Comparison of Two Differential Feedback Schemes for Beamforming

Comparison of Two Differential Feedback Schemes for Beamforming IEEE 802. 16 Presentation Submission Template (Rev. 9) Document Number: IEEE C 80216 m-09_0528 Date Submitted: 2009 -03 -07 Source: Qinghua Li, Yuan Zhu, Eddie Lin, Shanshan Zheng, Jiacheng Wang, Xiaofeng Liu, Feng Zhou, Guangjie Li, and Yang-seok Choi Intel Corporation Venue: Session #60, Vancouver , Canada Re: TGm AWD Base Contribution: None Purpose: Discussion and adoption by TGm AWD Notice: E-mail: guangjie. li@intel. com qinghua. li@intel. com This document does not represent the agreed views of the IEEE 802. 16 Working Group or any of its subgroups. It represents only the views of the participants listed in the “Source(s)” field above. It is offered as a basis for discussion. It is not binding on the contributor(s), who reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802. 16. Patent Policy: The contributor is familiar with the IEEE-SA Patent Policy and Procedures: <http: //standards. ieee. org/guides/bylaws/sect 6 -7. html#6> and <http: //standards. ieee. org/guides/opman/sect 6. html#6. 3>. Further information is located at <http: //standards. ieee. org/board/pat-material. html> and <http: //standards. ieee. org/board/pat >.

Outline • • • Background Two differential schemes Performance comparison Complexity comparison Conclusions Proposed text

Background • SDD supports differential feedback mode for SU and MU -MIMO precoding. • Only “rotation based schemes” are supported. • Two rotation based, differential schemes were proposed. – Scheme I: C 80216 m-09_0058 r 4. doc – Scheme II: C 80216 m-08_1187. doc • Comparisons are made.

System Model • is channel matrix of dimension • is beamforming matrix of dimension • is transmitted signal vector of dimension . . .

One-shot reset and differential feedback

Differential codebook

Scheme I • Actual quantization at SS: • Beamforming matrix reconstruction at BS: • Beamforming at BS: Note that quantization criterion here is better than that in C 80216 m-09_0058 r 4.

Scheme II • Actual quantization at SS: • Beamforming matrix reconstruction at BS: • Beamforming at BS: Note that quantization criterion here is better than that in C 80216 m-08_1187.

Codebook of Scheme I is compacter than Scheme II’s because of reduced dimension. • Scheme II always has unused codewords for each selected rank Ns. – Scheme II codebook is designed to cover every possible perturbations of identity matrix, while Scheme I codebook only covers necessary perturbations of e 1 and [e 1 e 2]. • Scheme I’s feedback matrix is Ntx. Ns e. g. 4 x 1 while Scheme II’s is always Ntx. Nt e. g. 4 x 4 regardless of rank. – Scheme I has a distinct codebook for each rank while Scheme II has a constant codebook for all ranks.

Other Differences • Stability – Scheme I can stay at optimum beamforming matrix because it includes e 1 and [e 1 e 2] in codebook. – Scheme II always vibrates around optimum beamforming matrix because it doesn’t have the identity matrix in codebook. • Adaptability – Scheme I has two codebooks for small and high correlation scenarios, respectively. The two codebooks are pre-defined and stored. – Scheme II varies codebook using measured correlation matrix and costly online SVD computation.

Scheme II’s complexity is more than triple of Scheme I’s because Scheme II uses 4 x 4 matrix operation rather than 4 x 1 or 4 x 2. 3 -bit Scheme I (No. of 4 -bit Scheme II (No. of real multiplications) 4 x 1 314 960 4 x 2 516 2352

System Level Simulations • Isolate effect of restart (or initial) feedback – No restart feedback – Only measure performance of differential feedbacks • No feedback error • Scheme I’s 3 -bit vs. Scheme II’s 4 -bit

