A unified approach to the selection of structures for digital filters

Any digital transfer function can be realized in a multitude of configurations (structures). Several procedures for analyzing the quantization characteristics have appeared in literature. This paper unifies these analysis procedures to select an optimum structure for a given transfer function. The procedure scales the filter modules to prevent overflow, selects the coefficient word lengths by sensitivity considerations, selects the quantizer step lengths to balance the quantization errors between the quantizers and uses the quantization error at the output as the quality factor for the structure. Simulation results are provided to prove the validity of the procedure.