• DocumentCode
    3497952
  • Title

    Evaluating alternative implementations for LDPC decoder check node function

  • Author

    Theocharides, T. ; Link, G. ; Swankoski, E. ; Vijaykrishnan, N. ; Irwin, M.J. ; Schmit, H.

  • Author_Institution
    Dept. of Comput. Sci. & Eng., Pennsylvania State Univ., University Park, PA, USA
  • fYear
    2004
  • fDate
    19-20 Feb. 2004
  • Firstpage
    77
  • Lastpage
    82
  • Abstract
    Low density parity checks (LDPC) are a method of error detection and correction that are able to achieve near Shannon-limit channel communication. LDPC decoders involve a series of computations between two units, the check node and the bit node. In this paper we propose the use of an approximation unit to perform the check node operation. Additionally, we propose a ROM based look-up table (LUT) as a function approximation technique, to be used with an LDPC decoder. The paper shows that a ROM based LUT achieves better performance than using a piecewise linear approximation method to approximate the LDPC computation function. Furthermore, this paper shows that the ROM LUT method can gradually take over as the selected function approximation technique for computationally intensive demanding VLSI designs as the technology shifts to the nanometer era.
  • Keywords
    VLSI; approximation theory; channel capacity; error correction codes; error detection codes; parity check codes; piecewise linear techniques; random-access storage; table lookup; LDPC decoder; ROM based look-up table; Shannon-limit channel communication; alternative implementations evaluation; approximation unit; bit node; check node function; check node operation; computation function; computationally intensive VLSI designs; error correction; error detection method; function approximation technique; low density parity checks; nanometer technology; piecewise linear approximation method; Bipartite graph; Function approximation; H infinity control; Iterative decoding; Parity check codes; Piecewise linear approximation; Read only memory; Sparse matrices; Table lookup; Very large scale integration;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    VLSI, 2004. Proceedings. IEEE Computer society Annual Symposium on
  • Print_ISBN
    0-7695-2097-9
  • Type

    conf

  • DOI
    10.1109/ISVLSI.2004.1339511
  • Filename
    1339511