MIMO and Mesh Networks

Mesh networks today generally depend on all nodes using omnidirectional antennas. They do so primarily to make setup as easy as possible for the end-users. Directional antennas have some advantages over omnis - they are cheaper on a dollar-for-dB basis, they pick up less interference, and they enable the node-owner to intentionally point high-gain signal beams in an intended direction. However, having to manually point a directional antenna is not suited to mobile nodes - nor does it suit fixed nodes that are intended to be part of a neighborhood mesh. These nodes need to be able to connect to other nodes in any and all directions of the compass. A mesh network takes advantage of any connection opportunity that presents itself - it does not rely on nodes appearing in a location according to a preset plan. So omnidirectional antennas are are the best practical option for meshing.

Enter the yet-to-be-ratified 802.11n WiFi standard with it's Multiple-In, Multiple Out (MIMO) technology and in particular, beamforming antennas on consumer-grade equipment. Beamforming offers the best of both antenna worlds - omnidirectional access with directional high-gain and interference-rejection. Basically, beamforming antennas are electronically steerable. They can direct their high-gain transmission and/or reception beams in any direction they want at any time, without moving parts. They can even block (null) reception of signals from any direction.

The implications for mesh networking are massive. There has already been much adademic research into this possibility - much of it published at the recent IEEE WiMesh-2005 workshop. The big problem for mesh networks today is that a bunch of nodes with omni antennas on the same channel tend to interfere with each other. When any node transmits, every other node in reception range is forced to listen to it. Even worse - a node which within interference range of another node has to stay silent while that node transmits, even though the nodes are not close enough to actually communicate with each other. If these nodes used beamforming antennas, they would not need to stay silent if they were communicating with other nodes - beamforming allows the same channel to be used by many nodes simultaneously, and allows much more efficient routing.

Packets in Ethernet networks can be unicast, multicast and broadcast - in a wired network, the network switch deals with the different kinds of packets and decides if they need to be sent to one, many or every other host (or node) on the network, respectively. In a wireless medium, using omnidirectional antennas means that all packets are broadcast, even if at the Ethernet layer they were intended to be unicast. Beamforming means that unicast packets can be truly unicast, even as they are transmitted via the radio medium.

At the moment, MIMO technology probably isn't quite ready for community wireless networks. 802.11n isn't a standard yet, and a public-access community network can really only operate with a cheap, ratified standard. The various manufacturer's pre-N offerings are quite expensive compared to their 802.11g stablemates, and different manufaturers are offering competing technologies. This makes hacking these devices a fairly fruitless exercise. Also, beamforming antennas rely on the precise placement of multiple antennas - something which isn't easily manufactured or modified in a hobbyists workshop - although we are sure to see hobbyists try. Nonetheless, if 802.11n is ratified with decent beamforming capabilities it could do great things for community mesh networks.