Outline Use phase/channel tracking, DFE, and interference cancellation techniques in combination with physics-base time reversal for the acoustic MIMO

Transcription

1 High Rate Time Reversal MIMO Communications Aijun Song Mohsen nbdi Badiey University of Delaware Newark, DE University of Rhode Island, Oct. 2009

2 Outline Use phase/channel tracking, DFE, and interference cancellation techniques in combination with physics-base time reversal for the acoustic MIMO channel Introduction MIMO Receivers MIMO Experiment (MakaiEx) Results and Discussions MIMO-AUV Summary

3 System Model Received signal is summation of ISI distorted transmissions from all TXs plus ambient noise

4 Existing MIMO receivers Challenge for underwater environments Severe time-varying inter-symbol interference (ISI)-doubly spread channel Co-channel interference (CoI) Multi-channel DFE for multi-user user comm to jointly deal with ihisi and dcoi I(Stojanovic & Zvonar, JOE 1996) Time reversal for multi-user user at 3-5 khz (limited CoI, channel li in-variant) for MIMO/multi-user, user, r(hcs (H.C. Song et al.,, JOE 2006 & 2007) Iterative MIMO processors for coded systems (Roy, Duman et al.,, JOE 2007 & 2009)-Based on multi- channel DFEs

5 Proposed Structure Core: Time reversal enhanced single channel DFE Doppler/phase/channel tracking to combat channel variations at high freq (30-40 khz) Interference cancellation (IC) techniques Channel estimators Low-complexity implementation ti No channel coding

6 MIMO receivers: Without IC Channel Estimation Phase tracking Core demodulator Each data stream is demodulated considering others as noise

7 Channel-estimation-based processor Processing the data in blocks in decision-directed directed fashion Several types of channel estimators implemented in the structure Least squares (LS) algorithm Basic matching-pursuit (MP) algorithm LS-MP algorithm

8 Phase tracking Stage 1: Average Doppler removed by time domain re-sampling Stage 2: Phase fluctuations at individual channels for all TXs Modeled as linear trend, same for all TXs Time-varying Doppler, but processed in the narrowband Stage 3: Residual phase tracking in the core demodulator for each TX

9 Core demodulator Output SNR (Inverse of MSE) Time reversal enhanced single channel DFE Aided by channel estimates Song et al.,, JASA Feb 2008

10 MIMO receivers: Serial IC Data streams demodulated in the order of signal strength

11 MIMO receivers: Multistage IC Iterate between MIMO channel estimation and interference cancellation 2-3 iterations

12 MakaiEx 2005 Kauai Island, HI Sept--Oct, 2005 Sept 100 m water depth maps.google.com

13 MakaiEx 2005

14 MakaiEx: Wind/Water temperature Rough sea surface (wind speed >20 Stratified water 23:00 Source/receiver above the thermocline

15 Drifting source/receiver in a dynamic underwater environment Wind speed > 20 m/s

16 Communication data packets 1, 2 or 4 transducers simultaneously signaling BPSK/QPSK packets Source separation: 6 m Training overhead: 34% (Training symbols not needed for optimal receiver configuration)

17 Results and Discussions Role of phase tracking The MIMO channel and two estimators Performance of the basic receiver (no IC) Performance of the receiver with multi-stage IC

18 Time-varying, sparse Impulse responses LS estimation: Doppler removed Fading for all arrivals Different intensity levels for different TXs

19 Phase fluctuations Linear trend (stage 2) Residual phase (stage 3) Without phase tracking (stage 2), most 2/4-Tx packets fail

20 LS as channel estimator Non- negligible nonzero estimates

21 LS versus MP as channel estimator 4-Tx QPSK packet: Similar output SNR at two strong TXs 1.5 gain by sparse MP estimator for two weak TXs

