A network-efficient nonbinary QC-LDPC decoder architecture

Chuan Zhang, Keshab K. Parhi

Research output: Contribution to journalArticlepeer-review

10 Scopus citations

Abstract

Nonbinary low-density parity-check (LDPC) codes are of great interest due to their better performance cover binary ones when the code length is moderate. However, the cost of decoder implementation for these LDPC codes is still very high. In this paper, a low-complexity VLSI architecture for nonbinary LDPC decoders is presented. By exploiting the intrinsic shifting and symmetry properties of nonbinary quasi-cyclic LDPC (QC-LDPC) codes, significant reduction of memory size and routing complexity can be achieved. These unique features lead to two network-efficient decoder architectures for Class-I and Class-II nonbinary QC-LDPC codes, respectively. Comparison results with the state-of-the-art designs show that for the code example of the 64-ary (1260, 630) rate-0.5 Class-I code, the proposed scheme can save up to 70.6% hardware required by switch network, which demonstrates the efficiency of the proposed technique. The proposed design for the 32-ary (992, 496) rate-0.5 Class-II code can achieve a 93.8% switch network complexity reduction compared with conventional approaches. Furthermore, with the help of a generator for possible solution sequences, both forward and backward steps can be eliminated to offer processing convenience of check node unit (CNU) blocks. Results show the proposed 32-ary (992, 496) rate-0.5 Class-II decoder can achieve 4.47 Mb/s decoding throughput at a clock speed of 150 MHz.

Original languageEnglish (US)
Article number6132390
Pages (from-to)1359-1371
Number of pages13
JournalIEEE Transactions on Circuits and Systems I: Regular Papers
Volume59
Issue number6
DOIs
StatePublished - Jan 23 2012

Keywords

  • Decoder architecture
  • VLSI
  • geometry properties
  • low-density parity-check (LDPC) codes
  • nonbinary

Fingerprint Dive into the research topics of 'A network-efficient nonbinary QC-LDPC decoder architecture'. Together they form a unique fingerprint.

Cite this