How Does MU-MIMO Work?

The second wave of 802.11ac introduces multi-user, multi-in multi-out (MU-MIMO) technology to support an increasing number of WiFi devices consuming ever-increasing amounts of bandwidth. Separating the hype from the technical reality, however, can be quite a challenge, even for those of us who are experts in WiFi. In a two-part blog series, I'll provide a reality check, starting with how MU-MIMO works. In this post, I examine technologies that set the groundwork for MU-MIMO and how MU-MIMO builds on them.

802.11n introduced the technique of multi-in multi-out (MIMO) to enhance WiFi throughput between access points and client devices. For MIMO to work, the two wireless stations in communication (i.e., both the access point and the client device) must each have multiple radio/antenna chains that are identical and physically separated from each other by a fixed distance so as to purposely be out of phase at the operational wavelength. A spatial stream is a data set, sent by a transmitting radio chain, which can be mathematically reconstructed by the receiver’s radio chains. In MIMO, each spatial stream is transmitted from a different radio/antenna chain in the same frequency channel as the transmitter. The figure below illustrates this for a two stream case.

The receiver receives each stream on each of its identical radio/antenna chains. Since the receiver knows the phase offsets of its own antennas, it can use signal-processing techniques to mathematically reconstruct the original streams.  To enhance throughput, each spatial stream contains unique data, and the number of independent spatial streams is therefore limited by whichever WiFi device has the least number of radio chains. Typically, this limit is the client device: As each radio/antenna chain consumes power and space, most mobile smartphones and tablets are only capable of single stream or dual-stream communication, and even high-end laptops and PCs generally only support up to three streams.

In 802.11ac Wave 1, throughput is not only enhanced by MIMO, but utilizes other improvements, including using even wider channel widths and the more complex 256-QAM modulation and coding scheme. These other mechanisms  have limitations, however. The total size of the 5 GHz band is "finite;" as a result, wider channels lead to fewer independent channels and are subject to larger interference.

While there are efforts by the Federal Communications Commission to open up more of the 5 GHz unlicensed spectrum for WiFi, 80 MHz channels are likely to be a hard practical channel size limit going forward. Furthermore, the new 256-QAM modulation and coding scheme (MCS) rates requires a minimum SNR of 37 dB, meaning that a really good signal is needed between the WiFi devices, which is only practically achievable at very close distances in very clean RF environments.

Accordingly, another method of enhancing throughput is to actually have the access point transmit data to multiple client devices simultaneously, and this is what multi-user MIMO (MU-MIMO) is all about.

However, in order to understand how MU-MIMO works, it's important to first know about another technology introduced but not widely implemented in 802.11n: transmit beamforming (TxBF). Unlike MIMO, which sends a different spatial stream on each antenna, transmit beamforming sends the same stream on multiple antennas with deliberate timing offsets to increase range. Beamforming thus requires the use of a phased antenna array, where there are multiple identical antennas at fixed separation distances (so as to be out of phase).

The phase of each data stream is transmitted by all antennas at different times (i.e. with different phase offsets) that are calculated so to have these different signals constructively interfere at a particular point in space (i.e. the location of the receiver), thereby enhancing the signal strength at that location. The signal can be enhanced by 2x (i.e. 3 dB) for every phased antenna. When using omni-directional antennas, the effective antenna pattern created becomes effectively directional.

 

The main caveat to using transmit beamforming in WiFi is that the transmitter (i.e., the access point) needs to know the relative position of the receiver (i.e., the client device).  The access point does this by sending out sounding frames, essentially independent signals from each of its antennas, and then the client device responds with a matrix indicating how well it heard the signal from each antenna. Based on this matrix data, the AP can compute the relative position of the client device, and the phase offsets on each of its antennas required to maximize constructive interference at the client device.

MU-MIMO takes this process one step further. By adding even more radio chains/antennas, it can control the phased antenna pattern to control both the areas of maximum constructive interference -- where the signal is the strongest -- and maximum destructive interference --where the signal is the weakest. With a sufficient number of antennas and knowledge about the relative positions of all associated client devices, it can actually create a phased pattern to talk to multiple clients both independently and simultaneously.

The overall process for MU-MIMO is as follows:

  1. The AP broadcasts a sounding frame
  2. Each MU-MIMO compatible client device transmits back matrix data to the access point
  3. The AP computes the relative position of each associated MU-MIMO compatible client device
  4. The AP selects a group of client devices for simultaneous communication
  5. The AP computes the necessary phase offsets for each data stream to each client in the group and transmits all of the data streams in the client group
  6. The AP sends a BlockAckRequest to each client device in the group separately to get confirmation as to whether the client device received the data
  7. The AP receives a BlockAck from each client device in the group that successfully received data

The maximum number of simultaneous clients is one less than the total number of available AP streams.  This is a mathematical limitation since the AP needs to control both the areas of maximum constructive interference, to direct the strongest signal to the desired client device, and the areas of maximum destructive interference, to minimize the signal at the other client devices in the group.  

Mathematically, the number of variables exceeds the number of unknowns, so one stream cannot be controlled independently. That last stream, however, can be set to align with another stream, which can be used for multi-stream MIMO clients.  Thus, for the current generation of 4x4:4 MU-MIMO capable 802.11ac Wave 2 access points, the following combination of groups are valid:

  • One 3x3:3 stream client device and one 1x1:1 stream client device
  • Two 2x2:2 stream client devices
  • One 2x2:2 stream client device and up to two 1x1:1 stream client devices
  • Up to three 1x1:1 stream client devices

Naturally, such an access point can do conventional MIMO to one client, all the way up to a four-stream client device. While I'm unaware of any manufacturer planning a four-stream client device, an AP can be operated in “client bridge mode” for such a purpose.

Stay tuned for the second part of this blog series next month, when I'll discuss the limitations of MU-MIMO and under what circumstances it can actually be useful.