390 likes | 397 Views
This document outlines the TensorCom physical layer proposal for IEEE 802.15 TG3C, including key features, channelization, spreading codes, and common preamble/frame format for dual-mode single carrier/OFDM.
E N D
Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [TensorCom Physical Layer Proposal] Date Submitted: [ 7 May, 2007] Source: [Ismail Lakkis] Company [TensorCom] Address [10875 Rancho Bernardo Rd #108, San Diego, CA, USA] Voice:[858-676-0200], FAX: [858-676-0300], E-Mail:[ ilakkis@tensorcom.com] Re: [This submission is in response to the TG3C call for Proposals (IEEE P802.15-07-0586-02-003c)] Abstract: [This document describes the TensorCom physical layer proposal for IEEE 802.15 TG3C.] Purpose: [For considereation and discussion by IEEE 802.15 TG3C.] Notice: This document has been prepared to assist the IEEE P802.15. 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 P802.15. Ismail Lakkis, TensorCom
TensorCom Physical Layer ProposalDual-Mode Single Carrier / OFDM Ismail Lakkis TensorCom 10875 Rancho Bernardo Rd, #108 San Diego, CA, 92127 May 7, 2007 Ismail Lakkis, TensorCom
Outline • PHY key features • Channelization • Spreading codes • Common Preamble/Frame format • SC & OFDM • Common mode • Selected responses to the selection criteria • Advnatges of each mode • Summary Ismail Lakkis, TensorCom
PHY Key Features • Dual-mode SC (Single Carrier) / OFDM for different classes of devices • Low-complexity interoperability common mode for interoperability between different devices/networks • Unified common frame format enabling a single HW supporting SC / OFDM • Link Adaptation & Unequal Error Protection via low –complexity Structured Turbo LDPC / RS • Balanced Channelization with multiple XTAL support Ismail Lakkis, TensorCom
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 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 Ismail Lakkis, TensorCom
Channelization • Support Cell phone XTAL: 38.4 MHz & Other High frequency XTALs: 30.72, 46.08MHz, … • Very good balance between 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 2211.84 MHz 1720.32 MHz 13 MHz 139 MHz 1228.8 MHz 1 2 3 4 57 58 59 60 61 62 63 64 65 66 fGHz Ismail Lakkis, TensorCom
Channelization: PLL Reference Diagram fX fS Oscillator Phase Detector LPF VCO fc XTAL ÷ R1 × P ÷ Q fM ÷ M ÷ N ADC/DAC options: 1720.32 MHz 2580.48 MHz 3440.64 MHz Example: fADC = 3440.64 MHz Phase Detector LPF VCO ÷ R2 ÷ 16x7 ÷ 64x7 Ismail Lakkis, TensorCom
Spreading Codes: Desired Features • Quasi-perfect code: Low SLL (Side Lobe Level) and wide ZCZ (Zero Correlation Zone) for improved Detection • Perfect code for channel estimation, i.e. zero SLL • Binary codes (1 bit DAC versus multi-bit DAC) • Zero-mean codes for improved DC offset cancellation • Selected code should support a parallel Low complexity matched filter architecture • Maximum code length of 128 for multiple XTALs support (up to 50 ppm, ±25 ppm @ Tx/Rx). • Should support SC & OFDM Ismail Lakkis, TensorCom
Spreading Codes • Golay complementary codes of various length N (aN ,bN) are the spreading codes of choice • Each code has a low SLL and a wide ZCZ • The combination of their periodic & aperiodic autocorrelation provides a perfect code • Only 1 bit DAC & 1 bit ADC • Admit a very low-complexity highly parallelizable architecture • Key enabler for a low complexity synchronization, channel estimation & above all a common mode engine Ismail Lakkis, TensorCom
Spreading codes • Each Delay vector D and weight vector W specify a pair of complementary Golay codes • Highly efficient Golay matched filter with only 14 adders for a length 128 code (“Budisin”) • It provides simultaneous matched filtering with the two complementary codes at once. • Enables same preamble for SC, OFDM & interoperability common mode + + + DD(0) DD(1) DD(M-1) input - - - + + + + + + Matlab Code 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; Ismail Lakkis, TensorCom
Spreading codes: Preamble • D = [64 32 8 2 16 1 4]; • W = [++++-++] • a = [05C99C5005369CAFFA3663AF05369CAF] • b = [F5396CA0F5C66C5F0AC6935FF5C66C5F] Ismail Lakkis, TensorCom
Spreading Codes: Common Mode • BPSK modulated data with code length 64 • An extra bit per code can be obtained by selecting code a or code b • “00” transmit +a64 ; “01” transmit –a64 • “10” transmit +b64; “11” transmit –b64 Parameters • D = [16 8 32 1 2 4] • W = [+-+-++] • a = [DE21212174748B74]; • b = [2ED1D1D184847B84]; • Max SLL = 8 • Rms SLL = 4.5 Ismail Lakkis, TensorCom
