A sequential constrained least-squares method to minimax design of stable IIR digital filters

Much attention has been paid to minimax design of infinite impulse response digital filters during past decades. In this paper, a design procedure that combines a sequential constrained least-squares method and a Steiglitz-McBride scheme is presented. In this procedure, the minimax design problem is converted into a sequence of constrained least-squares problems, each of which is then resolved with a Steiglitz-McBride iterative scheme. What we are confronted with in each iteration of the design procedure is a circularly constrained quadratic program, the solution of which can be found by a recently developed efficient Goldfarb-Idnani based algorithm. To demonstrate the effectiveness of the presented design procedure, two filters with similar specifications as the Deczky's `benchmark' filter are designed, and comparisons with several existing methods are given.