During the realization of digital FIR and IIR filters the finite word length in computers affects the accuracy of filter coefficients. This leads to change in actual desired poles and zeros of the system function. In this paper firstly the analog filter is converted to digital domain using bilinear transformation and then the filter coefficients are quantized to analyze the perturbation error and the finite word length effect.This paper analyzes the sensitivity to quantization of filter coefficients for 8-bit, 16-bit and 32-bit precisions up to order 7 for digital low pass Butterworth, Chebyshev (type I and type II) and Elliptic filters for direct form realization. In the end result shows that the Chebyshev type II filter yields the least perturbation error and Elliptic filter yields the maximum perturbation error for higher orders.
Read full abstract