FPGA-based Implementation of Signal Processing Systems, 2nd Edition

14 Conclusions

14.1 Introduction

The aim of this book has been to cover many of the techniques needed to create FPGA solutions for signal and data processing systems. Interest in FPGAs for such systems has grown since the first edition as vendors have targeted this market and introduced innovations into their technology offerings to make them more attractive. This has included dedicated DSP blocks typically comprising 18-bit MAC blocks and increased memory units but also increased parallelism in terms of resources and the ability to employ pipelining to improve speed of operation. To support this activity, FPGA companies have created a range of IP cores (for customers to use in their designs) and design tools to create such implementations.

The main attraction of using FPGAs is that the resulting designs provide a very high quality of performance, particularly if MSPS/mm²/W is considered. This is possible as the designer is able to create an architecture which is a good match to the algorithmic needs, rather than striving to map the requirements onto a fixed architecture as is the case in microcontrollers, DSP microprocessors or GPUs, even though vendors increasingly offer multicore platforms. The main design challenge is to create a suitable architecture for the algorithmic requirements.

The creation of this suitable FPGA architecture comes from employing the right level of parallelism and pipelining to match initially the throughput rate and then area and power consumption ...