Backroom Boys

by Francis Spufford

  So physical models reduce the computational load by stylising the landscape. Instead of dealing with it in all its fractal complexity, they search for geometrical equivalents to its most decisive features. A physical model is still in the business of constructing a real description of a place, only it frames the description in terms of more unrealistic yet more easily-computable elements, like spheres and cylinders and wedges. The difference between a good physical model and a bad one lies in the elegance and subtlety of the stylisation. By selecting the right features to attend to, a good one gets the maximum possible purchase on the reality of the landscape, at an acceptable cost in computability. What John Causebrook brought to Vodafone was a plan for a model that examined three different landscape effects in turn and layered them together to predict the final signal. As he had written in his book, ‘it does not seem possible to have a calculation technique which intrinsically integrates these different mechanisms. However, it is possible to make calculations based upon a variety of techniques and then use a combination method.’

  First he dealt with the bending (diffraction) of the signal as it collided with the rise and fall of the ground along the length of the profile. He did this by thinking of the space between the highest points along the profile as a succession of triangular wedges, triangles being a lot easier to process than the twiddliness of the profile itself. The Causebrook Correction came in, modestly unnamed by its inventor, where there were a lot of peaks in the profile. When the profile was especially smooth or rounded, he added a tincture of the mathematics of spherical diffraction, worked out just before the Second World War. It had been vital for keeping up radio communication with bombers on long missions that took them away around the curve of the earth. Now it would help people to make mobile phone calls from sailing boats. ‘I saw my mission as bringing academic thinking into the operational world,’ he told me.

  Then he dealt with the loss of signal along the profile caused by what covered the ground. Remember the limp string? Pull it tight, so it leaves the ground and stretches in a line from hilltop to hilltop, passing through all the objects in between that radio waves can more or less penetrate, like blocks of flats and Forestry Commission plantations. Causebrook proposed looking at a sausage-shaped zone around the string. He’d test how blocked each section of the sausage zone was by all the things it passed through. If anything caused total blockage of the signal, he’d know that that particular path made no meaningful contribution to the signal that arrived at the receiver, and would lift the string so it draped over the top of whatever-it-was and do the calculation again. A series of losses accumulated that gave him a number for the total loss caused by ground cover.

  Finally, he dealt with the surprisingly powerful impact on the signal of what happened just before it reached the mobile phone, in the very last few metres of the profile. Britain is a built-up country. It has its wildernesses, but statistically, almost everyone whose mobile rings is inside a building or between buildings or somewhere near a building. If you’re in a street when your Nokia chirps out two bars of Robbie Williams, the signal that comes over the house-top in front of you, having already survived diffraction by the terrain and blockages in its sausage zone, arrives at the handset itself from two different directions. Part of the signal comes straight down to you; part of it bounces off the building behind. These two strands of signal can reinforce each other. They can also interfere with each other, if the bounce travels a distance that puts the radio waves out of phase. How strong each part is depends on how high the buildings are and how wide the street is.

  So – putting it very, very crudely – at any particular point on the map, the predicted signal strength will be the free-space field strength, minus the loss over the terrain, minus the loss caused by obstructions, minus the effects of the final bounce. Easy! Except that to do the sum – the reassuringly computable sum – you need accurate information about ground heights and ground roughness right along the profile; and data about the objects that clutter the profile right along its length; and a measurement of street height and width at the receiver itself. And, moreover, you need this information not just for one point and for one profile, but for every conceivable point around a transmitter, at some reasonable degree of resolution; and, of course, not just for every conceivable point in one cell around one transmitter, but for every cell, large and small, in the entire network. This is the burden of moving from an empirical model to a physical one. You need even more geographical information, though now for the opposite reason. You are no longer sorting units of the landscape into categories. You are measuring units of the landscape in order to plug them into a stylised version of the laws governing the movement of radio waves. The laws embody timeless, placeless universal truths, but they don’t tell you a damn thing that’s useful unless you feed them with specific figures. Not easy. Difficult! Costly!

  *

  When John Causebrook started work in Newbury, he made prudent enquiries ‘as to what was really required of me’. His immediate boss was Dave Targett. Beyond him was the main board, now chaired by Gerald Whent’s successor and accomplice, Chris Gent. ‘They said: “When there’s a planning idea around, when there’s a planning requirement around, we want a computer system that says, this is what we ought to do, and says it now, within the time it takes to have a cup of coffee.”’ The speed with which PACE2 was supposed to work was matched by the speed with which they wanted it to be ready. Causebrook found this exciting. ‘It was a total culture change from broadcasting, which was rather staid. One which suited me fine. When you came to Vodafone, it was just go-go-go.’ You get the impression, talking to him, that he had spent his working life till then grinding along in second gear, and Vodafone finally allowed him to change up. Pretty much whatever he needed to make things happen fast, they were willing to provide. When he asked for a van with an adjustable mast on the roof, topped by a rotating video camera, there it was. When he asked for high-end networked Hewlett Packard workstations – the state of the art for number crunching in the early 1990s – there they were in their cardboard boxes a few days later, waiting to be liberated from their beds of polystyrene nodules. Above all, when he asked for data, he got it. Data from the Ordnance Survey, data from planes carrying the stereo cameras, data from on-the-spot measurements in specially awkward places, eventually data from satellites. At one point, with the Soviet Union falling, it even looked as if they might be able to get Russian spysat pictures to use. It came to nothing, but Causebrook vividly remembers what arrived when he asked the Russians for a sample of their photographic workmanship. ‘They sent the Pentagon! A very good image of the Pentagon where you could see the cars in the carpark …’

