Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Thesis
Universal FFT core generator
Master of Science (M.S.), Drexel University
Dec 2007
DOI:
https://doi.org/10.17918/etd-2535
Abstract
This thesis presents a special purpose processor for multi-dimensional discrete Fourier transform (DFT) computation. The processor, called a Universal fast Fourier transform (FFT) processor, is capable of computing an arbitrary multi-dimensional DFT with a fixed number of input points. The processor can be configured to compute different dimensional DFTs simply by rearranging the input data and using a different set of constants, called generalized twiddle factors, throughout the computation. The basic computational process remains the same independent of dimension. The processor does not rely on one-dimensional FFT computation as does the standard approach using the row-column algorithm, but instead utilizes an algorithm called the Dimensionless FFT, whose computation data flow is the same as the FFT algorithm due to Pease. Since the computation performed by the Universal FFT processor is the same as the Pease FFT, an implementation can be obtained by modifying a one-dimensional FFT processor based on the Pease algorithm. The implementation presented in this thesis is obtained in this way from the core generator by Nordin, Milder, Hoe, and Püschel. The core generator from Nordin, Milder, Hoe, and Püschel generates a variety of cores, with various space and time tradeoffs for implementation on field programmable gate arrays (FPGA). The performance obtained is comparable to vendor supplied cores but offers much greater flexibility of design choices with different area and throughput tradeoffs. The resulting Universal FFT core generator can be used to instantiate multi-dimensional FPGA cores with approximately the same performance and area obtained in the one-dimensional case by Nordin, Milder, Hoe, and Püschel.
Metrics
38 File views/ downloads
32 Record Views
Details
- Title
- Universal FFT core generator
- Creators
- Pranab Shenoy - DU
- Contributors
- Jeremy Russell Johnson (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Master of Science (M.S.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Resource Type
- Thesis
- Language
- English
- Academic Unit
- College of Arts and Sciences; Drexel University; Mathematics
- Other Identifier
- 2535; 991014631958904721