Design of variable fractional delay fir filters with CSD coefficients using genetic algorithm

This paper presents a new method for the design of variable fractional delay (VFD) FIR digital filters using Genetic Algorithm (GA). Each sub-filter of Farrow structure is designed individually with defined accuracy and bandwidth. A variable mutation probability is also employed, which improves the accuracy of the solution. Compared with existing methods, it reduces the computational complexity and enhances the design flexibility. Furthermore, a simple GA is used to compute the filter coefficients in canonic singed digit (CSD) representations. Since this algorithm selects the initial population in CSD representation and searches for the digits between [-1 1], most of the digits become zero. The filter multipliers can be built with simple shift registers and adders, resulting in a high-speed and low-power filter implementation. Multiplier-free structure typically shows 11.4% saving in power and 17.2% reduction in area consumption.