Backroom Boys

by Francis Spufford

  To many people on the outside, GSM looked like a catastrophe in the making. A worried European Commission hired PA Consulting in London to write an assessment. PA’s man for cellular was Garry Garrard. ‘I started off as sceptical as everybody,’ he told me. ‘They were just going along with it, they sort of thought they had to. I thought, “This is going to be a camel” – you know, a horse designed by a committee. We looked at everything – is it going to work? Will it give the increase in capacity? Can it be done at a reasonable price with foreseeable technology? Does anyone want it? And, surprisingly, at the end of it, we said, “Well – yeah! It’s got a pretty good chance.” We had this great big meeting in Brussels, all the MDs of all the manufacturers were there, and they suddenly clicked about it; instead of just being dragged, they were then keen.’

  The camel was a racing camel. Instead of being overspecified, GSM turned out to be a remarkably prescient basis for the next decade’s worth of technical development. Instead of making it clumsy, the array of bright ideas made it future-proof. And because all the European operators and all the European manufacturers and all the European regulators signed up to it, GSM created something brilliantly simple that didn’t exist for mobile phones anywhere else in the world: a continent-sized market, unified around a single standard. A person in Europe with a GSM phone could use it anywhere. Wherever you were, the phone found the network; wherever you were, your phone number found you. Equipment of glorious uniformity made sure that all the networks talked to each other, and billing systems of equally glorious uniformity ensured that every participant involved in handling the calls got paid. Which meant that an entrepreneurial company could exploit the uniformity to build an empire as big as GSM’s domain, first in Europe, then beyond, as GSM became the norm almost everywhere. GSM did for European manufacturers like Nokia and European operators like Vodafone just what Windows did for Microsoft. Competitive instincts and investment capital weren’t enough on their own. There needed to be an unexpectedly brilliant piece of industrial policy on top; the gift Thatcherite Britain could give, plus the gift of dirigiste Europe.

  And crucially, in this technology market, there wasn’t an American challenger, because in America the golden combination wasn’t available. Ideology forbade it. Investment finance was more easily available in the US than anywhere else on the planet, but there was no mechanism for agreeing a standard. The federal authorities had issued licences for 733 separate operating territories. Some had gone to the big existing phone companies. Others had been handed out in a lottery and had gone to tiny syndicates put together by hopeful dentists. Hundreds of intensely localised mobile operators were therefore slugging it out for custom in 733 distinct cities, towns and counties, using a bewildering variety of different systems. Mobile-phone usage grew fast, then stalled as customers discovered how difficult and expensive it was to work their phones away from home. Suppose you were a subscriber of Lansing Cellular Telephone Inc. of Lansing, Michigan: an old steel town, population 490,000, blue-collar setting for Roseanne. Your phone would function OK in Lansing, but if you drove out of town, the moment you passed the last boundary of LCT’s last cell, you were dealing with a map of territories as complicated and madly unpredictable as the medieval Holy Roman Empire, in only some of which did LCT have (very expensive) roaming agreements. In the States the mobile phone never became a dependable, universal instrument. By the mid-1990s, the continent of North America was a poor, benighted place for mobiles, compared to Europe. Rationality had to come the slow way stipulated by pure Darwinian capitalism. The companies ate each other until only a few middle-sized survivors were left. Unfortunately for them, by that time there was a British giant leaning on the edge of the blood-stained playpen, saying, Hello, little fellow.

  So Vodafone’s way was clear. It could fight with Orange for the unconnected 98 per cent, it could try for Europe, it could try for the world. But before any of that could happen, it had to upgrade the core of technology on which all the big dreams depended. It had to do something about its radio planning.

  *

  The empirical model had performed honourably. It had provided a good-enough fit to the landscape. It had used spectrum efficiently enough for Vodafone to run at a very healthy profit. It had kept the customers happy, aided perhaps by the uncritical delight they’d felt early on at being able to phone from the fast lane of the M6 at all, even if the reception was a bit patchy in places. But now it was reaching its limits. Vodafone was contemplating a huge expansion in capacity. The only way to do that – remembering the basic principle by which cellular systems reuse a limited pool of frequencies over and over again – was to divide up the existing cells into smaller ones, so the frequencies could be reused more often. The original cells on the analogue network had already been split wherever the traffic increased too much. Each of the sectors of the sectorised cells in London had become independent cells, and then been split again, sometimes several times over, as demand increased. Now it was going to go much, much further. Cells half the size of a county were going to be fissured into masses of cells only a couple of miles wide. In suburbs and cities, some cells were going to shrink to the point where they only enclosed one street.

