Solving methods for semi-symbolic analog simulation

Parameters of analog circuits are not exactly known as they are influenced by fabrication, aging or environment temperature. Commonly, these influences are considered by corner case or Monte Carlo simulations. By using these methods, inclusion of the circuits extremal values is not guaranteed. A circuit simulator which uses special range arithmetic called Affine Arithmetic was developed at the Institute of Microelectronic Systems. Simulation modes known from SPICE-based simulators like DC, AC or transient analysis were implemented using Affine Arithmetic. These method guarantees the inclusion of the worst and the best case. By this semi-symbolic simulation method the influence of parameters on the circuit behavior can be traced and the circuit can be optimized accordingly.

The current work aims at improvements of the affine solving method for implicit nonlinear circuit equations. In certain operating points it is necessary to split parameters, solve separately and merge the solutions to obtain a convergent result. These points are not known a priori and should be detected automatically during simulation. Different splitting strategies are considered concerning their applicability and performance. By these measures larger circuits than before can be analyzed using affine arithmetic.