Interaction between model problems and convergence

Convergence problems is known to be an integral part of the Newton-Raphson algorithm. The convergence of a Newton-Raphson type of simulator is influenced by many factors: time-step algorithm, integration and iteration method etc. However, the quality of the used models, has a big impact on the convergence or nonconvergence of a simulation. Probably the most common problem of this nature, at least for MOS designers, is the discontinuity between subthreshold and the linear region in standard MOS models. However there are may other model issues that influence the convergence of a particular circuit simulation. This paper addresses these issues and shows how to cure or avoid them