Simulation Parameters Parameter Names Parameter Values Network Topology 57 sectors wrap around, 10 MS/sector MS Channel ITU PB 3 km/h Frame Structure TDD, 5 DL, 3 UL Feedback Delay 5 ms Inter cell Interference Modeling One tap fading Antenna Configuration 4 Tx, 2 Rx Code book configuration 16 e (4, 1, 3), TF with Rtx, Diff(4, 1, 3, Ф) Tx Channel Correlation Matrix (only for TF code book) Ideal known to both MS and BS Q matrix update frequency Once for the whole simulation PMI error free PMI calculation ML with SVD precoding vector System bandwidth 10 MHz, 864 data subcarriers Permutation type AMC, 48 LRU CQI feedback 1 Subband=4 LRU, ideal feedback

4 Tx (0. 5λ), 2 Rx, 1 Stream SE gain over 16 e 5%-ile SE gain over 16 e Scheme I: 3 -bit, 5 o 10. 94% 32. 87% Scheme II: 4 -bit, 7. 09% 25. 65% 3. 60% 5. 75% 0. 9 ρ Scheme I over Scheme II

4 Tx (4λ), 2 Rx, 1 Stream SE gain over 16 e 5%-ile SE gain over 16 e Scheme I: 3 -bit, 20 o 6. 08% 18. 86% Scheme II: 4 -bit, 5. 86% 15. 12% 0. 2% 3. 25% 0. 9 ρ Scheme I over Scheme II

4 Tx (Uncorrelated), 2 Rx, 1 Stream SE gain over 16 e 5%-ile SE gain over 16 e Intel: 3 -bit, 20 o 4. 2% 14. 06% Samsung: 4 -bit, 0. 9 4. 31% 12. 48% -0. 1% 1. 41% ρ Intel over Samsung Scheme II

Conclusions • In all cases, for both spectrum efficiency and 5%-SE, Scheme I outperforms Scheme II, except 0. 1% SE loss in uncorrelated channel. – Scheme I’s overhead is less than Scheme II’s by 25%. – Scheme II’s codebook has unused codewords for each rank. – Since Scheme II’s codebook doesn’t have identity matrix, it vibrates around optimum point. • Scheme I’s complexity is three times less than Scheme II’s.

Proposed Text • Adopt proposed text in AWD. 15. 3. 7. 2. 6. 4. 4 Differential codebook-based feedback mode

Figure xxx. Computation of Q(t-1) for two streams with four transmit antennas.

Differential codebook for 4 x 1 D 1=[1 0 0 0]T. Remaining codewords for small correlation are D 2 D 3 D 4 D 5 D 6 D 7 D 8 0. 9397 0. 0450 0. 0244 i 0. 2175 + 0. 0434 i 0. 0049 + 0. 0470 i -0. 2801 + 0. 1283 i -0. 0576 0. 3291 i 0. 0203 + 0. 0636 i 0. 0495 + 0. 0708 i -0. 0738 0. 3282 i 0. 1094 + 0. 0388 i -0. 3055 + 0. 1415 i 0. 0776 0. 0189 i 0. 0374 + 0. 0630 i 0. 0479 + 0. 0372 i 0. 1063 + 0. 0663 i 0. 0093 + 0. 0338 i 0. 0815 0. 2184 i 0. 0351 0. 0133 i 0. 0657 0. 1068 i -0. 0000 + 0. 0000 i -0. 3285 + 0. 0309 i 0. 1367 + 0. 2741 i Remaining codewords for large correlation are D 2 D 3 D 4 D 5 D 6 D 7 D 8 0. 9962 0. 9962 -0. 0483 + 0. 0650 i 0. 0015 + 0. 0046 i 0. 0381 + 0. 0130 i 0. 0533 0. 0087 i 0. 0011 0. 0029 i -0. 0471 0. 0661 i 0. 0014 0. 0049 i 0. 0007 0. 0183 i 0. 0019 0. 0268 i 0. 0470 + 0. 0459 i -0. 0201 0. 0559 i -0. 0722 + 0. 0486 i 0. 0137 0. 0066 i 0. 0290 + 0. 0132 i -0. 0106 + 0. 0244 i -0. 0035 0. 0827 i -0. 0405 + 0. 0046 i -0. 0041 + 0. 0336 i 0. 0024 i -0. 0242 + 0. 0142 i 0. 0805 + 0. 0083 i

