Project: IEEE P Working Group for Wireless Personal Area Networks N

Transcription

1 Project: IEEE P82.15 Working Group for Wireless Personal Area Networks N (WPANs( WPANs) Title: Texas Instruments Impulse Radio UWB Physical Layer Proposal Date Submitted: 4 May, 29 Source: June Chul Roh, Anuj Batra, Sudipto Chakraborty, Srinath Hosur, and Timothy Schmidl Texas Instruments 125 TI Blvd MS 8649, Dallas, TX, USA {jroh, batra, schakraborty, hosur, schmidl}@ti.com Re: Response to IEEE call for proposals Abstract: This document describes the Texas Instruments impulse radio UWB physical layer proposal for IEEE Purpose: For discussion by IEEE TG6 Notice: Release: 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. The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P Slide 1

2 Texas Instruments Impulse Radio UWB Physical Layer Proposal June Chul Roh, Anuj Batra, Sudipto Chakraborty, Srinath Hosur, and Timothy Schmidl Texas Instruments May 29 Slide 2

3 Outline Motivation Details about the impulse radio UWB PHY: Frequency Band of Operation Frame Format: Preamble, Header, PSDU Symbol Structure Burst Position Modulation with Time-Hopping Time-Hopping Sequence FEC: BCH Codes System Parameters Performance Results: Link Budget and Receiver Sensitivity Simulation Results in AWGN and 15.3a CM1,2 Performance with Co-channel Interference Complexity and Power Consumption Summary and Conclusions Slide 3

4 Overview of Proposal Goal: Design a low-power, low-complexity UWB PHY for BAN Start by re-using some aspects of IEEE a PHY: Preamble structure Burst position modulation and time-hopping (BPM-TH) Add new features that reduce complexity and lower power consumption: More efficient symbol structure eliminate unnecessary overheads A new time-hopping sequence that supports new symbol structure Limit modulation scheme to BPM-TH simplifies receiver Limit systems to a single bandwidth of 512 MHz simplifies receiver Limit systems to higher frequency bands eliminates need for complex DAA algorithms Replace RS codes with low-complexity binary BCH codes Add support for simultaneous operation of at least 12 piconets Slide 4

5 Improvements over 15.4a New frequency band plan Use only the UWB high band does not require power-hungry DAA or LDC Each band has 512 MHz bandwidth New symbol structure and time-hopping sequence No fixed guard interval for improved PHY efficiency Time-hopping sequence is designed to avoid inter-symbol interference (ISI) Binary burst position modulation with time-hopping (BPM-TH) Binary BPM simple non-coherent receiver in mind BPSK of a is not supported in this proposal want ultra-simple receivers Low-complexity binary FEC codes BCH (31, 16, t = 3), BCH (63, 51, t = 2), BCH (63, 57, t = 1) Slide 5

6 WW Regulations on UWB Band Low Band* DAA or LDC is a must (except USA) after 21 DAA results in huge penalty on complexity and power for BAN transceivers High Band* DAA is not required. Ideal for low-complexity, low-power BAN Concern: only 1.25GHz bandwidth is common worldwide Solution: new proposed band plan * Tables from P PSD Frequency Bands Remarks Australia N/A N/A N/A EU dbm/mhz GHz LDC or DAA is needed GHz By Dec. 31, 21 Japan dbm/mhz GHz DAA is needed GHz By Dec. 31, 21 Korea dbm/mhz GHz LDC or DAA is needed GHz By Dec. 31, 21 USA -41.3dBm/MHz GHz Frequency Bands PSD Remarks Australia N/A N/A N/A EU GHz dbm/mhz Japan GHz dbm/mhz Korea GHz dbm/mhz USA GHz dbm/mhz Common GHz dbm/mhz Slide 6

7 Frequency Bands of Operation Channelization: Band Number Supported Region BW (MHz) Low Freq. (MHz) Center Freq. (MHz) High Freq. (MHz) 1 US, EU US, EU US, EU, Japan, Korea US, EU, Japan, Korea US, Japan, Korea All bands are located in UWB high band At least 3 bands available per country: 4 SOPs per band Center frequencies are integer multiples of 512 MHz: 512 [13, 14, 15, 16, 17] PLL is easier to implement than PLL for a EU US frequency (MHz) Japan, Korea Slide 7

8 PLCP Frame Format PPDU compromised of three components: PLCP Preamble: used for packet detection, timing acquisition, carrier frequency offset estimation, etc PLCP Header: convey information about to decode PSDU PSDU: MAC Header + MAC Frame Body (information) + FCS Structure: Slide 8

