doc IEEE 802 15 03109 r 0 March

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Project: IEEE P 802. 15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [Intel CFP Presentation for a UWB PHY] Date Submitted: [3 March, 2003] Source: [Jeff Foerster, V. Somayazulu, S. Roy, E. Green, K. Tinsley, C. Brabenac, D. Leeper, M. Ho] Company [Intel Corporation] Address [JF 3 -212, 2111 N. E. 25 th Ave. , Hillsboro, OR, 97124] Voice [503 -264 -6859], FAX: [503 -264 -3483] E-Mail: [jeffrey. r. foerster@intel. com] Re: [The contribution is in response to the Call for Proposals for a high-rate WPAN extension to be developed in the IEEE 802. 15. 3 a task group. ] Abstract: [This contribution details a proposal for a high-rate, short-range WPAN physical layer approach based upon a multi-banded UWB system architecture. The system has variable data rates to address numerous application requirements; flexible spectrum management techniques to adapt, either dynamically or statically, to different interference and regulatory environments; good performance in the presence of multipath and multiple access interference with several areas for improvement in the future; and scalable levels of complexity and power consumption to support devices with different device implementation targets. ] Purpose: [This contribution is given to the IEEE 802. 15. 3 a task group for consideration as a possible high-rate, short-range physical layer solution for WPAN applications. ] Notice: This document has been prepared to assist the IEEE P 802. 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 P 802. 15. Submission 1 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Intel’s Multi-band UWB PHY Proposal for IEEE 802. 15. 3 a Jeff Foerster, V. Somayazulu, S. Roy, E. Green, K. Tinsley, C. Brabenac, D. Leeper, M. Ho Intel Corporation Submission 2 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Overview of Presentation • Why UWB for 802. 15. 3 a and why spectrum agility? • Proposed multiband UWB PHY system architecture – Modulation, coding, pulse shaping • Link budget and supported data rates • Multiple access techniques and performance – Channelization methods • • • Multipath mitigation techniques and performance Coexistence and narrowband interference mitigation Acquisition and preamble definition Implementation feasibility Summary Backup Submission 3 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Why UWB and why spectrum agility? • Why UWB for IEEE 802. 15. 3 a? – UWB technology is uniquely suited for high-rate, short range access • Theoretical advantages for approaching high rates by scaling bandwidth (rather than power or complexity) • Newly allocated unlicensed spectrum (7. 5 GHz) that does not take away from other narrowband systems (licensed or unlicensed) • CMOS integration now possible at these higher frequencies for low cost implementations • Why spectrum agility for a UWB solution? – Just because the FCC allows UWB to transmit on top of other services does not mean we should! • Government regulations should be broader than industry requirements – Spectrum usage and interference environment changes by country location, within a local application space, and over time • Enable adaptive detection and avoidance strategies for better coexistence and possible noncontiguous spectrum allocations for flexible regulations in future – Allow for simple backward compatibility and future scalability Submission 4 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 What is meant by a Multibanded approach? • Divide spectrum into separate bands (BW > 500 MHz) • Allow devices to statically or dynamically select which bands to use for transmission Single Symbol – Decision based on device throughput requirements, interference environment, geographical location, etc. • Modulate data in an appropriate manner using a concatenation of these bands Frequency (GHz) Time (ns) Submission 5 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Proposed Multi-band UWB PHY System Architecture Submission 6 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 UWB PHY System Architecture • Transmitter and example receiver block diagrams – Coding/Interleaving/Modulation Submission 7 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 UWB PHY System Architecture • Proposed Coding and Modulation – M-ary Binary Orthogonal Keying (MBOK) + QPSK Modulation • Power efficient modulation • Orthogonal code (Walsh-Hadamard) with interleaving allows for symbol decision feedback equalization • Fast Hadamard Transforms exist with low latency and low complexity – Outer Reed-Solomon Code • Reed-Solomon used to correct burst errors • System architecture can accommodate any of these alternate coding options – – Submission Punctured Convolutional Codes Concatenated Convolutional + Reed-Solomon Turbo codes (convolutional or product code based) Low density parity check (LDPC) codes 8 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 UWB PHY System Architecture • Extended Time-frequency codes – Extension factor (N) = # of symbols Tx before hopping to new frequency (N=4 selected for this proposal) Submission 9 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 UWB PHY System Architecture • Mapping and interleaving of bi-orthogonal codewords – Block interleave 4 bi-orthogonal codewords (as shown below) – 6/3 byte interleaving delay (depending on I/Q interleaving strategy) Submission 10 4 read 32 write Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 UWB PHY System Architecture • Frequency mapping and waveform shape – 3 nsec pulse with rectified cosine shape (~700 MHz 10 -d. B bandwidth) – Frequency separation = 550 MHz – Center Frequencies – 1 st 7 bands: [3. 