31 October, 2002: White heat

[ Home page | Web log ]

Last weekend, I went to the Cambridge Museum of Technology. This was great fun and is highly recommended. Until 1968, the museum was the sewage pumping station for Cambridge; sewage was pumped from the city through a great big pipe up to the sewage works at Milton. Originally this was done using two enormous steam engines, which ran continuously from 1894 until 1968. These, naturally, are all finished out in red paint and polished brass, and are very pretty. Distributing steam to the two cylinders of the engine at the proper points in the cycle requires an insanely complex contraption involving numerous little pipes, valves, and two additional small steam engines. (Later, to cope with surges of runoff water during wet weather, internal combustion engines powered by gas from the next-door gasworks were installed. These are also pretty crazy.)

We should all be grateful that the people who came up with this stuff did not -- as any sensible person would -- throw up their hands in despair and say ``Sod this, let's just wait until somebody invents a microprocessor so that we can use that to control the engine.''

In fact, that seems to have been the antithesis of the attitude: wanting an electrically-lit engine room but lacking a connection to any mains electricity supply the people in charge of the pumping station installed an electric generator and a little steam engine to drive it. They got water from a well using yet another steam engine. ``If the only tool you have is a hammer, everything begins to look like a nail.''

As originally envisaged, the entire thing was run by steam produced from burning household rubbish, which was delivered by horse cart from around the city. This was fed into the gloriously-named `Destructor' furnaces, which are visible in a partially-deconstructed state in the museum. When they first started, household refuse apparently contained a lot of partially-burned coal from house fires, which made it an economical fuel; as time went on, households burned less coal and acquired more efficient hearths, so that the calorific value of the refuse dropped; simultaneously, labour costs rose, and since it was necessary to separate glass from the other rubbish -- to prevent it from melting and clogging up the furnaces -- this eventually made the `Destructors' uneconomical; from 1923 coke from the adjacent gasworks was burned, and the coke boiler is the one which operates the engines now.

The boiler -- which, let's face it, is basically a big tube with a fire underneath it -- does incorporate one really excellent invention, which is a water level gauge working on a simple optical principle. A transparent vertical pipe connects to the boiler at top and bottom in view of the stokers, and behind it is placed a pattern of diagonal black-and-white stripes. When the tube is empty, the stripes show through as normal; when it is full of water, it acts as a lens, and an inverted image of the stripes is seen. So the level of water in the boiler can be seen instantly by looking at the direction of the stripes behind the tube. Isn't that clever?

In another bit of the complex there's a collection of various ages of printing equipment, which is fascinating. Among numerous other clever devices was a machine invented by a German gentleman named Dr. Hell for making halftone printing plates from photographs. Rather than using a chemical procedure, this worked by placing the photograph underneath a sheet of metal which was drilled out to form the plate; the machine drilled a hole, measured the brightness of the bit of the picture underneath it, and expanded the hole accordingly. This thing dated from the 1950s and apparently took only about 20 minutes to make a plate from a photo.... Equally scary was a plate of type made up for a mathematical journal; a real labour of love in those dark, far-off days before TeX.

For some slightly more recent technology, yesterday I went to a talk by Peter Duffett-Smith of Cambridge Positioning Systems. These people make `technology' -- meaning, of course, `computer software' -- for determining where mobile 'phones are. It grew out of a project called `CURSOR' which used medium-wave radio signals to infer location data; the game here is that you build a radio interferometer, one end of which is at a known location, and try to find out the location of the other end. By comparing the phase information measured at the fixed and mobile receivers, a position can be obtained; we were shown a demo of this in which eight radio stations over about a 200km baseline were used to determine the position of a car to within about five meters. Apparently the British Airports Authority were interested in using this technology to track aeroplanes taxiing around Heathrow and other large airports. Duffett-Smith was, he said, rather relieved that GPS filled that niche instead.

The game with positioning mobile 'phones is sort-of-interferometry. Every 50 milliseconds, each base station transmits an `extended training signal' which a 'phone will lock on in order to synchronise its own transmissions. This apparently gives a spatial (radial) accuracy of about 50 meters, though the clocks in the base stations are not synchronised, so there's a timing uncertainty as well. So if you measure the timing differences between a fixed and a mobile receiver for three base stations -- whose positions are, of course, known -- you get to figure out your position. This is one of the technologies being used for the `E911' emergency-call location service being implemented by federal mandate in the United States. This mandate is apparently the source of most of CPS's income, since telecoms companies aren't exactly swimming in money any more. Adapting handsets to talk to the CURSOR system requires, apparently, about 2kB of code running on the 'phone; the remainder of the system is made up of a large collection of acronyms which, I think, basically constitute `a big computer and some software'. Since data have to be sent from the handset being located and all the relevant fixed `location measurement units', the whole location procedure takes a few seconds to do, although the measurements are being made twenty times per second or whatever.

The competing technologies for the E911 stuff are all based either on GPS -- a bit of a bummer if you're indoors when you want to report an emergency -- or on measuring the transmissions of the phone at a variety of fixed receivers, which has serious scalability problems. (Naturally, one shouldn't believe this stuff coming straight from the mouth of the company against which they're competing, but these limitations sound plausible.)

They have another technology which doesn't require the fixed stations at all. The game with this is that with two mobile receivers which can each see five base stations, you can infer the position of both receivers. With a bunch of receivers and base stations, you can continuously update a model of the positions of all of the handsets. We were shown a demonstration in which a hundred people had wandered around Cambridge with CURSOR-enabled 'phones; 95% of them were located to within 50 meters of their true positions.

Now, this is all very clever, but ever so slightly scary. Admittedly, the 'phone company already knows where I am to within one cell, but cells can be quite big; in particular, a lot bigger than 50 meters across. And, of course, in the event of an emergency, it would be nice if the emergency services could locate me to high accuracy. But a requirement of the no-fixed-stations scheme is that a large fraction of 'phones in use are always transmitting the timing data which can be used to determine their location; otherwise, there wouldn't be enough data to locate an emergency caller. So in order for this to work, the positions of all the 'phones in any given region will be recorded at all times. Will the data be logged? Who knows. Let's hope that questions of this sort are made by some higher authority than a system administrator concerned about disk space.... (Another thing to wonder about is what happens in a remote region when there are very few 'phones -- perhaps only one -- in a given cell; presumably in this case you're out of luck when it comes to location data....)

Copyright (c) 2002 Chris Lightfoot; available under a Creative Commons License.