INTRODUCTION
Although this book has hardly been reticent on the important role of digital signal processing (DSP) in modern radio systems, we have bypassed many of the details so far. In this chapter we circle back to the topic of digital implementation of some principal radio functions. There are essentially three broad categories of functions we must address: filtering, frequency conversion, and modulation.
Section 18.2 (“Single-Rate Filters”) provides a primer on digital implementation of lowpass filtering, concentrating on finite impulse response (FIR) filters. Such filters are used in a number of different ways. They are prevalent especially in direct conversion architecture and in architectures which employ direct conversion as the interface between complex baseband and a real-valued IF. Changes in sample rate are typically required as part of this operation, and are specifically addressed in Section 18.3 (“Multirate Filters”). Then Section 18.4 (“Quadrature Upconversion and Downconversion”) addresses techniques for combining filtering, rate changing, and frequency conversion in a computationally-efficient manner.
We are then in a position to tie up some loose ends from Chapter 6 (“Digital Modulation”): Specifically, how to go about implementing pulse shaping in transmitters, and symbol timing recovery in receivers. These topics are addressed in Section 18.5.
Finally, Section 18.6 addresses the hardware technologies available for DSP, including general-purpose computing, programmable logic devices, and various other possibilities. A major theme here is the need for parallelism and the resulting tradeoff between sample rate, reconfigurability, and size/power/cost.
SINGLE-RATE FILTERS
Digital filters can be classified according to two broad categories: Finite impulse response (FIR) and infinite impulse response (IIR). The distinction is exactly as implied by the nomenclature, but implementation considerations introduce some subtleties which are potentially confusing to the uninitiated.
To begin, let us assume we wish frequency response which is as close to “brick wall” performance as possible. Exactly this idea was considered in Section 6.4.2 (Figure 6.7 and associated text). We now re-introduce the concept as shown in Figure 18.1, with only a slight revision in that the lowpass bandwidth W is not necessarily related to any properties of the modulation within this bandwidth.