The Impact of Finite Wordlength Implementation - Digital Signal Processing: An Introduction with MATLAB and Applications - page 193

Note that if the filter does not satisfy the required specifications, then the order
should be increased.
Next,

convert the filter coefficient to fixed-point representation with the

required fixed-point parameters:

Display the fixed-point coefficients:

4.4 Coefficient Quantization Error in Digital Filters

When one seeks to implement a desired digital filter transfer function H(z) (FIR or
IIR) one generally implements a modified transfer function H ð z Þ , due to the effects
of quantization of the filter parameters. For this reason it is important to try and
design digital filter structures that exhibit minimal sensitivity to coefficient
quantization errors. The following subsections discuss this issue in more detail.

4.4.1 Coefficient Quantization Error in IIR Filters

Quantization of the coefficients of a rational function results in the movement of
the poles and zeros of that transfer function from their original, infinite-precision
locations. Consequently, the actual system frequency response will be different
from the desired one. An extreme scenario is that some of the poles overshoot the
unit circle boundary in the z-transform, resulting in an unstable IIR structure.
A very important question to answer is: How can the harmful effects due to
quantization of IIR filter coefficients be minimized?