Quasi-cyclic low-density parity-check convolutional code

This paper proposes a novel quasi-cyclic low-density parity-check convolutional code, and a two-stage construction algorithm with modified progressive edge growth (PEG) method is provided. We propose both encoder and decoder implementation architecture for this code. The quasi-cyclic form provides the parallelism for encoder and decoder, which can increase the throughput and decrease the delay significantly. The proposed modified min-sum decoding algorithm can speed up the process of convergence and reduce the hardware complexity. We also designed a GPU based simulation platform to speed up about 200 times against CPU to verify the code performance. Simulation results show the proposed code can get 0.5~1dB coding gain and lower error floor compared with the LDPC codes in WiMAX standard with the same code length, while the decoder only has 20 iterators.