An optimal filter design for variable sampling rates

Digital filters in future speech communications terminals should be alterable to allow for various services, bandwidths, customer loads, and digital line rates. A filter structure which maintains a constant frequency response as the sampling rate is varied is considered for such an application. The filter structure requires several times the number of design parameters than in the cascade canonic form of the digital filter in order to maintain the constant frequency response. The filter design problem is formulated as a minimax optimization problem and solved using direct search techniques. An example is then solved using a modified version of Rosenbrocks Rotating Coordinate System Method. The example illustrates the considerable compromises needed over standard computer aided design techniques when solving this class of problems.