Differential codebook for 4 x 2 Remaining codewords for small correlation are D 2 D 3 D 4 D 5 D 6 D 7 D 8 0. 7632 0. 6120 0. 8773 0. 4440 0. 8721 0. 3412 0. 6119 0. 741 2 0. 6783 0. 642 6 0. 9150 0. 3452 0. 7040 0. 6942 0. 3721 + 0. 4681 i 0. 4103 0. 6189 i 0. 1482 0. 4457 i 0. 259 0+ 0. 7820 i 0. 27 04 0. 2337 i 0. 6880 + 0. 6229 i 0. 6430 + 0. 4578 i 0. 46 60 0. 319 7 i 0. 61 94 + 0. 2977 i 0. 653 30. 306 4 i 0. 2124 0. 2666 i 0. 67 03 + 0. 5866 i 0. 0569 + 0. 6944 i 0. 046 90. 6331 i -0. 2161 + 0. 0474 i 0. 091 80. 0789 i 0. 040 0+ 0. 0657 i 0. 191 10. 0987 i 0. 01 92 0. 0470 i 0. 032 8+ 0. 0853 i 0. 0211 + 0. 0125 i 0. 218 50. 184 9 i 0. 0728 + 0. 2485 i 0. 040 7+ 0. 217 9 i 0. 15 81 0. 0545 i 0. 0650 0. 0997 i 0. 0673 + 0. 0963 i 0. 195 80. 2197 i -0. 1008 + 0. 0293 i 0. 0578 0. 2367 i 0. 035 2+ 0. 0512 i 0. 2588 + 0. 1046 i 0. 25 74 0. 2071 i 0. 067 80. 0965 i 0. 04 14 + 0. 0169 i 0. 214 10. 060 2 i 0. 01 42 + 0. 0180 i 0. 03 85 + 0. 125 5 i 0. 0479 + 0. 1276 i 0. 27 06 + 0. 0059 i 0. 00 87 0. 0707 i 0. 0117 + 0. 1684 i

Remaining codewords for large correlation are D 2 D 3 D 4 D 5 D 6 D 7 D 8 0. 6726 0. 7368 0. 8247 0. 5609 0. 1212 0. 9861 0. 6909 0. 7190 0. 6695 0. 7373 -0. 4177 0. 9018 0. 7603 0. 6466 0. 1854 0. 7155 i 0. 1657 + 0. 6449 i -0. 5290 0. 1993 i 0. 765 5+ 0. 2908 i 0. 9729 + 0. 1911 i 0. 1188 0. 0204 i -0. 1282 + 0. 7086 i 0. 116 00. 6771 i 0. 7303 + 0. 1094 i 0. 66 28 0. 0912 i -0. 8914 - 0. 1754 i 0. 4063 0. 0808 i 0. 476 20. 4393 i 0. 5576 + 0. 5080 i -0. 0022 0. 0126 i 0. 0581 + 0. 0663 i -0. 0056 + 0. 0013 i 0. 0982 0. 0589 i 0. 0422 + 0. 0181 i 0. 02 38 0. 0742 i -0. 0374 0. 0037 i 0. 0692 + 0. 0305 i 0. 0001 + 0. 0222 i 0. 00 01 0. 0495 i -0. 0069 + 0. 0059 i 0. 03 54 + 0. 1017 i 0. 012 70. 0244 i 0. 010 70. 0091 i -0. 0008 0. 0351 i 0. 0023 + 0. 0776 i -0. 0114 0. 0151 i 0. 037 50. 0167 i 0. 0019 0. 0089 i 0. 08 12 0. 0180 i -0. 0000 0. 0510 i 0. 0729 0. 0123 i 0. 024 5+ 0. 0736 i 0. 0665 + 0. 0425 i -0. 0006 - 0. 0026 i 0. 054 40. 0234 i 0. 0116 + 0. 0367 i 0. 0969 0. 0579 i

Backup

• Transition from restart feedback to differential feedback.