An algorithm for the design of optimal finite word-length FIR digital filters

The purpose of this work was to develop and implement an algorithm for the design of optimal finite word-length linear phase FIR digital filters. The algorithm proved to be quite practical for filters of length up to 60 or 80, depending on the word length. It is based on the branch-and-bound technique in which the Remez algorithm and an adaptive grid-growth technique are used. The algorithm also uses a new theoretical result on the lower bound of finite word-length cosine Chebyshev approximation error.