Project: 802.11n TG High Throughput WLAN Month 2002 doc.: IEEE 802.11-02/xxxr0 November 2004 Project: 802.11n TG High Throughput WLAN Submission Title: [11-04-1400-00-000n Comparison of proposals from TGn Sync and WWiSE with suggested evaluation additions ] Date Submitted: [November, 2004] Source: [John Egan, Rodger Tseng, Albert Liu] Company [Infineon Technologies] Address: [ ] Voice:[ Egan: 727.789.8857, Tseng & Liu: +886-3-5788879] E-Mail:[ Egan.external@infineon.com, \, albert.liu@infineon.com ] Re : [] Abstract:[This document proposes potential changes to what usage and applications should be considered for proposal analysis (effects document “11-03-0802-12-000n-usage-models” IEEE 802.11-03/802r12) and a brief analysis of differences between the proposals from TGn Sync and WWiSE] Purpose:[To improve the TG’s ability to determine the best route forward while also keeping market/end user requirements in mind as part of the analysis] Notice:[This document has been prepared to assist the IEEE P802.11n. It is offered as a basis for discussion and is not binding on the contributing or supporting individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Based on changes the list of supporters may increase or decrease.] Release: [The contributors acknowledge and accept that this contribution becomes the property of IEEE and may be made publicly available by P802.11.] John Egan, Rodger Tseng, Albert Liu, Infineon Technologies John Doe, His Company

November 2004 Overview There should be more Use and Application cases. Here, we suggest three areas for inclusion. Eventual analysis of proposals must include verification against agreed upon scenarios. The TGn Sync and WWiSE proposals were analyzed by Infineon. The other proposals were not analyzed due to resource and time constraints, not due to an Infineon view that they do not have merit consideration. John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Topics Usage Additions based on Applications November 2004 Topics Usage Additions based on Applications (Speaker : John Egan) TGn Sync & WWiSE Comparisons on PHY (Speaker : Rodger Tseng) TGn Sync & WWiSE Comparisons on MAC (Speaker : Albert Liu) Summary John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Use and Applications Additions November 2004 Use and Applications Additions Speaker: John Egan John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Use and Applications Additions November 2004 Use and Applications Additions PDA Low power receiver and transmitter Operating at edge of service area Requires sleep mode for 802.11n module Mobile while in session Must not lose session Streaming or random packets Unidirectional for most streaming Bidirectional for random John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Use and Applications Additions November 2004 Use and Applications Additions Wireless PAD Envisioned to be mobile screen pad, next generation of Tablet PCs of today Battery powered so must have power conservation mode Mobile while in sessions Video conference and streaming video required so must be bidirectional and requires error correction based on session type John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Session Mobility Has this been covered in simulations? November 2004 Session Mobility Has this been covered in simulations? Is this similar to “Measuring the Impact of Slow User Motion on Packet Loss and Delay over IEEE 802.11b Wireless Links” submitted to 802.11b by Christian Hoene, André Günther, Adam Wolisz of the Technical University of Berlin? Must include this in comparison simulations John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

