Quantization noise analysis for fixed-point digital filters using magnitude truncation for quantization

Abstract

Two methods are described for the determination of the quantization errors in digital filters using magnitude truncation quantization. The first method is very accurate but rather cumbersome for numerical analysis. The second method overcomes this disadvantage and uses a quasi-linearization technique. Results are given that indicate the very good agreement between computations and results of simulations, both for a second-order and for a tenth-order recursive digital filter.