  He spent about a third of his time working on the propagation algorithms which were to be the mathematical guts of the system. The rest of the days, he was working with Trevor Gill, the engineer who’d assembled the original PACE, and a team that eventually grew to about thirty people to procure the torrents of data and to find ways to process it within the requisite one-cup-of-coffee time limit. They broke the country down into 50 m x 50 m squares and collected a ground height for each, which gave them the raw material to calculate the profiles. (This represented a hundred-fold increase in resolution compared to the original 500 m x 500 m squares. There were a hundred of the new squares inside one of the old ones.) In a separate database, they collected estimates for the height of the clutter on the ground: the trees and buildings they needed to know about in order to hoist the imaginary string. Here the squares had only shrunk to 250 m x 250 m, though they had plans to reach 50 m x 50 m eventually. To check these clutter heights, the van with the mast would go out to the sites of possible base stations and film a 360° panorama. For the last element in the model, the bounce at the handset, they compiled average values for the width of the open space a caller might be standing in, throughout the country, and the likely height of the nearest buildings. According to their figures, a caller in a city would find themselves, o
n average, in a thoroughfare 19 m wide, sided by buildings 20 m high. The village caller’s road would be 25 m wide, but the sub-post office at the edge of it only reached 8 m. Approximately. It was pretty approximate, this, pretty empirical. But then nobody is pure. Every radio-planning model reaches a point where getting more precise information costs more than the marginal extra accuracy would be worth. The trick for John Causebrook was to balance cost and complexity and computability against one another and to settle for a compromise that at least moved the model substantially on, in the direction of physical reality.

  Then they integrated all of the information together. You used PACE2, as requested from on high, by asking it what would happen if you plunked a base station in a particular spot. It replied, as promised, before you’d finished your caffeine fix. It mapped the effect of the base station in the surrounding landscape by separately calculating the signal strength, according to John Causebrook’s model, at every single pixel on the screen, considering each one as a point at which a caller could be standing, connected to the base station by a unique profile cutting through the landscape in between. You could ask it to colour the varying signal strengths in different shades, producing a red-pink-orange-yellow-green-blue cell superimposed on the Ordnance Survey contours. You could zoom out to see the effects that many planned cells and microcells of different sizes would be having upon each other and use the results as a basis for computerised frequency assignments that owed nothing to the idealised repeating patterns of early cellular theory – just as the cell outlines PACE2 drew looked nothing like hexagons. They were deformed multicoloured bubbles.

  John Causebrook’s bosses did have one criticism of the system. They never complained ‘about the amount of money I spent’, Causebrook said, but they thought PACE2 was a bit austere. It was bell-less. It was whistle-free. They’d visit other companies, where the stuff on the computers had ‘what I used to call flashing lights’, and they’d come back and tell Causebrook wistfully how pretty it had all looked. But the other companies had bought in their prediction software. ‘I should think that Cellnet over the last fifteen years went through every possible commercial planning tool,’ Dave Targett told me with cheerful disrespect. Vodafone had made a strategic decision at the very beginning to develop theirs in-house, and they stuck to it. ‘If our quality of service went down’ – it’s a term with a mathematical definition – ‘it wasn’t, “Oh my God, we must get a better planning tool.” It was, “Well, let’s find out what’s really wrong.”’

  Beneath the high gloss and the twinkly animations, the commercial packages were effectively black boxes. You couldn’t tell exactly what was going on inside them, and if they stopped producing good results, or you had a propagation brainwave, you couldn’t get under the lid to alter them. You were effectively locked out of part of the design of your own network, if you used one. Also, none of them were written by John Causebrook. These things are difficult to estimate, but the extra realism of PACE2 clearly squeezed out significant efficiencies for Vodafone. It gave them the technological edge in the first half of the 1990s. Causebrook remembers conversations with Cellnet engineers in which they denied the possibility of doing things that PACE2 told him were perfectly feasible. ‘Oh,’ he says uncomfortably, not wanting to crow over the propagationally challenged, ‘Cellnet was … well behind.’ PACE2 might be a little drab, but it was brilliant. It let Vodafone move confidently into the new world of GSM and mass mobile use. It gave Vodafone’s new small cells that extra few per cent of carrying capacity that made the difference between a rough and a smooth experience for the user, and at the same time, the difference between an adequate return and a lucrative one for the operator. Fed with different data, it could turn out network designs for all the new territories where Vodafone was winning licences. It happily crunched the landscapes of Egypt, of China, of Australia. It radio-planned the world.

