An error cancellation digital-filter structure and its distributed-arithmetic implementation

This note presents a novel digital-filter structure which produces less roundoff noise than that which is produced by the conventional canonical form. This filter structure minimizes the roundoff-noise gain by introducing additional (or altering existing) zeros for the roundoff noise but not for the signal. Since the filter structure is based upon a general noise-canceling technique associated with the quantizer, it can be applied to a variety of forms of digital filters. A distributed-arithmetic ROM implementation of the new structure is presented as an application example to demonstrate its extreme simplicity.