1. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 1 Q and A of Proposed 3-Dimensional Joint Interleaver for 802.11n MIMO Systems Jeng-Hong Chen (jhchen2@winbond.com Pansop Kim (pkim@winbond.com) Winbond Wireless Design Center Torrance, CA, USA September 15, 2004 Q and A of documents: IEEE 802.11-04/934r2 and 802.11-04/1026r0

2. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 2 Purpose This document is NOT made for a presentation. This document does NOT change the contents of presented proposal. This document is trying to answer the same asked questions by members after the presentation. You are very welcome to discuss more details with me. After the end of this September meeting, please send questions directly to the email address of Chen, Jeng-Hong.

3. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 3 Q and A (1) Q1: What is the relations between the proposed 3D interleaver and the proposed circular transmission? A1: Each can be implemented independently to improve the system performance. A MIMO system can have a 3D interleaver and without circulation or vice versa. But best performance can be expected if one implements BOTH. Q2: What is your recommendation of the proposed interlever? A2: The 3D-A interleaver is recommended over 3D interleaver. Also, the proposed Sub_BC is recommended over S_BC unless RF considerations in implementation.

4. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 4 Q and A (2) Q3: What is HW complexity to implement the proposed 3D interleaver ? A3: NONE. Exact the same as SMX or STBC. Just change your interleaver mapping. Q4: What is HW complexity to implement the proposed circulation if compare (a) N OFDM xN SMX, (b) N OFDM (M)xN Sub_BC CSMX, and (c) N OFDM (M)xN S_BC CSMX A4: Interleaver size: (a)=(b)=N OFDM, (c)=N OFDM xnPattern, nPattern=C(M, N OFDM ) on slide 98 (934r2) Decoding delay in OFDM symbols: (a)=(b)=0, (c)= nPattern-1 # of matrix inverses required for SMX/CSMX: (a)=one N OFDM xN OFDM matrix inverse per sub-carrier per packet (b)=(c)=nPattern different N OFDM xN OFDM matrix inverses per sub-carrier per packet # of channel coefficient estimation: one H Nx2 per subcarrier in (a), nPttern different H Nx2 per subcarrier per packet in (b) and in (c).

5. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 5 Q and A (3) A4 (continue): Suppose a particular MIMO system (AP or STS) already has 4x4 (shared) TX/RX antennas and a built-in 4x4 matrix inverse HW, if the SNR can only support 2 data streams (2xN SMX), one may chose a 2 (4)xN CSMX to have the extra TX diversity (circulation) gain. If one gets lucky, even a 3(4)xN CSMX may be supported from the extra TX diversity gain. In this case, a 2xN requires one 2x2 matrix inverse per subcarrier per packet but a 2(4)xN requires nPattern=C(4,2)=6 different 2x2 matrix inverses. However, the complexity of six 2x2 complex inverses shall be smaller than the build-in one 4x4 inverse. Also, the HW cost is cheaper than the expected 5+ dB sweet diversity gain in 11n channel B.

6. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 6 Q and A (4) Q5: What is the performance comparison of interleavers 3D and 3D-A? A5: Generally speaking, interleaver 3D-A is better than 3D. However, PER performance is related to the chosen parameters: N SC, N column and d free. N column shall be carefully chosen according to d free of selected CC. In general, a trade- off between N column, N row and N I. Q6: What is your recommendation for 40MHz BW? A6: We dont have enough simulation results to recommend an optimal N column now. We will try N column =18 for the 40MHz BW with N SC =108 after the meeting, i.e., a 6 sub-carrier separation based 3D-A interleaver. Q7: What is the performance comparison of CSMX and SMX without circulation? A7: Both S_BC and Sub_BC CSMX are much better than SMX without circulation. Performances of S_BC and Sub_BC CSMX are almost exactly the same.

7. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 7 Q and A (5) Q8: What is the backward complexity to 11a? A8: Both interleavers 3D and 3D-A are backward compatible (if N I =1). However, both 1 (M)xN S_BC and Sub_BC CSMX are NOT backward compatible to 11a. In both cases, require ALL channel estimates from ALL TX antennas. However, after the HT-Header, one can apply S_BC or Sub_C. Also, both Intereleavers 3D/3D-A and S_BC/Sub_BC are applicable to 20MHz and 40MHz, BPSK/QPSK/16QAM/64 QAM, and CALA/CSMX. Q9: What are the relationships of simulation results in Part-I/Part-II and Interleaver/Circulation transission? A9: All results in Part-I applied 3D/3D-A inerleaver WITHOUT any TX circulation. ALL results in Part-II (TX Circulation) already applied 3D (NOT 3D-A) interleaver. ALL results in Part-I and Part-II are under ideal conditions.

8. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 8 Q and A (6) Q10: How to implement the 3D or 3D-A interleaver from the provided Eqs? A10: Provided Eqs are used for math/documentation purpose only. You can hardwire the output from CC encoder bit-by-bit directly into the corresponding (subcariier) QAM mapper of IFFT. Every bit (of the 3D-A Interleaver) from CC output can be found in the Excel file (1026r0) for 20MHz BW with N SC =48. No need to implement the interlever using the modulo operations defined by the Eqs.Use the visual tables in the Excel file as the bit-mapping (Interleaver 3D-A).

9. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 9 Q and A (7) Q11: Does your proposal support Adaptive Coding Modulation on each TX antenna? A11: Yes. Both the Sub-BC and S_BC do support ACM with or without a 3D interlever. The 3D interleaver does support ACM with different coding modulation on each TX ant. However, this new interleaver is not a perfect 3-D block anymore. EX: 4 TX, 4 th TX= 16 QAM, N CBPS =192=12 rows*16 columns=4 th slice of Excel file 16QAM 3 rd TX= QPSK, N CBPS = 96= 6 rows*16 columns= 3 rd slice of Excel file QPSK 2 nd TX= 64 QAM, N CBPS =288=18 rows*16 columns= 2 nd slice of Excel file 64QAM 1 st TX= BPSK, N CBPS = 48= 3 rows*16 columns= 1 st slice of Excel file BPSK Just remember, size of EVERY column is 3 subcarriers (3 bits for BPSK, …,18bits for 64QAM). Also, when finished one particular row, remember to write the next row in the next slice into the next subcarrier. Please check the difference of BSK/QPSK/16QAM/64QAM. Also, ONLY apply the 2 nd permutaion on the slice with MQAM which M>=16.

10. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 10 Q and A (8) Q12: You have recommended 3D-A and Sub_BC. How to implement both effectively. A12: The proposed 3D/3D-A interleaver is nothing but a bit-by-bit mapping from the CC outputs to inputs of subcarrier QAM Mappers (e.g., each bit of 48*4 subcarriers). The proposed Sub_BC is nothing but another bit-by-bit mapping from 3D/3D-A interleaver outputs into the inputs of subcarrier QAM Mappers (e.g., each bit of 48*4 subcarriers). Therefore, if one implements both 3D/3D-A interleaver and Sub_BC, just use the Excel file and slides 100 and 101 of doc934/r2. That is, combine BOTH mappings together and you only need ONE mapping from CC encoder to subcarrier QAM Mappers.

11. doc.: IEEE 802.11-04/1105r0 Submission September 2004 Jeng-Hong Chen, Pansop Kim, Winbond ElectronicsSlide 11 Thank You!! The explosive 11n world requires an innovative, systematic structure to explore all available diversities in space, frequency and timeA 3D interleaver and the circular transmissions fit this purpose in all dimensions!!