The hardware implementation of block digital filters.
T. J. Slegel · 1982
Block digital filtering is a relatively new signal processing method where the input samples are grouped into vectors and then, operations are performed on these vectors to carry out the filtering operations. If certain types of parallel processing are used, greater throughput rates can be achieved. This thesis first examines the number of raw computations between a block digital filter and a normal scalar digital filter. Here, the block filter is shown to be far worse than a scalar filter because of the matrix times vector multiplications required by the block algorithm. Next, both types of filters are examined on how they might be implemented on a pipelined array processor. In this case, the block filter gains in throughput rate but is still somewhat slower than a scalar filter. This is because a scalar filtering algorithm can also be executed efficiently on an array processor. Finally, a vector processing architecture, which is optimized for executing a block filter, is presented. The execution time of a given block filter is shown to be a constant and is not dependent on the order of the filter.