An Approach to Low-Power, High-Performance,
Fast Fourier Transform Processor Design


Ph.D. Dissertation
Bevan M. Baas
February, 1999

Abstract

The Fast Fourier Transform (FFT) is one of the most widely used digital signal processing algorithms. While advances in semiconductor processing technology have enabled the performance and integration of FFT processors to increase steadily, these advances have also caused the power consumed by processors to increase as well. This power increase has resulted in a situation where the number of potential FFT applications limited by maximum power budgets--not performance--is significant and growing.

We present the cached-FFT algorithm which explicitly caches data from main memory using a much smaller and faster memory. This approach facilitates increased performance and, by reducing communication energy, increased energy-efficiency.

Spiffee is a 1024-point, single-chip, 460,000-transistor, 40-bit complex FFT processor designed to operate at very low supply voltages. It employs the cached-FFT algorithm which enables the design of a well-balanced, nine-stage pipeline. The processor calculates a complex radix-2 butterfly every cycle and contains unique hierarchical-bitline SRAM and ROM memories which operate well in both standard and low supply voltage, low threshold-voltage environments. The processor's substrate and well nodes are connected to chip pads, accessible for biasing to adjust transistor thresholds.

Spiffee has been fabricated in a standard 0.7µm (Lpoly= 0.6µm) CMOS process and is fully functional on its first fabrication. At a supply voltage of 1.1V, Spiffee calculates a 1024-point complex FFT in 330µsec, while dissipating 9.5mW--resulting in an adjusted energy-efficiency more than 16 times greater than that of the previously most efficient FFT processor. At a supply voltage of 3.3V, it operates at 173MHz--a clock rate 2.6 times faster than the previously fastest.


Downloading a PDF copy

The content of both versions is identical--only front matter pages vary.
Please note: Material taken from this dissertation must be referenced appropriately.

Official one-sided version

Two-sided version

Identical to the official version except for even/odd margins, page breaks, and removed signature page.


B. Baas | Starlab | EE | Stanford
Last update: October 6, 2004