• DocumentCode
    2837259
  • Title

    A parallel architecture for FNT and modified Winograd algorithms to compute DFT

  • Author

    Rao, Tellakula Sreenivasa ; Gupta, Sumana

  • Author_Institution
    Wipro Syst. Ltd., Bangalore, India
  • fYear
    1989
  • fDate
    22-24 Nov 1989
  • Firstpage
    569
  • Lastpage
    572
  • Abstract
    Fermat number transforms (FNT) are a class of transforms which operate in the integer domain. This means there will be no complex number calculations. If the base of FNT is a power of 2, then the computation of the transform does not involve any multiplications at all. A discussion is also presented of computing a discrete Fourier transform (DFT) through Winograd algorithms, wherein the cyclic convolutions are computed by using the FNT. Three cases of the Winograd algorithm are considered
  • Keywords
    Fourier transforms; digital arithmetic; parallel architectures; pipeline processing; systolic arrays; transforms; 1D processor array; Fermat number transforms; cyclic convolutions; discrete Fourier transform; integer domain; modified Winograd algorithms; one dimensional processor array; parallel architecture; Arithmetic; Concurrent computing; Convolution; Discrete Fourier transforms; Discrete transforms; Frequency domain analysis; Parallel architectures; Roundoff errors; Sampling methods; Upper bound;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    TENCON '89. Fourth IEEE Region 10 International Conference
  • Conference_Location
    Bombay
  • Type

    conf

  • DOI
    10.1109/TENCON.1989.177004
  • Filename
    177004