Backroom Boys

by Francis Spufford

  Dave Targett was a young engineer in Racal’s research division. Immediately after the Christmas of 1982, mobile telephones took over his professional life and never let it go again. ‘The first thing that happened in January was that we had to work out how to start planning a cellular network. Given it was something Racal didn’t know much about, they did the standard thing: “Well, let’s give it to the Research guys, they’ll find out what we should do.” I’d just finished a project, I was a Project Leader, and I was available.’ The task was learning to plan a cellular network, because neither the original 1947 theory of cellular radio, nor the technical specs for the system they were going to install, engaged with the vital, local, dirty issue of how 900-MHz radio waves would behave inside each cell. For that you needed a discipline called ‘radio planning’, which investigated the radio characteristics of different landscapes and let you make predictions about how the signal would vary depending on where you put the base station. The radio planning had to be done from scratch for all of the particular environments where Vodafone was going to offer its service. Off-the-peg solutions wouldn’t work. Every mobile system had to be tailored to fit its surroundings, pieced together from cells that, despite Ring and Young’s beautiful analogy, weren’t really shaped like the hexagons of a honeycomb. Real cells looked more like melted Dali watches, stretched in whatever direction the radio waves found it easiest to travel across the air interface just there. Their edges ran. They slipped and slopped and flowed around. People at Racal had had a lot of experience, it was true, at making radio work for military clients from point to point across the world’s most difficult terrains. But to build that expertise into a consumer system that operated reliably and automatically was a challenge rather like the challenge of turning supersonic fighters into a supersonic passenger plane. From the moment when Dave Targett first knuckled down to the task to today, only a few people have ever been involved in Vodafone’s radio-planning department – a few tens, later a few hundred. But those were the conditions under which engineering thrived in Britain – small teams, no mass production, a bespoke ‘product’. And the radio planning was also to be Vodafone’s technological core, the basis on which it would compete with its rivals. Vodafone was not going to manufacture base stations, or handsets, or antennae, or computerised switching equipment. It was going to be a service business. It could buy in all those things. But the work of the radio planners would determine the quality of the experience that the company’s huge marketing machine then sold on to mobile users.

  What to do, what to do. Racal operated out of Reading. Gerald Whent believed that the new offshoot would perform best if it had its geographical independence, so he transported the infant Vodafone one town further along the railway and installed it in a small office building in Newbury, Berkshire, just behind a curry house. Then he sat down with the engineers, and they talked about what the Vodafone network should consist of, in its initial form; what there should be in place, when the network went live for public use at midnight on 31 December 1984, as per the terms of the licence. ‘Gerry Whent decided where we were going to cover and how many cells we were going to have. It was very Gerry,’ Dave Targett told me. ‘He liked very simple decisions, so we said, “Right, we’ll do a hundred sites” – a nice round number. And we said, “Well, how much can you do with that? We can’t do the whole UK. We’ll do London. We’d better do another city, so we’ll do Birmingham. We might as well go down to Bristol and Cardiff, and keep the Welsh happy. We’ll join it up, and do the motorways.” So, fundamentally: London, Birmingham, Bristol back to Cardiff, the M1, the M4, the M5. I think it was all planned on the map of the UK in the back of Gerry Whent’s Financial Times desk diary.’

  Did this mean that the radio planners had to go out and measure every bus shelter in four cities, and every bush along three motorways, in order to calculate how radio waves moved, thereabouts, through the ocean of the air? Not quite. There were ‘physical models’ in existence for some types of radio-planning problem that called for that kind of detailed data; but they tended not to be used by people who had one hundred cells to lay out in short order. Speed-reading through the literature, Dave Targett and his colleagues discovered a consensus among mobile-phone planners in favour of ‘empirical models’, which let you handle the complex specifics of an environment at one remove by taking one mathematical step back from it. The father of the empirical model was Professor Okumura of Japan, who proposed the first one in the late 1960s. It had been constantly reworked ever since, but the idea was still recognisably the same. It worked like this. Instead of directly investigating the mathematics of radio transmission through your chosen environment, you stuck a test antenna in the midst of it and started recording signal strengths at as many distances as possible from it, in as many directions as possible. Then you logged all your results onto a graph, a graph with signal strength going up the vertical axis and distance from the mast going along the horizontal axis. It didn’t matter about mixing up data you’d collected 1 km north of the mast with the data from 1 km south: the more the merrier. You knew the data was going to be full of noise and wild variation, since all of the different factors affecting radio waves had influenced the different bits of it to different degrees, without you asking why any particular bit of data came out the way it did. Nevertheless, once you had plotted enough points onto the graph, you could see that they formed a blob with a shape, a sort of buzzing, elongated cloud. If you then fitted a straight line through the cloud where it was thickest, you could derive an equation which worked as a sort of average mathematical description of all the radio waves’ complicated behaviours folded together. An empirical description, with no explanations sought or offered. Okumura showed that if you collected this kind of information for one specimen of each of the main types of environment you planned to operate in, the equations you got worked quite well as predictions for every other place of that type. This is what Vodafone needed, Dave Targett and his colleagues decided. Of course, they didn’t actually go and consult Japanese expertise. ‘Japan was too far away. To be quite honest, we were also a bit arrogant. We always sort of thought, “We can do this as well or better.” That’s the Racal culture. It was useful, if applied correctly.’ They decided that they’d knit their own Okumura model, substituting landscape types appropriate to Britain (Okumura’s types included ‘rice paddy’) and adding extra correction factors, as the model allowed, for the roll of the landscape and the height of buildings. They set to work.

