Intel's announcement of a dual-band silicon chip capable of supporting any and all Wi-Fi standards has been greeted as a major move towards the idea of smart radios that can morph easily to adapt to any task. That's overstating the case, but the new design has revived interest in this fast-changing and potentially very important field.
Radios convert wireless signals into data and vice-versa by a set of internal operations, each of which can be described fully in mathematical terms — in effect, they're a very specialised form of analogue computer. Amplifying weak signals to strong ones is a form of multiplication, changing a signal at one frequency to a different frequency is more complex but still boils down to multiplication, addition and subtraction. Replace the radio with a fast enough computer, and everything can be done in software — you'll need a smear of specialist circuitry to interface the antenna and some very fast analogue to digital converters, but the rest is just code.
Because radio technology is so well understood and highly developed — electronics as a whole grew from early wireless work — the cost advantages of doing things this way don't become attractive until you have a very large market. Ideally, you have just one chip that can be reprogrammed to any wireless standard: at that point, you can sell the same hardware to everybody who uses radio for anything. That's a global market in the billions.
The major problem with software radios is regulatory. In theory, there is no difference between a wireless LAN with a software radio and one built out of coils, capacitors and crystals. As long as they both conform to the regulations about spectrum use, frequency stability, bandwidth, noise and power, there's no way of telling which is which. In practice, though, the software radio is far easier to tweak. If you want to increase the range of an old-style transmitter, it's a matter of soldering irons and specialist knowledge. A software radio just needs new code, which can be downloaded and installed as easily as any other program.
Radio regulators have long experience of the troubles caused by equipment that's too easy to modify, and react with great suspicion to the idea that the airwaves will be controlled by software. Buggy and malicious software causes great trouble for everyone on the Internet; if the same sort of problems move to the wireless spectrum, the potential for disruption of established services will be huge. A hacking group could decide to take over a building of wireless networks and reprogram the lot to display dodgy images over the top of the local TV channels — or jam the local police systems.
However, this extreme flexibility may be the saviour of the idea. It's already feasible to have radios decide which of their various modes — GSM, Wi-Fi, Bluetooth — are most effective for communication in any particular circumstance. BT Fusion does just this, although its radio systems are entirely separate. True software radios will be able to talk to each other, sense their environment and mutually agree on the best way to communicate with a very fine degree of control — so called cognitive radio.
By setting power, frequency and modulation modes to consume the minimum amount of spectrum over the minimum distance, software radios will be able to reuse spectrum with a lot more efficiency than the current regulatory schemes allow. Moreover, because they will be very aware of their surroundings it will be very possible to build in detection and location systems so that if a member of the wireless community starts acting in a dangerous way it can swiftly be dealt with.
The world envisioned by software radio proponents is very different to that of today, a sea of rapidly mutating, very aware transceivers that react to local demands quickly and invisibly. Effectively self-regulating, they could do away entirely with the idea of standards such as 802.11 or GPRS; the system selecting the most appropriate mix of technical characteristics from an armoury of techniques. Such technology would also be automatically aware of the location of each of its components, both relative to each other and absolutely, by GPS and other location systems. The regulators' job ceases being one of making sure that different services can co-exist on the same spectrum, and starts to take on the same concerns about the various privacy and intellectual property rights that have dominated the world of the wired.
Intel's announcement today isn't a great step towards a smart wireless world, except inasmuch that it shows the company can use standard — i.e., low cost — chip manufacturing technology to do something that normally requires specialist processes. That also makes it easier and cheaper to integrate the soft of heavy-duty processing required to make a software radio work, even if the system is for now much closer to the traditional way of wireless. This chip will never be able to sprout Ultrawideband or WiMax without extensive additional hardware changes, for example, but it does give Intel the experience needed to go on to the next stage. Perhaps more prosaically, it also means that 802.11n — when it's finally agreed — will be in our laptops sooner rather than later, and given the enthusiastic reports from people using the pre-n systems this should keep us going until our radios finally become smarter than we are.
