doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 1 Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [mmWave OFDM Physical Layer Proposal (Previously TensorCom Physical Layer Proposal)] Date Submitted: [ September 19, 2007] Source: [ Ismail Lakkis (1), BeomJin (Paul) Jeon (2), Pascal Pagani (3), Ekhard Grass (4), Shuzo Kato (5), Seongsoo Kim (6), Edwin Kwon (6), John Marshall (7), Chiu Ngo (6), Jisung Oh (6) ] Company [ (1) Tensorcom, (2) LG Electronics Inc., (3) France Telecom, (4) IHP, (5) NICT, (6) Samsung Electronics, Co., Ltd, (7)SiBEAM, Inc.] Address [See Next Page for Contact Info.] Voice:[], FAX: [],[] Re: [This submission is in response to the TG3C call for Proposals (IEEE P c)] Abstract:[Physical layer proposal for IEEE TG3C.] Purpose:[For considereation and discussion by IEEE TG3C.] Notice:This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing 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. Release:The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 2 Contact Information Ismail Lakkis Tensorcom Corporation Voice: Beomjin (Paul) Jeon, & YongHoon Kim LG Electronics Inc. Voice : John Marshall SiBEAM, Inc. Voice : Echard Grass IHP Pascal Pagani and Wei Li France Telecome Voice: Shuza Kato NICT Voice: doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 3 Contact Information Seongsoo Kim, Edwin Kwon, Chiu Ngo, and Jisung Oh Samsung Electronics Co., Ltd 416 Maetan-3Dong, Youngtong-Gu, Suwon-Shi, Gyungki-Do , Korea seongsoo1.kim at samsung dot com, cy dot kwon at samsung dot com, chiu.ngo at samsung dot com, jisung0714.oh at samsung.com Gary Baldwin, Dengwei Fu, James P. K. Gilb, Jeff Gilbert, Ricky Ho, John Marshall, Steve Pope, Bernard Shung, Karim Toussi SiBEAM, Inc. 555 N Mathilda Ave Ste 100, Sunnyvale, CA 94085, gbaldwin at sibeam dot com, dfu at sibeam dot com, gilb at ieee dot org, jgilbert at sibeam dot com, kpho at sibeam dot com, spope at sibeam dot com, bshung at sibeam dot com, kntoussi at sibeam dot com Yasuhisa Nakajima Sony Corporation TV Business Group, Sony Corporation jim dot nakajima at jp dot sony dot com doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 4 mmWave OFDM Physical Layer doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 5 Channel Number Low Freq. (GHz) Center Freq. (GHz) High Freq. (GHz) OFDM Chip Rate (MHz) SC Chip Rate (MHz) MHz 120 MHz 1296 MHz 1728 MHz 2160 MHz f GHz Frequency Band Plan doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 6 PPDU Structure PLCP PreambleFrame HeaderPSDU PHY Header 28 octets MAC Header 10 Octets MAC SubHeader 80 Octets HCS 2 Octets RS Parity Symbols 16 Octets Frame Payload Octets FCS 4 Octets Pad Bits SCRAMBLER ID (4 bits) MCS (6 bits) FRAME LENGTH (16 bits) CES MODE (1 bit) PREAMBLE TYPE (2 bits) IFS MODE (2 bits) NUMBER OF SUBFRAMES (4 bits) TIME STAMP (24 bits) T0:T23 0:23 S0:S3 24:27 M0:M5 28:33 L0:L15 34:49 CE0 50 P0:P1 55:56 I0:I1 57:58 N0:N3 59:62 RESERVED BITS 151 bits R0:R150 73:223 58Mbps to 7.35Gbps 43Mbps, 171Mbps, 342Mbps, 684Mbps ANTENNA DIRECTION (8 bits) D0:D7 65:72 Tail Bits 0(6) Transmit Order (from left to right) HCS 2 Octets RS Parity Symbols 16 Octets CP MODE (2 bits) CP0:CP1 51:52 PCES MODE (2 bits) PC0:PC1 53:54 UEP-MAPPING INDICATOR (1 bit) U0 63 INTERLEAVER TYPE (1 bit) IT 64 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 7 PHY Header MAC Header MAC SubHeader HCS RS Parity Symbols Frame control 2 Octets PNID 2 Octets DestID 1 Octet SrcID 1 Octet Fragmentation control 3 Octets Stream index 1 Octet Subframe Length 12 bits MSDU Number 4 bits Fragment Number 4 bits CRC Present 1 bit MCS 6 bits Subframe 1 subheader 40 bits Subframe 2 subheader 40 bits Subframe 16 subheader 40 bits 10 octets Transmission direction HCS RS Parity Symbols Retransmission policy 2 bits Subframe Information 2 bits Selective ACK Request 1 bit Reserved 8 bits Subframe contains MSB/LSB MSB LSB Request Sel-ACK Follow ACK policy in MAC header MAC SubHeader doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 8 Subframes & Sel-ACK (Selective ACK) MSBLSB Frame type = Sel-ACK ACK policy = no-ACK PHY Header MAC Header MAC SubHeader HCS Subframe 1 MSBLSB Subframe n Indicate which part (MSB, LSB) has an error Subframe Subframe 1 MSB FCS LSB FCS doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 9 OFDM signaling Mode Outer Reed Solomon (K+16,K) Inner Hamming(12,8) PSDU Rates Header Rates doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 10 Frame Format PLCP PreamblePLCP HeaderPSDU Packet/Frame Sync Sequence 8, 4, or 2 symbols SFD Start Frame Delimiter CES Channel Estimation Sequence -s 512 u OFDM Data SymbolCPOFDM Data SymbolCPOFDM Data Symbol 512 chips ~ 197.5ns CP 32, 64, 96, & 128 Data BlockPCES Data BlockPCESData Block 2 symbols a CP v 512 b CP s , 256 or 512 symbols when PCES is present CES Channel Estimation Sequence Optional Extended CES (2 symbols) doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 11 Rate-Dependent Parameters (LDPC) RS encoded data bits with an outer RS(216,200) Data Rate(Mbps)ModulationMSB Coding rateLSB Coding Rate 1750 MbpsQPSKLDPC(672,336)LDPC(672,504) 2625 MbpsQPSKLDPC(672,504)LDPC(672,588) UEP Mode doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 12 Rate-Dependent Parameters (Convolutional) HRC mode index Coding Mode Modulation Convolutional Code Rate Net data rate (Gbps) MSBLSB [7][6][5][4][3][2][1][0] 0 EEP QPSK 1/ / / QAM 1/ / QAM 1/ / UEP QPSK 4/74/ QAM 4/74/ QAM 4/74/ MSB/(LSB)-only transmission QPSK 1/3(1/3) /3(2/3)2.0 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 13 Timing-Related Parameters * Allocated for PAPR and OOB power reduction doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 14 Frame-Related Parameters (1) Duration is based on the default CP of 64 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 15 The Preambles Three preambles are defined: Long preamble: 8 sync symbols, 1 SFD symbol, 2 CES symbols Medium preamble: 4 sync symbols, 1 SFD symbol, 2 CES symbols Short preamble: 2 sync symbols, 1 SFD symbol, 1 CES symbol Different preamble lengths reduces overhead and latency and enable efficient beamforming For data transmission, switching from long to medium or short preamble is upon device request. First packet shall use the long PLCP preamble, the remaining packets may use either one of the three preambles. When using medium or short preamble, packets shall be separated by MIFS. For beamforming, different preamble lengths are used to maintain a balanced spreading gain- antenna gain A unique preamble sequence set is assigned to each piconet within the same frequency channel (frequency & spatial reuse). PLCP Preamble Packet/Frame Sync Sequence 8, 4, or 2 symbols SFD Start Frame Delimiter CES Channel Estimation Sequence -s 512 u a CP v 512 b CP s 512 Long: T pre = 2.173s, N pre = 11 Medium: T pre = 1.383s, N pre = 07 Short: T pre = 0.790s, N pre = 04 CES Channel Estimation Sequence u a CP v 512 b CP doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 16 The Preambles Four preamble sequence sets (labeled by the parameter m) are provided for the purpose of frequency/spatial reuse A preamble sequence set consists of a base sequence s512,m and two CES sequences u512,m and v512,m. The length 512 base sequence s512,m is the Kronecker product of a length 4 cover sequence, c4,m and a length 128 modified Golay sequence u128,m. s 512,m [n] = c 4,m [floor(n/128)]u 128,m [n mod 128] n=0:511 The cover sequences and modified Golay sequences are listed in Tables 1 & 2 respectively. The base sequences occupy four non-overlapping frequency bin sets, and therefore are orthogonal in time and frequency domain. The mth base sequence occupies frequency bins m, m+4, m+8, m+12, Modified Golay sequences, are obtained from Golay sequences using time (or frequency) domain filtering to guarantee that only the used subcarriers are populated rather than the entire 512 subcarriers. doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 17 Golay complementary sequences (denoted a and b) of length N = 2 M are specified by: 1. A delay vector D of length M with distinct elements from the set 2 m with m = 0:M-1; 2. A seed vector W of length M with elements from the QPSK constellation ( 1, j); The receiver may use the efficient 2-levels (on I & Q) low-complexity Golay matched filter shown above for packet and frame detection input function [a,b] = golaySub(M,N,D,W); a = [1 zeros(1,N-1)];b = a; for m=1:M, ii= mod([0:N-1]-D(m),N); an = W(m)*a + b(ii+(1)); bn = W(m)*a - b(ii+(1)); a = an;b = bn; end; return; The Preambles doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 18 The length 512 CES sequences u 512,m and v 512,m are modified complementary Golay sequences derived from Golay sequences a 512,m and b 512,m. They are listed in Table 3. Modified complementary Golay sequences enable perfect channel estimation in both time and frequency domains The Golay matched filter (shown before) can be used to provide simultaneously matched filter outputs to codes a and b. Combining the two outputs appropriately provide a perfect channel estimation in time domain; Frequency domain channel estimation is done in the conventional way. The complementarity property is conserved in frequency domain. OFDM systems can benefit from time-domain channel estimation due to dimensionality (ranking) issue; The Pilot CES (PCES) field is an optional field. When present, it is equivalent to the CES field and is repeated periodically to allow channel tracking. Three periods are provided which correspond to pedestrian speeds of 1, 3, and 6m/s. The Preambles doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 19 Frame Header Encoding Process doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 20 MCS M0:M5 Rate (Mbps) Reserved Scrambler Seed S0:S3 Scrambler Initial State abcd a bcd CP CP0:CP1 CP Size PCES C0:C1 Type of Preamble Used for next packet 00PCES Not present 01PCES repeated every 128 symbols 10PCES repeated every 256 symbols 11PCES repeated every 512 symbols IFS Mode I0:I1 SIFS value MIFS value s1.0 s s2.5 s s0.25 s 11Reserved CES CE0:CE1 CES Size SCRAMBLER ID (4 bits) MCS (6 bits) FRAME LENGTH (16 bits) CES MODE (1 bit) PREAMBLE TYPE (2 bits) IFS MODE (2 bits) NUMBER OF SUBFRAMES (4 bits) TIME STAMP (24 bits) T0:T23 0:23 S0:S3 24:27 M0:M5 28:33 L0:L15 34:49 CE0 50 P0:P1 55:56 I0:I1 57:58 N0:N3 59:62 