Raman Pump Optimization for Maximizing Capacity of C+L Optical Transmission Systems

Within a context of C+L band transmission, this work proposes a design approach for Raman pumps in hybrid fiber amplifiers (HFAs) with the goal of maximizing the total system capacity. First, the optimization problem is constructed. The capacity of a system can be equivalently assessed by the mean generalized signal-to-noise ratio (GSNR) of all channels, which is chosen as the optimization objective. The powers of Raman pumps are chosen as the decision variable, and constraints on the Raman pump powers are added. Then, an optimization framework for Raman pump powers is proposed. An artificial neural network (ANN) is used to establish a differentiable model for GSNR and signal power after Raman amplification (RA). The gradient descent algorithm is adopted to perform the optimization. Simulations are conducted on a single-span link for modeling and on an 8-span link for optimization. Results show that the ANN can reach a high modeling accuracy with a prediction error of 0.16 dB for GSNR and 0.06 dB for signal power with specific link and signal parameters. Results also demonstrate the superior performance of the proposed optimization framework, and a gain of 0.54 dB on the mean GSNR can be achieved by the designed HFA compared with that of the design aimed at a flat gain profile.