1. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a1 Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [Edge Detection in dense multipath and heavy interference] Date Submitted: [15 May 2005] Source: [Zafer Sahinoglu, Mitsubishi Electric] Contact: Zafer Sahinoglu Voice:[+1 617 621 7588, E-Mail: zafer@merl.com] Abstract: [This document provides a technical recommendation on how the first arriving signal energy can be detected in dense multipath and heavy SOP interference] Purpose: [To point out basic requirements for a signal waveform to deal with multipath and SOP interference in edge detection] 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. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a2 Zafer Sahinoglu, May 11, 2005 Mitsubishi Electric Research Labs

3. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a3 Signal Parameters Signal Energy Conditioner Channel Characteristics Signal Energy Collector Signal TOA Estimate Signal Energy Edge Detector Generic Architecture for Ranging Received signal energy is collected Energy vector is processed to suppress noise artifacts and enhance signal containing parts Edge detection is performed channel

4. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a4 Outline Ranging signal waveforms SOP Interference –Deficiencies of coherent energy combining A look into signal energy conditioning techniques Edge detection for ranging

5. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a5 M chip times One Bit The Other Bit TH- freedom Always Empty M chip times Always Empty TH freedom Always Empty Signal Waveforms Always Empty Enough long not to cause IFI optional

6. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a6 SOP Interference Desired user signal Interference Edge information may not be recovered under SOP interference (by energy detecting receivers), if every piconet uses the same preamble pattern Received energy Deviation from the true the TOA

7. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a7 Implicit TH code: {1,3,3,1} – bits: {1,0,0,1} Implicit TH code: {1,1,3,3} - bits: {1,1,0,0} Example Acquisition Waveform Piconet-I Piconet-II Using the two specified bit waveforms (TH freedom = 0) Bit interval

8. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a8 SOP Interference (2) Strong SOP interference even with a different transmission pattern can be useless to coherent energy combining –Example simulation: CM2 (desired and interferer with different channel realizations) Energy Window Size = 4ns E B N 0 = 22dB (for both desired user and interferer – equidistant to the receiver) True TOA corresponds to window index “12” Detected TOA corresponds to window index “4”. TOA estimation error = (12-4)*4 = 32ns receiver desired transmitter interferer True TOA

9. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a9 How to Filter Out SOP Interference? Signal processing of “energy samples before coherent combining for non-coherent ranging” and of “sampled correlator outputs for coherent ranging” –Statistical multiplexing to increase SNR –Exploiting correlation properties of the samples –Frequency domain analysis

10. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a10 (1) (2) (M)(1) (2) (M) (1)(2) (N) E(N, TH(N))E(N, TH(N)+1)…E(N, TH(N)+M) ………… E(2, TH(2))E(2, TH(2)+1)…E(2, TH(2)+M) E(1, TH(1))E(1, TH(1)+1)…E(1, TH(1)+M) (1) (2) (M) E(1,1)E(1,2) E(1,M) E(2,1)E(2,2) E(2,M) E(N,1)E(N,2) E(N,M) Frame interval Generating an Energy Image

11. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a11 Generating an Energy Image (II) No change on the receiver radio architecture –Matrix is created only from sampled correlator outputs or from the output of square-law devices Added small processing complexity due to vertical image edge detection

12. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a12 Energy Image Illustrations To better understand how images are created, just shift the mask below over desired user and interferer signals one block at a time, and populate the matrix from open windows in the mask User Energy matrix Interference Energy Matrix frame interval mask Desired user code-2 {2,1,2,1} Interference code {1,3,1,2}

13. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a13 Pulse compression Narrower energy windows or Thicker vertical edges Energy Image Variations Thinner vertical edges Single Pulse Very large energy window or Signal waveform determines thickness and strength of vertical edges

14. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a14 Interference when received according to time hopping sequence TH2 (CM2-49) Desired user when received according to its own time hoping sequence TH1 (CM2-43) “Desired user (TH1) + Interferer (TH1)” when received according to the time hoping sequence TH1 (CM2-43 for desired user and CM2-49 for interferer) Frame index Energy window index “Desired user (TH1)+ Interferer (TH2)” when received according to the time hoping sequence TH1(CM2-43 for desired user and CM2-49 for interferer) Frame index Energy window index Frame index

15. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a15 Energy Image Superposition Simulation settings (single pulse): –CM2 (Residential NLOS) and no SOP interference –E B N 0 = 18dB, T F =200ns, W E = 4ns –Transmission duration 60µs (for 10 images) –Vertical edge detection with a “Prewitt” method (see Matlab image toolbox) –RAM: 1.5KB N 1 Energy window index Frame index N=1 1 30 1 50Energy window index Frame index N = 10 30 1 1 50 When pulse compression with M chips, edges will be only thicker and MN images needed

16. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a16 FFT Analysis of Detected Vertical Edges Desired user energy forms vertical lines in multipath channels Interference forms a pattern that repeats itself along a vertical line –(Left) – energy window size: 4ns, TH code length: 4

17. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a17 FFT Analysis of Energy Image Interference energy forms non-vertical patterns on the image Frequency analysis of a column can be used to further increase confidence level that a detected vertical edge is not due to interference Only need to apply 32-point (or less) FFT to the sample series of the detected image column

18. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a18 Recommended Edge Estimator Architecture Energy VectorEnergy Matrix Generator Vertical Edge Detector TOA estimate FFT Test on Samples of the Detected Vertical Edge Column index From the outputs of correlators or square- law device

19. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a19 Summary and Conclusion Coherent energy combining may not be sufficient to accurately detect leading edges Energy images provide more insight into whereabouts of leading edge even under dense multipath and SOP interference Signals should be transmitted with a distinguishable pattern for the energy detectors –This can be achieved by Coarse block time-hopping Pulse compression with very low PRF and block time hopping

20. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a20

21. April 25th 2005Doc: IEEE 15-05-0269-01-004a Zafer Sahinoglu, Mitsubishi Electric SlideTG4a21