Measurements and Modeling for the 5.9 GHz WAVE Vehicle to Vehicle Channel

1. January 2004 Measurements and Modeling for the 5.9 GHz WAVE Vehicle to Vehicle Channel Prof. Mary Ann Ingram School of Electrical and Computer Engineering Georgia Institute of Technology January 2004 Mary Ann Ingram, Georgia Tech

2. Objective and Approach • Determine the worst case for the 5.9 GHz DSRC mobile-to-mobile channel • Channel measurements and literature used to determine extreme channel parameters • Computer simulations and throughput of LINKSYS 802.11g link over a channel emulator used to rank channel harshness Mary Ann Ingram, Georgia Tech

3. Overview of This Talk • Channel parameterization and worst-case trends • Review of Results • Literature survey • Channel measurements • SIMULINK simulations • Emulator throughput tests • Conclusions • Discuss next steps Mary Ann Ingram, Georgia Tech

4. Channel Description • Vehicle-to-vehicle (V2V) • All vehicles going 80 mph • Range up to 1000m • Small-scale fading only • OFDM • 52 subcarriers (48 data, 4 pilots) • 1.6ms guard interval • 10 MHz bandwidth Mary Ann Ingram, Georgia Tech

5. Channel Parameterization N Number of paths Delay of the nth path Gain of the nth path Mary Ann Ingram, Georgia Tech

6. d + + ( f 2100 Hz ) d - ( f 2100 Hz ) t n P n d + + 0 . 5 ( f 100 Hz ) S ( f ) n d ( f ) d - 0 . 5 ( f 100 Hz ) + ¥ + ¥ + ¥ K n Current Test Parameters Parameter Description Multi - Doppler Amplitude Rician Path Spread Variation Test Test Test Test N No. of taps 2 1 1 1 Excess delay 200ns 0 0 0 th of n tap L Path loss Min sens. Min sens. + Min sens. + 3dB Min sens. + + 3dB 3dB 3d B Normalized Equal 0 dB 0 dB 0 dB tap power ( - 3 dB distribution each) th PSD of n Flat 6 dB tap Classical or (“bathtub”) ( f ) d +A spectrum Rice factor 10 dB th of n tap Mary Ann Ingram, Georgia Tech

7. Worst-case Parameter Trends Mary Ann Ingram, Georgia Tech

8. Definitions of Worst-case Channel • The channel that has the worst-case extreme values for every parameter • A harsh channel that causes a “well-designed” receiver to nearly fail at a range of 1000m Mary Ann Ingram, Georgia Tech

9. Overview of This Talk • Channel parameterization and worst-case trends • Review of Results • Literature survey • Channel measurements • SIMULINK simulations • Emulator throughput tests • Conclusions • Discuss next steps Mary Ann Ingram, Georgia Tech

10. Literature Survey • Most papers treat Roadside-to-vehicle • For vehicle-to-vehicle, • Path loss will be higher • Delay spread will be lower • Doppler will be higher • K factor will be lower • Consider overbound for delay spread vs. path loss Mary Ann Ingram, Georgia Tech

11. Path Loss • 2-ray model at 1000m • Roadside-to-vehicle, 2.6 GHz • Theory: 105 dB • Measured: 120 dB • Microcell, antenna ht = 3.7 m, 1.9 GHz: 112 dB • Tokoyo street microcells, 2 GHz: ~112 dB • Roadside-to-vehicle, 5.8 GHz, measured agrees well with 2-ray for uncluttered, LOS channel • Truck or bus diffraction loss: >= 20 dB Mary Ann Ingram, Georgia Tech

12. Intersection of Worst-case Environmental Features • Intersections with main-street urban canyons with numerous glass/metallic reflectors, including distant reflectors and clutter near transmitter or receiver • Tall, dumb directional antennas not directed at LOS with no downtilt • Relatively large TX-RX separations Mary Ann Ingram, Georgia Tech

13. Platoon Vehicle-to-vehicle Delay Spread Summary • Delay spreads are small because in the platoon application vehicles are crowded Mary Ann Ingram, Georgia Tech

14. Delay Spread Overbound vs. Path Loss • 1.9 GHz • Microcell • V2V delay spead will be lower Up to 13.3 m antenna height 3.7 m antenna height ~ 1 ms 112 dB Mary Ann Ingram, Georgia Tech

15. Doppler Spectra • Theoretical V2V flat-fading, based on circle-to-circle scattering model • Measured V2V flat-fading • Measured roadside-to-vehicle frequency selective (per-tap spectra) Mary Ann Ingram, Georgia Tech

