Black-Box Modeling of Digital Devices

The design for signal integrity and electromagnetic compatibility of modern fast electronic circuits heavily relies on numerical simulations and requires effective models of active devices. Black-box identification methods based on I/O data may offer a useful systematic approach to build such models. Black-box identification could be used for both the modeling of unknown components and for the refinement or the simplification of existing models. In this paper, we address the black-box identification of digital devices via NARX (Nonlinear AutoRegression with eXtra input) models. We apply a NARX identification algorithm to model a common highly nonlinear and fast electronic device, the CMOS inverter gate, and we perform different tests to assess the effectiveness of such an approach. We obtain accurate models with relatively simple structures and we also verify a low sensitivity of the identification process to the noise affecting the output data.