RESERVED BITS 151 bits R0:R150 73:223 ANTENNA DIRECTION (8 bits) D0:D7 65:72 CP MODE (2 bits) CP0:CP1 51:52 PCES MODE (2 bits) PC0:PC1 53:54 UEP-MAPPING INDICATOR (1 bit) U0 63 INTERLEAVER TYPE (1 bit) IT 64 The PHY Header doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 21 Scrambled PSDU Append & Scramble Scrambled Pad Bits N pad bits Scrambled FCS 32b Scrambled Frame Payload LENGTH octets QPSK/QAM Mapper Frame Payload FCS Pad Bits (zeros) FEC Coder, Puncturer & Interleaver Preamble, Pilot/DC Insertion OFDM Modulator Symbol Shaper MixedSig, Analog & RF LPF Tone Interleaver Transmitter Reference diagram doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 22 x n-1 xnxn x n-14 x n-15 snsn vnvn Serial Data InScrambled/Descrambled Serial Data Out DDDD Seed Value: x init = [x -1 x -2 x -15 ] PRBS out first 16 bits: [x 0 x 1 x 15 ] for a=b=c=d= a bcd matlab code function [dataOut] = tcScrambler(dataIn,a,b,c,d) shiftRegister = [ a b c d]; for k = 0:length(dataIn) -1, feedback= xor( shiftRegister(13+(1)), shiftRegister(14+(1)) ); dataOut(k+(1)) = mod(dataIn(k+(1))+feedback, 2); shiftRegister = [feedback shiftRegister([0:13]+(1))]; end; return; The Scrambler doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 23 matlab code (indexing from 1 rather than 0) r = ones(1,M); for k = 1:K, f = mod(d(k) + r(M),2); r = mod([0 r(1:M-1)] + f*g(1:M),2) end; r = xor(r(M:-1:1),1) r0r0 r1r1 r M-2 r M-1 g1g1 g2g2 g M-2 g M-1 g0g0 Message block Input: m 0, m 1, , m K-1 First to enter encoder CRC: r M-1, r M-2 , r 0 Last out from encoder XY X Y XY Encoding Operation Step 1. Reset Shift Register (SR) to all ones. Step 2. The 3 switches are placed in position X and the K bits are fed into the encoder. Step 3. After the last bit (m 0 ) has been fed into the Shift Register (SR), the switches are moved to position Y. At this point the SR contains the CRC bits. These bits are then shifted out of the SR and complemented. CRC MAC frame payload FCS generator polynomial (M = 32): g(x) = x 32 + x 26 + x 23 + x 22 + x 16 + x 12 + x 11 + x 10 + x 8 + x 7 + x 5 + x 4 + x 2 + x + 1 = [ ] PLCP Header HCS generator polynomial (M=16): g(x) = x 16 + x 12 + x = [ ] MAC payload or PLCP Header message polynomial: m(x) = m 0 + m 1 x m K-1 x K-1 With (m K-1 = lsb of first octet of MAC payload or PLCP Header) The FCS & HCS doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 24 The Reed Solomon FEC over GF(2 8 ) r0r0 r1r1 r2r2 r3r3 r T-2 r T-1 g1g1 g2g2 g3g3 g T-1 g0g0 Message block Input: m 0, m 1, m 2, , m N-T-1 First to enter encoder Code Word Output: m N-T-1, , m 2, m 1, m 0, r T-1, , r 0 Last out from encoder XY XY X Y Example matlab code (255,239) Data= round(rand(8,239)) data = (2.^[0:7])*data parity = rsenc(gf(data,8),255,239); parity= parity(:,end-15:end); parity= reshape(de2bi(parity,8)',1,128); code = [data parity]; RS(K+16,K) & RS(K+8,K) RS(K+16,K): N = 255, T = 16 RS(K+8,K): N = 255, T = 8 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 25 FEC Option I: RQ-LDPC No interleaving is required Supports rates , , and 7/8 Very low complexity systematic encoder Low complexity highly parallelizable decoder (gate count ~ 105Kgates) Throughput matched to that of RS 1 RS and 1 LDPC Decoder engine is needed for LDR devices Throughput of ~ 6 Gbps with Master clock of 324 MHz (BW/8) and 32 iterations Rate1/23/47/8 KK NN672 d min 14106 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 26 FEC Option I: RQ-LDPC doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 27 FEC Option I: RQ-LDPC Parity check matrix H is specified by an exponent matrix E, i.e. H = JE Matrix J is the cyclic shift of the 21x21 Identity matrix, i.e. J = 0; J0 = I; J21 = I E78: Rate 7/8 E34: Rate 3/4 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 28 FEC Option II: RQ-Convolutional Inner convolutional codes combined with outer Reed Solomon codes Outer-interleaver in-between : 4x224 bytes block interleaver Outer code rate : RS(224, 216) RS Encoder Outer Interleaver Convolutional Encoder MUX 8:1 RS Encoder Outer Interleaver Convolutional Encoder Bit Interleaver MSB LSB Convolutional Encoder 8 parallel encoders Demux 1:4 Demux 1:4 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 29 Convolutional encoder : R = 1/3, K = 7 Generator polynomial : g0=133 8, g1=171 8, g2=165 8 Convolutional Encoder & Puncturer Code rate Puncturing pattern Transmitted sequence 1/3X:1 Y:1 Z:1 X1Y1Z1X1Y1Z1 1/2X:1 Y:1 Z:0 X1Y1X1Y1 4/7X: Y: Z: X1Y1X2X3Y3X4Y4X1Y1X2X3Y3X4Y4 2/3X:1 1 Y:1 0 Z:0 0 X1Y1X2X1Y1X2 4/5X: Y: Z: X1Y1X2X3X4X1Y1X2X3X4 FEC Option II: RS-Convolutional doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 30 Data multiplexer combines data from all convolutional encoders EEP Mux/interleaver Bit interleaver size: 48 A1A6 come from encoder A, and similarly for BCDEFGH FEC Option II: RS-Convolutional doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 31 UEP-Mapping Mux/Interleaver Overall bit interleaver size 48 FEC Option II: RS-Convolutional doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 32 FEC Option II: RS-Convolutional UEP-Coding Mux/Interleaver Overall bit interleaver size 96 For first half cycle of 48 bits: A1A7 come from