Constructing short-length irregular LDPC codes with low error floor
IEEE Transactions on Communications
| Hi-index | 0.00 |
In this paper, one improved frame error rate (FER) estimation approach in high Signal Noise Ratio (SNR) region for low-density parity-check (LDPC) codes is reported. Firstly dominant trapping sets (TSs), accounting for most of FER in high SNR region, are picked out as decoding result, with input of decoder consisting of a group of biased bits. And how those biased bits are chosen is reduced to finding solution to an binary integer program. Then decoder termination conditions are slightly modified so that history of the tentative decoding could facilitate action of searching for dominant TSs effectively. Without no prior knowledge of shape of error regions for specific LDPC code, we implement weighted mean translation (MT) of importance sampling (IS) method to compute FER in high SNR region. Simulation results demonstrate estimated FER curve conforms well with that of standard monte carlo (MC) simulation in overlapped SNR region.