16. RX TX RX TX Circle-to-circle V2V Flat-fading Model Single-ring model Double-ring model The critical parameter is the ratio of vehicle speeds Mary Ann Ingram, Georgia Tech

17. Spectra for the Circle-to-circle Model equal velocity • The first few taps of our measured freeway spectra resemble the a=1 spectra V1=2V2 fixed- to-veh. Mary Ann Ingram, Georgia Tech

18. Measured Flat-fading Vehicle-to-Vehicle Spectra Urban Highway Mary Ann Ingram, Georgia Tech

19. Fixed-to-Vehicle Per-tap Doppler Spectra Mary Ann Ingram, Georgia Tech

20. Fading Distribution • Flat-fading vehicle-to-vehicle @ 5.2 GHz: Rician • Per-tap vehicle-to-vehicle @ 900 MHz: Rician 1.38<K<17.6, typically 5<K<11 • Per-tap fixed-to-vehicle • LOS: 2 dB<K<5.5dB on early taps, Rayleigh on late taps • NLOS: -1 dB on 1st tap, Rayleigh on late taps Mary Ann Ingram, Georgia Tech

21. Conclusions From Literature Study • Extreme vehicle-to-vehicle parameter values • Path loss @ 2 GHz, 1000m = 120 dB • Delay spread @ 1000m = 1 ms (pessimistic) • Doppler per tap – All shapes are possible • Fading distribution: K values as low as -1 dB on 1st tap, Rayleigh on rest Mary Ann Ingram, Georgia Tech

22. Overview of This Talk • Channel parameterization and worst-case trends • Review of Results • Literature survey • Channel measurements • SIMULINK simulations • Emulator throughput tests • Conclusions • Discuss next steps Mary Ann Ingram, Georgia Tech

23. Channel Measurements in Atlanta, Georgia • Phase I – OTS 802.11b sniffer • Delay only • 7 sites identified with large spreads (~300ns) • Reported in August 20, 2003 presentation • Phase II – Custom DSSS sounding at 2.4 GHz up to 300m range • Delay and Doppler • Reported in November 11, 2003 presentation Mary Ann Ingram, Georgia Tech

24. News Since November • Low-power intermodulation-like products apparent in post-collection testing of equipment • Phase II delay and most powerful Doppler components deemed reliable Mary Ann Ingram, Georgia Tech

25. Main Observations of Phase II • Majority of delay profiles are unimodal with no more than 3 significant 50ns taps • Delay spreads of 300ns confirmed when Phase I sites were re-visited • An absolute delay of 600ns was observed • Up to 8 consecutive taps within 7 dB of each other observed • Repeatable per-tap Doppler profiles, not previously published, measured on freeway Mary Ann Ingram, Georgia Tech

26. RMS Delay Spread of MA (L=10) of Sounding Samples Ignoring L Samples after Threshold = -70 dBm 350 300 ns 300 250 200 delay, nsec 150 100 50 0 0 0.5 1 1.5 2 2.5 3 3.5 4 Sounding samples 4 x 10 Example of Isolated High Values of Delay Spread From Phase I Peak occurred when passing the intersection in the middle Delay vs. sounding sample Mary Ann Ingram, Georgia Tech

27. Extreme Absolute Delay Spread of 600 ns (301 ns rms) Freeway exit ramp 10 dB 600ns (180m) Mary Ann Ingram, Georgia Tech

28. Extreme Number of Strong Consecutive Taps T-intersection with tennis courts, parking lots, and 4-story bldgs Power Delay Profile 10 dB Eight 50 ns taps Mary Ann Ingram, Georgia Tech

29. Two Passes Through a Freeway Channel Vehicles going in same direction in slow lane (55mph) Power Delay Profiles Pass 1 Pass 2 Mary Ann Ingram, Georgia Tech

30. Short-time (50 ms) and Long-time (0.7s) Spectra for Pass 1 +600Hz -600Hz +1000Hz -1000Hz Tap 1 Tap 2 Mary Ann Ingram, Georgia Tech

31. Taps 3 and 4 400 Hz component is transient Tap 3 Tap 4 Mary Ann Ingram, Georgia Tech

32. Path is closing at nearly 110 mph Stationary Reflector 55 mph 55 mph Suspected Origin of 400 Hz Component • Scaled to 5.825 GHz and 80 mph, this component is 1,383 Hz • Current Doppler shift in Std is 2100 Hz Mary Ann Ingram, Georgia Tech

