On the role of modeling in passive synthetic aperture processing

There is a conventional wisdom regarding passive synthetic aperture processing which states that it is not a tenable method of array processing. The reasons given are: (1) There is not enough temporal coherence in the signal to allow the full spatial aperture to be realized, (2) It is simply a scheme to convert temporal gain to spatial gain, so there is no new gain achieved. (3) The source frequency must be known a priori to compute the phase correction factor, otherwise the unknown Doppler will introduce an unacceptable bias in any bearing estimation problem. This paper shows that all of these objections assume signal models that do not faithfully embody the physics of the situation, and thereby basically doom the processor to failure. It is shown that by using more than one hydrophone the source frequency can be estimated simultaneously with the bearing, that by including the forward motion of the array in the signal model the bearing information intrinsic to the Doppler can be utilized, and that by applying a recursive (Kalman based) algorithm the errors due to changes in temporal coherence can be eliminated. Results using real data are presented that demonstrate the improvement over the conventional beamformer