On the performance of discrete-time hybrid filter banks with coefficient errors and ARMA stochastic process input

In the theory of filter banks, perfect reconstruction plays a key role. However, when the implementation is considered, the errors due to technological limitations severely limit the filter bank performance. This problem is of great concern, mainly in applications of high-speed, high-resolution A/D and D/A conversions, for which the hybrid filter banks are becoming attractive solutions. This paper presents a study about the performance, assessed by the signal-to-noise ratio (SNR), of A/D, D/A and purely analog filter banks subjected to coefficient errors, in the light of probability theory. Expressions for the SNR mean value and standard deviation are derived, valid for zero mean ARMA, MA, AR and white noise stochastic processes as inputs. Computer simulations support the results.