Machine Learning-based Model for Defining Circuit-level Parameters of VCSEL

Recently, many computationally efficient models have been introduced to accurately define the static and dynamic Vertical Cavity Surface Emitting Laser (VCSEL) behaviors. However, in these models, many physical parameters must be appropriately set to reproduce existing laser sources' behavior accurately. The extraction of these unknown physical parameters from experimental curves is generally time-consuming and relies mainly on trial and error approaches or regression analysis, requiring extra effort. In this scenario, we propose a machine learning-based solution to the problem, which can effectively extract the required VCSEL parameters from experimental data in real-time. The proposed approach predicts the parameters exploiting the light-current curve and small-signal modulation responses with two steps at constant and variable temperature, respectively. Promising results are achieved in terms of relative prediction error.