9 PLCP Preamble Reuse the a preamble signal structure Use the length 31 ternary codes (of a) with following band assignment Define 4 preamble codes per band Assign different preambles to adjacent channels, minimizes false alarms due to adjacent channel energy leaking into the desired band Code index Code sequence Band number 1, 3, 5 1, 3, 5 1, 3, 5 1, 3, 5 2, 4 2, 4 2, 4 2, 4 Slide 9

10 PLCP Header Proposed PLCP Header Structure (31 bits) 3 bits 8 bits 2 bits 1 bit RATE LENGTH Reserved BURST MODE PHY Header HCS BCH Parity Bits 14 bits 2 bits 15 bits PLCP Header Format the PHY header as shown in figure based on data provided by the MAC Calculate the 2-bit HCS value over the PHY header CRC-2 polynomial: g(x) = 1 + x + x 2 Apply a BCH (31,16) code to PHY header + HCS The resulting encoded bits are modulated using the lowest data rate Serial Data Input Serial Data Input D MSB Preset Registers to ONES CRC-2 D LSB Serial Data Output ONES Complement Serial Data Output (MSB First) Slide 1

11 Burst Position Modulation with Time-Hopping Basic concept: Binary PPM based modulation Multiple pulses are continuously transmitted in a symbol Time-hopping for multiple access (symbol-rate hopping) Random pulse polarity changes within a pulse burst Signal for k-th symbol interval may be mathematically expressed: p() t s + { 1,1} kn ( k ) cpb n d {,1} ( k ) h {,1,, N hop 1} N cpb T = N T T c burst cpb c TBPM = NhopTburst Ncpb 1 ( k) ( k) ( k) = kncpb + n BPM burst c n= x () t s p( t d T h T nt ) : transmitted pulse shape at the antenna input, : chip scrambling code used during the k-th symbol interval, : k-th data symbol carrying information, : time-hopping position for the burst during the k-th symbol interval, : number of chips per burst, : slot time (or burst time), : chip time, : BPM (burst position modulation) interval. Slide 11

12 a symbol structure: a Symbol Structure 5% of symbol duration is reserved as guard interval (GI): 5% of symbol is overhead! Why two guard intervals in 15.4a? 1 st GI avoids interference from symbol to symbol 1 region 2 nd GI prevents inter-symbol interference (ISI) GI is unnecessarily large compared to typical channel delay spread for data rates of interest Slide 12

13 Elimination of 1 st Guard Interval 1 st guard interval (GI) of 15.4a is unnecessary as BPM-TH inherently provides GI Since (N hop 1)T burst > τ max for data rates of interest (τ max : max expected delay spread of channel) Fixed-length 2 nd GI with T GI > τ max can be used to prevent ISI symbol # 93: bit = 1, th_seq = 15 Fixed Guard Interval symbol # 94: bit =, th_seq = time (in T_burst) time (in T_burst) Leads to a more efficient symbol structure, less overhead Q: Can we do better? Slide 13

14 Proposed Optimal Symbol Structure (1) A: Yes, we can! We only need a guard interval when transmitting a 1 on previous symbol at the end of the burst, and when transmitting a on current symbol at the beginning of a burst ISI Example: symbol # 458: bit = 1, th_seq = 15 ISI symbol # 459: bit =, th_seq = time (in T_burst) time (in T_burst) Can eliminate these cases from happening by designing the time-hopping sequence properly! symbol # 458: bit = 1, th_seq = 15 Embedded Guard Interval symbol # 459: bit =, th_seq = time (in T_burst) time (in T_burst) Slide 14

15 Proposed Optimal Symbol Structure (2) New proposed symbol structure: N hop possible burst positions for symbol N hop possible burst positions for symbol 1 T burst T c Symbol time, T s N cpb pulses per burst Completely eliminate the two fixed guard intervals of 15.4a Time-hopping sequence provides embedded guard interval only when necessary ISI can happen when two consecutive hop locations are the last slot and the first slot Design time-hopping to avoid the ISI condition Increased channel efficiency can be used for Increasing the overall possible data rates (increase channel efficiency), and/or Providing better interference mitigation capability by increasing N hop Slide 15

16 Time-Hopping Sequence Time-hopping sequence design constraint to avoid ISI: ( k) ( k 1) h h Nhop Nch k ( 1) for 1 (1) ( k ) h N hop τ max N ch {, 1,, 1} τ = Tburst max 1 : time-hopping sequence for the k-th symbol, : expected maximum delay spread of channel, An intuitive example: Let N hop = 8 and N ch = 4 Slide 16

