On practical aspects of optimal FSD filter design using extracted window method

Abstract

In this paper the practical aspects of the extracted window method applied to the nearly optimal fractional sample delay (FSD) filter design has been investigated. As the window method is one of the most numerically efficient digital filter design methods, this approach is well suited for variable fractional sample delay (VFSD) filter implementations which require frequent impulse response recalculation. On the other hand, the use of symmetric window extracted from the optimal filter (minimax or least squared error) solves the problem of window selection and results in nearly optimal high performance VFSD filters. However, the design procedure requires additional gain correction dependent on fractional delay. As the optimal gain correction factor computation should not be performed at runtime, low order polynomial approximation has been investigated in this paper. Additionally, performance loss resulting from limited word length in fixed-point arithmetic has also been discussed.