K-means clustering of single-shot pulses from gain-switched semiconductor lasers with Shannon entropy

We quantitatively analyzed shot-to-shot waveform characteristics in a palm-sized Fabry–Pérot (FP)-type gain-switched semiconductor laser diode (GS-LD) module by combining Shannon-entropy- and discrete Fourier transform (DFT)-based feature extraction with K-means clustering. The Shannon-entropy-based descriptors capture waveform complexity, whereas the DFT-based descriptors characterize the dominant pulse-train periodicity. The clustering revealed a fundamental-period pulse-train class with pulse intervals of 5.3 ps, an intermediate waveform class, a half-period-like dense pulse-train class with intervals of 2.7 ps, and a minor residual outlier-like class. A simple simulation based on a longitudinal-mode spectrum showed that these waveform clusters can be understood in terms of different degrees of mode-locked-like intermodal phase correlation among the FP longitudinal modes.