Error Analysis of Digital Filters Using Fixed Point Arithmetic

Authors

Abstract

Based on an electrocardiogram filter, measurement methods of magnitude and phase responses, quantization and overflow errors, as well as limit circles in digital filters for fixed number representation are presented. A computer library for SCILAB has been created to simplify simulations. Direct form II, cascade, and rotation structures performance has been compared. It has been shown that there is no the best structure but the rotation one is superior to classical structures except for quantization errors. However, due to its low overflow errors, quantization noise can be further minimised by relocation of integer bits to fractional part of fixed point number representation.

Additional Files

Published

2025-07-09

Issue

Section

Digital Signal Processing