Using Matlab to analyse finite word length effects in infinite impulse response digital filters

Infinite impulse response (IIR) digital filters are particularly sensitive to problems associated with using a finite wordlength. The aspects most affected include the frequency and phase response, stability and signal to quantization noise ratio. This paper aims to present an overview of the design techniques that may be employed to analyse finite wordlength effects using Matlab. Of particular interest is the possible use of recursive filters for soft modem technologies. A Gaussian low pass filter that could be employed for gaussian minimum shift keying (GMSK) has been investigated in order to determine whether there is an advantage in using an IIR filter rather than a finite impulse response (FIR) filter. A tutorial style comparison of the finite wordlength analysis required for this particular example and that used for classical digital filters is also presented.