Efficient segmented power control for C+L+S ultra-wideband fiber transmission using CMA-ES

Ultra-wideband C+L+S fiber transmission systems require efficient launch power optimization to mitigate inter-channel stimulated Raman scattering and fully exploit the available bandwidth. This work studies power optimization under practical per-band constraints using a segmented slope-offset model for both SNR flatness and total throughput objectives, and applies the covariance matrix adaptation evolution strategy (CMA-ES). Numerical results show that CMA-ES achieves comparable optimization performance to particle swarm optimization (PSO) with significantly faster convergence. Segmented power optimization strategy improves SNR flatness, achieving an SNR ripple of 0.335 dB while maintaining per-channel rates above 800 Gb/s, whereas capacity maximization reaches 161.9 Tb/s with limited benefit from further segmentation. The results also demonstrate that when using the triangular approximation to simplify the solution process, the optimal launch power distribution strongly depends on the optimization objective, and it is only applicable to maximizing capacity.