[x-1 x-2 … x-15] = [xx11 1111 1111 1111] Spreader seed ID = [0 0] or [0 1] or [1 0] or [1 1] D D D D Spread Data Out Serial Data In sn dn @ chip Rate @ rate R xn xn-1 xn-14 xn-15 Short Spreading Codes • For low spreading code length (8 and below), there are no good codes. • Use a varying spreading code generated by an LFSR • SC time spreading and • OFDM frequency spreading matlab code function [dataOut] = tcSpreader(dataIn,spreaderSeedId,Fast) shiftRegister = [spreaderSeedId ones(1,13)]; fork = 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; Ismail Lakkis, TensorCom
Preamble Structure Long: Tpreamble = 2.976ms Short: Tpreamble = 1.042ms • Long robust preamble for far reach; • Short low-overhead preamble for HDR • CES filed for perfect multipath estimation up to 150ns • For Sectored antennas / Beamforming, repeat sync sequence in each direction • Golay sequences for all fields for low complexity and HW reuse PLCP Preamble PLCP Header PSDU Packet/Frame Sync Sequence (Long Preamble: 32 Codes, Short Preamble: 6 Codes) SFD Start Frame Delimiter CES Channel Estimation Sequence 32 96 a128 a128 a128 -a128 a128 aCP a256 aCP bCP b256 bCP -a128 b128 For a channel of length 128 chips, a code length of 256 is needed to solve time ambiguity, i.e. start and end of channel Maximum possible code length is 128 due to frequency offset up to 3MHz. -b128 Ismail Lakkis, TensorCom
PLCP Header • Robust Common mode SC/OFDM Long Header spread by a pair of Golay codes & transmitted at the LDR of 52Mbps • Robust Short low-overhead mode specific header spread by a length 4 code (in time or frequency) transmitted at a MDR of 417 Mbps • Header is further protected by a systematic RS(255,247) Length 16b Rate 4b Preamble Type 1b Aggregation 1b #subFrames 4b Scrambler Seed 2b Sectorized /Beamforming 4b CP/UW mode 3b Reserved 5b PHY Header 4 octets MAC Header 10 octets HCS 2 octets RS(N,K) Parity Bits 8 octets PLCP Preamble PLCP Header PSDU Long: THDR = 3.72ms Short: THDR = 0.465ms Long: Rate = 52Mbps Short: Rate = 417 Mbps Ismail Lakkis, TensorCom
Unified Frame Format 64 chips ~ 37ms Common mode ±a64 ±b64 OFDM mode 512 chips ~ 300ms CP OFDM Data Block CP OFDM Data Block aM OFDM Data Block aM Variable length: 0, 16, 32, 64 for SC & 16, 32, 64, & 128 for OFDM SC modes 256 chips ~ 150ms aM SC Data Burst aM SC Data Burst aM SC Data Burst aM 32, 64, 128, or 256 Short CES Data Slot Short CES Data Slot Short CES Data Slot PLCP Preamble PLCP Header Payload FCS Pad Bits Ismail Lakkis, TensorCom
Unified Frame Format: HDR • A short CES field is transmitted periodically to reacquire the channel in both SC & OFDM • Variable length Golay codes are used for this field; • Preamble HW is reused during re-acquisition no extra cost • Mode specific frequency/timing tracking • Pilot tones for OFDM • CP Known Golay codes for SC • Highly complex channel tracking is no longer required • Channel tracking of large delay spreads would require a very dense pilot overhead in OFDM • The OFDM FFT(512) engine can be implemented as 2 small FFT(256) engines allowing HW reuse in SC mode with FDE (Frequency Domain Equalization) which requires FFT(256) than IFF(256). Ismail Lakkis, TensorCom
Frame Format: SC • The modulation of choice for low complexity low power devices • Support for 2 classes of devices: • Class I: Low-power, low-complexity Constant Envelope mode: limited p/2-BPSK with data rates 50Mbps-1.3Gbps • Class II: Quasi-constant envelope (BPSK, QPSK, & 8PSK) with data rates up to 4Gbps • Medium size FFT(256) & iFFT(256) for FDE is enough for all environments • Known Golay code of variable length will serve as CP. This puts the CP at works instead of being a Waste. • The Golay CP will be used for timing, frequency and channel tracking if desired. • Pilot CES are used to re-acquire the channel Ismail Lakkis, TensorCom
Frame Format: OFDM • The modulation of choice for HDR (16-QAM and above), • Data rates up to 5.3Gbps • Allows future data rates extension without RF HW change • FFT size of 512 allows operation in extremely harsh environments with very large delay spread • Periodic pilot would alleviate the channel tracking task and reduces the sync engine tremendously • SC with 16-QAM presents no advantages over OFDM • We need both SC & OFDM for different applications! Ismail Lakkis, TensorCom
