• machine learning
• linear regulator
• differential evolution
• AI
• VerilogA

This thesis's purpose is to develop a Verilog-A model of a linear regulator with reference to a 5V transistor-level DC-DC converter. Parameters are added to make the model versatile and re-usable for different use-cases, so that designer are supported in developing new devices and verification engineers in validating top schematics with a considerable saving of time. Generalized Differential Evolution is employed to fit the output waveforms: this powerful Artificial Intelligence algorithm provides all the tools to obtain the best solution under every requirements.
The final solution attains an overall mean accuracy error of less than 0.5% for DC analysis in regulating region, 25% for transient analysis but a computational time reduction of more than 90%. The model functioning is validated in a boost top schematic simulation.