Optimal Architectures for Massively Parallel Implementation of Hard Real-time Beamformers Literature Survey

Computationally intense spatial filtering, called beamforming, is central to acoustic imaging. Typically, in order to achieve an update rate similar to slow scan television, beamforming must operate in real-time, which leads to massive parallel processing. This paper surveys appropriate work in the field of synchronous data flow (SDF) as applied to the beamforming problem and defines the objectives of a research project. The project is focussed on the application of SDF techniques as a strategy to select the optimal architecture for a beamformer. In order to formulate the proper approach, the applicable theory of SDF is reviewed, including methods used to optimize designs within the SDF domain. The beamforming process is described, and two competing architectures of the beamformer are defined, along with the optimality criteria for selection.