6, 4. 15, 4. 7, 5. 25, 5. 8, 6. 35, 6. 9] GHz – 2 nd 6 bands: [7. 45, 8, 8. 55, 9. 1, 9. 65, 10. 2] GHz – Frequency offset of 275 MHz support for enhanced channelization – 1 st 7 bands: [3. 875, 4. 425, 4. 975, 5. 525, 6. 075, 6. 625, 7. 175] – 2 nd 5 bands: [7. 725, 8. 275, 8. 825, 9. 375, 9. 925] Submission 11 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Link Budget and Supported Data Rates Submission 12 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Link Budget and Supported Data Rates • Assumptions (see backup for more details) – – 7 d. B system noise figure 0 d. Bi Tx/Rx antennas 3 d. B ‘implementation margin’ 7 bands (3. 6 – 6. 9 GHz) Submission 13 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Link Budget and Supported Data Rates • Alternate rates can be supported using different number of bands MBOK rate 13 -band 7 -band 6 -band 3 -band 1 -band 3/3 1073 577 494 247 82 3/4 804 433 370 185 61 4/8 536 288 246 123 41 5/16 335 180 154 77 25 6/32 201 108 92 46* 15* 6/64 100 54 46 23 7 * Possible signaling schemes for Beacon • Allows for lower complexity devices to join the network • Bands could be located between 3. 1 -5. 1 GHz for easier coexistence with 802. 11 a Submission 14 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multiple Access Techniques and Performance Submission 15 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Channelization for multiple piconets • System uses a combination of DS/FH CDMA with optional FDM • Different users use different offset of long PN sequence – FH enabled through periodic Time. Frequency (FH) codes (7 bands numbered 0… 6) • 6 codes available – FDM enabled through piconet coordination Time slots in frame Piconet number – DS enabled through use of random PN mask applied to every chip + low rate code 1 0 1 2 3 4 5 6 2 0 2 4 6 1 3 5 3 0 3 6 2 5 1 4 4 0 4 1 5 2 6 3 5 0 5 3 1 6 4 2 6 0 6 5 4 3 2 1 • Receiver implementations – Rake receiver improves piconet isolation Submission 16 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multiple Access Performance in Multipath • Simulation results based on – 108 Mbps mode with 7 bands, RS encoder, 6/32 MBOK, 200 packets (221 byte packets), No AWGN (for simplicity) – CM 1(1) channel used for desired user (normalized total energy to one…take out effects of shadowing) – One interfering user tested using 25 CM 1, CM 2, and CM 3 channels (normalized total energy to one…channels selected were 2 -26) – Random propagation delay between desired and interfering user – No frequency offset (simulations show 2 -3 d. B ISR improvement with offset) – Metric used: Maximum Interference-to-signal ratio (ISR) • Results (% channels with 0 packet errors) ISR (d. B) 5 6 7 8 CM 1 CM 2 CM 3 Submission 92% 84% 96% 80% 9 10 11 12 13 88% 84% 80% 72% 28% 68% 44% 32% 16% 60% 28% 17 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multiple Access Performance in Multipath • Interpretation of results – Results based on single user interference yields total interference margin • Margin can be divided between small number of close-in interferers or larger number of further away interferers (correlation of random PN mask and long MBOK codeword makes interference look noise-like) • Example: Assume desired user operating at 5 m distance – ISR = 6 d. B allows one interferer at 2. 5 m distance or 4 interferers at 5 m distance – Results show 5+ d. B of protection for almost all channels tested • Many CM 3 channels with 7+ d. B of protection • Many CM 1 and CM 2 channels with 10+ d. B of protection Submission 18 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multiple Access Performance in Multipath • How much is enough? – Protection needed only when ‘simultaneous’ transmissions occur • Not all devices will be transmitting at the same time – Always cases where more protection is needed • Uncoordinated/Open-loop techniques: with increasing levels of SIR degradation due to MAI – use offset frequency bands (improves ISR by 2 -3 d. B) – reduce code rate – reduce number of occupied bands (drop heavily interfered bands) • Coordinated/Closed-loop techniques: Use “child” piconet mechanism to – Create time slots for the interfering piconets – Create frequency band-sets for the interfering piconets (FDM) • Piconets do not need time synchronization after coordination • Could help address severe near-far problems Submission 19 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multipath Mitigation Techniques and Performance Submission 20 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multipath Mitigation Methods • Multipath mitigated through 4 techniques – Interleaving MBOK chips over different frequencies provides frequency diversity • MRC of chips in MBOK decoder – Time-frequency codes results in 72 nsec