Machine Learning in Charge: Automated Behavioral Modeling of Charge Pump Circuits

Konferenz: SMACD / PRIME 2021 - International Conference on SMACD and 16th Conference on PRIME
19.07.2021 - 22.07.2021 in online

Tagungsband: SMACD / PRIME 2021

Seiten: 4Sprache: EnglischTyp: PDF

Grabmann, Martin; Glaeser, Georg (IMMS Institut für Mikroelektronik- und Mechatronik-Systeme gemeinnützige GmbH (IMMS GmbH), Ilmenau, Germany)
Landrock, Christian (X-FAB Global Services GmbH, Erfurt, Germany)

Behavior models of Analog/Mixed-Signal (AMS) components are used in today’s System-on-Chip (SoC) verification mainly for improving simulation speed. In addition, they can be used to enable verification scenarios including back-box intellectual properties (IP). Writing and maintaining such models is still time consuming and prevents widespread use. One class of essential building blocks are on-chip charge pumps (CP), which enable a variety of different features in SoCs e.g. embedded nonvolatile memory solutions. This contribution presents a novel concept for automating the generation of grey-box behavior models of charge pump circuits using a Machine Learning (ML) approach. Compared to classical modeling approaches, it is not necessary to formulate an analytic description specific to the used circuit topology. The applicability of the approach is presented in a case study using an industrial charge pump design.