Performance of reduced-complexity multi-channel equalizers for underwater acoustic communications

Underwater acoustic telemetry in shallow water environments is difficult due to large delay spreads, rapid fading, and reverberation clutter. Development of usable array equalizers remains an ongoing effort in the community. We evaluate three block-adaptive multichannel equalizer architectures that are effective for the underwater channel. These use a finite-window least squares (LS) approach for estimating both channel and equalizer. The channel identification model is constructed by decision-direction. The three equalizer designs are based on criteria of MMSE, zero-forcing, and space-time matched filtering justified by passive phase conjugation (PPC, or time-reversal mirror) theory. This architecture avoids committing to temporal parameter correlation models inherent in standard RLS adaptation methods. Robust numerical linear algebra technology (algorithm LSQR) iteratively solves the large linear systems involved. The performance evaluation compares MSE of the three equalizers in context of the adaptive architecture under channel estimation errors. We show an exact MSE expression for PPC equalization that includes channel identification error. The evaluations are based on data from underwater acoustic telemetry experiments made in Puget Sound, Seattle.