Area word-length trade off in DSP algorithm implementation and optimization

In this paper we propose a platform for high level synthesis of DSP algorithms while emphasising the differences between DSP systems and other digital systems. Accordingly, we allow variable word lengths within the system in order to optimize the system digital noise versus area. Using particular target architecture, a suitable cost function, together with a synthesiser and optimizer and intermediate data bases have been implemented. Optimization is based on a genetic algorithm.