1. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 1 Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: Ubisense UWB PHY Proposal to 802.15TG4f Date Submitted: 11 th September 2009 Source: Andy Ward, Ubisense Address: St Andrew’s House, St Andrew’s Road, Chesterton, Cambridge, CB4 1DL, ENGLAND Voice: +44 1223 535170, FAX: +44 1223 535167, E-Mail: andy.ward@ubisense.net Re: TG4f Call for Preliminary Proposals and Final Proposals, IEEE P802.15-09-0419-01-004f Abstract: UWB PHY-layer proposal for 802.15.4f Purpose:To be considered by 802.15TG4f 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.

2. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 2 UWB PHY-layer proposal for 802.15.4f Andy Ward Ubisense

3. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 3 Overview Discuss PHYs for supporting RTLS Relates to CFA responses: –IEEE802.15-09-0198-00-004f –IEEE802.15-09-0174-00-004f –IEEE802.15-09-0403-00-004f

4. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 4 Target system characteristics Industrial user base Mobile tags –Likely to be transmitter, for complexity/power reasons –Very likely to be most numerous system component –Very cost-sensitive –Off-the-shelf design desirable for manufacturers Fixed, managed infrastructure –Likely to be receiver, no complexity/power concerns –Connected to standard wired/wireless data network –No requirement for ad-hoc / peer-peer support –Likely to be few in number –Less cost-sensitive Differentiator from other RFID systems is accurate location Data transfer is at most a secondary concern

5. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 5 Primary PHY target characteristics Location-centric Global regulatory compliance Coexistence with existing spectrum use High performance Low cost / complexity Baseline PHY with optional extensions Flexible

6. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 6 UWB PHY Background

7. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 7 Background on UWB PHY Provides highest location accuracy –Required by highlighted CFA responses One-way asymmetric beacon architecture desirable –Lowest-complexity tag Proven potential for high performance –Proprietary systems already deployed worldwide

8. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 8 Global UWB bands 4.810.6 f / GHz US UWB (FCC Parts 15.517 / 15.519) US Wideband (FCC Part 15.250) EU Korea China Japan Canada Power At first sight, plenty of overlaps between UWB bands…

9. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 9 Practical problems with combining bands …but high attenuation often required at band edges –Want wide-bandwidth signal for best location accuracy & total power –Hard to achieve with low-complexity transmitter & low-cost filtering 4.810.6 f / GHz Power / dBm EU 2007/131/EC Typical EU UWB tag spectrum ‘Compromise’ tag spectrum => Narrow bandwidth, low total power! Typical US wideband tag spectrum FCC Part 15.250 -41.3 -51.3 -70 -65

10. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 10 Other restrictions on UWB systems UWB use often comes with onerous restrictions, e.g. –EU: No fixed outdoor transmitters (2007/131/EC) –Canada: 10s Rx ack needed for outdoor systems (RSS-220) –Singapore: 10s Rx ack needed (iDA TS-UWB) –Japan: No mobile beacon-only UWB (ARIB STD T-91) Not always possible for tag to be beacon-only Not always possible to do bidirectional UWB

11. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 11 Global compliance for UWB RFID systems Multiple distinct bands to cover all territories/applications –Band 0: 5.925-7.25GHz (US wideband) –Band 1: 6-8.5GHz (EU, China, US UWB, Canada) –Band 2: 7.25-10.2GHz (Korea, Japan, US UWB, Canada) Auxilliary non-UWB 2.4GHz PHY desirable for: –Band selection for multi-band devices –Regulatory compliance assistance for UWB PHY Use of auxilliary PHY not mandatory –E.g. US-only wideband class device Auxilliary PHY is described elsewhere

12. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 12 Location Enabler signal (1) In occluded environments, LOS propagation is rare This isn’t necessarily a problem for communications, but it is for location The more information we can get out of a LOS / near-LOS path, the better System architectureNumber of sensors with LOS/near-LOS to tag Other information required Result TDoA only4+ 3+ None3D position 2D position TDoA + AoA2+None3D position AoA-only2+None3D position Single AoA1Known height of tag2D position

13. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 13 Location Enabler signal (2) Measuring basic time-of-arrival of signal is easy Measuring angle-of-arrival takes a bit longer Measuring better time-of-arrival takes longer too Therefore, add an optional extra signal period for improved location measurements –Better location accuracy and robustness; but… –Higher power consumption and lower throughput; but… –Manufacturers/users can decide on the trade-off

14. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 14 UWB PHY Details

15. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 15 Modulation, pulse repetition rate, symbol mapping OOK modulation of data –Simple to implement in transmitter –Can be decoded by either coherent or non-coherent receiver 1.625Mpps pulse repetition rate –Will (just) be average power limited –Power-of-two division from standard 26MHz GSM crystal Cheapest source of really good timing for angle-of-arrival / coherent receiver support Many 2.4GHz auxilliary PHY chipsets use same 26MHz frequency for same reason Obviously can use a lower multiple of 1.625MHz / cheaper crystal if none of this is relevant OOK symbol to pulse mapping is 1:3 –An OOK ‘1’ is “pulse-pulse-pulse”, an OOK ‘0’ is “no pulse-no pulse-no pulse” –Non-coherent receiver wins by voting on received triplets (triple modular redundancy) –Coherent receiver wins by integrating bits for 4.8dB SNR improvement –Power overhead isn’t as high as it first looks for small packets Time taken to turn crystal on and get going is not insignificant

16. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 16 Band 0 PSD (US wideband) All figures e.i.r.p.radiated Operates under FCC Part 15.250 Frequency in MHzEIRP in dBm Below 960-41.3 960-1610-75.3 1610-1990-63.3 1990-3100-61.3 3100-5925-51.3 5925-7250-41.3 7250-10600-51.3 Above 10600-61.3

17. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 17 Band 1 PSD (EU, China, US UWB, Canada) All figures e.i.r.p.radiated Operates under –FCC Part 15.517/15.519 (US) –Decision 2007/131/EC (EU) –RSS-220 (Canada) –Singapore (iDA TS-UWB) –New Zealand (General User Radio Licence for Ultra Wide Band Communication Devices 2008) –China (MII draft proposal) Frequency in MHzEIRP in dBm Below 1600-90 1600-3600-85 3600-3800-80 3800-6000-70 6000-8500-41.3 8500-9000-65 9000-10600-70 Above 10600-85

18. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 18 Band 2 PSD (Korea, Japan, US UWB, Canada) All figures e.i.r.p.radiated Operates under –FCC Part 15.517/15.519 (US) –RSS-220 (Canada) –Korea (RRL announcement 2007-22) –Japan (STD-T91, assuming relaxation of minimum data rate regulation) Frequency in MHzEIRP in dBm Below 1600-90 1600-2700-85 2700-7250-70 7250-10200-41.3 10200-10600-70 10600-10700-85 10700-11700-70 11700-12750-85 Above 10600-70

19. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 19 SHR/PHR definition Suggested preamble /SFD sequence –Four octets of 11111111 (preamble / coherent RX acquisition) –One octet of 00010101 (coherent RX alignment) –One octet of 01001101 (true start of frame delimiter) These are symbols – remember that there are three pulses / pulse periods per symbol 5 Octets21Variable PreambleSFDFrame Length / Reserved PSDU

20. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 20 Required MAC changes Add “Blink” frame type for transmit-only tags using UWB PHY –See GuardRFID preliminary proposal IEEE 802.15-09-0591-00-004e –See Decawave proposal IEEE 802.15-09-0596-00-004e “Blink” transmissions should not perform CCA –Transmit-only tags –Already in 802.15.4a?

21. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 21 Simple blink PPDU 5 Octets2113 PreambleSFDFrame Length / Reserved PSDU Minimal zero-payload PDSU Total of 21 octets, 168 symbols Total packet time (3 pulses/symbol, 1.625Mpps) = 310.2us Source Address 8 CRC 21 Sequence Number 2 Octets Frame Control

22. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 22 Blink PPDU with Location Enabler Just enough payload to tell receiver that Location Enabler information is appended to packet ‘Postamble’ is used for more detailed signal measurements –Should go after the packet because it is not data, and isn’t covered by the frame check sequence Total of 23 octets (184 symbols) + LEI packet Total of 1064 pulse periods Total packet time = (1064 pulses / 1.625Mpps) = 654.7us 5 Octets2115 PreambleSFDFrame Length / Reserved PSDU Source Address Payload 8 CRC 221 Sequence Number 2 Octets Frame Control Location Enabler Information (512 pulses)

23. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 23 Using both UWB & 2.4GHz PHYs together Regulatory compliance –Many UWB regimes require some sort of two-way link UWB band change –Globally-compliant tags will have to switch bands (US wideband, US/EU/China UWB, Korea/Japan UWB) Security –Many security protocols involve two-way transfer Rate change –Variable tag update rates Mode change –Enable/disable extensions to UWB packet to aid location

24. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 24 2.4GHz link supporting UWB PHY (1) Regulatory support for UWB PHY 1: UWB – Part 15.517 2: Data request “Can you hear me?” 3: Data response “Yes – keep going”, or “No – cease transmission”

25. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 25 2.4GHz link supporting UWB PHY (2) PHY handover in global tag Probably needs bidirectional 2.4GHz –Can’t just transmit UWB since correct band is unknown! US UWB 2: 2.4GHz beacon frame 3: UWB – Part 15.517 JAPAN UWB 2: 2.4GHz beacon frame 3: UWB – ARIB STD T91 1: Beacon request

26. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 26 Coexistence Coexistence with general radio systems is assured by limitation of TX power by UWB regulations Only IEEE 802.15 system in same band(s) will be 802.15.4a-compliant

27. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 27 Example link budget Please note: Data not available at time of submission deadline – document will be updated with this information before presentation at IEEE meeting in Hawaii

28. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 28 Power consumption example 5 Octets2113 PreambleSFDFrame Length / Reserved PSDU Minimal zero-payload PDSU Total of 21 octets, 168 symbols Total packet time (3 pulses/symbol, 1.625Mpps) = 310.2us Average active power of UWB pulse generator: ~ 0.021mW –13mW peak, duty cycle of 1ns ‘on’ in 615.3ns pulse repetition period Average active power of controlling microprocessor + crystal: ~6.3mW –3mA current drain at 2.1V Startup time of crystal: ~400us Therefore, total energy consumption per blink = ((0.021mW+6.3mW)*310.2us)+(6.3mW*400us) = 4.5uJ For comparison, consider the same system but using one pulse per symbol. In this case, the total packet time is 103.4us, so the total energy consumption per blink = ((0.021mW+6.3mW)*103.4us)+(6.3mW*400us) = 3.2uJ. Therefore, the overhead of triple modular redundancy is only 41%, not 200% as might be expected. Source Address 8 CRC 21 Sequence Number 2 Octets Frame Control

29. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 29 ‘Blink only’ tag throughput Real-life throughput obviously depends on channel loading –Channel loading depends in turn on # tags, blink rate Device is only transmitting for a small fraction of the packet time –Quite possible for two tags to transmit overlapping packets successfully Assume 310.2us per blink –64-bit ID, no payload –No acknowledgement of blink required Assume 50ns delay spread Maximum possible theoretical throughput is ~40000 tag blinks / sec. For 1000 tags, using P-Aloha modified for low UWB duty cycle –1s blink rate:95% packet success rate –2s blink rate:98% message delivery probability For 5000 tags, using P-Aloha modified for low UWB duty cycle –1s blink rate:78% packet success rate –2s blink rate:88% message delivery probability

30. doc.: IEEE 802. 15-09-0617-00-004f Submission September, 2009 Andy Ward, UbisenseSlide 30 Summary We propose: –(Here) A primary UWB PHY for best RTLS performance –(Elsewhere) A complimentary 2.4GHz PHY Both PHYs are useful independently They can be used together advantageously We welcome collaboration with other proposers.