Nonlinear macromodeling with AWE

Asymptotic Waveform Evaluation (AWE) [1, 2] is an efficient and general technique for simulating linear(ized) circuits. This paper discusses strategies for macromodeling nonlinear circuits with AWE. One approach, multi-region AWE macromodels, represents an extension of piecewise linear models, with the addition of internal states. Each region represents an AWE approximation to a linearization (at some bias point) of the nonlinear circuit of interest. In addition a technique is presented for initializing the internal states when the model transitions from one linearization to another during a transient simulation. The second approach is to treat nonlinearity as a second order effect that is superimposed on a linear solution as a post-processing step. A relaxation algorithm that exploits the reuseable AWE solution is employed to modify the linear solution so that it accounts for the macromodel nonlinearity.