A New Early Termination Strategy for QC-LDPC Codes Based on the Layered Message Passing Decoding
Zhejun Zhang,Guanghui He,Jun Ma
DOI: https://doi.org/10.2991/wiet-13.2013.39
2013-01-01
Abstract:Low-density parity-check (LDPC) codes are one of the most popular linear block codes have ever been used in the communication system. The standard two-phase message passing decoding (TPMPD) algorithm and the layered message-passing decoding (LMPD) algorithm are widely used in LDPC decoder. However, the convergence speed of LMPD algorithm is much faster than that of TPMPD algorithm, particularly in the condition of high signal noise ratio (SNR). In this paper, we propose an early termination criterion for quasi-cyclic low-density parity-check (QC-LDPC) codes based on LMPD algorithm with fast convergence speed. The strategy is to dynamically reprocesses the uncorrected related layers while skipping the corrected ones. Simulations show that 3% reduction in terms of iteration numbers can be achieved with no performance loss compared to the traditional termination strategies, such as hard decision aided (HDA) strategy and parity check equations (PCE) strategy. Furthermore, the optimized early termination strategy achieves low complexity in hardware implementation. nodes and updates all the a posteriori probability (APP) values for the CNP in next sub-iteration. Compared with TPMPD algorithm, LMPD algorithm achieves better decoding performance and faster convergence speed, especially in the high Signal Noise Ratio (SNR) condition. The early termination strategy for the LMPD algorithm is an important technique since it can reduce the hardware complexity, the number of sub-iterations, and the power consumption significantly in hardware implementation. There are various algorithms such as hard decision aided (HDA) (11) strategy, parity check equations (PCE) strategy and partial parity check equations (PPCE) strategy proposed for LDPC termination strategies. However the HDA strategy consumes large memory while the PPCE and PCE strategies have high computation complexity. In this paper, an optimized early termination criterion is proposed with lower hardware complexity, faster convergence speed, and almost the same decoding performance compared with the traditional strategies. The rest of this paper is organized as follows. Section II describes LDPC decoding algorithms and the traditional termination strategies. Section III proposes a new early termination strategy for QC-LDPC codes. Section IV optimizes the early termination strategy for hardware efficient implementation. Section V concludes the paper.