encoder A, similarly for BCD; E1E5 come from encoder E, similarly for FGH For second half cycle of 48 bits: A1A7 come from encoder A, similarly for BCD; E1E5 come from encoder E, similarly for FGH doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 33 Optimized binary interleaver based on an iterative structure [1-4] Effectively maximizes both intra- and inter- symbols interleaving spreading Efficiently improves decoder performance Binary Interleaving doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 34 I k I I Initial block of K coded bits Successive iterations Iterative Block Interleaving doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide QAM mapping QPSK encoding table IkIk QkQk +1 b 2k b 2k+1 +1 Input Bit b 2k I-out I k Input Bit b 2k+1 Q-out Q k Input Bit b 4k b 4k+1 I-out I k Input Bit b 4k+2 b 4k+3 Q-out Q k b 4k b 4k+1 b 4k+2 b 4k+3 IkIk QkQk The Constellation Mapper doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide b 6k b 6k+1 b 6k+2 b 6k+3 b 6k+4 b 6k+5 IkIk QkQk 64-QAM mapping Input Bit b 6k b 6k+1 b 6k+2 I-out I k Input Bit b 6k+3 b 6k+4 b 6k+5 Q-out Q k The Constellation Mapper doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 37 Skewed Constellation for UEP-Mapping QPSK 16-QAM MSB = b0 LSB = b1 MSB = b0, b2 LSB = b1,b3 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 38 Normal FFT requires a bit-reversal operation before butterflying Bit-reversal interleaving pattern can be combined with FFT operation to reduce complexity Interleaving Rule : Before InterleavingAfter Interleaving The Tone Interleaver doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 39 The OFDM Modulator IFFT Guard Free NULL #-2 #-3 #-177 #177 #3 #2 Null Time Domain Outputs IFFT Inputs & Outputs Subcarrier frequency allocation: 16 groups, 22 subcarriers per group (21 data & 1 pilot) Frequency Domain Inputs Reserved Guard Free s 0 s 1 s 510 s 511 s 512-CP s 511 CP Default value is 64, Optional: 32, 96, & 128 Description#location Zero subcarriers3-1,0,1 Guard subcarriers141[-186:-256]U[186:255] Pilot subcarriers16[-166:22:-12]U[12:22:166] User defined subcarriers16[-185:-178]U[178:185] Data subcarriers336[-177:-2]U[2:177] [-166:22:-12]U[12:22:166] +1 Null Reserved doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 40 Optional Beamforming I doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 41 Beamforming Requirements A unified messaging protocol that supports : 1. Different antenna configurations on either side (Tx or Rx): Omni or quasi-omni antennas Directional single antenna Switch diversity antennas Sectored antennas Beamforming antennas Etc, 2. Both pro-active and on-demand beamforming 3. Different usage models Per packet beamforming from PNC to multiple DEVs and DEVs to PNC PNC to one DEV DEV-DEV Others, The unified messaging protocol should be independent of the specifics of the beamforming algorithm and antenna configuration implementation. Therefore, the actual beamforming algorithm will be left to the implementer. However, the tools enabling the beamforming should be defined. These tools should support all scenarios while enabling: 1. Reduced latency 2. Reduced overhead 3. Fast beamforming doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 42 Beamforming tools Four type of packets can be used during beamforming; These packets are SC packets based on the CM (common mode) and therefore can be decoded by all DEVs; Most packets have no body. i.e. preamble only; Usage example: Quasi-Omni transmission with 0~3dB antenna gain: uses type I Directional transmission with 3~6dB antenna gain: uses type II Directional transmission with 6~9dB antenna gain: uses type III Directional transmission with 9~12 dB antenna gain: uses type IV TypePreamble Length (# 512 symbols) Header Rate (Mbps) Data Rate (Mbps) I95843 II III IV doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 43 Pro-Active Beamforming Pro-active beamforming is useful when the PNC is the source of data to one or multiple DEVs. Usage model example: Kiosk, STB, Laptop: The PNC is the source of data to multiple DEVs; The PNC can send each packet in a different direction, optimized to the destined device. Beacon Period CAP (Contention Access Period) CTAP (Channel Time Allocation Period) Q-Omni Beacon #1 Directional Beacon #(m-1)N+1 Directional Beacon #(m-1) N+2 Directional Beacon #mN Q-Omni Beacon #2 Q-Omni Beacon #L Beacon Period for Superframe # m Superframe # m Packet To User #1 Direction #j 1 Packet To User #2 Direction #j 2 Packet To User #L Direction #j L MIFSSIFS or MIFS Q-CAP #1 Q-CAP #2 Q-CAP #L COMPA Requested to Add optional Directional-CAPs as well to align with their beamforning doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 44 Pro-Active Beamforming The first L transmissions in each superframe use Quasi-Omni (Q-Omni) beacons that together provide a Quasi-omni pattern; For a PNC capable of a Quasi-omni coverage, L = 1; For a PNC with sectored antennas, L would be the number of sectors; For a PNC with switching transmit diversity antennas, L would be the number of transmit antennas; It is assumed that the PNC can beamform in J = NM directions; A direction does not