Extending the OCATA Digital Twin to the Frequency Domain

We extend the OCATA time domain digital twin to the frequency domain. It is shown that deep learning-based models enable to predict the linear and nonlinear impairments affecting the optical constellations, whereas analytical models are suitable to estimate the filter penalties affecting the optical spectrums. Illustrative results show the benefits of the proposed approach for failure detection and identification.