Proposal for Coherent Pulse Compression

1. Project: IEEE P802.15 Working Group for Wireless Personal Area Networks Submission Title: [Pulse code proposal] Date Submitted: [June 2005] Source: [Dan Raphaeli, Gidi Kaplan] Company [SandLinks Ltd.] Address [Hanehoshet 6 Tel Aviv Israel] Voice:[], E-Mail: [danr@eng.tau.ac.il] Re: [] Abstract: [] Purpose: [Contribution to 802.15 TG4a] 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. Dani Raphaeli, SandLinks

2. A Burst Sequence Proposal for Coherent Pulse Compression Dani Raphaeli & Gidi Kaplan Sandlinks July 18, 05 Dani Raphaeli, SandLinks

3. Terminology • [Terminology - as agreed lately over the reflector] • Pulse – a single UWB pulse (on the order of 1-2 nsec) • Burst – a sequence of ‘L’ UWB pulses (each pulse possibly modulated, the whole sequence has some ‘code’). Possibly, L may be between 11 to 33. • Symbol - for data or ranging – comprises of M bursts. • Each pulse has energy of Ep= Es/(L*M) where Es is the symbol energy. Dani Raphaeli, SandLinks

4. Motivation to use a single short sequence for preamble and data • Simplify correlator having it fixed and short • Enable sliding correlator implementation(even analog) • Enabling CCA function of the MAC during the packet • Deterministic spectrum – for long spreading code with short packet we will have peaks in the spectrum. Need to quantify. • The shorter the sequence – the shorter the acquisition Dani Raphaeli, SandLinks

5. Burst Code Selection Criteria • Signal Spectrum Flatness. • Autocorrelation Goodness. • Length. • Complexity of generation and correlation. • Ability to provide SOP separation Dani Raphaeli, SandLinks

6. Optimization of Sequence • The sequences which give almost ideal autocorrelation, and therefore a flat spectrum are called “barker sequence”. • The largest known binary barker sequence is of length 11 [in binary form: 10110111000] • Larger sequences, even of length much larger, e.g. 32, cannot yield as good autocorrelation as the above. Dani Raphaeli, SandLinks

7. Consequences of not good autocorrelation • As a first result of ‘moderate’ autocorrelation, the spectrum will have peaks. • Since FCC limits measure the peak in the spectrum - the effect is a decrease in the allowed power level (at the transmitter). • Barker sequence have only 1dB peaks. Other sequences have larger peaks – up to 5dB (see 15-05-0240-02-004a) • Note – to assure this good spectrum, care must be taken to modulate the bursts with good sequence and random like data Dani Raphaeli, SandLinks

8. Spectrum of Barker-11 Dani Raphaeli, SandLinks

9. Consequences of not good autocorrelation • Another consequence of ‘moderate’ autocorrelation - the autocorrelation sidelobes will be high • The effect is a ghost multipath, leading to reduced performance or failures in ranging • Barker sequence have only -20dB sidelobes. • The above argument is irrelevant if different sequences will be used for ranging and for data (increasing the solution complexity) Dani Raphaeli, SandLinks

10. Using sequences with perfect periodic autocorrelation • There exist some sequences with good periodic autocorrelation: PN codes, Ipatov sequences • The problems is that in order to take advantage of this property, the sequence should be sent repetitively without a gap or without modulation, limiting their usefullness • Although such periodic sequence can be used in the preamble, it will make an adverse effect on the spectrum – resulting in peaks in the repetition frequency. This makes a lower limit on the sequence length of few uS Dani Raphaeli, SandLinks

11. Supporting multiple Piconets • Can one practically achieve physical piconet seperation using a short code? • Codes with no crosscorrelation cannot exist if codes are non synchronized • Due to Near-far effect one cannot use CDMA type separation in 4a application unless the code is very large (e.g. l000 pulses) or alternatively, there is effective time hopping code with very large discrete resolution. Both approaches lead to complicated receiver • Conclusion – SOP should be differentiated logically if more than 3 needed. Not a big issue in 4a, where traffic is low. • Nevertheless we suggest good SOP separation using time code as shown in the following slide Dani Raphaeli, SandLinks

12. Supporting multiple piconets using time code Dani Raphaeli, SandLinks

13. Integration with modulation proposal • The proposed sequence is fitted into type 2b in document 344/r1 “TG4a Review of Proposed UWB-PHY Modulation Schemes and Selection Criteria “ Dani Raphaeli, SandLinks

14. Integration with modulation proposal PPM For noncoh. BPSK Coding approach (see 15-05-240-02 for example) Systematic Conv. Code BPSK PPM **Coherent receiver use both bits + viterbi decoder Dani Raphaeli, SandLinks

15. Example parameters • Data rate = 0.935Mbps • Code rate =2/3 • Average frame duration = 713ns • Chip rate = 123.5MHz • Symbol duration= 89nS • Effective Prf=15.43MHz Dani Raphaeli, SandLinks

16. Summery • Short sequence simplify the demodulation and the acquisition • Using barker sequence allows keeping very low backoff and therefore extend range • Multiple SOP are possible • Either noncoherent, differential and coherent detection using PPM and BPSK. • Good autocorrelation allows using it also for the preamble, saving receiver complexity. Dani Raphaeli, SandLinks