Performance Analysis of Finite-Length LDPC Codes Over Asymmetric Memoryless Channels

In this paper, we analyze the waterfall performance of finite-length low-density parity-check (LDPC) codes for asymmetric memoryless channels by considering variations of channel parameters: the observed bit error rate P
_obs
and mutual information I
_obs
during the real-time transmission. Moreover, we propose to use the variation of a new channel parameter: the maximal achievable channel coding rate Robs at finite blocklengths to estimate the waterfall performance. Simulation results demonstrate that our estimations can well predict the waterfall performance of LDPC codes. In particular, the estimation based on R
_obs
can fit the waterfall performance for various code lengths. For estimations based on P
_obs
and I
_obs
, their accuracy increases as the codeword length increases and no curve fitting is required.