I want to get just a little technical and talk about how radios work. I'm going to avoid the detailed, low-level engineering stuff, apart from saying that designing radios is at least as much art as science, and the engineers who do it are a very bright bunch indeed.

I can't think of many electronic engineering activities as difficult as building a modern radio. Think about what we've accomplished over the past 15 years in wireless LANs alone, with huge advances in both throughput and cost. Price/performance has improved by hundreds of times, and the innovations keep on coming.

While there are many types of radio, they all work basically the same way. On the transmit side, we create a carrier wave at a given frequency, and then modulate it with the data we want to send. We do this by changing characteristics of the carrier wave. For example, we can change its amplitude, or power level, thus creating amplitude modulation, or AM. We can similarly change the frequency, implementing frequency modulation, or FM. And there are many other forms of modulation beyond these - hundreds, in fact. But they all work pretty much the same way. The change in the wave is called a symbol, and that's what the receiver at the other end looks for. Assuming a decent radio environment, communication can take place.

Dealing with the real world The big engineering challenge in radio, however, is that we're dealing with the electromagnetic spectrum - the real world, as engineers sometimes call it. And, like the rest of the real world, the spectrum is an analog quantity, described by continuous waves and differential equations. This is not the nice, neat world of digital electronics, where ones and zeroes are all that matter. On the contrary, the analog world is downright nasty. For example, just a change in the ambient temperature can play havoc with analog circuitry, which is why we often include heaters in radios permanently installed outdoors or otherwise operated in cold climates. Analog electronic engineering is notoriously difficult and complex, and engineers that master it have significant job security.

But suppose we could marry the nice, neat digital world of modern digital electronics to analog systems - in other words, suppose we could build a radio out of a computer. In fact, we can, and this is the world of software-defined radio, or SDR.

What's in an SDR? Using SDR techniques, we can replace most of the analog circuitry with digital electronics. The processing on the receiver end might look like this: First, we have an antenna that grabs the very weak modulated carrier out of the air. We then use a device called a low-noise amplifier, or LNA, to boost this very weak signal without adding noise or distortion. We then do a little basic filtering, to select just the band of frequencies we want, and then downconvert the signal to a lower frequency which is easier to process (even the 2.4 GHz. of wireless LANs can be tough to deal with in its native form). We then feed this signal to an analog-to-digital converter (A/D). Now we have a digital representation of what went over the air - not at all different from what's stored on a CD or DVD. And we can now use a high-speed processor and appropriate software to extract (demodulate) the signal and get the information back out.

Note that SDR has other benefits: We can easily fix bugs in the radio implementation - or add new features - just by loading new software. We could even change the nature of the radio quite dramatically - a phone that's on GSM (Global System for Mobile Communications) or UMTS (Universal Mobile Telecommunications System) in Europe could magically switch to CDMA (Code Division Multiple Access) or WiMax in the US [the agility of SDR has also suggested new approaches to radio licences - Ed].

We could have a single radio process both cellular and wireless LAN signals at the same time. And, while power consumption is an issue in current SDR designs, there are some good approaches to addressing this challenge under development today.

SDR products? What got me thinking about SDR today was an announcement by Airspan of its new HiperMax product line, which the company says is based on SDR techniques. They use SDR in this case to implement a flexible fixed/mobile architecture for WiMax base stations, and, given that WiMax is still evolving, this sounds like a very good idea to me.

I suppose the big question is when we'll see SDR techniques in handsets and other mobile devices. I think that's some way off, because we need to make progress on mobile processors and especially on their power-consumption characteristics. My guess is we're still five years, away, but it could be sooner.

In the meantime, two sites I recommend for more information on SDR are those of the SDR Forum, a trade association whose technical conference happens this week, and Vanu, a company run by SDR pioneer (and heck of a nice guy) Vanu Bose. Vanu builds software for SDR applications, and the site has a wealth of information on what may be the most important direction in radio design today.