Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANS) Submission Title: [General Atomics Call For Proposals Presentation] Date Submitted: [3 March 2003; 7 March 2003- rev1] Source: Naiel Askar, General Atomics- Photonics Division, Advanced Wireless Group, 10240 Flanders Ct, San Diego, CA 92121-2901, Voice +1 (858) 457-8700], Fax [+1 (858) 457-8740], E-mail [naiel.askar@ga.com} Re: [802.15.3a Call For Proposal, Spectral Keying™ UWB Multi-Band Technology] Abstract: [This presentation outlines General Atomics’ PHY proposal to the IEEE 802.15.3a Task Group] Purpose: [To communicate a proposal for consideration by the standards committee] 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 or organization. The material in this document is subject to change in form and content after further study. The contributor reserves 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.

Overview of General Atomics PHY Proposal to IEEE 802.15.3a Presented by: Naiel Askar www.ga.com/uwb

Outline of Presentation Description of Spectral KeyingTM (SK) SK parameters and operating frequencies Channelization scheme SK performance Implementation issues Interference and co-existence Preamble definition Self-evaluation Conclusions

Summary of Proposal Scalable data rates from 15-1300 Mbps Spectral KeyingTM modulation* Compliant with FCC 02-48, UWB Report & Order Multi-Band system, scalable from 4-12 bands, occupying 2 - 6 GHz total bandwidth Supports at least 4 co-located piconets * Spectral KeyingTM is a registered trademark of General Atomics

Key Features A new modulation scheme which has been optimized for UWB systems Low symbol rate with guard time between symbols Enhanced multipath immunity by limiting channel-induced inter symbol interference (ISI) Low duty cycle allows power saving features Minimizes collisions between colocated piconets Set of allowable symbols increases with the factorial of the number of frequencies Bit rate scalable with power consumption, cost and occupied frequency Enhanced co-existence with IEEE 802.11a

UWB Multi-Band Technology UWB spectrum divided into multiple bands One symbol will be composed of subpulses from multiple bands Excellent performance in multipath Scalability Bit rate Power consumption Range Complexity / Cost Coexistence IEEE 802.11a Regulatory Compliant with US FCC Flexibility for world-wide regulatory action

Spectral Keying™ Modulation UWB Symbol in Time Transmit 2 or more subpulses using different bands Order of bands defines symbol Voltage Time (ns)

SK Definitions Data encoded with Sequence of bands in the pulse Phase information on the subpulses

Spectral Keying™ General Case An SK symbol X, where can be defined in terms of the location in a MxT matrix, B and P where 0 means no transmission ±1 allows Binary Phase Shift Keying (BPSK) ± i allows Quadrature Phase Shift Keying (QPSK) M is # of frequency bands T is # of time slots B is # of non-zero entries P is # of polarity bits N is # of available bits where For Optimum BER Performance in SK use M=T=B

SK Rate Scalability Examples 2 bands no polarity 5 bands, with BPSK 8 bands with QPSK Sequence bits/sym. 1 ~6.5 ~15 Phase bits /sym. 0 5 16 Total bits/sym. 1 11.5 31 For sequence bits, the set of allowable symbols increases with the factorial of the number of bands

Data Rate Examples

SK Parameters For Base Rates

Transmit Sub-pulse Shaping A rectangular 2 ns pulse is low pass filtered (2nd order) to suppress out of band emissions 3 dB bandwidth 440 MHz 10 dB BW 700 MHz

Frequency Plan for 110/200 Mbps Piconet 1 Piconet 2 Piconet 3 Piconet 4 Define 20 bands centered 3.4-7.2 GHz Bands are spaced 200 MHz apart Piconets will have different bands 4 piconets will have 5 unique bands Bands in each piconet will have 800 MHz separation Other piconets will share some frequencies, less separation

4 Piconets at 110/200 Mbps Systems will be able to cancel or modify the frequency of one band to avoid interference Reducing receive filter bandwidth can reduce interference from adjacent piconets

Piconet Isolation Performance improved by 3 factors Frequency separation isolation Low symbol rate reduces collision rate between piconets; random or passively synchronized Coding gain of SK and channel coding

Passive Synchronization for Channels Time interleaving may be used by channels 3, 4 to minimize interference Passive scanning of bands will identify best time slots Improved performance with lower symbol rate Has reasonable margin for channel delay spread timing uncertainty due to near-far problem Clock synchronization can be avoided by repeated scanning

Performance Bounds for SK Case when M = T = B, P = 1 The Euclidian Distance (ED) when a frequency is in error has a value of 2 Similar to antipodal modulation BPSK SK will require lower EbNo for the same performance compared to BPSK because of the higher order modulation Where M is # of frequencies T is # of time slots B is # of non-zero entries P is # of polarity bits Ps is the probability of subpulse error Es is the energy per subpulse No is the noise spectral density EbNo is the ratio of bit energy to noise density

SK Error Rate Performance: Predictable Analysis vs. Simulation Results TM

BER Performance of 5 Band SK in AWGN Improvement Over BPSK TM

Channel Capacity in AWGN (Coherent Receiver) M is # of frequency bands T is # of time slots Q is # of non-zero entries P is # of polarity bits Operating Point

Channel Capacity in AWGN (Non coherent receiver) M is # of frequency bands T is # of time slots Q is # of non-zero entries P is # of polarity bits Operating Point