  Using an empirical model freed you from having to calculate what your radio waves exactly did in each cell, but since the model worked by dividing up the landscape into a set of categories (for instance, ‘dense urban’, ‘surburban’, ‘wooded’, ‘open’), you still had to study the landscape where you were going to operate and decide in which category every single unit of it belonged. They split the area where the first hundred cells would be into squares, 500 m by 500 m. Problem: there wasn’t enough data on Ordnance Survey maps, even the largest scale ones – and besides, maps used conventions that made them easy for humans to understand at the cost of grossly stylising what was actually on the ground. If drawn on maps at their actual width, roads would be hair-thin, unfindable red threads; work backward from the thick red snakes represented there instead, and if you took what they told you literally, you’d be making calculations on the basis of every road being a hundred yards wide. The Ordnance Survey was only just starting to digitise its map data in more realistic form, at an extortionate price of about £800 per sheet, but it also had what it called a Trade Database for sale. So Vodafone took it, and elaborating on it, compiled their own Geographical Information System, or GIS. To get the average heights of buildings, they sent out light planes with double cameras on them that took stereo photos. At the contractors they used, skilled interpreters ran double-lenses over the pictures, assessing the heft of this church spire, that branch of Woolworths, this set of tower blocks arranged in an asterisk on the outskirts of S
windon.* The building heights went into the database and were added onto the contour data from the OS, because in this early iteration of Vodafone’s system, human structures were just treated as solid outcrops of the earth, places where the ground suddenly went up thirty metres in the rough shape of an Odeon. And each 500 m2 of aerial photograph was assigned to a category. How many houses in a field turned it into a suburb? How dense did a suburb have to be before it was a town? Does a four-storey building count as low-rise or high-rise? It was an inevitably subjective process.

  But when all the subjective assessments on all the photocopied worksheets were reduced to numbers and input into the GIS at Newbury, along with the modified contour data and an estimate of the ‘clutter’ on the ground, the radio planners got a computerised map with which they could experiment. It was called PACE: Prediction And Coverage Estimation. Suppose we put a base station there, on that rounded hilltop in Wiltshire smelling of hot wheat chaff, this August of 1983, just after the harvest has been brought in; there, where there are Iron Age fortress walls under the sheath of topsoil. What does the Okumura model say will be the signal strength in each concentric zone heading out from the hilltop? Where will the power drop to the point where the cell effectively ends? What difference does it make if we increase the tower height of the transmitter? Or drop it? Or move the whole thing ten metres north or twenty metres east? Does that nudge the north-eastern boundary of our sloppy Dali-watch of a cell so it just has the reach to push over this ridge, here, and ooze down the other side? And what does putting the base station here imply about where we have to put the adjoining base stations? Does the electromagnetic jigsaw work? Or have we squeezed out the space where a vital piece has to go? For every question, every hypothetical layout of the network, PACE used the Okumura model to crunch through a new good-enough answer, and the planners tapped their teeth thoughtfully and gazed at the answers, all of which contained layers of approximation, none of which was the one right answer.

  And, of course, you couldn’t always plunk your base station there, anyway. Radio planning is the only applied science that has to be conducted with a team of estate agents standing by. A farmer had to sell you a piece of that hilltop. In those days, before the alarm about health risks began, it was much easier to do the deal. Dave Targett remembers, ‘You went along to a farmer and said, “Here’s a thousand pounds for the corner of your field,” and they bit your hand off to take it.’ But the land wasn’t always available, and the price wasn’t always right by any means, especially when city cells were being designed. So quite often the planners were working through the problem the other way around. OK: given that the cell has to cover this area here, more or less, and these five sites are the only locations the property department can offer us, which one is the least worst option? What’s the best we can do? How can we tweak it? Every variable in the calculation was slippery, every variable was subject to adjustment (including the budget). The different parts of the calculation jostled each other. You could fiddle for ever. It was a question of coming up with the best of the imperfect compromises on offer, the best approximation obtainable from a process you couldn’t wholly control.