33. Tap 6 • Except for DC component, spectrum becomes a shallow bathtub Mary Ann Ingram, Georgia Tech

34. Short-time (50 ms) and Long-time (0.7s) Spectra for Pass 2 Tap 1 Tap 2 Mary Ann Ingram, Georgia Tech

35. Taps 3 and 4 Tap 3 Tap 4 Mary Ann Ingram, Georgia Tech

36. Tap 5 Mary Ann Ingram, Georgia Tech

37. Channel Emulator Settings That Match Highway Pass 1 (Unscaled) Mary Ann Ingram, Georgia Tech

38. Extremes Found in Measurements • RMS delay spread: 301 ns • Number of significant 50 ns taps: 8 • Doppler shape: box (uniform) or shallow bathtub + DC component • Maximum Doppler (scaled): 1,383 Hz • K-factor: 1 in Highway, 0 elsewhere Mary Ann Ingram, Georgia Tech

39. Overview of This Talk • Channel parameterization and worst-case trends • Review of Results • Literature survey • Channel measurements • SIMULINK simulations • Emulator throughput tests • Conclusions • Discuss next steps Mary Ann Ingram, Georgia Tech

40. Simulation Description • Used SIMULINK standard block library • DSRC and 802.11a specs • Includes ½-rate code, interleaving and puncturing • Differential modulation • Receiver perfectly synchronized to LOS, regardless of delay or Doppler profiles • Gives pessimistic performance because real receiver would not always synchronize to LOS (e.g. when later tap is stronger) Mary Ann Ingram, Georgia Tech

41. Convolutional Encoder D-QPSK Modulator Puncturer Interleaver Viterbi Decoder D-QPSK Demodulator Deinterleaver Input Data OFDMTX Coding & Modulation De-coding & Demodulation Output Data Channel OFDM RX BER BER Calculator Block Diagram Mary Ann Ingram, Georgia Tech

42. Simulation Strategies • Vary an individual channel parameter, while keeping others fixed, to see which are the most critical channel parameters • Try various combinations of channel parameters, for example delay and max Doppler, to find combinations cause the highest BER • Emphasis BER in 10-5 to 10-4 range because these correspond to PERs of 10% Mary Ann Ingram, Georgia Tech

43. PER Definition • Simulated 1000-byte “packet” to get PER • Packet error declared if at least one bit in a packet is wrong • In 2- and 3-tap simulation with Rayleigh fading and rms delay of 300ns or 400ns, 10% PER 10-5 < BER < 10-4 Mary Ann Ingram, Georgia Tech

44. Observations from Simulations • RMS delay spread is the most critical parameter • More paths, with fixed rms delay spread is worse • Equal tap powers is worse • Max Doppler on second tap (2 taps total) not so important when powers are equal • Rayleigh per-tap distribution is worse than Rician Mary Ann Ingram, Georgia Tech

45. Overview of This Talk • Channel parameterization and worst-case trends • Review of Results • Literature survey • Channel measurements • SIMULINK simulations • Emulator throughput tests • Conclusions • Discuss next steps Mary Ann Ingram, Georgia Tech

46. Throughput Testing at 2.4GHz • NetIQ’s Chariot 4.3 software used to measure throughput of a LINKSYS 802.11b link over the TAS 4500 RF channel emulator • Usually high SNR (received power = -35 dBm) • Performed long (1 GByte) and short (10 Mbytes) tests Mary Ann Ingram, Georgia Tech

47. Rate and Modulation Differences Between 802.11g and DSRC • 802.11g • 1 & 2 Mbps DSSS • 5.5 & 11 Mbps CCK • 5.5 & 11 Mbps HR/DSSS/PBCC (optional mode replacing CCK) • 22 & 33 Mbps ERP_PBCC (optional) • 6, 9 12, 18 24, 36 48, 54 Mbps OFDM • 6, 9, 12, 18, 24, 36, 48, 54 Mbps DSSS-OFDM (optional) • DSRC • 3, 4.5, 6, 9, 12, 18, 24, and 27 Mbps, all OFDM Mary Ann Ingram, Georgia Tech

48. Throughput Test Set-up Mary Ann Ingram, Georgia Tech

49. Favorable 2-Path Channel Showing Repeatability 54 Mbps Rate is higher than highest DSRC rate (27 Mbps) • Essentially a 1-path model with zero Doppler Mary Ann Ingram, Georgia Tech

50. Unfavorable 2-Path Channel Showing Repeatability 2 Mbps Rate is lower than lowest DSRC rate (3 Mbps) and is not OFDM • Equal-powered paths, 150ns rms delay, Rayleigh with classic Doppler, fmax=1000Hz Mary Ann Ingram, Georgia Tech