Coefficient sensitivity of polynomial-predictive FIR differentiators: design for short word length
Jarno M. A. Tanskanen, S.J. Ovaska · 2003
In this paper, a genetic algorithm for designing polynomial-predictive FIR differentiators (PPFD) for short word length fixed-point realizations is presented. Critical design aspects are pointed out, and a fitness function for genetic algorithms is designed to strive for desired prediction and differentiation properties with filter coefficients quantized to eight bits. The genetic algorithm is in many cases found to produce better quantized coefficient filters than the original quantized coefficient PPFDs.