Frame Format: Common Mode • Common mode: necessary for interoperability between different devices & different networks • It requires no additional circuitry to that used during preamble detection; it comes for free! • Very low complexity with a single multiply and add (in serial implementation) • Requires only Reed Solomon Code, already needed for the header! Ismail Lakkis, TensorCom
X Y r15 g15 g3 r0 g0 r1 g1 r2 r3 r14 g2 FEC: Reed Solomon X Y • Systematic Encoding for an RS(255,138) over GF(28) • Primitive polynomial: P(z) = z8 + z4 + z3 + z2 + 1 • Root z = 00000010 • Generator polynomial: g(x) = ∏i=1:16(x-zi) matlab code 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]; X Message block Input: m0, m1, m2, … , m238 Y Last to enter encoder First to enter encoder Last out from encoder First out from encoder Code Word Output: m238, … , m2 , m1 , m0, r15, …, r0 Ismail Lakkis, TensorCom
STLDPC • Supports rate ½, ¾, and 7/8 • Very low complexity systematic encoder • Low complexity highly parallelizable decoder • Throughput matched to that of RS • 1 RS and 1 LDPC Decoder engine is needed for Class I devices • Throughput of 1720 Mbps with Master clock of 215 MHz (BW/8) and 64 iterations Ismail Lakkis, TensorCom
STLDPC • Parity check matrix H is specified by an exponent matrix E, i.e. H = JE • Matrix J is the cyclic shift of the 18x18 Identity matrix, i.e. • J = 0; J0 = I; J18 = I E78: Rate 7/8 E34: Rate 3/4 Ismail Lakkis, TensorCom
STLDPC E12: Rate 1/2 Ismail Lakkis, TensorCom
SC Parameters Mode 1 Mode 2 Mode 3 Mode 4 Ismail Lakkis, TensorCom
OFDM Parameters Ismail Lakkis, TensorCom
Simulation Results Ismail Lakkis, TensorCom
Simulation Assumptions • Channel Bandwidth = 1720.32 MHz • AWGN, CM13, CM23, CM31 (Golden Set) • Omnidirectional antennas at both ends • 50 ppm XTAL (±25 ppm @ each side) • Simulation includes • Coarse/fine frequency acquisiton & tracking • Channel estimation • Frequency domain MMSE Equalizer • Soft bit generation • TLDPC & RS decoding Ismail Lakkis, TensorCom
Long Preamble Miss Detection & False Alarm Ismail Lakkis, TensorCom
Simulation Results: AWGN Ismail Lakkis, TensorCom
Simulation Results: CM13 (CP=0) Ismail Lakkis, TensorCom
Simulation Results: CM31 (CP=64) Ismail Lakkis, TensorCom
Simulation Results: CM23 (CP=64) Ismail Lakkis, TensorCom
Link Budget: AWGN (8%PER) Ismail Lakkis, TensorCom
Link Budget: CM31 (8%PER) Ismail Lakkis, TensorCom
PHY-SAP Throughput • Assumptions: • MPDU (MAC frame body + FCS) length = 16384 Octets • SIFS = 2.5 ms • MIFS = 0.5 ms Ismail Lakkis, TensorCom
Aggregation Mode • PHY aggregation mode is highly efficient, minimizes the memory requirement at the device and is compliant with IEEE802.15.3b MAC • MAC should support very lengthy MSDUs (and consequently very long MPDUs) or aggregated MPDUs,; • PHY will fragment the frame into subframes, protect each subframe with its own CRC and allow retransmission of a subframe rather than the entire frame. • The number of subframes can be negotiated between different devices. Once these parameters are negotiated they stay the same during one session. This reduces the overhead and these parameters need not be transmitted every frame or before each subframe. • If errors occur at the receiving device, the receiving device will request from the transmitting device the retransmission of only those subframes in error and not the entire MPDU. This will increase the overall efficiency and capacity of the system. MPDU-1 MPDU-2 MPDU-M SIFS Subframe 1 Subframe 2 Subframe N Block ACK (corresponding To the N subframes) Preamble MAC & PHY Headers CRC-1 CRC-2 CRC-2 Ismail Lakkis, TensorCom
Summary • Dual-mode SC (Single Carrier) / OFDM for different classes of devices • SC is the mode of choice for low complexity medium data rate • OFDM is the modulation of choice of very high data rate • Low-complexity interoperability common mode for interoperability between different devices/networks • Unified common frame format enabling a single HW supporting SC / OFDM • Link Adaptation & Unequal Error Protection via low –complexity Structured Turbo LDPC / RS • Balanced Channelization with multiple XTAL support Ismail Lakkis, TensorCom
802.15.3c Early Merge Work • Tensorcom has agreed to create a joint submission with COMPA • A Formal Joint submission would be made in July Meeting in San Francisco • Objectives: • “Best” Technical Solution • ONE solution • Fast Time To Market • We encourage participation by any party who can help us reach our goal Ismail Lakkis, TensorCom