight. It is a humbling experience, he points out, to see the difference between trying to stay on beam as you land your plane, and watching the autopilot conform to the beam as it lands. The pilot is constantly losing the signal; the autopilot gracefully, surely, moves onto the beam with every incidental motion and mishap.
The rest of the technical story of blind landing is simply told. The SEP1, or RAF Autopilot Mk 9, was succeeded by the SEP2, or Mk10, and then by the Smiths Military System, which had provision for automatically feeding it with localiser beam information. Impressed by ongoing trials by the Blind Landing Experimental Unit, the Air Ministry authorised the implementing of automatic landing on the SMS-equipped Avro Vulcan (the prospect of landing with atomic bombs on board focusses attention on important things.) The first regular service automatic, blind landing by a commercial aircraft followed ten years later, on 10 July 1965. The Wikipedia article is quite interesting. I had no idea that the "three votes" made famous by NASA originated in the SMS, with the "votes" being taken with a torque analyser! It's also less than fully reassuring that modern airlines are stepping back from full autoland, essentially because it is too expensive to equip North American secondary airports with it.
The image of three feeds into SMS autopilots being "voted" by a measurement of torque in three driveshafts, however, tells us that there is more to the autopilot story.
Of course, it's not.
It can't be. The magic of the gyroscope is fundamentally mechanical. When their axis of rotation is pushed out of alignment with one of the three axes of an aircraft, they respond by exerting mechanical force. Some very small modern autopilots seem to use only accelerometers, and I suppose that an accelerometer could be entirely electronic, However, the Smith's gyros were tiny little electrical motors, spinning at 12,000rpm, so that the pickoff that detected their movement could be an eddy current damper, reducing the costs of manufacture, the maintenance burden. As well, the use of electrical current at all stages of the mechanism directly couples the toque shafts and servomotors to the gyros.
However, the very small magnitude of the current signal from the gyros means that it must be electrically amplified to run the servomotors that control aircraft systems. Depending on whether we are reading corporate brochures disguised as Flight articles, or listening to Meredith's disgusted comments, we can either be impressed with the use of three magnetic amplifiers linked to the three gyros,or upset that British industry can't deliver a sufficiently robust and reliable amplifier "valve."
Finally, I would be remiss not to point out that the "Pilot's Controller," in the picture above, although normally used to reset the autopilot without turning it off and then back on again, is a rudimentary fly-by-wire system. The actual aircraft systems that operate the rudder, planes and flaps may not normally be electrically actuated as yet, but it is a fundamentally electronic interface through which the aircraft can be flown. Electrical "feel" is a long way away, but the problem presents itself directly, especially in the last stages of a blind landing, when the autopilot is in control, but the pilot may have to intervene in unforeseen circumstances.
Or not quite finally, because, having already made the point, last week, that it was the search for solid-state amplifier that led Bell Labs to the first transistor, and patent-proofing that invention that led to the development or refinement of semiconductor theory, I'm going to remind everyone of that, here. It is particularly interesting that Meredith gave this talk a year after the transistor was announced. It's very much a clarion call for a British transistor, and I expect to see movement on that front as we come to better understand the failure of the British IT-industrial revolution. At the moment, I'm leaning towards dumping responsibility on the decision to cancel the TSR2 and its low-altitude, high-speed control system, but then when I follow up, I learn about the Elliott DEXAN system, the subject, amongst many others, of Simon Lavington, Moving Targets: Elliott-Automation and the Dawn of the Computer Age in Britain, 1947 67, a very interesting-sounding book that unfortunately fell into the hands of Springer, and is available on Kindle for a mere hundred pounds, six pounds less than the paperbook edition. On the bright side, UBC has collected it, so at some point in the near future I'll have a look.
If you notice a "Blog Comment Follow-Up" label, this is a response to Alex's long-ago challenge to have this series do more with early electronics-age journalism. I'm getting there!