Two Design Techniques for Digital Phase Networks

Two computer-aided algorithms for the design of all-pass digital filters are presented. The first technique is based on a linear programming approach to solving the approximation problem posed by the minimax design of an all-pass digital filter. A new iterative algorithm with stability constraints is offered for direct form design. The second technique implements a gradient search for those quadratic factors of an all-pass transfer function that lead to a locally optimal approximation (in the least-squares sense) of a desired phase function. New initial guess procedures and the parameterization of linear-phase offset enhance the least-squares design procedure. Examples illustrating the result of both procedures are included.