A modular pipelined implementation of large fast Fourier transforms

This paper presents a modular pipeline architecture for computing long discrete Fourier transforms (DFT). For an N point DFT, two conventional pipeline /spl radic/N point fast Fourier transform (FFT) modules are joined by a specialized center element. The center element contains memories, multipliers and control logic. Compared with a standard N point pipeline FFT the modular FFT reduces the number of delay elements required. Further, the coefficient storage is concentrated within the center element, reducing the storage requirement in the pipeline FFT modules. The centralized memory and address generator provide the storage and data reordering. The throughput of a standard radix-2 pipeline FFT is maintained with slightly higher end-to-end latency. A simulator has been built to analyze the proposed architecture. Results for DFTs of lengths up to 4M points are presented and compared with alternate algorithms.