• DocumentCode
    3080765
  • Title

    Coset-based quasi-cyclic LDPC codes for optimal overlapped message passing decoding

  • Author

    Dai, Yongmei ; Yan, Zhiyuan

  • Author_Institution
    Dept. of Electr. & Comput. Eng., Lehigh Univ., Bethlehem, PA, USA
  • fYear
    2005
  • fDate
    2-4 Nov. 2005
  • Firstpage
    762
  • Lastpage
    767
  • Abstract
    Due to the regular structures of quasi-cyclic (QC) low-density parity-check (LDPC) codes, the message passing decoding of these codes not only has efficient hardware implementation, but also can be overlapped to improve throughput and hardware utilization efficiency (HUE). In previous work, we proposed a scheduling scheme for the overlapped message passing (OMP) decoding of the SFT codes. In this paper, we first show that our scheduling scheme produces the minimum intra-iteration waiting time for QC LDPC codes. Then we propose a coset-based construction of QC LDPC codes, and study the girth and minimum distance properties of our new codes. Given the same parameters, the ensemble of the SFT codes is a subset of the ensemble of our coset-based QC LDPC codes. Our coset-based QC LDPC codes allow the OMP decoding to achieve greater throughput gain and higher HUE while maintaining the same theoretical performances as the SFT codes.
  • Keywords
    cyclic codes; decoding; matrix algebra; message passing; parity check codes; scheduling; coset-based quasicyclic LDPC codes; hardware utilization efficiency; low-density parity-check; minimum distance properties; minimum intraiteration waiting time; optimal overlapped message passing decoding; overlapped message passing; scheduling scheme; throughput improvement; Design methodology; Electronic mail; Galois fields; Hardware; Iterative algorithms; Iterative decoding; Message passing; Parity check codes; Performance gain; Throughput;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Signal Processing Systems Design and Implementation, 2005. IEEE Workshop on
  • ISSN
    1520-6130
  • Print_ISBN
    0-7803-9333-3
  • Type

    conf

  • DOI
    10.1109/SIPS.2005.1579967
  • Filename
    1579967