Digital-Filter Implementation

Abstract

Digital filters and signal-processing systems are implemented in two basic ways: software in general-purpose computers, and special purpose hardware. The software approach is relatively straightforward, given the computational structures derived in previous chapters. Questions such as the numbers of multiplications and adds, coefficient sensitivity, roundoff noise, in-place FFT computation, etc., must be addressed; but the hardware considerations in that case are restricted to how fast the computer must be, how much storage is required, whether to add a high-speed floatingpoint unit, and so forth. In the highest-speed applications or for production quantities, however, it is often necessary to design special-purpose hardware to satisfy the system specifications at an acceptable cost. The advent of custom VLSI (very-large-scale integration) has made this approach even more attractive and powerful, and has provided systems of ever-increasing complexity and speed. In this chapter, we described special-purpose hardware systems that take advantage of the highly parallel computational structure of these systems. This material assumes a basic knowledge of binary number systems and digital logic.