  The problem was, the smaller the cell, the less well the empirical model worked. The smaller the area you were looking at, the less average was the way radio waves travelled through it. Particular characteristics of a place, like an oddly shaped building or a very dense Leylandii hedge on the brow of a hill, could produce some very distinctive local effects. Vodafone’s empirical prediction system PACE had smoothed these all out statistically. It assigned 500 m squares to categories – field, suburb, town – then looked to see how the average field or suburb or town behaved. Which was fine, so long as you were still dealing with large cells. But you couldn’t zoom in. If you tried, it was like getting too close to a photograph. The image dissolved into uninformative greyscale blobs. This couldn’t be tweaked. It followed from the fundamental architecture of information in PACE, which was top-down rather than bottom-up (and not just because a lot of aerial photography went into it). It mapped down its categories onto localities, rather than building up local characteristics into a map. That needed to change. And then there were other consequences of having a multitude of small cells, such as the dwindling of the geographical distance between any cell and the next cell in the patchwork where the same frequencies were in use. At least one other cell would always be in between: that was a given of radio planning. Now, though, there were going to be callers potentially trying to use the same frequency for different calls when they were only five or six streets apart. PACE had never been especially good at coping with ‘co-channel’ interference. It would be a permanent, crackly presence in the life of Vodafone’s mass-market customers, unless a much more exact method for planning cell boundaries could be put in place. Finally, a new design was being proposed for very small cells, the ‘microcell’, in which the base-station antenna would be below the level of the cluttered urban scene, about lamppost high, rather than perched above it, tower-block high. In microcells, the signal would propagate to handsets by different paths, paths which empiricists like Okumura had never even thought about.

  For all these reasons, it was time to get away from the empirical model and engage directly with the properties of the environment. It was time to get physical.

  Vodafone needed a specialist. Around Britain, academics were increasingly studying propagation problems, usually in university departments of electrical engineering. Up in Liverpool, Professor Parsons and Dr Ibrahim had devised a ‘semi-empirical’ method which scored squares of ground in cities according to the three factors of Height, Urbanisation and Land Use. Down in Southampton, Professor Steele was exploring the theory of the microcell and diagramming the type of microcell that might be able to take over the task of coverage along motorways. As well as the vans sent out by the networks, there were Ford Transits on the road carrying university experiments too. Researchers drove
carefully in loops around London, took microscopic measurements at all the junctions along Harley Street, chugged across the green face of the fells above Keswick. Egg sandwiches were removed from brown paper bags. Pipes were lit. Ice lollies were purchased. But none of this work was exactly adjacent to what the networks were doing just then. It didn’t answer the pressing questions: it wasn’t locked into the pragmatic business of getting from the present to the future without ever ceasing to have a system that worked. Vodafone wasn’t looking for someone too immersed in pure theory. But as it happened there was another small group of radio engineers in Britain, pragmatically committed to making a system work, whose expertise predated the birth of the mobile phone by decades. Cellphone signals are broadcast on 900 MHz. Nearby, in the same UHF band, the slightly longer waves of the British analogue TV signal had been spreading across the landscape since 1962, facing the same problems of refraction through the atmosphere, diffraction over buildings, reflection off hillsides and tower blocks, and scattering through the branches of trees. The signal for The Clangers and Nationwide navigated the same invisible ocean. There were no cells in TV broadcasting, but the relay masts for the signal had to be placed with the same attention to the environment, and the BBC had trained a cadre of learned in-house engineers to tend and extend the system. The joke had it that the BBC’s engineers were the people to ask about the propagation of anything – even daylight. Three of them had just published a book together: Masts, Antennas, and Service Planning. Vodafone invited the third of them, the author of the ‘service planning’ section, for an interview.