separation between frequency ‘on’ times (allows for multipath to ring down) – ISI between 4 adjacent chips during ‘on’ time requires equalization • Interleaving MBOK codewords allows for effective decision feedback equalizer – Feed-forward filter can capture energy of multipath during 4 -chip ‘on’ time • Additional rake fingers could also be used Submission 21 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multipath Performance • Simulation results based on – – – – 108 Mbps mode with 7 bands, RS encoder, 6/32 MBOK 200 packets (221 byte packets) 100 realizations of CM 1, CM 2, and CM 3 (CM 4 in future) 333 MHz sample rate (one sample per chip) Fixed sample time between samples (sub-optimum sampling per band) Simple decision feedback equalizer + 4 -tap feed-forward filter No rake • Results (% channels with 0 packet errors) Eb/No (d. B) 8 CM 1 TBD CM 2 CM 3 Submission 10 12* 14 16 59% 75% 86% 92% 35% 54% 67% 76% 22 >18 *Margin available at 10 m 80% 86% Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multipath Performance • Interpretation of results – Link performance dominated by energy capture (shadowing + finite length rake) – Maximum Eb/No = 12. 8 d. B @ 10 m can be supported (includes the implementation margin…some implementation losses captured in sims) • Can close-the-link for all channels in which Eb/No~12 d. B yields 0 packet errors • Lots of room for improvement – Improved receiver design • Add more rake arms – Detect partial overlapping pulses within 12 nsec interval – Add parallel receiver branches to capture energy in 24+ nsec intervals • Improved sampling time by optimizing for each band • Improved equalizer + rake combining schemes – Alternate FEC schemes Submission 23 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Coexistence and Narrowband Interference Mitigation Submission 24 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Coexistence Strategies • Static Control – Pre-configure device (through software control) not to use a particular band • Based on geographic region or device usage • Dynamic Control – Allow device to detect presence of NBI and avoid – Device interoperability requirements could specify detection requirements to ensure adequate control – Similar methods used in 802. 11 h for WLAN coexistence with radar systems in Europe • UWB power emitted into 802. 11 a bands – Avoiding 5. 25 (5. 8) GHz band for lower (upper) UNII band coexistence: < -20 d. B attenuation from Part 15 limits at band edge • UWB power emitted into 4. 9 GHz WLAN band in Japan – Avoiding 4. 7 (4. 975 using frequency offset channels) GHz band: -10 d. B (<-20 d. B) attenuation from Part 15 limits at band edge Submission 25 < Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Narrowband Interference Strategies • RF Front End Implications – All UWB systems must deal with strong interference at antenna (not unique to multi-band solutions) • Can be handled through filters, component linearity requirements, and power consumption • For strong NBI – Detect and avoid use of band via signaling to PNC – Rely on adjacent channel rejection of filters + receiver signal processing • For moderate or weak NBI sources (SIR < X d. B) – Let link design and receiver implementations mitigate interference • UWB pulsed signaling + MBOK + RS coding • Interference suppression and/or cancellation techniques Submission 26 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Acquisition and Preamble Definition Submission 27 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Preamble Definition • Goal: Pfa and Pmd ~ 10% of 8% PER target, i. e. < 0. 008 • Simulations in multipath so far show estimated preamble lengths to be quite conservative • Preamble divided into two parts • CCA/packet detection + coarse timing acquisition • Fine timing adjustment + channel estimation + SIR estimation Step 1 CCA/packet detect, Coarse Timing 5. 4 ms Step 2 Fine timing; channel, SIR estimation 4 ms Total proposed preamble time: 9. 4 ms • Beacon packets use the basic preamble structure shown • Actual preamble sequence discussed in back-up based on concatenation of CAZAC sequences • Shorter preamble options can be used for higher throughputs Submission 28 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Implementation feasibility Submission 29 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Implementation feasibility • Multiband approach yields – Non overlapped timing • Shared pulse generator, ADC, correlator, … – Reduced power consumption via duty cycle of bands – Reused circuits = smaller die area • Many possible transceiver architectures Submission 30 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Implementation feasibility Submission 31 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Implementation feasibility • Analog Power Estimates* RX Oversampling factor #bands 1 x 333 MHz 2 x 333 MHz TX 3 130 m. W 153 m. W 32 m. W 6 156 m. W 202 m. W 43 m. W 7 164 m. W 218 m. W 47 m. W 10 192 m. W 292 m. W 59 m. W 13 215 m. W 315 m. W 70 m. W *0. 