Interactive constrained min-max optimization of multi-rate digital filters over the canonical signed-digit coefficient space

Abstract

This paper presents an overview of the development of a graphical software environment called Papillon DSP OptiStation for the design and constrained min-max optimization of multi-rate FIR and IIR digital filters. The optimization engine is required to handle simultaneously multiple objective functions and multiple arbitrary equality and inequality constraints. Moreover, it is required to handle not only infinite-precision optimization, but also finite-precision optimization over the canonical signed-digit transfer function coefficient space. In addition, it is required to have guaranteed convergence, i.e. it is required to find a solution if one exists. The Papillon DSP OptiStation environment is based on a multi-threaded multi-process architecture facilitating co-operative online interaction between the user on the one hand and the optimizer on the other, through a Tcl/Tk graphical user interface. It is required that the user be able to change, interactively, any time-domain and/or frequency domain design specifications throughout the course of the optimization.