1. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 1 Project: IEEE 802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [Body Area Network UWB channel modeling update] Date Submitted: [14July2004] Source: [Andrew Fort and Bart Van Poucke] Company [IMEC] Address [Kapeldreef 75, Leuven, Belgium 3001] Voice:[+32(0)16 28 12 11], FAX: [+32(0)16 22 94 00], E-Mail:[forta@imec.be] Re: [Channel model proposal] Abstract:[Update on channel model for communication around the body] Purpose:[Contribute to low power air-interface definition for body area applications] Notice:This document has been prepared to assist the IEEE 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 802.15.

2. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 2 BAN UWB Channel Model Update Andrew Fort Bart Van Poucke IMEC, Wireless Research forta@imec.be

3. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 3 Experiment setup Path loss versus distance Path loss versus frequency Power delay profile Matlab channel model code Outline

4. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 4 Model is Based on Over 500 Measurements Taken Around the Torso 9 Simulations were made along the height of the torso. Each simulation measured several positions around the torso. Measurement were spaced at least 4 cm or approximately ½ the center frequency wavelength. UWB pulse: 3-5 GHz Gaussian Pulse (10 dB bandwidth)

5. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 5 The measurements were divided up between four areas defined by angle. Area 1 Area 2 Area 3 Area 4 Area 1: 0°-50° Area 2: 50°-100° Area 3: 100°-150° Area 4: 150°-180° Channel parameters changed as signal travels around the body.

6. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 6 Small-scale fading averaged out to extract path loss versus distance. Measured data Small-scale fading removed (area 1 – area 4) Best fit path loss model Distance (m) Path loss (dB)

7. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 7 This path loss model is not the same as the classical path loss model. Path loss modelMSE 1.32.5 2.5.4 3.0.8 Possible reasons Extreme close range Path loss mechanisms close to body are different

8. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 8 Body area channel was not frequency dependent in the 3-5 GHz band Distance (m) Path loss (dB) To be confirmed…

9. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 9 Power Delay Profiles were extracted according to 802.15.4a Guidelines

10. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 10 Statistics for each channel tap were tabulated in each area. Tap dB K-S 2 15.327.33PassInconclusive1 Tap dB K-S 2 114.4712.73Pass 1 226.5412.00Pass 0.87 Tap dB K-S 2 125.8715.97PassFail1 233.5013.99Pass 0.91 342.8711.15Pass 0.76 447.1010.93Pass 0.70 558.5611.46Pass 0.85 Tap dB K-S 2 121.3313.53PassFail1 226.7112.18Pass 0.89 330.9613.10Pass 0.78 435.119.73Pass 0.78 551.5811.35Pass 0.86 Area 1 Area 2 Area 3 Area 4

11. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 11 Area 2 CDF Area 3 CDF The model statistics provides a good match to measured data.

12. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 12 Preliminary results indicate Body area channel statistics were not the same as classical indoor statistics. Nakagami-m distribution failed the goodness of fit tests in all cases. Log-normal distribution was clearly superior. Very strong correlation between taps (70-90%)

13. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 13 Our parameters and conclusions may change before August. Problems we have identified and will rectify soon: The boundary conditions were not well adjusted: some reflections off of the edge of our simulation environment could have influenced measurements taken on the back. We encountered some errors when simulating frequency selective materials.

14. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 14 We wrote Matlab code to simulate the channel. Red = Random channels form Matlab Blue = measured data Distance (m) Path loss (dB)

15. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 15 Conclusions We have developed an UWB channel model for 3-5 GHz band including the following: –Path loss versus distance –Path loss versus frequency –Small scale fading statistics –Power delay profile This model has been implemented in Matlab. Resulting model matches measured results closely.

16. doc.: IEEE 802.15-04/0371r0 Submission July 2004 Andrew Fort & Bart Van Poucke, IMECSlide 16 Future Work Refine our results by eliminating the influence of boundary reflections and correct any problems with frequency selective material definitions. Extend simulation to 2-6 GHz pulses Include the impact of the floor. Confirm our results with actual measurements Measure the path loss at a reference distance close to the antenna. Incorporate the influence of surrounding obstacles (if time).