MU-MIMO improves efficiency for WiFi networks from the “spatial” perspective, but what does “spatial” mean? On a typical WiFi device, the number of physical transmitters generally equals to the number of “spatial” streams the WiFi device can support if it’s not indicated otherwise in the product specification. For example, a 4x4 AP generally means that it supports four “spatial” streams. In “SU-MIMO”, the “spatial” imbalance between the AP and client STAs restricts the efficiency of WiFi networks. For example, an 8x8 AP serving 2x2 clients with “SU-MIMO” has the efficiency limited to 25%; ideally, if the 8x8 AP serves four 2x2 clients (a total of 8 “spatial” streams) simultaneously with “MU-MIMO”, the efficiency can be boosted to as high as 100%.

“DL MU-MIMO” and “UL MU-MIMO” will become buzzwords with emergence of WiFi 6 (11ax) as “DL MU-MIMO” was first introduced with WiFi 5 (11ac) and WiFi 6 (11ax) extends the MU “grouping” number from 4 to 8.

• “SU-MIMO beamforming” vs “DL MU-MIMO”

  The same as “SU-MIMO beamforming”, the fundamental technology of “DL MU-MIMO” is “beamforming”. The main processes to execute “beamforming” include “sounding”, “CSI feedback” and “pre-coding” before generating the energy “beams” to client STAs. The advantages of “SU-MIMO beamforming” and “DL MU-MIMO” are different though they adopt the same fundamental technology. “SU-MIMO beamforming” generates a “single” beam and effectively direct the radio energy to the single served client STA; the advantage is the extended service coverage but no improvement on efficiency. On the other hand, “DL MU-MIMO” generates multiple orthogonal beams and steers these beams to the “group” of the client STAs being served simultaneously. As “DL MU-MIMO” can improve efficiency in a significant way, but, improving service coverage like “SU-MIMO Beamforming” is not its main purpose.

  Though both of them are dubbed “MU-MIMO”, “DL MU-MIMO” and “UL MU-MIMO” are in fact based on very different technologies: “UL MU-MIMO” doesn’t adopt “beamforming” as “DL MU-MIMO” does.

• “Multi-point to Point” secret

  Physically “DL MU-MIMO” performs by “Point to Multi-point” behaviors, in which timing, frequency and transmission power, etc. are all controlled by a single AP; “UL MU-MIMO” performs by “Multi-point to Point” behaviors in contrast. Synchronizing all independent client STAs to work orderly is the major concern for “UL MU-MIMO”, and “UL MU-MIMO” has to address some issues such as “CFO” (central frequency offset), “received power variance” in AP and “timing synchronization”. They are very important to mitigation of the potential ICI (inter carrier interference) and ISI (inter symbol interference) effects due to simultaneous transmissions from multiple client STAs if they are not properly synchronized.

• The rise of MIMO

  The WiFi 5 (11ac) supports only “DL MU-MIMO”, and its uplink traffics still follow “SU-MIMO”. For example, a 11ac 4x4 AP simultaneously serves four 1x1 STAs with “DL MU-MIMO”, but, the four STAs still need to compete for the air time and send “ACKs” back sequentially with “SU-MIMO” to confirm successful communications – a very inefficient way. “UL MU-MIMO” provides an opportunity for the “grouped” STAs to send “ACKs” back simultaneously to significantly improve uplink efficiency and to reserve more air time for effective data transmissions.

  Aside from improving efficiency, MU-MIMO can natively reduce the latency as well – though it is not a major advantage comparing to OFDMA introduced in next section.

OFDMA improves efficiency for WiFi networks from the “spectrum” perspective, and UL & DL OFDMA is one of the most notable and well-known WiFi 6 features among service providers. Its variant applications are now running over WiFi network, however some of the services such as AR/VR and 4K/8K streaming need higher throughputs while some services like gaming and VoIP need lower latency. Some services like IoT require higher reliability. To most services, large spectrum bandwidths such as 80/160 MHz are often more than necessary. With SU-MIMO/MU-MIMO, the whole “spectrum” bandwidth only serves a single STA at a time, which is inefficient utilizing the large spectrum in many application scenarios. OFDMA can divide a large spectrum bandwidth into several smaller RUs (resource units) to carry multiple services for different needs simultaneously. The 11ax defines seven RU sizes constructed by “26”, “52”, “106”, “242”, “484”, “996” or “2x996” SCs (sub-carriers). With the combinations of OFDMA, 8x8 MIMO, different MCS rates and GI (Guard Interval), the 11ax can support a range of data rates from 0.4 Kbps to 9.6 Gbps for different application needs. It is flexible to dynamically allocate RUs in different sizes within the large-spectrum bandwidth according to the STA needs to be fulfilled.

The 11ax defines a maximum of 74 RUs (26-SC) in a 160 MHz channel bandwidth, which means that as many as 74 STAs of different services can be served simultaneously by an AP. Comparing to SU-MIMO/MU-MIMO, the OFDMA can significantly improve the utilization efficiency of the large spectrum a WiFi network can offer. In addition, OFDMA brings other improvements such as lower latency and better reliability as well.

Higher allocation hit-rates can reduce latency.

Smaller RUs with narrower bandwidth can natively provide better SINR (Signal-to-Interference-plus-Noise Ratio). In general, better SINR means wider service coverage or more reliable communications with less retransmissions. It also benefits IoT applications and reliable “ACK” confirmations that don’t require high bandwidths.

For some real-time applications like gaming, lower latency is usually more important than large bandwidth from the user experience perspective. From the technology viewpoint, OFDMA is more capable of having higher grouping/scheduling hit rates per time slot for STAs running time-sensitive services; along with the better SINR thanks to the smaller RUs, the 11ax by OFDMA is expected to significantly improve the latency issue comparing to SU-MIMO/MU-MIMO.

Smaller RUs with narrower bandwidth can natively provide better SINR (Signal-to-Interference-plus-Noise Ratio). In general, better SINR means wider service coverage or more reliable communications with less retransmissions. It can benefit IoT applications with less power consumption.

Similar to DL & UL MU-MIMO, the “DL OFDMA” operations are “Point to Multi-points”, however the “UL OFDMA” operations are “Multi-points to Point”. In synchronization with “CFO”, the “received power variance” in AP and “timing” is also necessary for “UL OFDMA”. Just like “UL MU-MIMO”, “UL OFDMA” can especially improve uplink efficiency by grouped “ACK” confirmations from multiple STAs simultaneously and reserve more air time for effective data transmissions.

Now we have talked about many advantages of MU-MIMO and OFDMA, but they have disadvantages as well. In a typical WiFi network constructed by a “strong AP” (e.g. 4x4/8x8, 80/160 MHz) and several “thin client STAs” (e.g. 1x1/2x2, 20/40/80 MHz), MU-MIMO can well utilize resources of the “spatial” domain, but not the “spectrum” domain.

As OFDMA can flexibly utilize the resource of the “spectrum” domain rather than “spatial”, MU-MIMO-OFDMA is the way to inherit their advantages, discard the disadvantages and fully utilize the “spatial” and “spectrum” resources as well as possible. According to IEEE 802.11ax, MU-MIMO-OFDMA can support max. 128 RUs to service 128 client STAs in a group/schedule. Unfortunately, MU-MIMO-OFDMA is too complicated to implement. We don’t see the clear roadmaps from chipset vendors that MU-MIMO-OFDMA will happen in the short term.