On the optimal number of finite-duration impluse-response (FIR) filter coefficients under a memory size constraint

A method is given for efficiently assessing the performance of a finite-duration impulse-response (FIR) digital filter under the constraint of fixed (number of coefficients) × (word length) product. The tradeoff of impulse-response truncation error against coefficient quantization error is readily evaluated. The method yields a quantitative estimate of the optimal solution and of the sensitivity of this solution without the need for extensive filter simulations.