Channel effects on DSSS Rake receiver performance Paul Hursky, Michael B. Porter Center for Ocean Research, SAIC Vincent K. McDonald SPAWARSYSCEN KauaiEx Group Ocean Acoustics Conference, San Diego, 4 March 2004 Work sponsored by ONR
Outline Will discuss DSSS modulation with Rake receiver, originally designed for multi-user access Have tested this at several shallow water sites while simultaneously measuring the state of the ocean will show results of this testing Will discuss how various channel phenomena have impacted the performance of this system and what needs to be modeled to predict performance in general
Acknowledgments DSSS Rake receiver from: E. M. Sozer, J. G. Proakis, M. Stojanovic, M. Hatch, J. A. Rice, A. Benson, Direct Sequence Spread Spectrum Based Modem for Under Water Acoustic Communication and Channel Measurement, Oceans 99 Recent work: M. Stojanovic, L. Freitag, Hypothesis-Feedback Equalization for Direct-Sequence Spread Spectrum Underwater Communication, Oceans 00. F. Blackmon, E. Sozer, M. Stojanovic, J. Proakis, Performance Comparison of RAKE and Hypothesis Feedback Direct Sequence Spread Spectrum Techniques for Underwater Communication Applications, Oceans 02.
Direct-Sequence Spread Spectrum Designed for multi-user applications, DSSS is an alternative to frequency-hopped FSK Rather than assigning a frequency hopping pattern to each user, assign distinct bi-polar sequences to different users that have minimal cross-correlation (Gold, Kasami): User 1: 1, 1, -1, 1, etc User 2: 1, -1, -1, -1, etc These sequences are used to modulate the phase of a carrier at the chip rate each information bit conveyed by L chips: L the spreading rate
How information is transmitted c1 c2 c3 c4 c5 c6 c7 c8 c9 c10 Bit 1 Bit 2
Why use more bandwidth? Would ordinarily transmit at information bit rate Instead, send L chips for every info bit This spreads spectrum by factor of L At receiver, we do matched filter on known sequence nad use this to make decision Because different chip sequences are uncorrelated, this effectively decorrelates multiple access (multi-user) interference, and isolates different multipath arrivals
Spreading gain (chips/bit)
Synchronize on 3-pulse preamble
Modem parameters 8-16 khz band, 4000 chips/sec, and spreading factors of 10, 20, 40, 80 This yields info bit rates of 50 (4000/80), 100, 200 and 400 bps
DSSS M and Gold sequences M sequences used in ocean acoustic tomography for their auto-correlation properties Gold sequences used in multi-user spread spectrum wireless communications for their cross-correlation properties Note that the desirable properties of these sequences are the result of circular correlation, which rarely happens, given sign changes across bit boundaries and multipath
Rake receiver multipath recombination Z -1 Z -1 Z -1 p n p n p n Integrate, threshold Integrate, threshold Integrate, threshold Z -1 * Z -1 * Z -1 * To detector
DSSS Rake receiver timing loop Synchronization with incoming chip sequence can drift compare matched filter outputs of slightly delayed and slightly advanced in time versions of signal If in synch, delayed and advanced versions will be equal otherwise, use difference to correct for drift (shift the reference point in the data where we start matching on the spreading sequence)
Simulation example for illustration
KauaiEx configuration Receive VLA Transmit Testbed
KauaiEx sound speed profile
Performance at KauaiEx site 50 bps 100 bps 200 bps
200 bps BER vs SNR
200 bps BER vs channel (depth)
Near surface Impulse responses Mid-array Near bottom
Bathymetry at ElbaEx-North site
Bathymetry at ElbaEx-South site
Elba sound speed profiles North site South site
Impulse response functions vs depth North site South site
Impulse response function vs time North site South site (receive element nearest the surface)
Comparing two Elba sites: 200 bps North site South site
Comparing two Elba sites: 100 bps North site South site
200 bps BER % vs channel North site South site
Impact of multipath Mechanism to combat ISI due to multipath is spreading gain Because we are taking a small excerpt from Gold sequence, we are not realizing touted low cross-correlation properties However, increasing number of chips per bit means longer integration times phase may wander and system will lose spreading gain Increasing bandwidth (more chips/second) would result in better performance at higher bit rates
Need propagation-based simulator Acoustic communications is not always predictable (in the field) Cannot plan missions without being able to predict what configuration will achieve desired levels of performance With simulator, can synthesize arbitrary environments - determine performance and optimal configuration before deployment, using knowledge of ocean state and weather (inputs to acoustic propagation models)
Modeling for acoustic comms Acoustic propagation modeling a mature technology at low frequency, but acoustic modeling at high frequency for acomms mission planning is in its infancy: How do we handle dynamics? What phenomena are important? What are the statistical characterizations of various ocean phenomena that enable us to adequately reproduce the performance of acoustic communications systems? Deterministic modeling seems out of reach...
Doppler due to motion and surface
Need statistical characterization to generate realizations
Conclusions DSSS modulation not taking full advantage of Gold or m sequence properties multipath not isolated by Rake structure Fluctuations exacerbate problems caused by multipath Modeling a frozen ocean only covers the most basic failure modes (does not address synchronization very well) Need to extend current acoustic models to include dynamics, especially due to surface motion