TGn Sync & WWiSE Comparisons on November 2004 TGn Sync & WWiSE Comparisons on PHY Speaker: Rodger Tseng John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Throughput enhancement November 2004 Features T W Bandwidth extension 20MHz  40MHz mode (M) 20MHz mode (M) 40MHz, whenever regulatory domain permits this extension (M) 20 MHz mode (O) 40 MHz mode MIMO-OFDM-SDM (M) 2 spatial streams @ 20MHz mode (M) 2 spatial streams Guard interval (GI) shortening ( 0.8us  0.4us ) (M) (N) Higher code rate (R) (M) R= ½, 2/3, ¾, 7/8 (M) R= ½, 2/3, ¾, 5/6 Higher order modulation scheme (O) 256 QAM (ABF-MIMO mode) Adaptive modulation (O) Bit loading (+ 256 QAM) + power weighting (ABF-MIMO) Reserve more data tones (M) 48 (4 pilots) @ 20MHz (M) 108 (6 pilots) @ 40MHz (M) 54 (2 pilots) @ 20MHz (O) 108 (4 pilots) @ 40MHz (M) Mandatory (O) Optional (N) Not available John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Throughput comparison (Maximum achievable uncoded data rate @ 64QAM) November 2004 Throughput comparison (Maximum achievable uncoded data rate @ 64QAM) T W 20 MHz BW + 2 Tx (M) 108 (R=3/4) (M) 140 (R=7/8 with ½ GI) (M) 121.5 (R=3/4) (M) 135 ( R=5/6 ) 20 MHz BW + 4 Tx (O) 280 (R=7/8 with ½ GI) (O) 270 ( R=5/6 ) 40 MHz BW + 2 Tx (M) 243 (R=3/4 ) (M) 315 (R=7/8 with ½ GI) (O) 243 (R=3/4) (O) 270 (R=5/6) 40 MHz BW + 4 TX (O) 630 (R= 7/8 with ½ GI) (O) 540 (R=5/6 ) (M) Mandatory (O) Optional Observations: (1). T > W, between “R=7/8 with ½ GI” for T and “R=5/6” for W. (2). W >= T, at “ R=3/4 & 2Tx” John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Performance improvement November 2004 Performance improvement Features T W Bandwidth extension (M) (O) Mandatory coding scheme (M) Convolutional code (M) Convolutional code Advanced Coding scheme (O) LDPC (O) RS+Conv (removed) (O) LDPC (O) Turbo Duplicate format @ 40 MHz (O) 6Mbps BPSK, (R = ½) (N) Spatial processing Eigenvector steering (ES) Spatial spreading (SS) (O) SVD_MIMO (O) Walsh+CS Space Time Block Code (STBC) (M) Mandatory (O) Optional (N) Not available John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Observations & conclusions before further detailed analysis November 2004 Observations & conclusions before further detailed analysis Features T W Preamble format Neat More Efficient (shorter) MIMO Channel estimation Simple Needs more adders for interpolations Spectrum Fewer null sub-carriers around DC @ 40MHz mode Wider @ 20MHz mode Throughput (for long packets under identical MAC efficiency) Higher, if “R=7/8 & ½ GI” is employed. Higher, at “R=3/4 & 2 Tx” claimed Performance for reference (@ 2X2, 64QAM, R=3/4, 20MHz) 0.3dB Better @ ch-D (full GI, CC67) 0.3dB Better @ AWGN (CC59) 0.8dB Better @ ch-B (T: 1/2GI , CC67) Gate count (Mandatory mode) Slightly lower due to simpler channel estimation Slightly higher Optional features Many Few Proposal stability & Completeness Sensitivity & EVM proposals are still missing Better John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Suggestions & comments November 2004 Suggestions & comments Further extensive simulations are needed to identify performance under identical conditions to enable fair comparisons. As both proposals employ mandatory and basic MIMO-OFDM-SDM architecture in constructing their modems, there exist possibility for harmonization and convergence in the future. John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

TGn Sync & WWiSE Comparisons on November 2004 TGn Sync & WWiSE Comparisons on MAC Speaker: Albert Liu John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Feature comparisons November 2004 TGn Sync WWiSE NOTE New definition in general frame N Y TGn Sync has higher design complexity. New control frames New data frame New mgt frame Aggregation Several frames can be put together MP(S)DU Control frames can be also aggregated in TGn Sync. A-MSDU aggregation Several A-MSDUs can be aggregated in WWiSE. Scheme Complex Simple TGn Sync needs IAC/RAC to initialize aggregation. WWiSE needs one bit in QC to initialize aggregation. John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Feature comparisons (Con’t) November 2004 Feature comparisons (Con’t) TGn Sync WWiSE NOTE receivers multiple direction unidirection/bi-direction unidirection Bi-direction can have little interactive latency Performance (no header compression) normal better (w/ header compression) A-PPDU length limited 264143 Bytes unlimited WWiSE uses one bit in PLCP header to identify which PPDU is the last one. Power saving modes good TGn Sync uses MRAD, receivers NOT inside aggregated frame can shut down their RF for power saving. John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Preamble + PLCP headers + SIFS will be saved November 2004 What is aggregation ? Legacy Burst PSDU1 PSDU2 PSDU3 Preamble PLCP header MPDU Header MPDU Payload Preamble PLCP header MPDU Header MPDU Payload Preamble PLCP header MPDU Header MPDU Payload FCS FCS FCS Perform aggregation SIFS SIFS Preamble + PLCP Header A-PSDU Preamble + PLCP headers + SIFS will be saved Some overheads will be induced to identify each MP(S)DU John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

