Automatic passive mode-locked fiber laser based on nonlinear polarization rotation using support vector regression and sliding window correlation

We propose and experimentally demonstrate a novel, to our knowledge, technique to achieve automatic mode-locking in nonlinear polarization rotation-based fiber lasers. The polarization dynamics of the fiber laser are studied in the framework of a support vector regression model. The feature vectors for the support vector regression are obtained by downsampling the photocurrent spectrum, acquired by an electrical spectrum analyzer, after optical to electrical down conversion of the laser output. Combining support vector regression for anomaly detection along with sliding window correlation of optimal window size, we obtain an empirical model that predicts the precise voltage to be applied to the electronic polarization controller to produce the required polarization state for automatic mode-locking at the fundamental frequency. The proposed algorithm offers fast operation with low computational complexity, making it practical for hardware realization in real-time laser control. Using this method, we demonstrate stable mode-locking in an Er-doped fiber ring laser operating at 1.55 µm having a fundamental repetition rate of 185 kHz.