WIRED ON FRIDAY: Even after nearly a century of advances in radio, it seems much more can be done and it has become the hotspot of innovation and investment in Silicon Valley. Danny O'Brien reports.Danny O'Brien

Despite his famously impenetrable genius, the most magical technology for most people during Einstein's life was still the radio.

And so, as the world's "cleverest" man, he was constantly asked to account for it. Frustrated, he is once said to have provided this definitive explanation: "Wire telegraph is a kind of a very, very long cat. You pull his tail in New York and his head is meowing in Los Angeles. Do you understand this? And radio operates exactly the same way: you send signals here, they receive them there. The only difference is that there is no cat."

These days the secret of radio seems comfortingly intelligible, at least by comparison to the exotic, invisible-cat-driven tech of Silicon Valley. So it is all the more surprising that radio - and that nitty gritty hardware problem of getting data wirelessly from one place to another - seems to be the hotspot of innovation and investment in the Valley at the moment.

Mr Craig Barrett, chief executive of Intel, describes it as "the new world" and has set aside half a billion dollars for investing in start-ups in the sector.

Mr Kevin Werbach, editor of the influential Release 1.0 newsletter, talks of the technology reaching a "crucial point", ushering in "the new wireless paradigm". There's a new name to fit the new paradigm: they call it "cognitive" or "smart radio".

Clearly, this is not your old-style, Albert-level radio. While most of us aren't exactly sure what replaces the cat in Einstein's equation, we do have some idea of how traditional radio works in practice.

You have a transmitter. You pick a frequency to broadcast on - because otherwise the lad at the other end of the cat won't know where to tune his radio to hear you.

You can't have too many people broadcasting on the same frequency, otherwise you'll jam them. Certain frequencies are better than others. Signals travel further on some; there's less interference on others. So someone has to allocate those frequency slots.

Here the government has traditionally made a quiet cough, and stepped in to help divvy up the spectrum. So, of course, by now all the best slots are taken - by the military, the broadcasters and the mobile telephone companies which, you will recall, spent quite a sum snagging frequency licences across Europe.

Which is exactly what has accelerated the innovation in the science parks of Silicon Valley. As the radio frequency spectrum has grown exhausted - or at least overly expensive - those rag-tag remains have to be exploited in a far cannier fashion.

It's the innovation of the underdog that the start-ups here manage so well. In a rash of theoretical innovations (some many years old but all coming together to create new products in the past few months), engineers are taking the very disadvantages of the bands they have been left and, with a bit of artificial intelligence, turning them into advantages.

Here are the problems the innovators are saddled with. The free, unlicensed bands are small. Pick a channel within it and you're likely to hear a neighbour trying to talk on that frequency too. Any frequency these days is full of noise, either natural or created by all the other machines (like microwaves) that leak into the ether at that point. Any receiver has to filter that out somehow. Any transmitter has to raise its signal above it.

And, by and large, the remaining channel's radio waves have unpleasant characteristics. Unlike the fancier, more expensive waves, they tend to bounce off buildings, trees, hills and people, instead of going through them. So not only is it difficult to get a message 50 yards down a street without it being swallowed up, when the signal does arrive, it's full of echoes. Those weaker, repeat signals come from the time-delayed radio waves of the same signal taking a more scenic route, bouncing and ricochetting off all and sundry.

Enter the gadgets of cognitive radio solutions. First, smart aerials: By taking into account all the echoes and delayed whispers from a far away source, and compensating by sending one signal out a few microseconds earlier, and far stronger than the main signal, you can actually use all those dozens of ricochetting paths to your advantage.

Can't get through a tree? Then bounce around it using the surrounding hills. Aerials like this have been around for a while (the satellite TV squarials are an example), but as computers get faster and cheaper, the calculations grow more precise, and the signal stronger.

The latest transmitter aerial, announced by Silicon Valley start-up Vivato this week, can calculate the precise direction of dozens of users, and recalibrate to allow a WiFi signal to travel four miles, instead of 300 metres.

Channels full of noise? Ultrawideband - a new technique that has finally been given approval by the United States Federal Communications Commission (FCC), after extensive lobbying by fledgling Valley companies - lives for the noise. It sends tiny signals that look like normal noise to current radios but can be detected by its UWB receivers. To prevent real interference, either with existing stations or other ultrawideband systems, it scatters its message across almost all of the spectrum. To re-piece together the stream at the other end, fantastically precise timers are used.

To use only a slightly fanciful analogy, both transmitter and receiver know, for instance, that if it's 12:45:04.5748 in the afternoon, they should be tuned to 93FM. But they're only tuned there for a microsecond, so RnaG need never know.

And what about overcrowding? Well, if your neighbour is squatting in your frequency slot, perhaps he can pass on the message? The idea here is that all the cognitive radios co-operate: if the mobile phone cell is half a mile away, but there's another mobile phone user a quarter of a mile nearer, perhaps he can relay your voice as well as his own to the tower? That's called mesh networking and is based less on Marconi's spark-generator than how the internet transmits a message.

Mr Werbach estimates that, with techniques like this, you could roll out a 3G-level network for a 10th of the traditional cost of putting up a dense grid of cell towers.

And there's more ideas where these came from, to the point where Michael K Powell (Colin Powell's son and current head of the FCC) has intimated that the traditional idea of allocating frequencies may need to be completely rethought around the frequency-hopping, bottom-feeding techniques of cognitive radio.

Which just goes to show, even after nearly a century of advances in radio, there's more than one way to skin a non-existent cat.