Differential group delay prediction in optical fiber links

Area of Science:

• Optical communications
• Signal processing
• Time series analysis

Background:

• Differential Group Delay (DGD) is a critical parameter affecting signal integrity in optical fiber links.
• Accurate prediction of DGD evolution is essential for adaptive optical systems.
• Existing forecasting methods may not fully capture the complex dynamics of DGD.

Purpose of the Study:

• To simulate and forecast the time evolution of DGD in fiber links using various time series forecasting techniques.
• To evaluate the performance of different forecasting methods, specifically focusing on the prediction horizon.
• To investigate the impact of measurement noise on the prediction accuracy and horizon.

Main Methods:

• Hinge model simulation of DGD time evolution.
• Application of time series forecasting procedures: Autoregressive (AR) models with Kalman filter, pattern imitation, and Taylor expansion.
• Analysis of prediction horizons achieved by each method.
• Assessment of measurement noise influence on prediction performance.
• Validation using measured DGD data from 40 Gb/s fiber links.

Main Results:

• The Autoregressive (AR) model combined with the Kalman filter procedure achieved the furthest prediction horizon among the tested methods.
• Measurement noise was found to significantly influence the prediction horizon, generally reducing it.
• The applied methods were successfully demonstrated on real-world 40 Gb/s fiber link DGD data.

Conclusions:

• AR models with Kalman filtering offer a robust approach for predicting DGD evolution in optical fiber links.
• Understanding the impact of measurement noise is crucial for deploying effective DGD forecasting systems.
• The simulation and analysis provide valuable insights for mitigating DGD impairments in high-speed optical communication systems.