Reconfigurable optical add-drop multiplexers (ROADMs) based on wavelength selective switches (WSSs) are crucial devices in optical networks, facilitating reconfigurable optical routing. In practical systems, soft failures in WSSs may occur, leading to connection disruptions if not addressed promptly. For rapid network diagnosis and maintenance, it is essential to quickly localize the failed WSS and estimate the failure magnitude. However, current approaches rely on scarce real-system failure data or require the installation of additional monitors. To address these limitations, in this paper, we propose a physics-based learning approach (PBLA) to localize the failed WSS and estimate the failure magnitude with two types of failure: filter shift (FS) and filter tightening (FT), without using historical failure data and additional monitors at the repeater nodes. The method leverages the power spectrum (PS) extracted from the chromatic dispersion compensation (CDC) module in receiver digital signal processing (DSP) as input. It formulates an optimization problem to minimize the discrepancy between the extracted PS and the PS calculated through a theoretical model. By solving the optimization problem, the method can effectively localize the failed WSS and estimate the failure magnitude in WSSs. The method is validated through extensive simulations, exhibiting a high localization accuracy and a small estimation error. Furthermore, we thoroughly explore the influence of various factors, including the measurement of PS, failure location, magnitude, the increasing number of WSSs, different symbol rates, and model inaccuracy, on the performance of our method, demonstrating its adaptability across diverse scenarios. Finally, an experimental demonstration further substantiates our method's performance in localizing and estimating WSS-related anomalies in optical networks.
A new ISI-free frequency domain M-shaped pulse is proposed and experimentally verified to have better SPM tolerance than both RRC and optimized pulses in a 14Gbaud 56Gbit/s DP-QPSK system over dispersion unmanaged links.
A constellation independent look-up table (LUT) method for transceiver nonlinearity predistortion in digital-analog radio-over-fiber system is proposed and experimentally demonstrated, achieving a SNR gain of 1.04dB. The table size can be reduced to 0.3% of the conventional LUT.
We propose a physics-informed EDFA gain model based on the active learning method. Experimental results show that the proposed modelling method can reach a higher optimal accuracy and reduce ~90% training data to achieve the same performance compared with the conventional method.
A coherent receiver based on silicon photonics is presented. Its performance is confirmed by single-channel and WDM transmission tests with PM-QPSK modulation at 28 Gbaud up to 4800 km. The packaged core is very compact with dimensions of 6 mm × 8 mm.
We experimentally demonstrate a 4-dimensional modulation direct detection system based on Stokes vector receiver and a novel reduced-complexity SSBI cancellation algorithm for signal linearization. A 480 Gb/s data rate over 80 km SSMF is achieved.
We demonstrate a transmission of 8×256 Gb/s WDM 64 Gbaud QPSK signals over 2880 km of standard single mode fiber (SSMF) with erbium doped fiber amplifier (EDFA)-only amplification and coherent detection.
A reliable and accurate meta-learning-assisted training framework is proposed for deploying machine-learning-based applications to optical networks. The framework provides prediction confidence interval and makes better utilization of noisy data.
With the advance of elastic optical networks, optical communication systems are becoming more flexible and dynamic. In this scenario, soft failures are more likely to occur due to various link impairments, of which the filter impairment caused by WSS is a major one. If these soft failures are not handled properly and timely, the quality-of-transmission (QoT) will degenerate, even leading to service disruption. During this process, it is important to know the accurate location and the magnitude of soft failure. However, it is difficult for traditional methods to accomplish this target. Fortunately, with the fast progress of the powerful machine learning (ML) algorithms, a new promising way is provided to address this problem. In this article, we propose to use artificial neural network (ANN) and Gaussian process regression (GPR) to localize the soft failure location and estimate anomaly value. The input of the ANN and GPR is extracted from the power spectrum density (PSD) and the equalizer taps, which can be easily obtained from a coherent receiver. We explore two types of soft failure caused by the WSS, including the offset of WSS's center frequency, i.e., frequency shift (FS), and the tightening of WSS's 3-dB bandwidth, i.e., frequency tightening (FT). To validate the proposed method, we perform extensive simulations. The localization accuracy of the ANN can reach 95%, and the mean-absolute-error (MAE) of the GPR can reach 0.1-GHz, demonstrating the effectiveness of the proposed method. Besides, we validate the generalization of the proposed method under different link conditions. Finally, the importance of each input feature is explored, showing the effectiveness of the selected features.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
