JultikaUniversity of Oulu repository

Abstract

Digital filters are an essential component of Digital Signal Processing (DSP) applications and play a crucial role in removing unwanted signal components from a desired signal. However, digital filters are known to be resource-intensive and consume a large amount of power, making it important to optimize their design in order to minimize hardware requirements such as multipliers, adders, and registers. This trade-off between filter performance and hardware consumption can be influenced by the quantization of filter coefficients. Therefore, this thesis investigates the quantization of Finite Impulse Response (FIR) filter coefficients and analyzes its impact on filter performance and hardware resource consumption. A method called dynamic quantization is introduced and an algorithm for step-by-step dynamic quantization is provided to improve upon the results obtained with the classical fixed point quantization method. To demonstrate the effectiveness of this approach, the dynamic quantization of filter coefficients for a Low-pass Equiripple FIR filter is examined and a comparative study of the magnitude response and hardware consumption of the generated filter using both the classical and dynamic quantization methods is presented. By understanding the trade-offs and benefits of each quantization method, engineers can make informed decisions about the most appropriate approach for their specific application.