MPDU Delimiter + PAD induce 4B ~ 7B per MPDU November 2004 TGn Sync Aggregation Pad to 4 bytes MPDU1 MPDU2 MPDU3 Preamble + PLCP Header A-PSDU Max PSDU Length = 264143B RSV/LEN/CRC/UP One mechanism can find next correct MPDU delimiter when the current one is broken 4B MPDU Delimiter + PAD induce 4B ~ 7B per MPDU John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

TGn Sync Header Compression November 2004 TGn Sync Header Compression Suggestion: No need HID FC/DRU/a1/a2/a3/SC/QC 26B Uncompressed header MPDU1 MPDU2 MPDU3 Perform header compression CHDATA header FC/SC/HID/RSV MHDR MPDU 30B FC/DRU/a1/a2/a3/HID/RSV/QoS/FCS 6B CHDATA header can save 26B-6B = 20B per MPDU John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

TGn Sync Aggregation + Header Compression November 2004 TGn Sync Aggregation + Header Compression MHDR MPDU CHDATA MPDU1 CHDATA MPDU2 CHDATA MPDU3 CHDATA Header CHDATA Payload CHDATA Header CHDATA Payload CHDATA Header CHDATA Payload MHDR MPDU A-PSDU Induced 4~7B Saved 20B Saved - induced = 13B~16B per MPDU are saved John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

SF-Header can save 26B-14B = 12B per MSDU November 2004 WWiSE Aggregation SF-Header A-MSDU A-PSDU MSDU1 payload MSDU2 payload MSDU3 payload A-MSDU Header Preamble + PLCP header LEN/SA/DA ZIFS+PLCP header (8K) 14B Suggestion1: enlarge A-MSDU Suggestion2 : SF-header needs CRC if suggestion 1 is applied MSDU headers 26B are saved  NOTE: PLCP header can NOT be efficiently saved SF-Header can save 26B-14B = 12B per MSDU John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Aggregation Comparison Table November 2004 Aggregation Comparison Table TGn Sync WWiSE 1) Saved preamble Yes 2) Saved PLCP headers 3) Saved SIFS 4) Saved MP(S)DU headers 5) Extra IAC MPDU Delimiters PAD HID compressed headers sub -frame headers 6) Saved MPDU header is replaced by CHDATA header MSDU header is replaced by Sub-frame header John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

November 2004 Case Study – TGn Sync case 1 : Assume 24 MPDU (2KB), 1TX, 64QAM, 3/4 , --> 54Mbps , No aggregated consumes : 24 preambles + 24 PLCP headers + 24 MPDU headers + 24 MPDUs + 23 SIFSs = 8006 us Aggregated + header compression are applied: aggregated consumes : 1 preambles + 1 PLCP headers + 24 MPDU delimiters + 1 MPDU header + 1 HID + 24 compressed headers+ 24 MPDUs + {23 PAD octets} + IAC (40B) 1) 23 preambles are saved = 23 * 16us = 368us are saved. 2) 23 PLCP header are saved, 23*4us = 92us are saved. 3) 23 SIFSs are saved = 23 * 10us = 230 us are saved. 4) 23 MPDU header are saved = 23 * 26 = 598 bytes are saved 5)extra (1 HID) and ( 24 compressed headers) and (24 * MPDU Delimiters) and (23 PAD octets) + IAC are consumed = 1 + ( 6 * 24) + ( 4 * 24) + ( 3 * 23) +40 = 350 byte are wasted analysis 4) and 5) : ((598 - 350)*8)/216 = 10 symbols = 40 us are saved conclusion : 368 + 92 + 230 + 40 = 730 us are saved Performance improved : 730us / 8006us = 9.12% John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