necessarily mean a single beam; it can be any number of beams. The directional beacons are distributed over M superframes with N directional beacons per superframe; The structure is periodic of period M superframes; The CAP is divided into L sub-CAP periods corresponding to the L Q-omni beacons. During the l th Q-CAP, the PNC antenna is in the same direction it used to transmit the l th Q-Omni beacon. Q-Omni Beacon #1 Directional Beacon #1 Directional Beacon #2 Directional Beacon #N Q-Omni Beacon #2 Q-Omni Beacon #L Q-Omni Beacon #1 Directional Beacon #N+1 Directional Beacon #N+2 Directional Beacon #2N Q-Omni Beacon #2 Q-Omni Beacon #L Q-Omni Beacon #1 Directional Beacon #(M-1)N+1 Directional Beacon #(M-1)N+2 Directional Beacon #MN Q-Omni Beacon #2 Q-Omni Beacon #L Beacon Period for Superframe # 1 Beacon Period for Superframe # 2 Beacon Period for Superframe # M doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 45 Pro-Active Beamforming The first L beacons may be of any packet type. For example: Omni beacons shall use packet type I with long preamble; Q-Omni beacons using sectored antennas or antenna arrays with say 3-6dB gain may use type I or type II; Q-Omni beacons using sectored antennas or antenna arrays with say 6-9dB gain may use type I, type II, or type-III; When used, the packet type shall be indicated in the SFD. Upon SFD detection, DEV knows the header and data rates and can decode the packet; Each Q-Omni shall carry the beamforming IE (Information element); indicated in the next slide; in addition to the default information required by b; The new information element would inform all devices listening to the PNC about the exact structure of the beamforming beacons; After a DEV decodes any one of the Q-omni beacons during any superframe, it is capable of understanding the entire beamforming cycle; Each directional beacon consists of a preamble only, i.e. it contains no header or data body; doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 46 Pro-Active Beamforming Octets: PNC address PNC response Piconet Mode MAX TX Power Level Duration per Q-CAP Number of Q-CAP Periods CAP end time Superframe duration Time token Piconet synchronization parameters field format: 21+3n octets Element ID hex valueElementSubclausePresent in beacon 0x14Beamforming InformationAdd as In every beacon Beamforming Information Element DIRECTIONAL PACKET TYPE CURRENT DIRECTIONAL BEACON IDENTIFIER NUMBER OF SUPERFRAMES PER BEAMFORMING CYCLE NUMBER OF DIRECTIONAL BEACONS PER SUPERFRAME CURRENT Q-OMIN BEACON IDENTIFER (4bits) NUMBER OF Q-OMNI BEACONS (4bits) Length (=5) Element ID Beamforming Information Element Format PLCP Preamble Type I, II or III Directional Beacon doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 47 O-DEV (Omni DEV), fast way: An O-DEV may chose to listen to the Q-omni beacons of only one superframe; Upon detection of the Q-omni beacons, the O-DEV stores a Link-Quality Factor (LQF) for each of the Q-omni beacons, than it sorts these L LQFs [LQF(1), , LQF(L)] and identify the best PNC direction l corresponding to the highest LQF; l = arg{max[LQF(i)]} i=1:L DEV associates with the PNC during the l th CAP of the current superframe, and informs the PNC that all further communications should happen with the PNC using its l th Q-omni direction; DEV can still choose to track the set of Ls best directions by monitoring the corresponding Q- omni beacons every Q superframes. If a direction, say the r th Q-omni direction, is found with a better LQF, DEV can choose to inform the PNC to transmit the next packet using the r th Q-omni direction by encoding it in the NEXT DIRECTION filed in the PHY header. Q-DEV (Single Directional Antenna): An SD-DEV may choose to listen to the entire cycle of M superframes; If a DEV starts listening during the m th superframe, than upon detection of one of the Q-Omni beacons, it will learn that this is the m th superframe, and will listen to superframes number: m, m+1, , m+M-1; During this cycle of M superframes, SD-DEV can measure, store, and sort J LQFs corresponding to the J directional PNC directions. During the same cycle, it measures the L LQFs corresponding to the L Q-omni PNC directions. Denote by j the best directional direction and by l the best Q- omni direction. DEV associates with the PNC during the l th CAP of the (m+M-1) th superframe, and informs the PNC that all further communications should happen with the PNC using its j th directional direction; DEV can still choose to track the set of Js best directions by monitoring the corresponding directional beacons every QM superframes. If a direction, say the r th directional direction, is found with a better LQF, DEV can choose to inform the PNC to transmit the next packet using the r th directional direction by encoding it in the NEXT DIRECTION filed in the PHY header. Pro-Active Beamforming Examples doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 48 D-DEV (Directional DEV) capable of transmitting/receiving in at least one Q-omni direction and I directional directions: D-DEV starts with its antenna in one of the Q-omni directions; If the D-DEV starts listening during the