22 Basic receiver- without IC: 1-Tx BPSK Dt Data rate: rt:4kilbit/ kilobits/s BER=0/5568

23 Basic receiver- without IC: 2-Tx BPSK Data rate: 8 kilobits/s BER=2/10880

24 Basic receiver- without IC: 4-Tx BPSK Dt Data rate: rt:16kilbit/ kilobits/s BER=459/20960(2%)

25 Basic receiver- without IC: 1-Tx QPSK Dt Data rate: rt:8kilbit/ kilobits/s BER=0/11152

26 Basic receiver- without IC: 2-Tx QPSK Dt Data rate: rt:16kilbit/ kilobits/s BER=121/21760 (0.6%)

27 Basic receiver- without IC: 4-Tx QPSK Dt Data rate: rt:32kilbit/ kilobits/s BER=3086/41920 (7%)

28 Multi-stage IC (MP): How many iteration needed? 2-TX QPSK: converge in two IC stages 3-4 db improvements for both TXs

29 Multi-stage IC (MP): 4-TX QPSK: converge in three IC stages 3-5 db improvements

30 Scatter plot for Tx-1 Without IC Multi-stage IC (k=3)

31 Multi-stage IC (LS versus MP): Reduced d improvement with LS estimates t LS estimator 3 db worse than MP for weak TXs

32 Optimized performance Multi-stage IC with sparse MP estimation 2-Tx BPSK, 8 kilobits/s, all most error-free 4-Tx BPSK, 16 kilobits/s, BER<.6% 2-Tx QPSK, 16 kilobits/s, BER<.2% 4-Tx QPSK, 32 kilobits/s, BER<2% Maximum bandwidth efficiency=4.57 i bits/s/hz /H for 7 khz use Periodic training should not be needed (decision- directed)

33 MIMO Acoustic Transceiver System camera battery DVL INS Technical Specifications-Hafmynd Gavia kg; 200 m depth rating; 20 cm diameter T da Advanced navigation system (GPS Inertial Navigation + DVL) Surface communications (WiFi & Iridium Satellite) Acoustic Sensors- # tx channels=4, # rx channels= 8 Electronics DAQ 187 db source level l at khz band 2-30 khz receiving band, sample rates=80 khz Mission Capability- Speed=2 m/s max.5 m/s min (Max, and dmi Min) Length over which data can be collected=25-30km per charge depending on speed and configuration Endurance=4-5 5hrs per charge for single battery pack (could double with the purchase of a 2nd set) Towed Array MIMO

34 Acoustic Modem on AUVs Signal Acquisition/Transmission Module

35 Summary: Rates (6-7 khz bandwidth) Data Rate Range Overall Exp/Time No. Tx/Rcx BER 6 kilobits/s 2 km 0 MakaiEx 2005: 1Tx/7Rcx SIMO 16 kilobits/s 2 km <0.02 MakaiEx 2005: 4Tx/8Rcx MIMO 32 kilobits/s 2 km <0.04 MakaiEx 2005: 4Tx/8Rcx MIMO 16 kilobits/s 4 km KAM08: SIMO 1Tx/16Rcx 12 kilobits/s 8 km KAM08: SIMO 1Tx/16Rcx 16 kilobits/s 4 km KAM08: MIMO 4Tx/16Rcx

36 Summary: Extended period Data Rate Range Overall Period No. Tx/Rcx BER 4 kilobits/s 3 km hours 1Tx/4Rcx KauaiEx kilobits/s 4 km hours 1Tx/4Rcx KAM08 5 kilobits/s 8 km hours 1Tx/4Rcx 4 kilobits/s km KAM08 < hours KAM08- TOWED 1Tx/4Rcx

37 Acknowledgement Thanks to: M. Porter (Heat, Light & Sound Research Inc.) VK V. K. McDonald MD ld(spawar) T. C. Yang (Naval Research Lab) Participants i t of MkiE MakaiEx Support from ONR 321OA (Robert Headrick)

38 Summary High rate MIMO receivers Core demodulator: time reversal enhanced by single channel DFE Doppler and phase compensation (in three stages) Interference cancellation: serial and multi-stage Two channel estimators: LS and MP algorithms Experimental results Drifting source/receiver in dynamic ocean environment Significant improvements by parallel interference cancelation 16 kilobits/s and 32 kilobits/s using 7 khz bandwidth MIMO-AUV