  PACE was able to spit out the provisional radio plans in cartographic form, using one of those old-fashioned plotters where a needle nib jerked to and fro – skree, skree, skree – on greaseproof paper. The radio planners laid the greaseproof sheets back out on the OS map on the map table, being extremely careful to line up the different sections of the overlay correctly, and contemplated the latest prediction. They were combining the maths of radio propagation with the intensely local geography of Chingford and Tiger Bay. The empirical model they were using couldn’t deliver very great accuracy, but it worked well enough, for now, and they worked it as hard as they could. The business depended on them. It wasn’t that a network badly fitted to the environment would plain fail to function. It would function, in a way. You could stick an antenna on a randomly chosen tall building, and people around it would get some kind of phone service. But the service would be full of holes, constantly frustrating any customer who wanted to hold long conversations on their mobile and pay Vodafone handsomely for the privilege; and it would use the pool of available frequencies so inefficiently that there’d be a very low ceiling on the number of users the system could support, perhaps such a low ceiling that you couldn’t run the system at a profit.

  In the second half of 1984, Ericsson started to deliver the hundred base stations, and Vodafone’s engineers feverishly started checking the predicted signal strengths against the real thing. That’s what the van was doing on Park Lane, but there were white vans patrolling in all directions. Panasonic sent the first consignment of handsets, and Gerry Whent laid claim to handset 001. Dave Targett had 002 for a while. He and his colleagues spent December of 1984 driving round and round the capital, in their own cars this time, testing the ability of London’s five cells to handle calls to every type and location of telephone exchange they could think of, down every different path through the BT network backbone. They had a list of people’s friends and relatives who didn’t mind being rung up, over and over again, by tired men in cars with nothing much to say. Officially, the first call on the Vodafone network was made on New Year’s Eve with pomp and circumstance by Ernie Wise, with the cameras on him and St Katharine’s Dock behind him, the refurbished ex-industrial shininess of the dockland view providing one Thatcherite vista, the future represented by the phone opening another one. But the real first calls were the ones the engineers had made. Every so often, someone at the other end was so surprised that it gave the engineer a little start and made them realise anew how odd this technology was, judged by normality as it had been up till now. Dave Targett remembers ringing his mother from a lay-by, and her finding it hard to believe her son was telephoning from a car, a car; and that, in turn, awakening his own sense of wonder for a moment. He drove past Fulham delicatessens, under soot-stained viaducts in Vauxhall, through the creditworthy purlieus of Threadneedle Street, out along the Westway, listening to this astonishing thing: a voice out of the air. He was a radio man, so he didn’t perceive the object in his hand as a telephone with the cord snipped off: instead, it was a radio ingeniously patched into the phone network. He dialed; a phone rang; for a few minutes a pair of voices radiated past each other in the previously silent 900-MHz band, travelling as coded wrinkles in wavefronts that richocheted off DIY warehouses at the speed of light, and shimmered through the bare branches of trees in dogshit-flecked parks.

  Hello. Is that, er, Fred in Walsall? Yeah, it’s us again. Can you hear me all right? Any crackling on the line, any buzzing or clicking or anything? Great. Kkhhrrrr … What? Oh, that was a bridge, I just went under a bridge, it interferes with the signal. Yes, I really am in a car. Yeah, you’re right, I suppose it is strange, if you think about it. Well, cheers. Thanks for putting up with us. Happy Christmas.

  The Plessey engineer who told me the story about the indestructible Racal backpack radio, Garry Garrard, became a consultant in the 1980s, specialising in the development of the mobile-phone market. Having read his book about it, I went to see him in Bed-ford, among his collection of different editions of The Rubaiyat of Omar Khayyam. ‘Look,’ he said, showing me a diagram. ‘Take-up of mobile phones from 1985 to now has followed an S-curve, slow and shallow at the beginning, steep and quick in the middle, slower again now. It went through four stages. First, a mobile was a status symbol: there was one per millionaire. Then it was a business tool: there was one per businessperson. Then it became a household luxury: and there was one per family. Now it’s a household commodity: and there’s pretty much one per person in the UK.’

  Vodafone’s very first users when the system went live were company directors in Bentleys, followed by yuppies in Porsches. Mobile phones to them were declarations of excess, statements of a power to consume that cheerfully transcended need, just as the cars they were in cheerfully exceeded any of the requirements of ordinary driving. In fact,
the cars and the phones were inextricably entwined, because for the first few months mobile phones were only available in cars and had to be installed there by experts. True ‘handportables’ came a little later; the famous Motorola brick may be remembered now as the original brontosaurus of mobiles, but in its time it was itself a breakthrough. In the spring of 1985 you had to send off your fire-engine red 911 Turbo to be fitted with your new toy. The Porsche came back with a transceiver the size of a car battery in the well between the two front seats and a chunky telephone sitting on top. Powering up the M1, idling at a red light in Bishopsgate, you could chat away ostentatiously, showing onlookers that your redundantly powerful vehicle had this other redundantly powerful capability too; one as gloriously wasteful as a stone-flagged farmhouse kitchen installed in a corporate jet, for who really needed to make phone calls from a car? For a while, its role as a chief symbol of conspicuous consumption set the public meaning of the mobile phone. It was loved or loathed as people loved or loathed wealth itself. Luckily for Vodafone and Cellnet, the mid-1980s were a time when there was a lot of new money around, eager to claim boasting rights.