m th superframe, than upon detection of one, or the best, Q- Omni beacon (say Q-omni beacon number l), it will learn that this is the m th superframe; DEV can choose to do one of the following options: Lazy approach: DEV listen to IJ superframes, J superframes for each of its I directions as follows: DEV sets its directional direction to number 1, than listens to M superframes number: m, m+1, , m+M-1 and store and sort the corresponding J LQFs, LQF(1,1)LQF(1,J) where the first index refers to the DEVs direction whereas the 2 nd index refers to the PNCs direction; DEV sets its directional direction to number 2 and listen to the next superframes and store the J LQFs: LQF(2,1)... LQF(2,J); DEV sets its directional direction to number I and listen to the next M superframes and store the J LQFs: LQF(2,1)... LQF(2,J). DEV finds the best directional combination (i,j) referring to DEV using its i th directional direction and PNC using it j th directional direction. DEV associates to PNC during the l th Q-CAP period and informs the PNC of the best direction (or best Js combinations). Fast approach: DEV sets its directional direction to number 1, than listens to the N directional beacons during the current superframe. If a direction, say j th PNC directional direction, with adequate LQF is found than DEV will associate to the PNC during the l th Q-CAP period and informs the PNC to use its j th direction for data communication. DEV can still choose to scan for better directions. If one is found it can inform the PNC to switch to the new direction by encoding appropriately the field NEXT DIRECTION in the PHY header. If no adequate direction is found, than DEV switches to another direction, for example direction r, which is orthogonal to direction 1. and listen to the next superframe. If no adequate direction is found, DEV can switch to another direction and continue the search until an adequate direction is found. Pro-Active Beamforming Examples doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 49 DEV starts on Direction i=1 and set m=0, and timer t=0 While t < Tmax, DEV looks for a Q-Omni BF DEV reads Beacon information. (DEV now knows L, J, N, M, SF timing and all timing parameters of all BFs) Knowing the timing of directional Beacons, DEV listens and detect directional BFs on SF#m. Device stores Link quality Factors: LQF(i,(m-1)N+1), , LQF(i,(m-1)N+N) m = M ? m = m + 1 Detection Successful? Yes: Locked to Q-Omni #S No: Device starts PNC or go to sleep Optional: DEV sorts matrix LQF And keep the best Q directions in decreasing order (i1,j1), (i2,j2), (iQ,jQ) Optional: DEV listens to another IM SFs And rescan these best Q directions For verification DEV transmits during Q-CAP listening period #l on SF #y & requests that all further communications are exchanged using direction j1 for PNC Device switches to Direction i1, and ADD is SUCCESSFUL Device waits for an ACK from PNC Successful ACK ? DEV switches to Direction 1, set y=0 y = Y ? y = y + 1 NO Yes DEV restarts Scan NO YES i = i + 1 & before start of next SF, DEV switched to direction i i = I ? NO YES Pro-Active Beamforming: Lazy ADD Algorithm doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 50 On-Demand Beamforming On-demand beamforming is the beamforming of choice between two DEVs or between PNC and one DEV. Since the details are the same, we describe the case of the two DEVs; On-demand beamforming will take place in the CTA allocated to the link between the TWO DEVs; When a DEV is talking to multiple DEVs, the same messaging protocol as the pro-active beamforming messaging protocol is used. In this case, the CTA will play the role of the beacon period during the beamforming phase, and will be used for data communication thereafter. For the case of two DEVs only, since the CTA is a direct link between them, it is possible to devise a more collaborative and interactive on- demand beamforming messaging protocol; doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 51 Q-omni phase: DEV1 starts its first transmission with L 1 Q-omni packets followed by L 1 corresponding Q-omni listening periods; DEV1 keeps repeating this section until DEV2 sends back a response Each Q-omni training packet would contain the Q-omni training packet IE; DEV2, capable of L 2 Q-Omni directions, will set its reception direction to one of the L 2 directions and listens to DEV1 first L 1 transmissions and stores L 1 LQFs. DEV2 moves to a new direction and listen to DEV1 second period of L 1 transmissions. DEV2 continue to do so until an adequate LQF. Alternatively, DEV2 may choose to listen using all L 2 directions than find the best LQF. At the end of this phase, DEV1 and DEV2 would know the best Q-Omni directions combination to exchange further data. Using the Q-omni training response packet IE, DEV2 would inform DEV1 of its Q- omni capabilities, i.e. L 2, as well as its own best 1 st direction and 2 nd direction that it will use for all messaging. Furthermore, DEV2 would inform DEV1 of the best 1 st and 2 nd direction it found out from the L 1 direction. DEV1 best Q-omni direction would be labeled l 1, and DEV2 best Q-omni-direction would be labeled