  *

  But it began by mapping the propagation qualities of Britain, and it was on the British landscape that its strange gaze first fell.

  Imagine Romford. No, go on, imagine Romford; or if you can’t quite bear that, at least imagine the approach to Romford, in the north-eastern corner of London where thinning city is shading over into built-up Essex. Imagine yourself coming up out of Gants’ Hill tube station on a windy grey day to a roundabout on the Eastern Avenue. Buildings in pillared maroon-brick 1930s retail-classical form an incomplete ring. There’s a bank there; a car-alarm shop; a mobile-phone showroom (naturally). The main alignment is east. Facing that way, with the wind out of Essex in your face, you look along a crowded dual-carriageway with scrubby bushes on the central reservation, set between 1930s semis planted deliberately wide apart to accommodate the thundering traffic artery: this is old town planning, a little wooden model once in a steel and glass pre-war architectural practice, where the men wore dark wool suits and each hung a hat on the hatstand in the corner. The brow of the hill is a couple of hundred metres further east. Before it, there’s a small upgrowth of taller buildings: a mock-Tudor red-brick pub, a flat-fronted four-storey barn of an old cinema with a vaguely Italianate band of red and blue plaster foliage across it, and, just beyond the low-lying shopfront of the Chabad Lubavitch Centre, a fourteen-floor office block of the 1960s, a little scruffy, its window glass frosted here and there, as if the building were turning milky with age. Pizza-related litter blows about underfoot. Everything you are presently looking at, PACE2 also looked at, back in the early 1990s. It took no notice of the styles of the buildings, it came to no human conclusions about the layers of history that went into the making of this piece of the urban periphery, it didn’t form an opinion about what it would be like to live here. But it did take exact note of the characteristic shapes contributed by the buildings and therefore of the physical matrix that the different strands in the history of Gants’ Hill had combined to create. It pointed out that the office block near the brow of a hill was an obvious site for a base station, and then calculated the profiles raying away from it. It took Gants’ Hill down as stylised geometry, as a reef on the floor of the invisible ocean, given its particular form by processes that were none of its concern. And the radio planners, tending PACE2 as it calculated this and every other local environment in Britain, all reduced to incurious numbers, in effect became arm’s-length, completely non-judgemental connoisseurs of what Britain actually looked like as the 1990s began. Not what it thought it looked like, not what it hoped it looked like: what it actually looked like. They didn’t care why it was or how it was that the car culture had created new zones of automotive space round the outside of every town, where the knotting and looping of the fast roads enclosed smooth null acreages landscaped like golf courses and studded with car parks and Little Chefs, a dollop of South Mimms for every locality. They just registered the fact of it. They took the radio wave’s eye view.

  The big irony, of course, was that the radio planners’ intense attention to place allowed mobile-phone users to feel that place had been abolished. Once the revolution had happened that John Causebrook and his colleagues and his opposite numbers prepared for, and 50 per cent, 60 per cent, 70 per cent, 80 per cent of adults signed up, telephones floated free of houses. Your phone number became something directly connected to your body; it was no longer the address of a container that might or might not happen to have you inside it. By the late 1990s, some homeless people had mobiles. Their phone gave them the only address they had. It didn’t seem to matter so much where people were any more. All sorts of relationships smeared out across expanses of geography that would have severed the connection before. Threads between people loosened without snapping. Parents gave their teenage sons and daughters mobiles and made them call home at 11 p.m. Women sitting in broken-down cars knew that the RAC was on its way. Lovers who couldn’t often meet in the flesh kept up a near-constant murmur nonetheless. Adulterers used mobiles to deceive their spouses at greater distances. ‘Hello, dear, I’ve just reached Coventry,’ I heard a man standing in
the middle of Cambridge market say into his Nokia once. Hundreds of thousands of commuters asserted domesticity while they were still three stops from home, enduring Railtrack’s rail track. ‘I’m on the train …’ People who felt themselves slipping away out of connection, who thought that catastrophe of one kind or another was now driving them into the dark where all speech ceases, used their phones to say goodbye from the blizzard-struck mountain top or the hijacked aeroplane. People whom society feared it had lost, or was losing, were summoned forlornly back: it became a grim ritual for the police to send text messages to the mobile phones of missing children. PLS COME HOME UR MUM & DAD R WORRIED. Who knew where in the air interface that text arrived? Location had changed its meaning. In functional terms, the cellphone system was not much more decentralised than the old fixed-line system. It used the same central switching and trunking facilities, and it originated the calls from many points at once at the edge of the network. No change there. But it felt more decentralised, it felt as if the structure had changed, because it had gone fluid, it had gone into free motion. All the points where calls originated were now in unpredictable self-governing orbit around the unmoving infrastructure. The network had become an unending maypole dance. The air was filled with a cloud of witnesses.