• DocumentCode
    393928
  • Title

    A low-power architecture for maximum a posteriori decoding

  • Author

    López-Vallejo, Marisa ; Mujtaba, Syed Aon ; Lee, Inkyu

  • Author_Institution
    ETSI Telecommunicacion, Univ. Politecnica de Madrid, Spain
  • Volume
    1
  • fYear
    2002
  • fDate
    3-6 Nov. 2002
  • Firstpage
    47
  • Abstract
    In this paper, we present a novel architecture for soft-input soft-output (SISO) maximum a posteriori (MAP) decoding. The architecture leverages an ASIP (application specific instruction-set processor) structure, where the datapath has been designed to achieve high-speed performance and low power dissipation. Salient features of this architecture include: (a) delayed renormalization of the /spl alpha/ metrics with register bypassing to reduce latency, (b) use of register files to minimize power dissipation, and (c) microprogrammed control to achieve flexibility. The resulting architecture for the SISO-MAP decoder achieves a maximum throughput of 10.9 Msymbols/second, operating at 142 MHz and dissipating 21 mW in the datapath.
  • Keywords
    application specific integrated circuits; codecs; computer architecture; digital signal processing chips; instruction sets; low-power electronics; maximum likelihood decoding; turbo codes; /spl alpha/ metrics; 142 MHz; 21 mW; ASIP structure; SISO-MAP decoding; application specific instruction-set processor; delayed renormalization; high-speed performance; low-power architecture; maximum a posteriori decoding; microprogrammed control; power dissipation; soft-input soft-output; turbo-decoding; Application specific processors; Bit error rate; Delay; Iterative decoding; Jacobian matrices; Power dissipation; Registers; Telecommunication standards; Throughput; Turbo codes;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Signals, Systems and Computers, 2002. Conference Record of the Thirty-Sixth Asilomar Conference on
  • Conference_Location
    Pacific Grove, CA, USA
  • ISSN
    1058-6393
  • Print_ISBN
    0-7803-7576-9
  • Type

    conf

  • DOI
    10.1109/ACSSC.2002.1197147
  • Filename
    1197147