1. doc.: IEEE 802.11-11/1484r4 Submission January 2012 Hongyuan Zhang, et. Al.Slide 1 11ah Data Transmission Flow Date: 2012-01-16 Authors:

2. doc.: IEEE 802.11-11/1484r4 SubmissionHongyuan Zhang, et al Authors continued: Slide 2 January 2012

3. doc.: IEEE 802.11-11/1484r4 SubmissionHongyuan Zhang, et. Al.Slide 3 Abstract This presentation proposes the data transmission flows for 11ah PPDUs, including 1MHz, 2MHz, and 4 / 8/ 16MHz packets. January 2012

4. doc.: IEEE 802.11-11/1484r4 Submission Overview TGah decided to have 2/4/8/16MHz signals that are basically down clocked versions of 11ac 20 / 40 / 80 / 160 MHz, also an 1MHz signal based on 32FFT was introduced. –Both 2MHz and 1MHz receptions are mandatory. We propose the 11ah data transmission flow in this presentation. –Flow for regular MCS. –Flow for a new MCS0- 2x Repetition mode for range extension. Hongyuan Zhang, et. Al.Slide 4 January 2012

5. doc.: IEEE 802.11-11/1484r4 Submission I. Transmission Flow for Regular MCSs Apply the same Tx flow in 11ac in the data tones for the data field. Hongyuan Zhang, et. Al.Slide 5 January 2012

6. doc.: IEEE 802.11-11/1484r4 Submission Tone Plans >=2MHz follows the same tone plans as the corresponding FFT sizes in 11ac. Define the 1MHz tone allocation as: 24 Data tones, 2 Pilot tones at tone indices +/-7, 3 Guard tones on left and 2 guard tones on right, and 1 DC tone. –24 Data tones does not impose any MCS exclusion as seen in 11ac, suitable for all MCSs. The next “good number” is 18 tones, but the spectrum waste would be too high. –2 and 3 guard tones are not worse than 64FFT (3/4 guard tones) regarding the filter designs. –2 pilot tones seems sufficient for phase/frequency tracking (see Appendix). January 2012 Hongyuan Zhang, et. Al.Slide 6 012 ….. 21314-13-2 ….. -14 Data (pilot at -7)Data (pilot at +7) 1415-15-16

7. doc.: IEEE 802.11-11/1484r4 Submission Discussions Each block is the same as or similar to the corresponding part of 11ac. –The 11ah stream parser is the same as 11ac. –The 11ah encoder parser and segment parser (16MHz only) are the same as 11ac; N ES in the MCSs of 2/4/8/16MHz is the same as the corresponding values in 11ac; and N ES = 1 in all the MCSs of 1MHz. Up to 4 space-time streams (refer to [3]). The draft specification shall support following MCSs: –For 1MHz, 11ac MCS0~9, as well as an MCS0-rep2 mode. –For >=2MHz, the MCS tables for BCC are the same as the corresponding tables in 11ac before downclocking, i.e. same MCS exclusions for BCC as in 11ac. In 1MHz, interleaver is applied for all MCSs incl. MCS0-Rep2, Ncol=8, Nrot TBD. (refer to [4]) Hongyuan Zhang, et. Al.Slide 7 January 2012

8. doc.: IEEE 802.11-11/1484r4 Submission II. MCS0 Rep2 Mode in 1MHz An MCS0-Rep2 mode in 1MHz is needed to improve the range (refer to [1]). –Lead to the lowest data rate ~150Kbps, reach the longest range –~6dB gain compared with MCS0 of 64FFT mode. Similar to the gap between 1Mbps (DSSS) and 6Mbps (11g) in 2.4GHz. We propose a simple 2x repetition mode that imposes minimal Tx and Rx design change from regular MCSs. November 2011 Hongyuan Zhang, et. Al.Slide 8

9. doc.: IEEE 802.11-11/1484r4 Submission MCS0 Rep2 Tx Flow Hongyuan Zhang, et. Al.Slide 9 Scrambler Interleaver FEC 2x Block-Wise Repetition BPSK mapper Spatial Mapping IDFT :::: GI & Window Analog & RF GI & Window Analog & RF After repetition, N CBPS = 12 (same as 1SS-MCS0) bits per symbol January 2012 Ncol = 8

10. doc.: IEEE 802.11-11/1484r4 Submission Discussions MCS0 Rep2 is applied only for single space-time stream. –N SS =1, no STBC The “2x block-wise repetition” performed on a per-OFDM symbol basis: –C out =[C 1 ….C 2N DBPS, C 1 ….C 2N DBPS ], where [C 1 ….C 2N DBPS ] are the FEC output bits per symbol. Interleaver parameters are the same as regular MCS0. –Ncol = 8 Receiver may conduct MRC combining to improve SNR. Hongyuan Zhang, et. Al.Slide 10 January 2012

11. doc.: IEEE 802.11-11/1484r4 Submission Performance in AWGN, 2MHz Receiver Hongyuan Zhang, et. Al.Slide 11 Assumptions: 4 LTF1 repetitions in 1MHz Preamble (refer to [1]) 1MHz with 24 data tones, Intlv N col =8 SNR is time domain: avg received power over noise in 2MHz January 2012

12. doc.: IEEE 802.11-11/1484r4 Submission Performance in SCM-UMa, 2MHz Receiver November 2011 Hongyuan Zhang, et. Al.Slide 12

