Radio silence betrays industry secrets

In this business, you soon learn that what people say is far less interesting than what they don't. Take chips. Intel and AMD will fall over themselves to tell you how their new processors work. Press conferences and white papers are full of the fascinating mechanisms of hyper-threading, cache design, bus architectures and much else too darn dull to list. They tell us all this in an attempt to prove that they and they alone have the best bang per buck, the fastest chip bar none, or just because we should feel that they're very clever chaps with lots of bright ideas. Who are we to argue?

But it's all so much flummery, like a church claiming it's much holier than the rest because its bishops have the tallest mitres. When you're sitting at your computer getting work done, how many times do you sigh to yourself and say "Oh, if only the processor was just that little bit faster?" Outside the greasy worlds of oil and Hollywood, we don't care. We haven't cared for years.

What matters to us these days is communication. Networks. I want my data here and now. When I pick up my laptop and walk I want my data there and now too. Most of my computing delays of late have been due to wireless hiccoughs, cable modems not coping with bad connections, networks not delivering what I wanted when I wanted.

So how much information do we get from the chip companies about our network components? We're expected to understand the details of processor arcana, but not to care about the stuff that shovels the data in and out -- here, the great engines of publicity mumble and cough.

I have a suspicion why. Take wireless, the masked superhero carrying the hopes of the industry on its wide yet ethereal fairy wings. A friend reports from a huge conference run by an enormous and very famous software company that the wireless bits just weren't working properly: the state of the art installation had hundreds of 802.11a/b/g access points scattered across the site, and on paper it should have been what that company likes to call a compelling user experience. In reality, it was what us compellingly experienced users like to call frustration. Real pig-kicker class.

"What did you expect?" said an engineer to whom I related the story. "You've got hundreds, thousands of transmitters in a small space, and the receivers are tosh. They haven't got a hope in hell of working. Couple of access points, handful of laptops? No problem. More than that? Might as well use Morse code. Ever had two cards refuse to work together? Ever have any idea why? That's why."

I know what he means. Receivers have a simple job to do -- get as much information as possible out of the wanted signal, and thoroughly reject all the others. But that's a tough job, especially when you're trying to winkle a weak little whisper of a thing out from a band crammed with brutish great thugs yelling away from nearby transmitters. Fortunately, engineers have been tackling this for more than a century and in the hands of an expert, receiver designs can do wonders.

After all, it's just engineering. You have to understand things like headroom, intercept points, image rejection, noise floors, dynamic range, rolloff and ripple -- the list goes on, but it's finite and well understood. To be sure, a lot of these things are particularly difficult in the low voltage, low power environment of the average PC card, let alone Compact Flash. Making it all happen in a tiny space with a Bluetooth transmitter on the next chip along is even harder.

But hold on -- don't the chip companies take great delight in telling us what clever chaps they are, how their seams fairly burst with bright ideas? If we can cope with out-of-order microinstruction scheduling, we can cope with high level local oscillator mixing. Bring it on, you radio frequency experts. Show us what you've got.

Silence. Hm. Could it be that their radios are rubbish? But they can get away with it because not only are radios hard to design, they're hard to test. Any old spod can flip a benchmark disk into a PC and come up with figures that give some idea of how well the thing works: with wireless, you have to battle entire armies of invisible foes before bringing back the goods. You need test equipment that can scare the chrome off a shiny robot, a test environment that blocks extraneous signals to military specification, and a test methodology that correctly identifies and reproduces a large set of real world conditions. Including ones where hundreds if not thousands of transmitters are battling for space.

Again, it can be done -- but not for the pennies that people normally expect to spend on testing. But until a way is found of properly analysing and reporting on how well wireless networks work, they won't work well. I don't know where the money will come from: there are cases to be made for government dosh, industry consortia, even a speculative commercial effort by a far-sighted company. Some form of independent judgement is absolutely necessary if the wireless industry is to fulfil its potential -- and the manufacturers need to start telling us the truth. We can handle it -- can they?

Let the editors know what you think in the Mailroom.