Multiplier-less floating point 1D DCT implementation

Abstract

Multimedia applications are becoming even more demanding. Hence, the next generation codecs invariably should be floating point compliant. With the field programmable gate arrays (FPGAs) technology getting mature, more and more signal processing applications are finding their niche in FPGAs. Current generation FPGAs have got hardware multipliers. However, these are general purpose multipliers and cannot be used for specific purposes. The present paper presents a novel FPGA implementation of one dimensional (8 times 1) point, multiplier less, floating point Discrete Cosine Transform. Distributed Arithmetic, parallelism and pipelining are exploited to produce a DCT implementation on a single FPGA. Two implementations are presented, one using single LUT and second using 2 parallel LUTs utilizing 68% and 89% area respectively with a maximum clock frequency of around 50 MHz.