Feedforward equalizer with selective noise decorrelation for bandwidth-limited signal

The performance of the current high-speed optical transmission system could be limited by the bandwidth of optical transceivers. Linear equalization is a feasible way to mitigate the inter-symbol interference (ISI) caused by the bandwidth limitation, but it enhances high-frequency noise, thereby degrading receiver performance. One effective way to cope with noise enhancement is based on the noise-whitening approaches at either the transmitter side or receiver side. However, conventional noise whitening requires a maximum likelihood sequence estimation (MLSE) equalizer to correct the ISI introduced intentionally, while its computation complexity is beyond the power of cost-sensitive applications. In this paper, we propose a novel selective noise decorrelation (SND) equalizer positioned after the linear equalizer to correct most of the errors in a feedforward manner. By leveraging the noise patterns to predict the likelihood of symbol errors and applying a weighted average of neighboring noise for decorrelation only when it is necessary, the SND equalizer provides an alternative approach to mitigate the noise enhancement issue without complicated MLSE equalization or noise whitening. We demonstrate its capability through proof-of-concept experiments at various bandwidth limitation ratios. The proposed SND equalizer outperforms the feedforward equalizer as well as the decision feedback equalizer in all cases. Moreover, it achieves comparable performance as the MLSE-based approach with a much lower computational complexity. Based on analysis, the proposed SND equalizer needs a mere 6.25% of the PF-MLSE equalizer's computational load.