Dadic M., Sruk V., Somek B.

Linear phase response and flat magnitude response are highly desirable characteristics of any filter in audio signal processing. This paper proposes a simple and efficient novel procedure for the design of finite impulse response (FIR) maximally flat filters with linear phase. The procedure is based on Butterworth polynomials, bilinear transformation and adaptive inverse modelling. Instead of using prescribed magnitude response through a set of discrete values, the designed filter has magnitude response equal to the squared magnitude response of a Butterworth filter. Instead of multisine excitation, random noise is used as the excitation signal for the adaptive process The delayed noncausal impulse response component is approximated using the causal impulse response of a finite impulse response filter. Possible applications of newly proposed design method are in restoration of old analog audio recordings, acoustical measurements or audio engineering.