November 2004 Case Study – WWiSE case 1 : Assume 24 MSDU (2KB) = 6 * A-MSDU, 1TX, 64QAM, 3/4 , --> 54Mbps , No aggregated consumes : 24 preambles + 24 PLCP headers + 24 MSDU headers + 24 MSDUs + 23 SIFSs = 8006us Aggregated is applied : aggregated consumes : 1 preambles + 6 PLCP headers + 6 MSDU headers + 24 sub-frame headers + 24 MSDUs 1) 23 preamble are saved if all are ZIFS = 23 * 16us = 368us are saved 2) 18 PLCP headers are saved = 18 * 4us = 72us are saved 3) 23 SIFSs are saved = 23 * 10us = 230us are saved. 4) 18 MSDU headers are saved. 5) extra 24 sub -frame headers. Analysis (worst case): One A-MSDU = ((14B*4- 26B*3 ) *8 )/216 = 0 symbols, total 6A-MSDU = 0 symbols = - 0us are saved conclusions : 368 + 72 + 230 + 0 = 670 us are saved Performance improved = 670/8006 = 8.37% John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Comparison table for Case Study November 2004 Comparison table for Case Study TGn Sync WWiSE 1) Saved preamble 23 2) Saved PLCP headers 18 3) Saved SIFS 4) Saved MP(S)DU headers 5) Extra (1 HID) and ( 24 compressed headers) and (24 * MPDU Delimiters) and (23 PAD octets)+ IAC extra 24 sub -frame headers 4) – 5) ((598 - 350)*8)/216 = 10 symbols = 40 us are saved One A-MSDU = ((14B*4- 26B*3 ) *8 )/216 = 0 symbols, total 6A-MSDU = 0 symbols = - 0us are saved Total saved time and % 730us 9.12% 670us 8.37% Item 4 and 5 need to be balanced John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Performance Comparison November 2004 Performance Comparison 1) TGn Sync performance is better than WWiSE when aggregation + header compression is performed. I.e. (case1)125*2KB packets => TGn Sync/WWiSE = 1.100 I.e. (case2) 24* 2KB packets => TGn Sync/WWiSE = 1.090 I.e. (case3) 120*1KB packets => TGn Sync/WWiSE = 1.044 I.e. (case4) 24* 1KB packets => TGn Sync/WWiSE = 1.034 I.e. (case5) 8* 1KB packets => TGn Sync/WWiSE = 1.000 2) TGn Sync performance is worse than WWiSE when only aggregation is performed. i.e. (case1) 125*2KB packet => TGn Sync/WWiSE = 0.997 I.e. (case2) 24* 2KB packet => TGn Sync/WWiSE = 0.982 I.e. (case3) 120* 1KB packet => TGn Sync/WWiSE = 0.946 I.e. (case4) 24* 1KB packet => TGn Sync/WWiSE = 0.932 I.e. (case5) 8* 1KB packet => TGn Sync/WWiSE = 0.900 John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Conclusion When there are more aggregated packets, regardless of size November 2004 Conclusion When there are more aggregated packets, regardless of size When higher air speed (more bits inside one symbol) The dominant factor will be the number of PLCP headers. Improvements to Aggregation need to be in the final 802.11n definition to achieve best performance page15 John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Suggestions to these 2 campaigns November 2004 Suggestions to these 2 campaigns (1) To TGn Sync: Topic : Remove HID in MHDR MPDU and CHDATA MPDU. Reason: For implementation feasibility, one MHDR MPDU should be often followed by same a1/a2/a3’s CHDATA MPDU. Benefit: save one byte HID may possibly save 4 bytes of padding in each MPDU. Drawback: continuous MHDR MPDU with different HID are forbidden. (2) To WWiSE: Topic1 :Enlarge A-MSDU size to 256KB. Reason: save more PLCP headers in A-PSDU Benefit : save more PLCP headers mean more air time is saved. Drawback: different address set (a1/a2/a3) should be separated by another A-PSDU. Topic2: Sub frame header should have CRC and develop one mechanism can find next correct sub frame header when current one is broken if topic 1 is applied. Reason: make sure this header is right. One sub frame header is fail, all the A-MSDU is fail. Benefit: One sub-frame header is fail, but the rest of sub-frame can still help. Drawback: Implementation overload. John Egan, Rodger Tseng, Albert Liu, Infineon Technologies

Month 2002 doc.: IEEE 802.11-02/xxxr0 November 2004 Summary More Use and Application cases need to be considered for a full analysis of proposals. This analysis needs to be scheduled. Both TGn Sync and WWiSE proposals have strengths and weaknesses. We recommend an effort to resolve these through some form of merger. The market should get as strong and technologically advanced a standard as possible to promote the next wave of consumer and in-premise distribution beyond traditional LAN devices John Egan, Rodger Tseng, Albert Liu, Infineon Technologies John Doe, His Company