Performance of LDPC codes with various decoding algorithms and schedules

This paper presents a systematic analysis and comparison of LDPC decoding performance for combinations of representative check node update algorithms, including Sum-Product (SP), Normalized Min-Sum (NMS), and Offset Min-Sum (OMS), and scheduling strategies such as Flooding, Layered Belief Propagation (LBP), and Residual Belief Propagation (RBP). We use the GPS L1C irregular ( 1200 , 600 ) LDPC code as well as the irregular ( 1128 , 528 ) LDPC code derived from the cropped 5G NR BG1 for the evaluation. For NMS and OMS, we identify the scaling factors that yield the best performance under each scheduling strategy. Using these scaling factors, we evaluate the FER and the average number of decoding iterations for the above combinations and compare the required E b / N 0 to achieve an FER of 1 0 − 4 . Experimental results show that, at a fixed E b / N 0 , LBP-based combinations require about 40% fewer iterations than Flooding-based combinations, while RBP-based combinations require about 30% fewer iterations than LBP-based ones but incur higher overall computational complexity. From these comparisons, we conclude that OMS with LBP provides the best trade-off between FER performance and computational complexity.