Forward error correction (FEC) is a key element of modern optical transmission systems. This technique introduces a small overhead to the information (typically, 7% to 13% for high bit-rate implementation) that is used to correct the transmission errors and thus makes error-free high bit-rate long distance transmissions possible. To cope with the constant bit-rate increases and the resulting signal-to-noise degradation caused by numerous propagation impairments (polarization mode dispersion, cross phase modulation, etc.), more powerful FECs are required.
LDPC codes, invented by Gallager in 1962, rediscovered by MacKay in the 1990's, and commonly used in wireless communications, are now receiving all the attention of the telecommunication community. A 2-dB coding gain is theoretically expected—compared with the actual FECs used in optics. This means that LDPC codes can provide the same performance (in terms of bit error rate) for a noise level 2-dB higher. However, this amazing efficiency comes at a cost of tremendous implementation complexity. Indeed, the LDPC decoding algorithm is both iterative and based on soft-decision inputs. Thus, the processed data are not the received bits but their reliabilities (the probability of a bit to be a 'zero' or a 'one'), which are in principle not quantized. In practice, these reliabilities are of course quantized. By using only four values for the reliabilities (two quantization bits), the decoding performance is slightly degraded, but the decoding complexity is highly reduced.
Previous research has focused on improving the code designs or the decoding algorithms, but the contribution of Sakib et al. is different. They propose an efficient and original technique to generate the quantized decoding input data. Instead of realizing this operation entirely electronically, a hybrid method is presented in which some of the processing is done in the optical domain, so that a few power-consuming electrical components can be replaced with a simple passive optical coupler.
The authors demonstrate experimentally the feasibility of their method on a 12.5 Gsample/s WDM transmission with off-line LDPC decoding. A slight performance degradation is observed, but the technique is less power consuming and thus more cost effective. The proposed scheme is compatible with direct-detection systems already deployed and is a very appropriate solution for the future upgrade of these systems with LDPC codes.
You must log in to add comments.