Cost-effective OSNR monitoring with large chromatic dispersion tolerance using random forest for intermediate nodes

• Published in: Optics communications Vol. 479; p. 126469
• Main Authors: Chai, Jia, Zhao, Yan, Chen, Xue, Du, Jinhao, Li, Jian, Yang, Tao, Wang, Liqian, Zhang, Zhiguo
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.01.2021
• Subjects: CD tolerance; Cost-effective; Intermediate nodes; Machine learning; OSNR monitoring; Random forest; Intermediate nodes; CD tolerance; Random forest; OSNR monitoring; Machine learning; Cost-effective
• ISSN: 0030-4018; 1873-0310
• DOI: 10.1016/j.optcom.2020.126469

A cost-effective optical signal-to-noise ratio (OSNR) monitoring scheme with large chromatic dispersion (CD) tolerance is proposed. By leveraging the low-bandwidth coherent detecting with both highly sensitive optical to electrical conversion and small bulk wavelength scanning, no CD compensation and the random forest trained with asynchronous amplitude histograms, the proposed technique could serve as a cost-effective option to realize OSNR monitoring at intermediate nodes of elastic optical networks. By comprehensive simulations in 8 and 16 GBaud polarization multiplexed (PM)-4/16/32/64 quadrature amplitude modulation (QAM) coherent transmission systems, the relationship between the monitor bandwidth and the mean absolute error (MAE) of OSNR estimation are investigated without CD compensation. Meanwhile, under the optimal monitor bandwidth, the fluctuation of MAE is smaller than 0.1 dB when CD varies from 1600 to 16,000 ps/nm, proving the tolerance to large CD. Experiments of 8 and 16 GBaud PM-16/32QAM systems are conducted, the results also demonstrate the feasibility of the proposed scheme for CD tolerance. The coexistent advantages of cost-effectiveness and large CD tolerance make it quite suitable for the applications at intermediate nodes. •The proposed OSNR monitoring scheme for intermediate nodes has a large CD tolerance.•The low-bandwidth coherent detecting with wavelength scanning leads to be cost-effective.•The use of the random forest algorithm trained with AAHs makes the scheme’s complexity low. complexity.