How to simplify ultra wide band radio channel models? Alain Sibille

Transcription

1 How to simplify ultra wide band radio channel models? Alain Sibille Telecom ParisTech

2 Outline Introduction Complexity: why? What is a good channel model Generic/specific UWB channel models Antennas contribution One case study: backscattering based UWB RFID Summary & conclusion page 1 / 53

3 Introduction page 2 / 53

4 Introduction Ultrawideband means At least 500 MHz -10 db BW or 0.2 fractional BW Between 3.1 and 10.6 GHz (US) Which means that a channel model should be able to accommodate such large BW, either fully for a generic model or partly for a specific one page 3 / 53

5 Introduction UWB targets a number of niche or widely used applications High throughput / short distances Low power Location and tracking Use cases and propagation environments are quite diverse Multimedia for home/professional needs Health Industry Transports Logistics Smart homes / cities/ areas... page 4 / 53

6 Introduction One challenge is the variety of cases Another, really UWB specific, is technical: Large frequency variations frequency non uniformity Fine temporal phenomena waveform distortions How can channel models cope with these two challenges? Radio channel page 5 / 53

7 Complexity: why? page 6 / 53

8 Complexity: why? 1980 page 7 / 53

9 Complexity: why? 2012 page 8 / 53

10 Complexity: why? WINNER, D5.4, 2005 page 9 / 53

11 Complexity: why? Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain Close environment domain Mobility (macro) domain page 10 / 53

12 Complexity: why? response (arb. units s) Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain response (arb. units s) Close environment domain Mobility (macro) domain time (ns) time (ns) page 11 / 53

13 Complexity: why? Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain Close environment domain Mobility (macro) domain page 12 / 53

14 Complexity: why? Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain Close environment domain Mobility (macro) domain page 13 / 53

15 Complexity: why? Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain Close environment domain Mobility (macro) domain page 14 / 53

16 Complexity: why? Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain Close environment domain Mobility (macro) domain page 15 / 53

17 Complexity: why? Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain Close environment domain Mobility (macro) domain page 16 / 53

18 Complexity: why? Frequency domain Delay domain Angular domain/multiantenna domain Polarization domain Far/medium distance environment domain Close environment domain Mobility (macro) domain page 17 / 53

19 What is a good channel model? page 18 / 53

20 What is a good channel model? Why do we need channel models? For system deployment: prediction of coverage, engineering rules For the testing of competing PHY schemes (and beyond) For the testing of terminal performance (e.g. OTA) To be used as standards Preferably a good model! Can a single model achieve all that? Maybe but the price to pay is enormous Dimensionality curve! page 19 / 53

21 What is a good channel model? How can we justify the quality of a channel model? What metric used to measure the quality? How representative is the data used to construct the model? M. Landmann et al. On the influence of incomplete data models on estimated angular distributions in channel characterisation, COST 2100 TD(07)321, 2007 Given the uncertainty on the value of parameters, how worth is a complicated model? page 20 / 53

22 What is a good channel model? Model construction aided by first principles from information theory Maximize entropy Criteria for model quality (Akaike, MDL, goodness of fit ) Example: tap power statistics, inter-tap correlations (beyond US) U.G. Schuster et al. Ultrawideband Channel Modeling on the Basis of Information-Theoretic Criteria, IEEE TWC, 2007 page 21 / 53

23 What is a good channel model? Example: tap power statistics, tap correlations (beyond US) page 22 / 53

24 What is a good channel model? Example: tap power statistics, tap correlations (beyond US) page 23 / 53

25 What is a good channel model? Example: tap power statistics, tap correlations (beyond US) A. Menouni Hayar et al., Eurasip, 2005 page 24 / 53

26 What is a good channel model? Two approaches: generic Π 1 Phenomena 1 Radio channel Π 2 Π n Phenomena 2 Phenomena n... page 25 / 53

27 What is a good channel model? Two approaches: specific Phenomena 1 Radio channel Phenomena 2 Phenomena N... page 26 / 53

28 What is a good channel model? Two approaches: generic Π 1 Phenomena 1 Radio channel Π 2 Π n Phenomena 2 Phenomena n... page 27 / 53

29 Generic/specific UWB channel models page 28 / 53

30 Generic/specific UWB channel models IEEE a (generic) Cluster based (Saleh-Valenzuela) Multiple environments Frequency dependence of the path gain Mixed Poisson times of arrival, delay-dependent cluster decay Path small scale fading What is lacking? Antennas Directionality, Polarization mobility A.F. Molisch et al., 2006 page 29 / 53

31 Generic/specific UWB channel models IEEE a (generic) CM2 (residential NLOS) CM8 (industrial NLOS) A.F. Molisch et al., 2006 page 30 / 53

32 Generic/specific UWB channel models IEEE a (generic) A.F. Molisch et al., 2006 page 31 / 53

33 Generic/specific UWB channel models Room electromagnetics (specific) page 32 / 53

34 Generic/specific UWB channel models Room electromagnetics (specific) page 33 / 53

35 Generic/specific UWB channel models Room electromagnetics (specific) page 34 / 53

36 Generic/specific UWB channel models Room electromagnetics (specific) t 4V 1 W = W exp τ r D D ηa 0 = = d 1 2 τ cηa 2 J. Bach-Andersen et al., 2006 page 35 / 53

