Analysis and Design of Efficient and Flexible Fast-Convolution Based Multirate Filter Banks

Abstract

In this paper, we investigate fast-convolution based highly tunable multirate filter bank (FB) configurations. Based on FFT-IFFT pair with overlapped processing, they offer a way to tune the filters' frequency-domain characteristics in a straightforward way. Different subbands are easily configurable for different bandwidths, center frequencies, and output sampling rates, including also partial or full-band nearly perfect-reconstruction systems. Such FBs find various applications, for example, as flexible multichannel channelization filters for software defined radios. They have also the capability to implement simultaneous waveform processing for multiple single-carrier and/or multicarrier transmission channels with nonuniform bandwidths and subchannel spacings. This paper develops an effective model for frequency response analysis and optimization for such filter banks, explores various practical issues of fast-convolution processing and reports the optimized frequency response characteristics versus direct raised cosine based designs. The fast-convolution approach is shown to be a competitive basis for filter bank based multicarrier waveform processing in terms of spectral containment and complexity.