Learned perturbation-based digital backpropagation with low complexity for nonlinearity compensation

Optical fiber communication systems play a crucial role in providing high-speed data transmission, forming the backbone of modern digital services and computational infrastructure. Further increases in data rates and transmission distances require higher signal power, which in turn amplifies the detrimental effects of fiber nonlinearity. However, developing a method that effectively compensates for nonlinear signal distortions while balancing performance and computational complexity remains an open challenge. Traditional approaches, such as digital backpropagation (DBP) and perturbation-based models (PBM), offer certain advantages but also have drawbacks that limit their practical implementation. In this work, we propose a low complexity perturbation-based digital backpropagation (PB-DBP) method to compensate for intrachannel nonlinearity. We introduce a novel approach that combines the DBP structure with an advanced PBM-based nonlinear effects model, using machine learning techniques to optimize compensation scheme parameters and improve the trade-off between accuracy and computational complexity. We present experimental results that demonstrate the performance of the proposed method, together with a comparative analysis based on data from a 20×100 km dispersion uncompensated fiber link with a dual-polarization QPSK signal. The results show that the proposed PB-DBP achieves an 1.6 dB signal-to-noise ratio improvement compared to compensation of only linear distortions, and an improvement of 0.26 dB compared to the enhanced DBP method.