18 um mixed signal CMOS, 5 -bit ADCs, digital processing excluded, estimates for smaller # of bands not optimized. Submission 32 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Summary Submission 33 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Summary • Proposed UWB multi-band system architecture provides spectrum flexibility for – Good coexistence with narrowband systems – Adapting to different regulatory environments – Future scalability of spectrum use (don’t need to occupy all 7. 5 GHz of spectrum today) • Good performance with multiple access interference and multipath – Additional back-off modes for improved robustness – Room for improvement in receiver implementations • Next steps – Work with IEEE 802. 15. 3 a members to merge ideas towards a single UWB PHY Submission 34 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 802. 15. 3 a Early Merge Work Intel will be cooperating with: • • • Time Domain Discrete Time General Atomics Wisair Philips Focused Enhancements Objectives: We encourage participation by any party who can help us reach our goals. • “Best” Technical Solution • ONE Solution • Excellent Business Terms • Fast Time To Market Submission 35 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Backup Material Submission 36 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Backup Material • • Self-evaluation matrix Example Link Budget Calculation Piconet setup example for selecting channels Simulation results for multiple access interference with multipath Simulation results for single user in multipath Preamble definition and detection characteristics Ranging techniques Channel characteristics vs. pulse bandwidth Submission 37 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Self-evaluation Matrix: General Solution Submission 38 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Self-evaluation Matrix: PHY Protocol Submission 39 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Self-evaluation Matrix: MAC Enhancements Submission 40 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Example Link Budget Calculation Submission 41 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Piconet Setup Example PNC scans for beacons from other PNCs No Beacons found SIR estimation over 7 bands to determine which bands to occupy If possible, choose different T-F code or band offset Else Generate beacon message encoding number of bands supported, etc. Transmit (3 -band or 1 -band) beacons with chosen T-F code Use child piconet mechanism to create a separate piconet, using FDM or TDM Submission Scan using all permissible T-F codes for 3 -band/1 -band beacons 42 DEVs scan for beacons and join piconet Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multiple Access Performance Simulations • CM 1(1) desired path, CM 1(2 -26) interfering path Submission 43 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multiple Access Performance Simulations • CM 1(1) desired path, CM 2(2 -26) interfering path Submission 44 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Multiple Access Performance Simulations • CM 1(1) desired path, CM 3(2 -26) interfering path Submission 45 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Single-user Multipath Performance Simulations: CM 1 TBD Submission 46 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Single-user Multipath Performance Simulations: CM 2 Submission 47 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Single-user Multipath Performance Simulations: CM 3 Submission 48 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Preamble Definition Step 1 CCA/packet detect, Coarse Timing Step 2 Fine timing, channel estimation, SIR estimation Step 1 Total 16 reps of CAZAC-16 sequence per band x 84 ns frame time = 5. 4 ms Frequency band s 0 s’ 0 s’’’ 0 s 1 s’ 1 s’’’ 1 s 15 s’’’ 15 12 ns 84 ns Submission CAZAC-16 sequences: {s 0 s 1…s 15}, {s’ 0 s’ 1…s’ 15}, {s” 0 s” 1…s” 15} {s”’ 0 s”’ 1…s”’ 15}, 49 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Preamble Definition Step 2 12 reps of CAZAC-16 sequence per band x 84 ns frame time = 4 ms s 0 s’’’ 0 s 1 s’’’ 1 Frequency band s 0 s’’’ 0 s 1 s’’’ 1 s 15 s’’’ 15 12 ns 84 ns CAZAC-16 sequences: {s 0 s 1…s 15}, {s’ 0 s’ 1…s’ 15}, {s” 0 s” 1…s” 15} {s”’ 0 s”’ 1…s”’ 15}, Submission 50 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Detection Characteristic for Packet Detection/Coarse Timing Goal: Pfa and Pmd ~ 10% of 8% PER target, i. e. < 0. 008 • Simulations so far show derived preamble lengths to be quite conservative Submission 51 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 UWB Ranging via Two-Way Time Transfer* tp (unknown) Device clocks are offset by to (unknown) A T’A = T’B - to + tp Devices A & B swap two range messages M and M’ B TB = TA+ to + tp Two equations in two unknowns yield: tp = ½ ( T’A – TA + TB – T’B ) Accuracy & Precision Is independent of Tx/Rx “turn-around time”. Can rely on sub-ns Tx/Rx clocking circuits. Is nearly independent of chosen UWB pulse width. * US Naval Observatory, Telstar Satellite, circa 1962 http: //www. boulder. nist. gov/timefreq/time/twoway. htm Submission 52 Jeff Foerster, Intel Corporation

doc. : IEEE 802. 15 -03/109 r 0 March 2003 Channel Characteristics vs. Pulse Bandwidth Total energy capture greater for narrowband pulses Channel fading greater for narrowband pulses Results for 1 -arm rake and averaged over all CM 1 -4 channels Submission 53 Jeff Foerster, Intel Corporation