17 Time-Hopping Sequence Generation 1. Generate a random number z (k) {,1,, N hop 1} by tapping m = log 2 (N hop ) shift registers of the a LFSR. For each symbol interval, the LFSR shall be clocked N cpb times. z (k) 2. Calculate related parameters: α N = N α ( k 1) = h γ, reduced hop where γ = N N 1 is known (pre-calculated) for each data rate. hop 3. Generate TH sequence as follows: ch where k is symbol index. h ( k ) ( k) ( k 1) z, if h γ = ( k) ( k 1) ( z k) mod Nreduced + + α, if h > γ Slide 17

18 BCH Encoder BCH (31,16) code: gx ( ) = 1+ x+ x + x+ x+ x + x+ x+ x + x + x Low-complexity, low-power implementation: BCH (63, 51): gx ( ) = 1+ x+ x + x+ x+ x + x BCH (63, 57): gx ( ) = 1+ x+ x 6 Slide 18

19 Process for BCH Encoding 1. Compute the number of bits in the PSDU: N = ( N + N + N ) 8 PSDU MACheader payload FCS 2. Calculate the number of BCH codewords: N CW N k PSDU = 3. Compute the total number of shortening bits * : Nshorten = NCW k NPSDU 4. Calculate the number of shortening bits needed per codeword: 5. Distribute shortening bits uniformly over codewords: a. Each of the first rem(n shorten,n cw ) codewords have N spcw + 1 shortened bits b. Remaining codewords have N spcw shortened bits N N shorten spcw = NCW 6. Shortened bits are not transmitted on-air, but receiver will re-insert them into known locations * Shortened bits are message bits that are set to zero Slide 19

20 System Parameters MCS number Chip rate (MHz) Chip time (ns), T c Modulation BPM-TH BPM-TH BPM-TH BPM-TH BPM-TH BPM-TH BPM-TH BCH code rate, r 16/31 16/31 16/31 16/31 51/63 57/63 57/63 # bursts in symbol, N burst # hop bursts, N hop # of chips in burst, N cpb # chips per symbol, N cps Burst length (ns), T burst Symbol period (ns), T s Symbol rate (ksps), R s Data rate (kbps), R b Average PRF (MHz) N ch for TH sequence Slide 2

21 Energy-Detection Based Non-coherent Receiver Low complexity and low power-consumption receiver Other non-coherent receiver structures are also possible Slide 21

22 Link Budget and Receiver Sensitivity Parameter Value Value Value Unit Bit rate (R b ) kbps Center frequency (f c ) MHz Bandwidth (B) MHz Average Tx power dbm Tx/Rx switch loss db Average Tx power before Tx Ant (P T ) dbm Tx antenna gain (G T ) dbi Distance (d) m Path loss at d meter (L) db Rx antenna gain (G R ) dbi Rx power (P R = P T + G T + G R L) dbm Average noise power per bit (N = *log 1 R b ) dbm Rx noise figure (N F ) db Total noise power per bit (P N = N + N F ) dbm Received SNR db Minimum required E b /N (S) db Implementation loss (I) db Link margin (M = P R P N S I) db Proposed min Rx sensitivity level dbm Slide 22

23 Justification for IEEE a Channel Model (1) CM3: Average Power Decay Profile CM4: Average Power Decay Profile Average Power Decay Profile Average Power Decay Profile Average power (db) -3 Average power (db) Delay (nsec) PDP decays 3dB at τ = 2 ns Mean excess delay: 26.3 ns RMS delay spread: 19 ns Delay (nsec) PDP decays 3dB at τ = 18 ns Mean excess delay: 4.9 ns RMS delay spread: 42 ns Slide 23

24 Justification for IEEE a Channel Model (2) a CM1 ( 4m, LOS): Average PDP a CM2 ( 4m, NLOS): Average PDP Average Power Decay Profile Average Power Decay Profile Average power (db) -3 Average power (db) Delay (nsec) PDP decays 3dB at τ = 4 ns Mean excess delay: 5.2 ns RMS delay spread: 6 ns Delay (nsec) PDP decays 3dB at τ = 5 ns Mean excess delay: 9.6 ns RMS delay spread: 8 ns Slide 24

