Design and Implementation of Discrete-Time Filters For Efficient Rate-Conversion Systems

Oversampled A-to-D converters commonly rely on sharp-cutoff, discrete-time filters that operate at fast input sampling rates. Filter design techniques for such filters typically use the length of the impulse response as an indicator of computational cost, assuming that each filter tap requires a multiplier in Parks, T. W., et al, (1972). This paper describes methods for designing zero-phase filters with sparse impulse responses, i.e., with many zero-valued coefficients, and presents rate-conversion structures for efficiently implementing these designs. By combining polyphase methods and delay-line folding, the structures presented require only one multiplier for each unique value in the impulse response of the filter.