Ultrawideband has long been promoted as the future of wireless personal-area networks (WPAN). But are WPANs still needed?
To begin, personal are network (PAN) is perhaps a bit of a misnomer since personal-area connections traditionally have been mostly point-to-point using, primarily, serial and parallel connections. But with the advent of USB and Firewire (IEEE 1394) technologies, personal-area networks gain the potential for a network topology. Extending these with wireless seems natural and, indeed, the first wireless USB products are now appearing on the market.
The most widely used WPAN technology so far has, of course, been Bluetooth, although it's rare to encounter those who actually use Bluetooth for anything more complicated than connecting a headset to a cell phone handset. However, Bluetooth is, in fact, a rich set of applications that support lots of useful personal functions, from printing and faxing to connectivity for file synchronisation. But for these applications, Bluetooth can seem redundant compared to Wi-Fi.
Bluetooth software on Wi-Fi would be terrific, I believe [And it is on the cards - Editor]. Bluetooth on IP would be ideal. But, for now, Bluetooth has instead been ported to ultrawideband adapters, which are also the basis of wireless USB. We're talking here about, theoretically, amazing levels of wireless throughput on the order of the 480 Mbit/s of USB 2.0. That 480Mbit/s represents a peak speed, so we'd expect to see throughput of 150-200 Mbit/s on both wired and wireless USB. But there's always a caveat when we talk about wireless throughput, and that is the inverse relationship between distance and throughput, especially with indoor, low-power systems operating in the unlicensed bands.
Ultrawideband got its name because each UWB channel is on the order of 500 MHz. That's very wide compared to, say, the 5 MHz. of UMTS, the 10-20 MHz of WiMax, and the 20 Mhz of Wi-Fi. More bandwidth usually means more throughput. However, because UWB operates over such wide channels, and because these channels are also used by other services, UWB must operate at very low power in the range of millionths of a watt. This means that UWB might have terrific throughput, but performance will fade rapidly with distance. That's not a bad thing, though, since PANs are indeed personal and usually operate only over a distance of a few meters, tops. As a result, UWB could be quite useful in room-size applications.
But there are two problems here. First, 802.11n is gunning for this kind of throughput and can operate at much higher transmission power levels. If we can get this kind of throughput from Wi-Fi, the demand for UWB might be quite limited indeed, assuming roughly the same cost and price points for both. And could we implement both USB and Bluetooth over Wi-Fi? Of course.
The second problem, however, is more serious: Initial testing of UWB products is showing pathologically low throughput. I spoke with two engineers, both of whom have significant design and testing experience, who are evaluating UWB products. They mentioned that throughput of less than 20 Mbit/s at a range of a meter or so is common, although up to 60Mbit/s has been measured. The exception to this is in reference-design products from Pulse-Link, which does not follow the certified Wireless USB spec. Here, throughput of over 500 Mbit/s has been observed - Gigabit Ethernet speeds on wireless.
So, why the limited performance of wireless USB? I debated this with the engineers doing the testing, and we discussed three possibilities. First, there could be a fundamental flaw in UWB or WUSB. This seems unlikely, since a lot of companies invested a lot of money in the technology and likely would not have done so without effective due diligence. Second, there could be flaws in today's hardware implementations, since first-generation chips are usually not all they could be. And finally, it could be a simple matter of immature firmware and drivers, a common problem. New drivers and firmware can often work wonders, as we saw in the case of upgrades of Draft 1 802.11n products to Draft 2.
And, as we saw with Draft 1 802.11n products, vendors often bring products to market before their time. I'm confident that the UWB community will achieve performance levels commensurate with those of wire, at least at short range. If not, 802.11n is going to play an even bigger role than it does today.
Craig J. Mathias is a principal with Farpoint Group, an advisory firm specialising in wireless networking and mobile computing. This article appeared in Computerworld.