Multi-rate bandpass filter bank implemented in QRNS complex arithmetic using parallel multiple DSP chips or ASICs
Michael A. Soderstrand, Dong Chu, W. Chan, M. Lazkani, H.H. Loomis · 2002
In decoding spread-spectrum BPSK signals in the presence of narrow-band interference, a perfect reconstruction multi-rate bandpass filter bank can be used with a very simple adaptive algorithm to eliminate narrow-band interference broadband signals. Inherent in this structure is the need to do quadrature processing to eliminate images that occur due to the up-sampling required to obtain the narrow-band bandpass filters. We propose to use parallel processors (microprocessors, ASICs, or DSP chips) to implement complex FIR filters needed to eliminate the upsampled images. The processors implement a very simple complex FIR filter in one modulus of a generalized quadratic residue number system (GQRNS). By keeping the moduli small and the filter order low, it is possible to use standard binary within the processor to implement a GQRNS number system. This allows us to obtain the advantages of parallel processing and simplified complex number arithmetic inherent in GQRNS arithmetic while maintaining standard binary arithmetic in the processors.>