Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures

Q. Zhang, P.T. Wolkotte, Gerardus Johannes Maria Smit, Arnaud Rivaton, Jérôme Quevremont

    Research output: Contribution to conferencePaper

    12 Downloads (Pure)

    Abstract

    The DRM standard for digital radio broadcast in the AM band requires integrated devices for radio receivers at very low power. A System on Chip (SoC) call DiMITRI was developed based on a dual ARM9 RISC core architecture. Analyses showed that most computation power is used in the Coded Orthogonal Frequency Division Multiplexing (COFDM) demodulation to compute Fast Fourier Transforms (FFT) and inverse transforms (IFFT) on complex samples. These FFTs have to be computed on non power-of-two numbers of samples, which is very uncommon in the signal processing world. The results obtained with this chip, lead to the objective to decrease the power dissipated by the COFDM demodulation part using a coarse-grain reconfigurable structure as a coprocessor. This paper introduces two different coarse-grain architectures: PACT XPP technology and the Montium, developed by the University of Twente, and presents the implementation of a Fast Fourier Transform on 1920 complex samples. The implementation result on the Montium shows a saving of a factor 35 in terms of processing time, and 14 in terms of power consumption compared to the RISC implementation, and a smaller area. Then, as a conclusion, the paper presents the next steps of the development and some development issues.
    Original languageUndefined
    Pages465-469
    Publication statusPublished - 2005
    Event16th Workshop on Circuits, Systems and Signal Processing, ProRISC 2005 - Veldhoven, Netherlands
    Duration: 17 Nov 200518 Nov 2005
    Conference number: 16

    Workshop

    Workshop16th Workshop on Circuits, Systems and Signal Processing, ProRISC 2005
    CountryNetherlands
    CityVeldhoven
    Period17/11/0518/11/05

    Keywords

    • IR-59553

    Cite this

    Zhang, Q., Wolkotte, P. T., Smit, G. J. M., Rivaton, A., & Quevremont, J. (2005). Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures. 465-469. Paper presented at 16th Workshop on Circuits, Systems and Signal Processing, ProRISC 2005, Veldhoven, Netherlands.
    Zhang, Q. ; Wolkotte, P.T. ; Smit, Gerardus Johannes Maria ; Rivaton, Arnaud ; Quevremont, Jérôme. / Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures. Paper presented at 16th Workshop on Circuits, Systems and Signal Processing, ProRISC 2005, Veldhoven, Netherlands.
    @conference{f7005ceb0f544e758b5b9c0e1dee7133,
    title = "Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures",
    abstract = "The DRM standard for digital radio broadcast in the AM band requires integrated devices for radio receivers at very low power. A System on Chip (SoC) call DiMITRI was developed based on a dual ARM9 RISC core architecture. Analyses showed that most computation power is used in the Coded Orthogonal Frequency Division Multiplexing (COFDM) demodulation to compute Fast Fourier Transforms (FFT) and inverse transforms (IFFT) on complex samples. These FFTs have to be computed on non power-of-two numbers of samples, which is very uncommon in the signal processing world. The results obtained with this chip, lead to the objective to decrease the power dissipated by the COFDM demodulation part using a coarse-grain reconfigurable structure as a coprocessor. This paper introduces two different coarse-grain architectures: PACT XPP technology and the Montium, developed by the University of Twente, and presents the implementation of a Fast Fourier Transform on 1920 complex samples. The implementation result on the Montium shows a saving of a factor 35 in terms of processing time, and 14 in terms of power consumption compared to the RISC implementation, and a smaller area. Then, as a conclusion, the paper presents the next steps of the development and some development issues.",
    keywords = "IR-59553",
    author = "Q. Zhang and P.T. Wolkotte and Smit, {Gerardus Johannes Maria} and Arnaud Rivaton and J{\'e}r{\^o}me Quevremont",
    year = "2005",
    language = "Undefined",
    pages = "465--469",
    note = "null ; Conference date: 17-11-2005 Through 18-11-2005",

    }

    Zhang, Q, Wolkotte, PT, Smit, GJM, Rivaton, A & Quevremont, J 2005, 'Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures' Paper presented at 16th Workshop on Circuits, Systems and Signal Processing, ProRISC 2005, Veldhoven, Netherlands, 17/11/05 - 18/11/05, pp. 465-469.

    Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures. / Zhang, Q.; Wolkotte, P.T.; Smit, Gerardus Johannes Maria; Rivaton, Arnaud; Quevremont, Jérôme.

    2005. 465-469 Paper presented at 16th Workshop on Circuits, Systems and Signal Processing, ProRISC 2005, Veldhoven, Netherlands.

    Research output: Contribution to conferencePaper

    TY - CONF

    T1 - Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures

    AU - Zhang, Q.

    AU - Wolkotte, P.T.

    AU - Smit, Gerardus Johannes Maria

    AU - Rivaton, Arnaud

    AU - Quevremont, Jérôme

    PY - 2005

    Y1 - 2005

    N2 - The DRM standard for digital radio broadcast in the AM band requires integrated devices for radio receivers at very low power. A System on Chip (SoC) call DiMITRI was developed based on a dual ARM9 RISC core architecture. Analyses showed that most computation power is used in the Coded Orthogonal Frequency Division Multiplexing (COFDM) demodulation to compute Fast Fourier Transforms (FFT) and inverse transforms (IFFT) on complex samples. These FFTs have to be computed on non power-of-two numbers of samples, which is very uncommon in the signal processing world. The results obtained with this chip, lead to the objective to decrease the power dissipated by the COFDM demodulation part using a coarse-grain reconfigurable structure as a coprocessor. This paper introduces two different coarse-grain architectures: PACT XPP technology and the Montium, developed by the University of Twente, and presents the implementation of a Fast Fourier Transform on 1920 complex samples. The implementation result on the Montium shows a saving of a factor 35 in terms of processing time, and 14 in terms of power consumption compared to the RISC implementation, and a smaller area. Then, as a conclusion, the paper presents the next steps of the development and some development issues.

    AB - The DRM standard for digital radio broadcast in the AM band requires integrated devices for radio receivers at very low power. A System on Chip (SoC) call DiMITRI was developed based on a dual ARM9 RISC core architecture. Analyses showed that most computation power is used in the Coded Orthogonal Frequency Division Multiplexing (COFDM) demodulation to compute Fast Fourier Transforms (FFT) and inverse transforms (IFFT) on complex samples. These FFTs have to be computed on non power-of-two numbers of samples, which is very uncommon in the signal processing world. The results obtained with this chip, lead to the objective to decrease the power dissipated by the COFDM demodulation part using a coarse-grain reconfigurable structure as a coprocessor. This paper introduces two different coarse-grain architectures: PACT XPP technology and the Montium, developed by the University of Twente, and presents the implementation of a Fast Fourier Transform on 1920 complex samples. The implementation result on the Montium shows a saving of a factor 35 in terms of processing time, and 14 in terms of power consumption compared to the RISC implementation, and a smaller area. Then, as a conclusion, the paper presents the next steps of the development and some development issues.

    KW - IR-59553

    M3 - Paper

    SP - 465

    EP - 469

    ER -

    Zhang Q, Wolkotte PT, Smit GJM, Rivaton A, Quevremont J. Low Power Implementation of Non Power-of-Two FFTs on Coarse-Grain Reconfigurable Architectures. 2005. Paper presented at 16th Workshop on Circuits, Systems and Signal Processing, ProRISC 2005, Veldhoven, Netherlands.