On the Generation of SPICE-Compatible Nonlinear Behavioral Macromodels

This paper presents a complete framework for the generation of behavioral macromodels of a wide class of nonlinear components, devices, and systems. The model structure and related identification algorithms are based on the Volterra series formulated in the frequency domain through multivariate Generalized Transfer Functions. A multivariate rational model is first estimated in pole-residue form from sampled responses and then converted to a bilinear state-space form. The main novel contribution of this work is a SPICE-compatible circuit synthesis, which enables the usage of nonlinear macromodels within circuit simulation environments as part of more complex system-level simulations. Examples are provided for a Low DropOut voltage regulator and for a system-level Power Distribution Network embedding integrated regulators. For such examples, the proposed SPICE equivalents offer speedup factors ranging from 12X up to 650X with negligible loss in accuracy.