AI-Assisted Online Compensation of SRS Effects in OTDR Traces for Super C+L-Band Systems

The emergence of wideband optical networks, particularly those utilizing super C+L-band transmission, introduces new challenges for in-service optical time-domain reflectometer (OTDR) due to the effects of stimulated Raman scattering (SRS). As OTDR probe pulses propagate through a fiber carrying high-bandwidth traffic, they experience wavelength-dependent amplification or depletion from SRS, resulting in distorted traces that no longer accurately reflect the fiber’s attenuation profile. In this work, we present a Transformer model-based online approach to correct OTDR trace distortions caused by super C+L-band traffic to restore the fiber loss profile. Notably, the model operates solely on the distorted OTDR trace without requiring knowledge of the fiber launch power distribution or any auxiliary input. The approach is experimentally validated over a 12 THz super C+L-band transmission system by varying the total fiber input power to induce different amount of SRS gain in the OTDR traces. By applying the correction method, the mean absolute errors (MAE) between the OTDR traces distorted by the traffic and the traces when there is no traffic were reduced from 4.7, 3.24, 2.18, and 1.51 dB to 0.072, 0.062, 0.063, and 0.09 dB, respectively, across four launch power settings.