l 2. DEV2 informs DEV1 of its directional capability as well. On-Demand Beamforming doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 52 Beacon Period CAP (Contention Access Period) CTAP: Channel Time Allocation Period CTA 1CTA 2CTA nCTA k Q-Omni Packet #1 MIFSMIFS Q-Omni Packet #2 MIFSMIFS Q-Omni Packet #L SIFSSIFS Q-Omni Listening Period #1 MIFSMIFS Q-Omni Listening Period #2 MIFSMIFS Q-Omni Listening Period #L SIFSSIFS Q-Omni Packet #1 MIFSMIFS Q-Omni Packet #2 MIFSMIFS Q-Omni Packet #L SIFSSIFS Q-Omni Listening Period #1 MIFSMIFS Q-Omni Listening Period #2 MIFSMIFS Q-Omni Listening Period #L SIFSSIFS Q-omni training section Q-Omni Packet #1 MIFSMIFS Q-Omni Packet #2 MIFSMIFS Q-Omni Packet #L SIFSSIFS Q-Omni Listening Period #1 MIFSMIFS Q-Omni Listening Period #2 MIFSMIFS Q-Omni Listening Period #L SIFSSIFS Superframe Structure On-Demand Beamforming doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide Q-OMIN LISTENING PERIOD DURATION NUMBER OF Q-OMIN LISTENING PERIODS CURRENT Q- OMIN PACKET IDENTIFIER Length (=5) Element ID Q-Omni Training Packet IE: DEV1 DEV DEV2 DIRECTIONAL CAPABILITY DEV1 PREFERRED DIRECTION #2 DEV1 PREFERRED DIRECTION #1 DEV2 PREFERRED DIRECTION #2 DEV2 PREFERRED DIRECTION #1 DEV2 NUMBER OF Q-OMNI DIRECTIONS Length (=11) Element ID Q-Omni Training Response Packet IE: DEV2 DEV1 On-Demand Beamforming doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 54 Directional phase: DEV1 uses an R-cycle procedure to do the beamforming; The R cycles may happen within one CTA or may be distributed over M superframes; Each cycle is made from K sub-cycles, where N and K can change from one cycle to another. This will allow for different search algorithms such as random and binary search, and differentiating between acquisition and tracking; Each cycle is preceded by an Q-omni transmission outlining the structure of the current cycle; Each sub-cycle consists of N directional preambles followed by an Q-omni listening period; During data communication, DEV1 may still choose to transmit a sub-cycle every superframe to allow DEV2 to continuously track the beams. If DEV2 finds a better direction, it can inform DEV1 to transmit next packets using the new direction by encoding the field Antenna Direction in the header appropriately. On-Demand Beamforming doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 55 Beacon Period CAP (Contention Access Period) CTAP: Channel Time Allocation Period Superframe Structure CTA 1CTA 2CTA nCTA k Q-Omni Packet SIFSSIFS Directional preamble #N Q-Omni Listening Period CTA k MIFSMIFS MIFSMIFS SIFSSIFS Directional preamble #1 Directional preamble #2 Directional preamble #2N MIFSMIFS Directional preamble #N+1 Directional preamble #N+2 SIFSSIFS Directional preamble #(r-1)N+1 MIFSMIFS Directional preamble #(r-1)N+2 Directional preamble #rN SIFSSIFS Q-Omni Listening Period Q-Omni Listening Period Q-Omni Listening Period SIFSSIFS SIFSSIFS SIFSSIFS SIFSSIFS SIFSSIFS MIFSMIFS Sub-cycle DIRECTIONAL PREAMBLE TYPE NUMBER OF SUPERFRAMES PER BEAMFORMING CYCLE NUMBER OF SUB- CYCLES PER SUPERFRAME NUMBER OF DIRECTIONAL PREAMBLES PER SUB-CYCLE Q-OMIN LISTENING PERIOD DURATION CURRENT Q- OMIN PACKET IDENTIFIER CTA END TIME Length (=11) Element ID Q-Omni Packet Information Element 11111 LQF-NLQF-2LQF-1 Length (=N) Element ID Beamforming Report Information Element On-Demand Beamforming doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 56 Modified Golay Sequences doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 57 Preamble Golay Sequences Preamble modified Golay sequences are generated from the four Golay sequences specified above. The 3 rd sequence is of type b whereas the 1 st, 2 nd and 3 rd are of type a; Each sequence is specified by a delay vector D and a seed vector W that together specify a pair of complementary Golay sequences of type a and b along the corresponding matched filter; These four sequences were optimized to have low sidelobe levels as well as low- cross correlation Delay and Seed Vectors for Golay sequences a1, a2, b3 & a4 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 58 Modified Preamble Golay Sequences doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 59 Modified Preamble Golay Sequences doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 60 Modified Preamble Golay Sequences doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 61 Modified Preamble Golay Sequences doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 62 Optional Beamforming II doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 63 Two major antenna configurations identified Phased antenna array: beams can be formed with most versatility to almost arbitrary directions Switched antenna: beams can be formed toward only a finite number of pre-defined directions Single antenna can be viewed as a special case of the switched antenna Antenna training protocol should support different antenna configurations Antenna Configuration doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 64 Number of Antenna Elements In both cases of phased antenna array (PAA) and switched antenna array (SAA), number of antenna elements in the array