1. DL MU-MIMO MAC efficiency issue
Month Year
doc.: IEEE yy/xxxxr0
DL MU-MIMO MAC efficiency issue
Date:
Authors:
John Doe, Some Company

2. Abstract
MU-MIMO technology essentiality for TGac is no longer to demonstrate
We need to make sure that this technology is efficient and performant.
MU-MIMO process can be divided into two phases
The feedback phase and the transmission phase
Previous works have shown that the feedback phase should be made approximately every 20 to 25ms (coherence time of 800ms) [1]
In this presentation we want to point out that such a low feedback period can strongly reduce the MAC efficiency of MU-MIMO, and reduce its usage.
To efficiently solve this problem,
Using only explicit feedback, compression of the CSI report needs to be supported
If we use explicit and implicit feedbacks, we could also update the feedbacks during transmission phase
Slide 2

3. MU-MIMO feedback phase duration
MU-MIMO process can be divided into two phases, the feedback phase and the transmission phase
If we assume explicit CSI feedback, the time duration of the feedback phase can be quite important
For the feedback phase of a group of 4 users with 2 antennas each and an AP with 8 antennas
with 6 bits quantization
MCS2 applied on UL CSI reports:
Feedback phase time duration
40MHz
80MHz
160MHz
Size of CSI report
4x2,7 kB
4x5,4 kB
4x10,9 kB
12Mbps / MCS2
2,8 ms
5,1 ms
9,6 ms
Slide 3

4. MU-MIMO feedback phase duration
Example for 80MHz, with a feedback phase (overhead) of 5,1 ms (MCS2 for CSI report) every 25ms
Feedback TXoP
Data TxOP
25 ms
Overhead
Data
4 users
1 MU-MIMO group
8 users
2 MU-MIMO groups
12 users
3 MU-MIMO groups …
16 users
4 MU-MIMO groups
time (ms)
Slide 4

5. Impact on MU-MIMO MAC efficiency
The MAC efficiency of MU-MIMO is strongly affected by the increase of the number of users with 25 ms feedback period
Calculations with explicit feedback, 80MHz, MCS2 for CSI reports,
"8 antennas AP" and 4 "2antennas STA"
TxOP of 3ms for transmission phase
Slide 5

6. MU-MIMO MAC efficiency analysis
The MAC efficiency is strongly affected by a low duration between feedbacks
For high number of users (MU-MIMO full load)
For higher bandwidth
In highly loaded environnement
This will reduce its usage: in our example, after 10 users, it seems not worth performing MU-MIMO
We therefore need to improve MU-MIMO MAC efficiency
The MAC efficiency gains provided by an increase of the feedback period are significant if the feedback period increase is significant
Interference cancellation's benefits may not be sufficient
With only explicit feedback, we need to provide strong compression mechanisms
With explicit and implicit feedbacks, we can extend the feedback period by refreshing the CSI during the feedback phase
Slide 6

7. One of the potential feedback architecture solution
During the feedback phase, implicit feedback could be used, but it seems that explicit feedback is favored
We therefore assumed that explicit feedback would be used during the feedback phase
During the transmission phase, implicit feedback could then be performed using Sounding PPDU (staggered pilots) for BlockAck UL transmissions.
The implicit feedback, which follows the time variation of the channel, would be used to update the CSI and re-compute the MU-MIMO beams
This would increase the period of the explicit feedback phase
Slide 7

8. One of the potential feedback architecture solution
Calibration performed during the first TxOP of the transmission phase
CSI updates and beam-recomputation performed on the following TxOPs
Feedback phase
Transmission phase

9. One of the potential feedback architecture solution
Calibration performed during the feedback phase
CSI updates and beam-recomputation performed on TxOPs in the transmission phase
Feedback phase
Transmission phase

10. Conclusion
Analysing MAC efficiency for MU-MIMO with explicit feedbacks leads to strong inefficiencies with a feedback period of 25ms
We believe that this problem needs to be solved, for the quality of user experience
With explicit feedback as the only feedback mechanism, strong compression of CSI report is needed
With explicit and implicit feedback mechanism, CSI updates during the transmission phase could extend the feedback period
Slide 10

11. References
[1] Ermani, S and Van Zelst, A, Interference cancellation for Downlink MU-MIMO, IEEE /1234r1, Mar. 2010
[2] Ishihara, K. and Yasushi, T., CSI Report for Explicit Feedback Beamforming in Downlink MU-MIMO, IEEE /0332r0, Mar. 2010
Slide 11