Analysis of errors in residue number system (RNS) based IIR digital filters

The problem of analyzing errors in Residue Number System (RNS) based IIR Digital filters is considered in this paper. There are basically three types of errors in RNS based digital filters, They are coefficient quantization errors(1), scaling errors and data errors. In particular the problem of roundoff errors which plagues fixed point digital filters is absent in RNS digital filters because of the exact nature of its arithmetic. A model to predict the scalino errors has been developed and exoerimental results which verify the model's validity are presented. Using this model, an aporoach to eliminate scaling errors is described. Finally, a technique by which random data errors may be detected and corrected is presented. This method is both hardware and temporally efficient.