Tamer Fares wrote:
what's meant by the stability problem if they are quantized excessively ?
Quantisation errors can mean that you can end up with a set of filter coefficients which correspond to having some poles outside the unit circle, and therefore a filter which will be unstable (e.g., when excited by an impulse, you could get a response with ever-increasing amplitude).
To avoid this, for applications which involve manipulating the filter coefficients (e.g., quantising them for speech coding, transforming them for speaker modification, synthesising them from a model, or even just storing them with finite precision) it is better to work in other representations in which it is simpler to guarantee a stable filter (e.g., LSFs are one such common representation that is used widely in speech coding and synthesis).
Simon