Error Correction Coding Approach Coding algorithm: Turbo Convolutional Code (TCC) Best performance Manageable cost and power consumption Best in flexibility in selection of code rate, on the fly code change Already selected for 3GPP, DVB, etc. Cost: Estimated power consumption = 35 mW. Estimated chip area = 3 mm2 in 0.13 mm CMOS Parameters Overall code rate 4/5 Memory size (4k bits), larger packets will be concatenated Number of iterations = 4, 3 bits of quantization EbNo = 3.6 dB for BER = 1e-5 Turbo code simulation (DLL) and performance supplied by iCODING Technologies

Turbo Code Scalability Range of Performance 1 iteration matches performance of K=7 convolutional code 1.5 or more iterations exceeds performance of K=7 convolutional code 3 or more iterations substantial performance gains (2-4 dB) Code rate is adjustable for longer range mode(s) Power consumption can be reduced by early stopping Larger frame size (up to 8K) can further increase performance Extreme low cost, low power & low latency option K=4 constituent code can be used as stand alone FEC option. Uses same encoder components as full Turbo Code Area less than 0.25 mm^2 in 0.18u process 1/8th the complexity of a K=7 CC with upwards scalability built in Very high coding gains in frequency selective fading channel 2-4 dB gain in AWGN can translate to 4-7 dB gain in frequency selective fading channels over non-iterative techniques.

Bit to Symbol Mapping Maximum # of frequency bits in SK symbol (M=T=B=5, P=0) = log2 120 = 6.9 (excluding polarity bits) Simple mapping will produce 6.5 bits,13 bits from 2 symbols The 3rd frequency of 2 symbols are combined to produce 3 bits Reserved symbols for preambles are available Time slot number No of choices Available bits Used bits T1 5 2.3 2 T2 4 T3 3 1.6 1.5 T4 1 T5

SK Simulation in AWGN with Channel Decoder 8% PER 1e-5 BER

Link Budget

Transceiver Block Diagram

Example of a Spectral Keying™ Transmitter

Example of a 5-Band SK Receiver

Unit Manufacturing Complexity* Preliminary area estimates ~3 mm2 for RF ~7.0 mm2 for digital The target is to have a one chip solution First implementation may have separate RF and digital chips The receiver has one signal chain per band (5 total) Allows implementing Rake receiver without extra hardware Having multiple receive chains increases area ~0.5 mm, but has little impact on overall complexity Allows tracking signal peak on each band individually giving improved performance Low risk * Estimates based on collaboration with Philips Semiconductors

Power Consumption* Power consumption will be dominated by Oscillators: trade performance for low current ADCs: limit number of bits to 3 Front end receiver: dominated by NF/11a interference requirements Minimize by designing with adaptive linearity/power tradeoff Low symbol rate gives low duty cycle allowing power saving techniques to be applied * Estimates based on collaboration with Philips Semiconductors

Manufacturability & Technical Feasibility Use of proven technology and processes No high risk components or technology Immune from distortion or ringing from antennas or filters owing to relatively long subpulse time Relaxed antenna characteristics Modules already tested in the lab TRDA / Taiyo Yuden Antenna Size 10 x 8 mm General Atomics Spectral Keying™ Transceiver Module

Experimental Results of SK Validate Simulations

Scalability Power consumption Scalable from 127 to 425 mW based on rate (55-200Mbps) Data rate Scalable from 23 – 1300 Mbps Range: Scalable with more rakes, more coding, lower symbol rate Complexity Lower complexity, lower performance system possible

Interference from 802.11a Flexibility in choice of bands is key to performance Bands centered at 5.0, 5.2, 5.4 GHz will be avoided Table below based on selection criteria parameters Bands centered on 4.8, 5.6 will be marginal at 1m Bands centered on 4.6, 5.8 will be OK at 1m separation between interferer and victim, marginal at 0.3m separation All other bands are OK

802.11a Co-existence Flexibility in removing bands from or moving to adjacent bands improves co-existence Interference from UWB is much lower than an 802.11a device at same distance SIR levels are based on 802.11a minimum sensitivity of -82 dBm for 6 Mbps rate

PHY Preamble Utilize the same preamble as the 15.3 PHY for each band separately Composed of 16 Constant Zero Autocorrelation (CAZAC) symbols The pattern is repeated 10 times Last symbol will have inverted polarity Total 160 symbols lasting 12.3 µsec. Detection miss probability and false alarm rate < 10-3 in multipath are achievable Detection is declared when a threshold is exceeded in 2 out of the 5 bands

General Solution Criteria (1/2)

General Solution Criteria (2/2)

Conclusions A UWB system based on Spectral KeyingTM will meet or exceed all selection criteria Spectral KeyingTM is a Multi-Band scheme Good multipath performance Flexibility in assigning bands for regulatory and interference avoidance Unique high order modulation allows low symbol rate with long guard time between symbols Minimizes ISI - At maximum data rate no equalizer needed Off period is 75% at 13 MHz symbol rate - Allows power conservation Efficient spectrum utilization allows frequency based channels Provides scalability for power consumption, rate, range and complexity Technology proven with demonstrations Letter of assurance for essential patents submitted to the IEEE 802.15.3a leadership The May 2003 presentation will focus on analysis and simulation

General Atomics will be cooperating with: 802.15.3a Early Merge Work General Atomics will be cooperating with: Discrete Time Focus Enhancements Intel Philips Samsung Time Domain Wisair Objectives: “Best” Technical Solution ONE Solution Excellent Business Terms Fast Time To Market We encourage participation by any party who can help us reach our goals.