Computationally efficient physics-informed difference-symmetric nonlinear equalizer for C-band DML-DD links.

Abstract

Volterra nonlinear equalizer (VNE) is widely used in intensity modulation and direct detection (IM/DD) systems because it employs multi-order operations to effectively capture the nonlinear characteristics of signals as a generic tool. In the specific directly-modulated laser with direct detection (DML-DD) link, the interaction between the chirp of DML and chromatic dispersion (CD) can be modeled as composite second-order (CSO) distortion. By incorporating the CSO model into the nonlinear equalizer, it is possible to better extract the feature of the end-to-end channel, achieving superior performance with lower complexity. In this work, we propose a computationally efficient physics-informed difference-symmetric nonlinear equalizer (DSNE) based on the analytical formulation of the CSO. Additionally, we provide a thorough comparison of the computational complexity and bit-error-rate (BER) performance of various equalizers. Compared to the conventional VNE, the DSNE provides a 1-dB improvement in receiver sensitivity while reducing computational complexity by 51%. It is shown that the model-assisted DSNE structure enhances the matching to channel nonlinearity by omitting the less cost-effective taps in the conventional VNE and applying difference operations to the symmetric taps. The DSNE incorporates difference-symmetric terms, in contrast to the quadratic nonlinear equalizer (QNE), which uses only diagonal terms. This addition leads to a 56% reduction in BER while incurring only a 12% increase in computational complexity. The proposed DSNE technique demonstrates significant potential for low-cost, high-performance DML-DD optical transmission systems.