Gaussian Noise (GN) Model Experimental Validation on Short Span Optical Fiber Transmission

Abstract

Nonlinear signal impairments are among the most critical phenomena in long-haul optical communication system transmission. Therefore, it is important to include a nonlinear impairment model in the Quality of Transmission (QoT) prediction algorithms. While many models to account for the fiber nonlinearities in optical fiber transmission have been proposed, the practical value of a model is also a consideration for implementation and calculation time is a particularly important factor. The Gaussian noise (GN) model accounts for the nonlinear signal impairment impact using a simple closed-form formula, reducing the complexity of the computation process and computation time. With reasonable accuracy, the GN model has become a powerful tool that has been implemented in many physical environment based QoT predictions [7][10], including a Python implementation that is being developed by an industry group, known as GNPy[10]. GN model performance in predicting long-haul transmission effects has been experimentally validated by many researchers on various types of networks. In many emerging networks, however, it is common to have non-homogeneous links and short span lengths, such as in a metro network, which are not accounted for in the original GN model. Furthermore, many optical networking lab experiments do not allow for recirculating loop methods and therefore long distance, many hop experiments are problematic. Since nonlinear fiber effects are power dependent, long distance transmission can be emulated by artificially increasing the signal power. In such scenarios, the conditions that allow for a Gaussian noise statistics approach might not be satisfied and the accuracy of the GN model may be compromised. Therefore, the performance ofthe GN model is tested on a coherent optical communication system testbed using 1 and 2 spans for high signal powers. In this experiment, it is shown that the calculated 𝑃𝑁𝐿𝐼 using GN model is overestimated results in the deviation of the calculated gOSNR from the measured gOSNR. The 𝑃𝑁𝐿𝐼 is also measured using several other models to increase its accuracy, such as using GN model with correction factor, model for DM transmission, and gnpy. However, these models only show slight improvement to the gOSNR calculation accuracy. The deviation of the calculated gOSNR is possibly due to the lack of signal dispersion in the first few spans results in the failing gaussianization of the nonlinear signal. Also, the strong SPM and XPM noise combined with the interaction of the nonlinear noise with the ASE produced by the last amplifier introducing a strong nonlinear phase noise and contribute to the significant error in GN model prediction