MIMO Sharing Question

[PinkFloydEffect]

When choosing an AP we have the option between SU-MIMO and MU-MIMO. My question is if you have a 3x3 SU-MIMO AP and all your clients are 1x1 mobile devices will it work like a MU-MIMO since it has 3 antennas and each client only needs one antenna to communicate?

My current AP is a SU-MIMO 3x3 and the next model up is a MU-MIMO 4x4 and I would like the 4x4 bandwidth to use with a 4x4 PCI card but have no need for MU-MIMO so I am paying for something I do not need just for the extra bandwidth.

Really curious though how SU-MIMO devices with multiple antennas work with multiple single antenna clients.

 

[sfx2000]

It will work just fine - and for most, SU-MIMO is a good place to be with most clients.

MU does improve overall channel utilization, but it's been problematic across different vendors, and even then, individual client performance might suffer.

MU comes into bigger play with WAN technology like LTE, where things are a bit different... MU was kinda patched into 11ac, even there, it's limited in scope/capability as it's DL only, and the UL feedback to the AP is per client in a SU scheme.

I'm keeping an eye on what's going on with 11ax, as it's a completely different MAC layer, and with an OFDMA access method, and uplink/downlink, MU might have a chance.

 

[Internet Man]

My question is if you have a 3x3 SU-MIMO AP and all your clients are 1x1 mobile devices will it work like a MU-MIMO since it has 3 antennas and each client only needs one antenna to communicate?

No, that's the whole point of MU-MIMO. A SU-MIMO access point only talks to one client at a time regardless of whether the available spatial streams are saturated.

It seems like newer hardware which happens to support MU-MIMO also performs better for other reasons though.

 

[PinkFloydEffect]

Thanks a lot guys, valuable information. I was comparing the Ubiquity AP AC Pro with the HD version and that is where this question was born. Just installed an AC Pro in my home with about 20 devices, the Chromecast would give intermittent bad picture quality using the ISP's AP-all-in-one unit. Will see how this works out

PS- Love this forum!

 

[sfx2000]

It seems like newer hardware which happens to support MU-MIMO also performs better for other reasons though.

Fun thing about MU-MIMO client stations that are broadcom based - esp. with SmartPhones - they'll go to single stream...

Many thought this was a bug, it's actually intentional to reduce interference impacts with MU...

 

[thiggins]

Fun thing about MU-MIMO client stations that are broadcom based - esp. with SmartPhones - they'll go to single stream...

Many thought this was a bug, it's actually intentional to reduce interference impacts with MU...

So two stream MU is a bad thing?

 

[sfx2000]

So two stream MU is a bad thing?

Not at all - each chipset vendor models the channel, and what works best for them.

In a DL MU group, each STA has to monitor the streams intended for itself, and the other STA's in the DL transmission to minimize the interference - single stream makes the math much easier for the client and the AP.

Remember with MU - the AP has to model the beam forming picture in relation to the client channel feedback, and still form the MU groups accordingly. That, along with the appropriate rates (MCS) for each client sta... one interesting outcome of MU is that channel noise in the MU frame is going to be higher, much higher, than in a SU frame -- which impacts things even more of a challenge to maintain high data rates.

Keep in mind that the AP sets the MCS based on the overall noise in the channel, along with feedback from the client, and the RSSI and CINR of the uplink from the client stations. Even with STBC's, MU starts running into Shannon limits, being noise limited due to inter-symbol interference within the channel when the frame is eventually transmitted.

Conducted path, or in a controlled radiated environment like an RF chamber is one thing, but in the real world, one also has to consider multiple paths - which can be either beneficial or destructive, and adjacent BSS's - throw in the SU/Legacy STA's, and the AP's job becomes that much more of a challenge - QCA has done the best job on that one, IMHO, and I can appreciate the challenges that the other chipset vendors have faced (Broadcom, Quantenna for example)

It's a wonder that MU works at all in the field

There's still benefit, as aggregate channel utilization within the BSS is higher, but each client is likely going to get lower bandwidth because of the things I mentioned above.

Keep in mind that like I mentioned earlier, it's not a flaw... it's physics, and some chip vendors will go single stream to get better bandwidth per frame.

Apologies if this is over the head of some of the forum members.

11ax does take steps to remove some of the limits that we've seen with 11ac DL-MU, and some of those steps are the same ones we used for 802.16m and LTE...

 

[thiggins]

one interesting outcome of MU is that channel noise in the MU frame is going to be higher, much higher, than in a SU frame

Sorry, but I'm not seeing where the increased channel "noise" is coming from. Higher channel use/congestion yes. Noise, no.

 

[sfx2000]

Sorry, but I'm not seeing where the increased channel "noise" is coming from. Higher channel use/congestion yes. Noise, no.

Tim, I think you know the answer already, but it's nice to ask the leading question, as you know I'll respond in kind

Really goes down into the PHY itself - the frame is orthogonal, but the streams within the frame are only quasi-orthogonal...

If we have three spatial streams, even with four radios, we have a limited set of choices, and all three streams are transmitted to all clients _at the same time_ - so if we have 3 clients in the frame, even with STBC's on each stream, the symbols for all three are present. Each STA has to pick out the relevant bits from that frame - the non-relevant bits are not useful, thus are noise... and that's a 6 dB hit at the PHY, by the times info gets up past the baseband, we have a string of bits that might or might not be a frame - if the CRC fails there, then the frame is dropped...

At the client side - I can drop from 2 streams in MIMO to one stream, and get 3 dB back, and that can be an MCS step up by one or two even, depending on the channel state at the time, and have a better chance of getting good bits.

SU in 11ac is fairly accurate, even when Beamforming is in place, and we see error rates at the PHY of around 2 to 4 percent - without beamforming, it's close to zero - in MU, packet error rates can approach 30 percent or higher - and the AP needs to sort out if a member should be dropped from a MU group or not based on PHY feedback, MAC layer feedback, and even network layer.

(note - it's hard to see this unless you have deep access into the WiFi chipset itself, or on a lab benchtest gear with the appropriate libraries, but it can be inferred to some degree just looking at the 802.11 traffic in wireshark - even with a MU capable client, one only sees the client itself if the driver allows, there are test/development builds at the AP side that can dump more data there)

That's the challenge with MU and commodity chips - inaccurate channel state information can cause problems like this in the simple MU that 11ac does. Part of the issue with the CSI loop is that full CSI is a lot of overhead to work with, so vendors use an estimate, and sometimes it works, and sometimes it doesn't.

802.16m and LTE addresses this by additional work, as does 11ax.

 

[thiggins]

Thanks, SFX. I really didn't know. Thanks for the lesson.

 

[sfx2000]

Thanks, SFX. I really didn't know. Thanks for the lesson.

No worries - and good to pick a thread to share info.

sfx

 

[dfarning]

Does anyone have any pointers to intro level information on how spacial stream works at the RF level? I have figured out how beam forming working by timing the signals from different radios to create constructive and destructive interference patterns

But, for some reason I have not found a good introduction to the physics behind spacial streams.

 

[degrub]

https://www.google.com/search?clien...ommunication&sourceid=opera&ie=UTF-8&oe=UTF-8

try "spatial"

 

[dfarning]

Thanks, I was searching for "spatial streams" and was getting buried in results that only said "more streams = higher data rate"

Found a great reference in a section titled 'Multiple Antenna Technologies' in a Coursera course on wireless communications.

https://www.coursera.org/learn/wireless-communications/home/week/4