A Method of Developing Frequency-Domain Models for Nonlinear Circuits Based on Large-Signal Measurements

We describe a method of generating models for nonlinear devices and circuits, based upon measurements of travelling-wave voltages at a periodic large-signal excitation and its harmonics using a nonlinear vector network analyzer. Utilizing a second source, we use multiple measurements of a nonlinear circuit to train artificial neural network models that yield portable, nonlinear large-signal scattering parameters. We obtain an independent check by comparing an example diode-circuit model generated by means of this methodology with a harmonic-balance simulation.