  Dr John Causebrook lived and breathed radio propagation, and quite possibly dreamed it too. It had been his focus his whole working life, ever since the RAF taught him the radio basics, by the numbers, during his National Service. The modification to his house that declares to his neighbours, just outside Winchester, that an engineer resides within, is a custom TV aerial, self-built, aligned exactly on the Isle of Wight transmitter, thirty-five miles or so over the horizon to the south-west. He gets very good reception. Like other engineers, he prefers to have things just so, intellectually as well as technologically. ‘I like clean definitions,’ he told me when I visited him (he’s just retired). ‘I’m a definitions man, really, I like to have each item that I’m dealing with well defined.’ He was the originator of the Causebrook Correction, a factor to be applied when calculating the path of radio waves over multiple diffracting edges. But he also possessed the indefinable sense of judgement about what will and won’t work that develops from long absorption in calculation and can sometimes, weirdly, outrun it, throwing up a conviction of the right answer long before the reasons for the answer arrive. ‘A bit of a black art,’ Dave Targett calls radio planning. John Causebrook had spent a lot of hours standing on steep hillsides in Snowdonia and the Scottish Highlands, trying to see where to site the booster mast that would let a remote community get their TV, and visualising the UHF signal as it came gliding up the valley, silvery wavefronts recoiling differently from the differently shaped lens of each hillside, and overlapping and interfering with each other, and coming clearest about … there. He was a tidy-minded radio necromancer. Over his thirty years in television, he had steadfastly done his job as it had to be done, given the constraints on the BBC’s budget and the limits of contemporary computer technology, all the while building up in his mind an alternative vision of how propagation work ought to be done, if only it were realistically possible to dispense with the empirical fudging. When Vodafone called him, he was ready to make the jump. He wasn’t attached to TV in itself; it made no difference to him that Vodafone’s network transmitted phone calls instead. ‘With the sort of things I do,’ he told me, ‘the science doesn’t care what’s being propagated …’ The Newbury he arrived in, the day of the interview, was already well on the way to being a Vodafone company town, as much devoted to serving the Big Red Comma as nineteenth-century Swindon had been to serving Brunel’s Great Western. The logo was splashed on the side of building after building; it was engraved five metres high into the shiny grey granite backside of the main stand at Newbury racetrack. They sat John Causebrook down in the boardroom and looked at him: not a bright spark, not a young man at a cutting edge; an expert in an unglamorous field, whose specialism had suddenly taken on urgent commercial life. What would you do, they asked, if we put you in charge of radio engineering? Well – this, he said, opening up his book. In six concise pages, he had laid out a toolkit of his favoured concepts. The theoretical literature of radio propagation is huge. This was a personal selection from it, honed for use; in effect, a kind of technical manifesto. In the book, the discussion of each technique included a stoical acknowledgement of the practical obstacles there were to using it. The words ‘difficult’ and ‘costly’ appeared several times. But as he knew very well, quietly developing his argument in the boardroom at Newbury, to Vodafone ‘difficult’ and ‘costly’ were not at all the same thing as ‘impossible’. ‘Can I see that?’ asked the Director of Engineering, coming round the end of the table. They gave him the job.

  *

  To construct a physical model, you begin by imagining a completely blank piece of space, like the holodeck in Star Trek before anything is projected on it. There’s nothing but empty air. If you put a transmitter in the middle of this zone of artificial blankness, the strength of the signal you receive from it will diminish with beautiful regularity the further away you go, just because the radio waves are gradually dispersing their energy into a larger and larger volume of air. It’s easy to predict how strong the signal is at any given point in the space. All you need to know is how far away from the transmitter you are. Then the nice smooth downward curve on the graph tells you how much of the original power of the signal is left at your point. This is the ‘free-space field strength’. It’s a real figure, determined by the real phenomenon of wave energy dissipating. If any piece of space in the real world were as vacant as this laboratory thought-experiment of a space, the free-space field strength would be all you needed to make perfect propagation predictions. But no real space is blank. So at any one point you want to know about in the real world, the figure for the free-space field strength is modified by all the various things that have happened to the radio waves as they made their way from the transmitter.

  Put the real world back into place now. Back comes volume, back comes texture, back comes density. Fill out the blankness with a section of genuine, three-dimensional landscape, centred around the location of the transmitter, on which we gaze down from the viewpoint of a god or a radio planner. The point at which we want to predict the signal strength is now a location upon that crowded, folded, complicated surface. Draw a line from it to the transmitter. It isn’t a taut geometrical line. It runs straight, but it’s a piece of limp string. It goes up when the ground goes up, it goes down when the ground goes down. It goes up the side of any house standing in the way, over the roof and down the opposite wall, before arrowing limply away through the shrubbery. Now swing around and look at the completed line from sideways on. It’s become a two-dimensional slice through the landscape, a flat silhouette of the ground along that one line between the point you’re interested in and the transmitter. This is called a ‘profile’. It’s what the radio waves will be contending with as they travel out from the transmitter. It’s what you will be contending with when you predict the signal strength at your chosen point.

  How are you to calculate what the profile does to the signal? Well, in theory, you could take Maxwell’s equations, which govern wave movement in all circumstances, and apply them, step by step, to every single part of the profile. It can be done; John Cause-brook has in fact done it, once or twice, in the interests of experiment. But solving Maxwell’s equations over a whole profile produces mathematical expressions so mountainously vast and repetitive that a computer asked to crunch them takes days to come back with an answer. (‘They’re tremendously time-consuming, you see,’ he told me. ‘Even now?’ I aske
d, thinking of the difference between the computing power available in the early 1990s and machines today, seven iterations of Moore’s Law later. ‘Even now! Most decidedly!’) It isn’t a viable option. The point of the exercise, after all, is not just to come up with a method that will allow you to calculate the effect of one profile or a few profiles. You’re not making a prediction now for one fixed link between a transmitter and a receiver. There are set-ups in which a dish beams a steady stream of microwaved data to another dish on a distant hill, and there it makes sense to pay exhaustive mathematical attention to the single relevant profile of the ground between. But you’re planning a cellphone system. You need to be able to make an adequate prediction for every point in the landscape that surrounds the transmitter – to crack every possible profile raying out from the transmitter – because the caller with the phone could be standing anywhere.