Low Complexity Layered Decoding for LDPC Codes

This paper presents methods to reduce the complexity of the layered decoding for LDPC codes with negligibly impacting the error-correcting performance. We accomplish this by deactivating variable nodes. The LLR updates associated with the deactivated variable nodes are skipped. To determine which variable nodes should be deactivated, the most reliable bit-level channels, derived from the M-ary input channel, are identified. Then, from the variable nodes assigned to the most reliable bit-level channels, we select those with initial LLRs above a threshold as the deactivated variable nodes. We establish an analytical relationship between the threshold and a desired error probability. The threshold can be adjusted accordingly by specifying the desired error probability. The standard 5G LDPC codes are simulated. By setting the desired error probability to 10^-3, more than 20% of the variable nodes are deactivated while the error-correcting performance is preserved.