Design of Mutual-Information-Maximizing Quantized Shuffled Min-Sum Decoder for Rate-Compatible Quasi-Cyclic LDPC Codes

In this paper, we propose a finite alphabet iterative decoder (FAID) named rate-compatible mutual-information-maximizing quantized shuffled min-sum (RC-MIM-QSMS) decoder, for decoding quasi-cyclic low-density parity-check (QC-LDPC) codes with various code rates. Our proposed decoder exchanges the coarsely quantized messages represented by symbols from finite alphabets and adopts single-input lookup tables (LUTs) to implement the node updates. To construct the LUTs used for decoding, we first propose a modified density evolution by considering the shuffled schedule to generate the LUTs which vary with different layers and iterations. Furthermore, to reduce the memory requirement for storing the LUTs, we optimize the constructed LUTs into a unique set of LUTs that only change with different decoding iterations. To the best of our knowledge, the RC-MIM-QSMS decoder is the first one to integrate the rate compatibility of LDPC codes with the shuffled decoding schedule. Simulation results show that the proposed RC-MIM-QSMS decoder outperforms the floating-point shuffled belief propagation decoder in the high signal-to-noise region and achieves comparable convergence speed to other state-of-the-art FAIDs. Moreover, the RC-MIM-QSMS decoder is able to save up to 93.22% memory requirement compared to the benchmark MIM-FAIDs