摘要
This paper investigates analysis and design of Low-Density Parity-Check （LDPC） coded Bit Interleaved Coded Modulation （BICM） over Additive White Gaussian Noise （AWGN） channel. It focuses on Gray-labeled 8-ary Phase-Shift-Keying （8PSK） modulation and employs a Maximum A Posteriori （MAP） symbol-to-bit metric calculator at the receiver. An equivalent model of a BICM communication channel with ideal interleaving is presented. The probability distribution function of log-likelihood ratio messages from the MAP receiver can be approximated by a mixture of symmetric Gaussian densities. As a result semi-Gaussian approximation can be used to analyze the decoder. Extrinsic information transfer charts are employed to describe the convergence behavior of LDPC decoder. The design of irregular LDPC codes reduces to a linear programming problem on two-dimensional variable edge-degree distribution. This method allows irregular code design in a wider range of rates without any limit on the maximum node degree and can be used to design irregular codes having rates varying from 0.5275 to 0.9099. The designed convergence thresholds are only a few tenths, even a few hundredths of a decibel from the capacity limits. It is shown by Monte Carlo simulations that, when the block length is 30,000, these codes operate about 0.62-0.75 dB from the capacity limit at a bit error rate of 10s.
This paper investigates analysis and design of Low-Density Parity-Check (LDPC) coded Bit Interleaved Coded Modulation (BICM) over Additive White Gaussian Noise (AWGN) channel. It focuses on Gray-labeled 8-ary Phase-Shift-Keying (8PSK) modulation and employs a Maximum A Posteriori (MAP) symbol-to-bit metric calculator at the receiver. An equivalent model of a BICM communication channel with ideal interleaving is presented. The probability distribution function of log-likelihood ratio messages from the MAP receiver can be approximated by a mixture of symmetric Gaussian densities. As a result semi-Gaussian approximation can be used to analyze the decoder. Extrinsic information transfer charts are employed to describe the convergence behavior of LDPC decoder. The design of irregular LDPC codes reduces to a linear programming problem on two-dimensional variable edge-degree distribution. This method allows irregular code design in a wider range of rates without any limit on the maximum node degree and can be used to design irregular codes having rates varying from 0.5275 to 0.9099. The designed convergence thresholds are only a few tenths, even a few hundredths of a decibel from the capacity limits. It is shown by Monte Carlo simulations that, when the block length is 30,000, these codes operate about 0.62–0.75 dB from the capacity limit at a bit error rate of 10–8.