13. doc.: IEEE 802.11-11/1484r4 Submission III. Short GI Support In [1][2] we propose a green field type of preamble for single user packets. Note that in 11n Green Field preamble, Short GI is not allowed if N STS =1, due to the decoding delay for the HTSIG field. In 11ah, most of the traffic will be SU 1SS, not allowing Short GI loses about 10% of peak throughput. –It is also easier to support SGI than in 11n/ac, because even SGI (4us) is much longer than a typical indoor channel multipath delay. Propose that in any 11ah short GI packet, short GI starts from the 2nd Data symbol, and the 1st Data symbol is always long GI. –Include Multi-stream or MU packets. –For a unified receiver processing. November 2011 Hongyuan Zhang, et. Al. Slide 13 HT-STF HT-LTF1HT-SIGDATA1DATA2DATA3 ….. Decoding delay STF LTF1SIG Data1Data2Data3Data4 ….. Indicates SGI (LGI)(SGI)

14. doc.: IEEE 802.11-11/1484r4 Submission Straw Poll – 1 (from Nov meeting) Do you support the general transmission flow for 11ah regular MCSs as in slide 5? Hongyuan Zhang, et. Al.Slide 14 January 2012

15. doc.: IEEE 802.11-11/1484r4 Submission Straw Poll – 2 (new in r4) Do you support the 11ah tone plans as specified in slide 6? Hongyuan Zhang, et. Al.Slide 15 January 2012

16. doc.: IEEE 802.11-11/1484r4 Submission Straw Poll—3 (new in r4) Do you support the followings regarding parsers? –The 11ah stream parser is the same as 11ac. –The 11ah encoder parser and segment parser (16MHz only) are the same as 11ac; N ES in the MCSs of 2/4/8/16MHz is the same as the corresponding values in 11ac; and N ES = 1 in all the MCSs of 1MHz. January 2012 Hongyuan Zhang, et. Al.Slide 16

17. doc.: IEEE 802.11-11/1484r4 Submission Straw Poll –4 (new in r4) Do you agree with the followings regarding 11ah MCS table? –For 1MHz, 11ac MCS0~9, as well as an MCS0-rep2 mode. –For >=2MHz, the MCS tables for BBC are the same as the corresponding tables in 11ac before downclocking, i.e. same MCS exclusions for BCC as in 11ac. January 2012 Hongyuan Zhang, et. Al.Slide 17

18. doc.: IEEE 802.11-11/1484r4 Submission Straw Poll – 5 (from Nov meeting) Do you support the transmission flow for MCS0-Rep2 mode as specified in slides 9~10? Hongyuan Zhang, et. Al.Slide 18 January 2012

19. doc.: IEEE 802.11-11/1484r4 Submission Straw Poll – 6 (from Nov meeting, with a small change) Do you support that in any 11ah short GI packet, short GI starts from the 2nd Data symbol, and the 1st Data symbol is always long GI? –Include Multi-stream or MU packets. Hongyuan Zhang, et. Al.Slide 19 January 2012

20. doc.: IEEE 802.11-11/1484r4 Submission References [1] 11-11-1482-00-00ah-preamble-format-for-1-MHz [2] 11-11-1483-00-00ah-11ah-preamble-for-2MHz-and-beyond [3] 11-11-1275-01-00ah-spatial-stream-support [4] 11-12-0113-00-00ah-32FFT-Interleaver Hongyuan Zhang, et. Al.Slide 20 January 2012

21. doc.: IEEE 802.11-11/1484r4 Submission Appendix Simulations to compare 2 pilots versus 4 pilots. –Assume 40ppm CFO, 1x1, DNLOS or AWGN, MCS7, pilots are used to track phase and correct CFO. –First set of simulations is to compare the gap of CFO on/off for 64FFT (4 pilots) and 32FFT (2 pilots). –Second set of simulations is to directly compare the gap between 2 pilots and 4 pilots in 32FFT. Fix to 24 Data tones, compare the PER of 2 pilots, and 4 pilots. –The 4 pilot case is a “fake” scenario where 2 extra guard tones are used for pilots—just for a fair PER comparison (rule out impacts from different encoders, interleavers, etc). –Tx power is adjusted to make sure per-tone SNR is the same as the 2 pilot case, for a fair comparison. January 2012 Hongyuan Zhang, et. Al.Slide 21

22. doc.: IEEE 802.11-11/1484r4 Submission AWGN, MCS7, 2MHz Receiver Marvell Slide 22

23. doc.: IEEE 802.11-11/1484r4 Submission DNLOS, MCS7, 2MHz Receiver Marvell Slide 23

24. doc.: IEEE 802.11-11/1484r4 Submission DNLOS, MCS7, 1MHz Receiver Marvell Slide 24

25. doc.: IEEE 802.11-11/1484r4 Submission Discussions CFO tracking effect by 2 pilots in 32FFT is same or better than that with 4 pilots in 64FFT. In 1MHz, 4 pilots show some 0.5~1 dB PER gain over 2 pilots. When using 4 pilots in reality, Ntones becomes 22 tones, so peak throughput drops 8.3%. It is then interesting to look at the “Goodput”: T = R*(1-PER). –See next page, where the 2 pilots lead to higher goodput. Assume select an MCS at 10% PER. Assume 22+4 Tones case gets the same PER as the 24+4 Tones case we simulated—for the same per-tone SNR, i.e. ignore impacts from different coding and interleaving. Limit to SNR range for MCS5~7. Marvell Slide 25

26. doc.: IEEE 802.11-11/1484r4 Submission Cont’d Marvell Slide 26 2 Pilots (24 Tones) shows higher goodput over 4 pilots (22 tones) at the considered SNR range, and considered channel.