A Fresh Look At 'Broadband ' Passive Sonar Processing

Traditional passive broadband sonar processing has been constructed by performing spatial decomposition (beamforming) followed by a temporal filter (Eckart weighting) structured to maximize deflection for signal present while minimizing deflection under noise only conditions. This processing, while approaching optimality in the stationary single signal-stationary noise case, is not well suited for clutter rich environments and spectrally diverse signal sets. Techniques such as the Smoothed Coherent Transform (SCOT) and the PHAse Transform (PHAT) as well as a new ad hoc construction referred to as Sub-band Peak Energy Detection or SPED are finding increased usage in modern sonar systems. This paper discusses sources of the performance gains over traditional processing. A conclusion that 'Passive Broadband ' is fundamentally a Direction of Arrival (DOA) estimation problem is developed from this analysis. Based on this conclusion higher resolution techniques including, Quadratic Spectral Capon and Spectral MUSIC are applied to generate a similar broadband output. A comparison of performance to the Navy's current Broadband schemes against a common real data set is presented. 1