Chromatic Dispersion Monitoring Scheme Based on Inverse Radon Transform with Large Angle Intervals
Objective Optical performance monitoring is crucial for enhancing the reliability and intelligence of optical communication networks.Chromatic dispersion(CD)monitoring is essential for adaptive,precise dispersion and nonlinear compensation in dynamic optical networks,where even small errors can significantly affect receiver performance.Traditional CD monitoring methods,which typically rely on clock phase information,autocorrelation functions,signal peak-to-average ratios,and signal amplitude and phase information,operate primarily in the time or frequency domains.However,these approaches often overlook information in joint time-frequency domain distributions and are generally limited to a monitoring range within 1000 km.The fractional Fourier transform(FrFT),an extension of the traditional Fourier transform,provides flexible analysis and extraction of time-frequency domain features by adjusting the transform order.While FrFT-based schemes can improve the range of dispersion monitoring,they typically suffer from monitoring errors above 100 ps/nm.In this study,we propose a chromatic dispersion monitoring scheme that uses an inverse Radon(iRadon)transform with large angular intervals to improve accuracy.Our approach effectively reduces noise interference and enhances subtle variations in time-frequency maps,increasing sensitivity to dispersion changes and reducing the complexity of network parameters.Methods We introduce a training sequence processed by FrFT into the transmitted signal for dispersion monitoring.This sequence can be generated by applying a fixed-angle FrFT to the DC signal.At the receiver,the received signal is downsampled,and the training sequence is processed through FrFT to obtain a two-dimensional time-frequency map.A cubic operation is then performed on this map to emphasize the energy peak position and mitigate the influence of nearby linear components.The iRadon transform is then applied to convert the two-dimensional time-frequency map from polar to Cartesian coordinates.Unlike traditional iRadon transform schemes that use small angular intervals,we propose a discrete iRadon transform with large angular intervals.This approach enhances peak shifts and increases the sensitivity of the time-frequency map to changes in dispersion characteristics.Using a 20° angle interval as an example,we demonstrate the effectiveness of the proposed scheme in improving the dispersion monitoring accuracy.The features extracted from the iRadon transform at large angular intervals are then fed into a lightweight convolutional neural network(CNN)for cumulative dispersion monitoring.Results and Discussions The results show that,for a dispersion monitoring range of 1600-24000 ps/nm,the proposed scheme consistently achieves monitoring errors below 30 ps/nm,with a mean absolute error(MAE)of 19.82 ps/nm.This represents a significant improvement over the maximum residual cumulative dispersion value of 350 ps/nm that can be compensated by the constant mode algorithm(CMA).Compared to traditional FrFT methods with small angle interval iRadon transforms,the monitoring performance is improved by a factor of four.The enhanced dispersion monitoring function,which focuses on the height variation between linear components in the time-frequency map,enables the lightweight CNN to achieve high-precision dispersion monitoring.Notably,the CNN in the proposed scheme requires only 15960 training parameters,compared to 6653703 in traditional methods.In addition,we explore the influence of different signal-to-noise ratios(OSNR),transmission power levels,and differential group delay(DGD)on dispersion monitoring in fiber optic links.We find that when the OSNR exceeds 19 dB,the dispersion monitoring error is kept within 50 ps/nm for transmission distances up to 800 km.As the OSNR decreases,the monitoring error increases.When the input optical power ranges from-2 to 2 dBm,the nonlinear effects are weak,and the dispersion monitoring error remains below 100 ps/nm.However,as the power increases,the nonlinear effects intensify,leading to a gradual increase in the monitoring error.Finally,we analyze the influence of DGD on the dispersion monitoring accuracy,with OSNR set to 30 dB and DGD varying from 0 to 50 ps in 10 ps increments.We observe that for DGD values between 0 and 40 ps,the overall error remains small,with a maximum error of-133.81 ps/nm at 40 ps,indicating robust performance against DGD variations within this range.However,when DGD reaches 50 ps,the overall prediction error deviates significantly from the centerline,with the maximum monitoring error reaching-319.99 ps/nm.Conclusions In this study,we propose a novel CD monitoring scheme based on the large angle interval iRadon transform.The approach effectively suppresses noise,enhances dispersion sensitivity,and reduces the required neural network parameters,achieving high-precision and low-complexity CD monitoring.Simulation results demonstrate that for accumulated chromatic dispersion values ranging from 1600 to 24000 ps/nm,the proposed scheme achieves an MAE of 19.82 ps/nm.This represents a more than fourfold improvement in accuracy and a two order-of-magnitude reduction in neural network parameters compared to the FrFT method with small angle interval iRadon transform.
NETL
NSTL
万方数据
