Friday, February 3, 2012

802.11n 450 Mbps just came out, but is it already the past?

The race for speed is a never-ending quest. Ethernet already reaches 100 Gbps, and 802.11 struggles to maintain the pace, defined by the IEEE, that says that data transmission speed doubles on average every 5 years. But how to increase speed in a half-duplex medium, plagued with interferers of all kinds, without dramatically changing the way we send radio waves?

802.11n offered two brilliant solutions: increase the signal size (use 40 MHz-wide channels instead of the old 20 MHz channel), and be more clever. Increasing the channel size is an obvious improvement. It works like a water hose: the larger the pipe (the larger the RF channel), the more water you can send at the same time (the more data you can send at the same time). Being more clever is actually harder. The idea is also to use larger pipes, but in space rather than in frequency. Instead of sending one signal from one radio (on one frequency), you send 2 signals from 2 radios (one signal from each radio), still one one frequency... and the receiver, having also 2 radios, nicely gets the first signal on its first radio, and the second signal on the second radio. Nice. Nice, but incredibly hard to achieve. In reality, the RF signals bounces on many objects around us (walls, ceilings, furniture, etc). Most objects absorb the RF signal, but some of them reflect it. This is called multipath. The result is that the receiver gets a signal that seems to come straight out of an echo chamber: multiple copies of the same signal, reaching the receiver at different points in time. Sorting and cleaning out all these waves to recover the original intended message is hard when there is ONE original signal. When there are 2 original signals, it becomes an engineering nightmare. But most vendors solved the equation, and use 2 streams APs and clients...

Can we add more? Well, 802.11n nicely allows up to 4 streams, 4 signals sent in parallel. Up to 2009, most vendors were whispering that 3 streams would be close to impossible, then a few big names took the bet, and finally made it work over the last few months. Intel and  Apple came out with 3 stream clients. Aruba came out with a 3 stream AP, then  Cisco also announced a 3 stream AP (the 3600, that actually uses 4 radios to shoot 3 signals right in the direction of the client, which makes it really efficient). With a Cisco 3600 and an Intel or Apple client, you can now get up to 450 Mbps. This is the speed of the link, not the actual throughput. In this half duplex environment, this rate is shared between traffic to the AP and traffic from the AP, and between clients of the cell... but it allows a throughput of up to 270 Mbps in ideal conditions... wow.

Can we do more, please? Can we get the 4th stream promised by 802.11n? Don't ask this question to a RF engineer, you would get bitten (or slapped). Many still say that 4 streams is (this time, for real), impossible in practice... Yes, they already said that for 3 streams, and still made it work. But this time, they may just give up on 802.11n...

Why? Because 802.11ac is a new amendment that should come by the end of 2012 or in 2013. This amendment takes another approach, and uses larger channels (up to 180 MHz!), among other techniques. It has been slowed down for a while, because major players could not agree on what was the right way of dealing with the side effects of an increased speed... but a lot of progress was made over the last few months (or, should we say, some major solutions won over some others), which is speeding up the development of the amendment. 802.11ac will come soon, and will allow data rates above 1 Gbps. So why bother with 802.11n?

A clear sign of this trend lies in all these announcements released during the CES earlier this month. Among the big players, Apple announced their coming support for 802.11ac. Of course, Apple likes to be at the edge of new technology. But they are not the only ones to make the announcement. MediaTek, Netgear and a few others also announced 802.11ac products for this year, months before the amendment expected publication date. This may seem surprising, but keep in mind that most vendors started selling 802.11ndraft products 2 to 3 years before the amendment was released. New technologies boost sales, and being among the first also helps.

If you just bought your new 3 stream 802.11n AP, take a picture of it now: it will look like an antique in 6 months or less. If you work in Wi-Fi, keep smiling. After the great 802.11n wave, 802.11ac will keep you in business for several more years!