Abstract:The rapid growth of online services and the global transition to digital platforms, accelerated by the COVID-19 pandemic and the emergence of AI-based services, have led to a surge in demand for high-capacity optical communication networks (OCNs). However, this transition exacerbates the issue of fiber nonlinearity, notably four-wave mixing (FWM), which can significantly impact signal integrity in long-distance OCN transmissions with multiple channels. To address this challenge, this paper presents a novel approach leveraging a convolutional neural network (CNN)-based digital signal processing (DSP) model to mitigate the nonlinear effects of FWM on signal quality. The proposed CNN model, designed with three convolutional layers and trained on a comprehensive dataset of simulated OCN transmissions, demonstrates significant improvement in reducing nonlinear distortions caused by FWM. Our method is compared with traditional compensation techniques, such as digital back-propagation and Volterra series-based equalizers, showing a reduction in power penalties by 1 dB and an enhancement in the power budget by 2 dB. Simulation results indicate that the bit error rates (BERs) remain consistently below the critical threshold of 10 −9 across various transmission distances, outperforming conventional methods. The simulation analysis, conducted on a 100 km multi-channel OCN testbed, confirms the CNN equalizer's efficacy in reducing crosstalk and improving power efficiency, with a demonstrated 15% reduction in required launch power. Moreover, the analytical models used to evaluate the CNN-based DSP model show high accuracy in predicting performance, underscoring the model's capability to achieve high-quality transmission with lower power requirements. These findings advocate for the integration of our CNN-based DSP model into OCNs, offering a fruitful solution for managing multi-channel, long-distance, and high-data-rate optical transmissions in real-world systems.

Compensation of Turbulence-Induced Wavefront Aberration with Convolutional Neural Networks for FSO Systems

Investigation of Convolution Neural Network-Based Wavefront Correction for FSO Systems

Convolutional Neural Network-Assisted Optical Orbital Angular Momentum Recognition and Communication

Recognizing Fractional Orbital Angular Momentum Using Feed Forward Neural Network

Aberration Correction Method For Fourier Ptychographic Microscopy Based On Neural Network

Elimination of Varying Chromatic Aberrations Based on Diffractive Optics

Atmospheric Turbulence Aberration Correction Based on Deep Learning Wavefront Sensing

Experimental demonstration of deep-learning-enabled adaptive optics

Coherently Demodulated Orbital Angular Momentum Shift Keying System Using a CNN-based Image Identifier As Demodulator

A deep-learning approach for turbulence correction in free space optical communication with Laguerre–Gaussian modes

Improving Optical Transmission Performance in Multi-Channel Networks Utilizing Convolutional Neural Network-Enabled Digital Signal Processing to Mitigate Four-Wave Mixing

High-speed Modulating Retro-Reflectors with Optical Phase Conjugation Compensation

Atmospheric turbulence recognition with deep learning models for Sinusoidal hyperbolic hollow Gaussian beams-based free-space optical communication links

An Atmospheric Turbulence Compensation Algorithm Based on FSM-DM Cascaded AO Architecture for FSO Communications

Improved Machine Learning Approach for Wavefront Sensing

Fast single pixel modal wavefront sensing using neural networks

Demonstration of photonic neural network for fiber nonlinearity compensation in long-haul transmission systems

Research on Orbital Angular Momentum Multiplexing Communication System Based on Neural Network Inversion of Phase

Automatic Mitigation of Dynamic Atmospheric Turbulence Using Optical Phase Conjugation for Coherent Free-Space Optical Communications

Turbulence compensation with pix-to-pix generative adversarial networks in vector vortex beams

Joint estimation model for FSO channel parameters and performance evaluation based on CNNs