We might be running out of things to do with faster processors on the desktop, but nobody feels the same way about speedier networks. Wireless data in particular can easily do with more: as ethernet speeds head on up to a gigabit a second, even the fastest radio system looks laggardly. Now, a new technique called MIMO -- multiple in, multiple out -- antennas promises to give a huge fillip to radio speeds without needing any more radio spectrum.
MIMO is just the latest development of a respectably ancient idea -- more wire in the air means more signal. The Yagi, your basic rooftop TV antenna, has a series of parallel elements that act like lenses, focusing the incoming signal and making it stronger. You can get better results still by stacking two or more Yagis, as common sense would suggest. But multiple receiving antennas have an unexpected benefit -- they give you diversity.
One of the constant demons of wireless is multipath, where a signal comes in from a single transmitter from several directions at once. This happens because buildings, hillsides and many other objects either side of the main path between transmitter and receiver can act as reflectors and refractors. The result is like shining a torch through a chandelier. As the many scattered signals take different paths and thus different times to arrive, they turn up with all sorts of different phases to the main signal and can cancel out or boost it seemingly at random. The result at any one point is a mess: that's the reason FM transistor radios can sound so rough at one point in the bathroom yet give perfect results if you just reposition the antenna.
Two or more antennas can easily sort out this mess. At its most basic, the receiver just has to pick the one with the strongest signal at any one time, although there are cleverer ways to analyse the signals and reconstitute the original. This is called diversity reception, and it's been in use since the Second World War: it's also the reason some wireless LAN base stations have two antennas.
For completely different reasons, it's also a good idea to have multiple antennas at the transmitter. By sending signals with slightly different delays to an array of antennas, the resultant transmission can form a tight beam: by changing those delays, the beam can be electrically steered at high speed. Called phased array antennas, these devices are frequently reinvented and given names such as smart antennas, but once again the original ideas can be traced back at least fifty years.
Lately, the idea of using multiple antennas on both transmitter and receiver -- hence Multiple In, Multiple Out -- has come into fashion. This is because while there are some extremely tempting extra benefits to be had by combining the two ideas, they take a lot of signal processing at the receiver to work, and until recently this was too expensive to be worthwhile. Now we have million-transistor chips running at gigahertz and costing thruppence ha'penny, the fiscal equation has changed. The maths of the physics is exciting too -- the potential gain is a simple multiple of the total number of antennas used
The fundamental innovation is the way multiple receive antennas can power phase processing, in effect using the phase differences from incoming signals to differentiate between them. Again, far from being a new trick, this has been around for the odd hundred million years: animals with two ears, including us, can not only tell what direction a sound is coming from but can concentrate on that sound even when there's a lot of extraneous noise appearing from other directions. We can even screen out sounds on the same frequency and with very similar information: the famous cocktail party effect is a good example, where curious socialites can earwig on interesting gossip coming from one direction while the rest of the room is also awash with drunken conversations.
What's really exciting about this is the ability to use space to differentiate between signals on the same frequency. Spatial multiplexing is a cracking way to get around the standard bandwidth limitations of only having so many megahertz of spectrum to use: if you can transmit multiple signals on the same frequency, you're able to cram much more in without breaking your licence restrictions. Moreover, if you encode the transmissions so that information on each can be used to help reconstruct the information on the others -- space time block coding -- you can increase robustness as well as pure throughput.
MIMO systems work best with lots of multipath, as they can treat each arriving signal as an independent link from which information can be extracted -- many of the benefits of the approach are lost if you've got a classically good line-of-sight link. Fortuitously, at the microwave frequencies used by wireless LANs and broadband data links, the world is stuffed full of multipath: it feels like MIMO is the right technique coming along at the right time.
There are restrictions. To work well, the antennas have to have a reasonable physical separation. Exactly what depends on the frequencies: you can easily build a MIMO system for, say, 30GHz broadband links, but there may not be the room in a PDA or mobile phone. A laptop with a MIMO-enhanced 2.4GHz or 5 GHz wireless LAN? Well, probably. They're working on it. Also, there's still plenty of work to be done in finding the best ways to model MIMO systems, finding out the best modulation and encoding schemes, and finding the most effective tradeoffs between cost and performance.
However, there's a good chance that MIMO will be incorporated in many forthcoming wireless standards, including 802.11n -- which will take 802.11-style networks up to as much as 300Mbps -- and 802.20 for wireless broadband. Toshiba is also showing an 802.11g network at the ITU Telecom World conference in Geneva this week, which uses MIMO techniques to increase the data rate to 100Mbps. Having found it, nobody's going to let MIMO go -- and the results should be more than worthwhile.