Design of a family of stable causal perfect-reconstruction IIR uniform DFT filter banks with good numerical properties

A family of stable causal perfect reconstruction (PR) infinite-impulse response (IIR) discrete Fourier transform filter banks is presented, where a tradeoff between frequency selectivity and numerical properties is made. Some multipliers in both the analysis and synthesis filters could be made powers or sums of powers of two, in particular for feedback loops. Moreover, a simple scaling procedure that results in good sensitivity behavior is described. For these reasons, and also for the use of low-order IIR filters, the overall digital filter-bank structure is efficient for high data-rate applications. A design example, including quantization effects, is provided.