Qualcomm is launching eight parts that all support 801.11ac Release 2 including MU-MIMO or Multi-User MIMO. This beamforming tech can effectively triple the throughput of a Wi-Fi device allowing for three times the users without any speed loss.
The current release of 801.11ac or ‘Gigabit Wi-Fi’ is not exactly slow but it is still not widely adopted in the field. Each channel is wider than before at 80MHz and will support 1.3Gbps in 3×3 MIMO but usable bandwidth does not scale linearly with stream count. The new Release 2 802.11ac adds three optional features, MU-MIMO, 4-stream support, and bonding two 80MHz channels to a 160MHz channel. Much of this was hinted at when 802.11ac first hit the market and now you are seeing Qualcomm support one of the three new optional features.
MU-MIMO is a fascinating technology, it uses beamforming to separate groups that each can benefit from nearly the full 802.11ac bandwidth. The way it works is easy enough, starting with the need for at least 3×3 MIMO on the transmitter/router side but 4×4 is better. All (hopefully) 4 antennas will send out a training sequence to the receiver which responds with a packet containing the received signal timing and strength from each antenna.
This means that the receiver only needs a minor firmware update to support MU-MIMO, it is the router/transmitter end that does all the heavy lifting. The more technical among you may also realize that MU-MIMO is downstream only at the moment, not surprising because there really aren’t 3×3 or 4×4 802.11ac client radios on the market, not to mention spatial separation headaches in a laptop. While it is possible to implement upstream MU-MIMO, don’t expect it any time soon.
Three times three doesn’t always equal 1/9th
Once up and running, MU-MIMO will put devices into groups and beamform the signal they receive so that each group can use the full 802.11ac bandwidth that the router is capable of. If you could do 910Mbps without MU you could do three groups of 910Mbps or 2.7Gbps aggregate throughput. As you can see below, if you have three devices in a group you will get a little less than 350Mbps per client theoretically possible but the aggregate throughput goes way up. An MU-MIMO device can support one bandwidth heavy device, three bandwidth light devices, or anything between per channel. That would be two for the math averse, please post on our forums if you require a further explanation of the math.
Not quite three times the bandwidth but really close
So how does this work. Easy, well not really easy, beam forming is a really interesting mathematical problem related to what noise canceling headphones do. If you have fond memories of wave experiments in high school physics, you will like the physics involves in beamforming. The vastly simplified science behind it all is that when two waves meet they are additive so when two signals of strength 1 and .5 meet, you will get an effective signal of 1.5 strength. A 1 and a -1 crossing will end up with a 0 or no signal and so on and so forth. If a signal from one antenna has a strength of +1 to -1, two antennas will have overlapping points from +2 to -2, and so on to the maximum transmitter count.
What beam forming does is locate the client rather precisely from signal strengths and timing data in the training packet. It will then shift the phase on the transmission antennas to cancel out the signal at points where non-group devices are and be additive where the devices in group are. In group sees hopefully 1+1+1+1, out of group sees 1+(-1)+1+(-1) or something else that adds up to zero. There is a horrendous amount of number crunching that goes on to make this work but it does actually work.
The net result is up to three groups of up to three devices can each have nearly the full bandwidth of a non-MU 802.11ac device. Nearly is in there because the training packets have overhead but the net result is far better than not using MU-MIMO. Better yet the devices don’t have to be spatially close, they can be anywhere within the signal radius of the transmitter, it doesn’t set groups by angles, they can be interleaved. Distance between devices is not a problem either, Qualcomm says that devices a few CM apart work fine, something you would expect given the wavelengths in the 5GHz spectrum. It is a very impressive piece of on the fly wave cancellation trickery no matter how you look at it, noise canceling headphones on a massive scale.
That brings us back to products and as we mentioned earlier, Qualcomm is launching eight of them. There are four infrastructure chips and four client chips, all aimed at different things and sporting different capabilities. The QCA9980 is a 4×4 4-stream 1733Mbps, that bandwidth is per stream mind you, residential gateway device. QCA9982 is a 3×3 3-stream 1300Mbps version of the QCA9980, and the QCA9990 and QCA9992 are higher security enterprise versions of the first two.
On the receiver side you have the QCA6174 which sports 2-stream receive for 867Mbps and Bluetooth 4.1 if those speeds are too much for your delicate constitution. It is meant for the Snapdragon 805 connecting via PCIe. The WCN3680 is a 1-stream, BT4.1, and FM version integrated in to the Snapdragon 801. Although they didn’t state this we expect all higher end Qualcomm Wi-Fi radios to support MU-MIMO from here on out.
For dedicated consumer electronics, think TVs and audio units, there is the QCA9378. It is the same thing as the QCA6174 but adds USB and SDIO interfaces. This should be obvious because 9378/6574 = 1.518950437, 1.xx8950xx is an internal Qualcomm numerological code for added USB and SDIO. Last but not least is the QCA6574, an automotive variant of the QCA6174. The spec sheets don’t have a PCIe interface but it probably has some way to connect with the outside world, a radio with no I/O isn’t all that useful. Automotive designs are long life rated, hardened packaging against temperature extremes and vibration, and a whole lot of other changes to make it last longer than a car usually does.
In the end Qualcomm has eight new chips that all implement the MU-MIMO option of 802.11ac Release 2. It will allow three groups to each simultaneously share the bandwidth available by a 3×3 or 4×4 802.11ac radio, that is between a lot and silly levels of aggregate bandwidth from a single base station. It does all this through beamforming, a silly amount of wave cancellation, and an enormous amount of heavy math. To the user it just works and most will never have a clue about why.S|A
Have you signed up for our newsletter yet?
Did you know that you can access all our past subscription-only articles with a simple Student Membership for 100 USD per year? If you want in-depth analysis and exclusive exclusives, we don’t make the news, we just report it so there is no guarantee when exclusives are added to the Professional level but that’s where you’ll find the deep dive analysis.
Latest posts by Charlie Demerjian (see all)
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018