25 PSDU = 256 bytes Simulation Parameters Transmit pulse: root-raised cosine (f cutoff = 24 MHz and α =.6) Channel AWGN Multipath channel: a CM1 and CM2 ( 4m, LOS, NLOS) PER results in multipath channel are averaged over 95% best channels Receiver Energy-detection based non-coherent demodulator Assume perfect packet detection and header decoding Ideal timing, zero carrier-frequency offset Slide 25

26 Packet Error Performance in AWGN AWGN results: ED MCS 1 MCS 2 MCS 3 MCS 4 MCS 5 MCS 6 MCS 7 PER SNR (db) Slide 26

27 Packet Error Performance in Multi-path CM1 CM ED MCS 1 MCS 2 MCS 3 MCS 4 MCS 5 MCS 6 MCS ED MCS 1 MCS 2 MCS 3 MCS 4 MCS 5 MCS 6 MCS 7 PER 1-2 PER SNR (db) SNR (db) Slide 27

28 Performance in SOP Co-channel Interference (1) 4 SOPs in a band 3 interfering piconets: Each piconet uses a unique time-hopping sequence Asynchronous between signals from multiple piconets 3 interferers continuously transmitting All users transmit at 1Mbps Interferers d Intf from reference receiver d Ref Tx d Intf Interferers Tx Path loss model: Free-space path loss model (exp α = 2) SIR = 1 log 1 (d Intf /d Ref ) α [db] for a single interferer Rx d Intf d Intf Tx Tx Channel: Each signal passes through an independent multipath channel (15.3a CM1) Receiver: non-coherent receiver based on energydetection Slide 28

29 Performance in SOP Co-channel Interference (2) SOP results: d_ref =.5 m d_ref = 1. m d_ref = 1.5 m d_ref = 2. m d_ref = 2.5 m d_ref = 3. m PER Distance ratio, d_intf / d_ref Results: d Intf /d Ref = 1.55 (to maintain a PER = 1%) Slide 29

30 Power Consumption Data rate kbps kbps kbps Analog: Tx Peak power (mw) Idle power (mw) Average power (mw) Analog: Rx Peak power (mw) Idle power (mw) Average power (mw) Tx Total (mw) Rx Total (mw) * Power analysis is based on low-voltage, low-leakage 13 nm CMOS technology. Slide 3

31 Comparison Criteria Criteria 1. Regulatory 2. Raw PHY data rate 3. Transmission distance 4. Packet error rate 5. Link budget 6. Power emission level 7. Interference and coexistence 8. Security 9. Reliability 1. Quality of Service 11. Scalability 12. MAC transparency 13. Power Efficiency 14. Topology 15. Bonus Point Proposed Capability Compliant with TG6 regulatory document in UWB frequency band 129 kbps to 9.65 Mbps supported between node and hub PER and link budget shown to support 1% PER for 256 octet PSDU at 3 meters within all operating frequency bands proposed dbm maximum EIRP Channelization: 5 channels total, at least 3 frequency bands available in each region 4 SOP supported per band, at least 12 SOP piconets supported in each region Time-hopping and pulse polarization scrambling used to mitigate interference Can be combined with MAC providing security Link margin sufficient in a UWB channel model. - Scalable data rate from common symbol rates. - To be added Star topology, broadcast beacon supported. Maximum number of nodes supported via multiple access mechanisms. - Slide 31

32 Summary and Conclusions Reuse the strengths of a PHY as much as possible Proposed a new frequency band plan simplifies receiver, no DAA requirements New symbol structure, time-hopping sequence eliminates ISI w/o needing a GI Low complexity and low power-consumption standard Binary burst position modulation with time-hopping (BPM-TH) non-coherent Rx Low-complexity binary FEC codes Wide range of data rates are supported: 128 kbps to 9.65 Mbps Supports for 12 simultaneous operating piconets Slide 32

33 Backup Slide 33

34 Better Channel Efficiency with Proposed Symbol Structure 15.4a symbol structure Proposed symbol structure: N hop doubled Proposed symbol structure: data rate doubled * For all the cases, the number of chips per burst N cpb is the same. Slide 34

35 Time-Hopping Sequence Generation (2) Conditional distributions from simulation: N hop = 8 and N ch = 4.2 PDF of h(k) given h(k-1) =.2 PDF of h(k) given h(k-1) = h(k) PDF of h(k) given h(k-1) = h(k) PDF of h(k) given h(k-1) = h(k) PDF of h(k) given h(k-1) = h(k) PDF of h(k) given h(k-1) = h(k) PDF of h(k) given h(k-1) = h(k) PDF of h(k) given h(k-1) = h(k) h(k) Slide 35