WFA VHT Study Group Consolidation of Usage Models V7

July 2010 doc: IEEE 802.11-10/0807r0 PHY Abstraction for MU-MIMO Date: 2010-07-11 Authors: Phone email Riichi Kudo +81-46-859- 3140 [email protected] Yusuke Asai +81-46-859-3578 [email protected] Koichi Ishihara +81-46-859- 4233 Name Masato Mizoguchi Submission Affiliations NTT Corporation Address Hikarino-oka Yokosukashi Japan Slide 1

+81-46-859- 3758 [email protected] o.jp [email protected] t.co.jp R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 Introduction Two approved strawpolls in TGac suggest to add PHY abstraction model to the FR-EM document [1] Do you think PHY abstraction method for MU-MIMO is needed? Do you agree to include PHY abstraction methods as a recommendation in FR&EM document? Show problems of PHY abstraction for MU-MIMO Present PHY abstraction method using a look-up-table Submission Slide 2 R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 Overview of PHY / MAC simulation One possible example of PHY/MAC simulation Channel model Channel Generation Precoding scheme

Channel PER Calculation MAC Protocol PER Data Set Data Set MAC Simulation MAC/SAP throughput This presentation focuses on Major problems of PHY abstraction for MU-MIMO Large computational load imposed by PER calculation How to reflect performance dependence of user combinations How to consider performance degradation caused by aged CSI Submission Slide 3 R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 PHY abstraction in TGn Look up table (LUT) for PER reduces calculation complexity. In [2] and [3] for IEEE 802.11 TGn, the relationships between PER and average SNR are shown for various MIMO configurations and MCS indices. We calculated the relationship between PER and average SNR

in MU-MIMO. Submission Slide 4 R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 PER versus average SNR PER versus post processing SNR is shown for MU-MIMO. PER inaccuracy is very large at certain SNR value. Number of channel conditions for multiple users is also large. 0 10 10 -1 PER PER 10 -2 10 -3 10 -4 10 10

20 25 Average SNR [dB] 30 35 PER of 500-byte packets in 2 Tx 1 Rx 2 user 40MHz, ZF precoding, Ch.C, and NLOS. Submission 10 -1 10 -2 10 -3 64QAM 3/4 BCC 15 Slide 5 0 10 -4 16QAM 3/4 BCC 10 15

20 25 Average SNR [dB] 30 35 PER of 500-byte packets in 4 Tx 2 Rx 2 user 40MHz, ZF precoding, Ch.C, and NLOS. R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 PHY Abstraction for MU-MIMO We present a look up table for PER calculation block PHY abstraction using LUT holding PER and average SNR Channel Calculate postCalculate Average Calculate PER using processing SNR SNRs for all LUT of PER versus average SNR in all subcarriers i,j subcarriers PER Proposed PHY abstraction using LUT holding PER and BER Channel Calculate postCalculate BER for Calculate PER using processing SNR each subcarrier LUT of PER versus i,j in all subcarriers with error function P BER

i,j: SNR of j-th data stream at i-th subcarrier : Average SNR P: Calculated BER using error function N N 1 P i , j SS N SS N SD Submission SD j 1 i 1 Slide 6 1 N SS N SD PER N SS N SD Aerfc j 1 i 1 R. Kudo et al., NTT i , j B July 2010 doc: IEEE 802.11-10/0807r0 PER versus Calculated BER in Proposed PHY Abstraction PER versus uncoded BER shown. Variance of PER for estimated BER is smaller than that for SNR

0 10 10 -1 PER PER 10 -2 10 -3 10 0 10 -1 10 -2 10 -3 64QAM 3/4 BCC -4 10 -3

10 -2 -1 10 10 Estimated raw BER 10 0 PER of 500-byte packets in 2Tx1Rx2user 40MHz, ZF precoding, Ch.C, and NLOS. Submission Slide 7 10 16QAM 3/4 BCC -4 -3 10 -2 -1 10 10 Estimated raw BER 10 0

PER of 500-byte packets in 4Tx2Rx2user 40MHz, ZF precoding, Ch.C, and NLOS. R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 Calculation Complexity The required time for PER calculations is shown. Bit amount: 2.16108 , Coding rate: 3/4, Mod.: QPSK Direct calculation of PER PHY abstraction using average SNR Proposed PHY abstraction using uncoded BER 556 sec 3.5 sec 3.6 sec System: Microsoft Windows XP, Intel(R) Xeon(R) CPU X5472 @ 3.00GHz 2.99GHz Submission Slide 8 R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 Conclusion Showed accuracy of PER estimation using average SNR and uncoded BER. Proposed PHY abstraction on the basis of uncoded BER; has higher accuracy than that possible with average SNR. Computational load of proposed PHY abstraction much lower than that of direct calculation. Submission

Slide 9 R. Kudo et al., NTT July 2010 doc: IEEE 802.11-10/0807r0 References [1] 11-10/0334r0 PHY Abstraction for MU-MIMO in TGac [2] 11-04/0895r6 TGn Sync TGn Proposal MAC Simulation Methodology [3] 11-04/0877r9 WWiSE Proposal Response to Functional Requirements and Comparison Criteria Submission Slide 10 R. Kudo et al.,(NTT) July 2010 doc: IEEE 802.11-10/0807r0 Straw poll Do you agree to add the following sentences in section 3 of FREM document to allow TGac proposal to use PHY abstraction? Each TGac porposal may use a PHY abstraction method. If a PHY abstraction method is used, the method must be described and disclosed. Yes/ No / Abstain Submission Slide 11 R. Kudo et al., NTT