can not be pre-fixed Standard simply does not specify how many antenna elements a device should implement Number of antenna elements thus should be exchanged over the air inside the training process doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 65 Octets: 1122 13 IE indexLengthSC mode supportOFDM mode support Explicit / implicitAntenna Support Bits: 8844 Number of TX elementsNumber of RX elementsTX antenna typeRX antenna type Antenna info (antenna type and number of antennas) are exchanged in the association stage Example: Antenna Info Exchange doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 66 Depending on the antenna types of STA1, and STA2; 4 possible combinations 1) PAA to PAA 2) PAA to SAA 3) SAA to PAA 4) SAA to SAA Apply optimized training sequence for each combination Possible Combinations doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 67 Channelization Desired Features Use free spectrums of Japan, USA, Korea & EU Support for 4 channels in the available spectrum Channel Separation in the order of 2 GHz Single/Dual integer PLL that generates all necessary frequencies using direct synthesis Support of multiple PLL architectures (Direct conversion, double conversion) High Frequency Dividers should be in power of 2 : low-frequency dividers can be programmable Support of multiple crystals including at least one cell crystal & one high frequency crystal doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 68 Channelization doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 69 Channelization Support Cell phone XTAL: 15 MHz, 18 MHz, 19.2 MHz & 24 MHz & Other High frequency XTALs: 22.5, 27, 30, 33.75, 36, 45, 54MHz, Balanced margins to 57/66 GHz & Good roll-off factor Supports Multiple PLL Architectures even with the Cell phone XTAL Dual PLL:High frequency PLL that generates carrier frequencies Low frequency PLL that generates the ADC/DAC & ASIC frequencies Channel Number Low Freq. (GHz) Center Freq. (GHz) High Freq. (GHz) 3 dB BW (MHz) Roll-Off Factor MHz 120 MHz 1296 MHz 1728 MHz 2160 MHz f GHz doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 70 Direct Conversion PLL Reference Diagram f c (GHz)f X (MHz)f Q (GHz)f N (MHz)f M (MHz)R1PQNMR 3 2 3 3 2 3 3 2 3 3 2 351 64x3 XTAL Oscillator Phase Detector LPFVCO P fXfX fcfc fMfM fNfN N M Phase Detector LPFVCO Example: f ADC = 3456 MHz 512 R1 R2 ADC/DAC options: 1728 MHz 2592 MHz 3456 MHz 128 Q fQfQ 128x3 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 71 Heterodyne PLL Reference Diagram: Variable IF f c (GHz) f X (MHz) f s (GHz) f IF (GHz) f N (MHz) f M (MHz) R1PQNMR x3x x2x x3x 3 2 3 3 2 3 3 2 351 64x9 XTAL Oscillator Phase Detector LPFVCO Q fXfX f RF-Mixer P fMfM fSfS N M Phase Detector LPFVCO Example: f ADC = 3456 MHz R1 R2 2/4 f IF-Mixer I Q %4 fIF ~ 6-7 GHz %2 fIF ~ GHz P = 1 fIF ~ GHz ADC/DAC options: 1728 MHz 2592 MHz 3456 MHz fNfN 16x3x3 128 4x3x3x5 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 72 Heterodyne PLL Reference Diagram: Fixed IF f c (GHz) f X (MHz) f RF (GHz) f IF (GHz) f Q (MHz) f N (MHz) f M (MHz) R1PQNMLR x2x2x x x 64x3 XTAL Oscillator Phase Detector LPFVCO fXfX fMfM f RF N M Phase Detector LPFVCO f ADC R1 R2 ADC/DAC options: 1728 MHz 2592 MHz 3456 MHz fNfN 16x3x3 128 L f IF Q fQfQ P 4x3x3x5 doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 73 Complementary Antenna Patterns doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 74 Antennas Antennas can be for example: Omni-directional antennas; Single-directional antennas; Sectored antennas capable of transmitting in L directions 1D (1 Dimensional) Antenna array capable of transmitting narrow-beams in J directions; 2D (2 Dimensional) Antenna array capable of transmitting narrow-beams in J directions; phased antenna arrays are a potential candidate for 60GHz. All antenna elements will have the same PA with transmit power = overall transmit power/number of antenna elements; The simplest phased antenna array is an antenna array implementing the following phases 0 o, 90 o, 180 o or 270 o for each antenna element. Using a quadrature transmitter with I & Q (In-phase and Quadrature), this means each antenna element will transmit I (phase 0 o ), -I (phase 180 o ), Q (phase 270 o ), Q (phase 90 o ). An equivalent set of signals would be: I+Q, I-Q, -I+Q, -I-Q. A phased antenna array cannot generate an omni pattern. However, it is possible to generate a Quasi-omni pattern, or a set of Q-omni complementary patterns that together provide an omni-coverage. doc.: IEEE c Submission November 12, 2006 Various Authors, TG3c ProposalSlide 75 Complementary Patterns Example Consider for example an 1D phased antenna array with 8 elements spaced by /2. The pattern [ ] corresponding to the transmission of [+I +I I I +I I +I I] on the 8 elements is maximum at = 0 o, has a HPBW of 98 o and a maximum gain of 3dB. This pattern will be referred as Main Q-Omni pattern. The pattern [ ] is maximum at = 90 o, has a HPBW of 41 o and a maximum gain of 3dB. This pattern will be referred as Complementary Q-Omni pattern. These 2 patterns are exactly complementary in the sense that the sum of their power gain is a constant = 2 (or 3dB) x y z