37 Generic/specific UWB channel models Mobility and moving obstruction (generic specific ) J. Kunisch et al., 2006 P. Pagani et al., 2006 page 36 / 53

38 Generic/specific UWB channel models An add-on approach of specific effects to complete a generic model By impacting the statistics of the multipath powers / angles By impacting the statistics of the multipath Doppler spectrum By impacting the statistics of the multipath correlations page 37 / 53

39 Antennas contribution page 38 / 53

40 Antennas contribution What happens with realistic (use case related) antennas? frequency dependence temporal dispersion Angular characteristics Can such (complicated) effects be simply assessed? page 39 / 53

41 Antennas contribution Is it possible to combine simply an antenna neutral channel model with an antenna model? Answer: yes, in the heavy way page 40 / 53

42 Antennas contribution Is it possible to combine simply an antenna neutral channel model with an antenna model? Answer: yes, in an alternative manner Analyze the effect of antennas, given the Tx-Rx processing scheme Include an effective antenna model, parameterized by the Tx-Rx processing scheme page 41 / 53

43 Antennas contribution cumulated probability 1 monocone-monocone Ideal templ. Adapt. templ. Incoherent OFDM cumulated probability 1 monocone-monocone Ideal templ. Adapt. templ. Incoherent OFDM SNR/ideal antenna SNR (db) SNR/ideal antenna SNR (db) LOS low multipath density 0.5 LOS dense multipath Amplitude Amplitude Time (ns) Time (ns) A. Sibille et al., 2006 page 42 / 53

44 Antennas contribution cumulated probability horn-horn Ideal templ. Adapt. templ. Incoherent OFDM SNR/ideal antenna SNR (db) 1 cumulated probability horn-horn Ideal templ. Adapt. templ. Incoherent OFDM SNR/ideal antenna SNR (db) LOS low multipath density 0.5 LOS dense multipath Amplitude Amplitude Time (ns) Time (ns) A. Sibille et al., 2006 page 43 / 53

45 Antennas contribution Observation Angular filtering by the antenna pattern, affecting multipath capture (not UWB specific) Effective filtering depends on the system PHY (incoherent less sensitive to distortion better energy capture) A simple way to account for antennas Apply a per-path filter - Related to the path direction, iff directional channel model - Related to the path delay otherwise, assuming a delay-direction dependence - In all cases parameterized according to the physical layer scheme page 44 / 53

46 Antennas contribution Problem How to deal with the variety of antennas? Answer Antennas can be seen as a stochastic part of the channel, propagation likewise stochastic antenna modeling an emerging research topic see e.g. the special session Statistical methods and applications in electromagnetism at AES 2012, Paris page 45 / 53

47 One study case: backscattering UWB RFID page 46 / 53

48 Backscattering UWB RFID page 47 / 53

49 Backscattering UWB RFID Issues for an RFID BS based channel model: UWB propagation in relevant environments (industrial, commercial ) RFID intrinsic radioelectric behaviour for a variety of potential designs Disturbances brought by the variety of attached objects Mobility for localization and tracking A nice challenge! page 48 / 53

50 Backscattering UWB RFID How to develop a simple UWB BS channel model? On one hand: energy aspects (integrated over 3-5 GHz) are dealt with through statistical modeling of the tag BS, given a subset of the propagation channel parameters (path powers, angles) The impulse response (IR) is treated separately, based on IEEE a and convolved with the tag IR (where necessary) high level parameters (environment type, tag use case..) are consistently used throughout tag mobility is injected as an add-on to IEEE a Specific phenomena are treated as add-on - Clutter, based on an adaptation of IEEE a - Tag structural scattering (bistatic) page 49 / 53

51 Backscattering UWB RFID Environed tags database Propagation database Environm ment and use case parameters antenna/structural mode tag statistical model A priori Directional model at tag IEEE based SV channel model Mobility model Simplification Simplification BS structural mode path loss model BS antenna mode path loss model BS tag impulse response BS clutter model and reader-reader channel model page 50 / 53

52 Backscattering UWB RFID A quadruple level of variability accounted for: tags, objects, rotations, propagation Isolated tags CD DF rotated Disturbed tags 0.2 vertical See also: R. D Errico et al., Eucap 2012 Z. Mhanna et al., Eucap 2012 page 51 / Backscattering Effective gain (db) A. Sibille et al., ICUWB, 2011

53 Summary & conclusion UWB channels contain even more complexity than ordinary channels and we require more and more from them The trade-off between complexity and accuracy is exhausting: a universal model doesn t exist, or is inaccurate Specific models are suited to well identified needs and demand controlled modeling efforts. Alternatively we may complement a generic model by ensuring that the targeted phenomena are properly modeled Don t ask a model more than it can do page 52 / 53

54 Acknowledgements This presentation is based on the work of a team at ENSTA-Ptech then Telecom-Ptech, which has been active in UWB for nearly 10 years: C. Roblin, F. Guidi, Z. Mhanna, A. Mellah, M. Sacko, E. De Mur, M.A. Yousuf, S. Bories, R. D Errico, J. Braga, H. Ghannoum, X. Zeng, Y. Wei, V. Casadei Particularly in collaboration with Univ. Bologna and CEA-LETI Thank you page 53 / 53