No abstract

The paper presents a family of architectures for FFT implementation based on the decomposition of the perfect shuffle permutation, which can be designed with variable number of processing elements. This provides designers with a trade-off choice of speed vs. complexity (cost and area.). A detailed case study is provided on the implementation of 1024-point FFT with 2 processing elements using 45nm process technology, including area, timing, power and place-and-route results.

No abstract

Scalable architectures were proposed for Discrete Cosine Transform (DCT). Number of processing elements (PE) can be reduced significantly using partial column structure for computing the DCT transform. This feature is very desirable for multimedia applications usage in handheld devices. As per transform computation, data reordering is required between stages (columns) where intermediate computed values are saved in memory-like temporary locations called FIFO's. A scalable interconnect network for both global and local data reordering and its implementation is presented in this paper. Scalability is based on transform size and desired number of processing elements (PE). The structure gives choice flexibility of throughput vs. complexity (cost and area.) of the overall system.

No abstract