Automated extraction of broadly applicable nonlinear analog macromodels from SPICE-level descriptions

Abstract

Automated techniques for generating macromodels from SPICE-level circuit descriptions are rapidly gaining importance as a sustainable methodology for the design of large, complex mixed-signal SoCs and SiPs. We demonstrate the efficacy of a novel macromodel extraction technique, dubbed piecewise polynomial (PWP), for extracting broadly-applicable general-purpose macromodels from SPICE netlists. A key advantage of PWP over other methods is that it can generate a single macromodel that captures linear, weakly nonlinear and strongly non near system dynamics. We demonstrate the application of PWP using a current-mirror op-amp, comparing simulations of the macromodel against those of the original SPICE circuit using DC, AC, harmonic balance and transient analyses. We also illustrate how PWP-generated macromodels can be used for system-level simulation using a simple analog-digital converter example. We confirm excellent accuracies, relative to full SPICE circuit simulation, while achieving order-of-magnitude speedups.