Erweiterte gebietsarithmetische Verfahren zur Simulation analoger Schaltungen

authored by
Oliver Scharf
supervised by
Erich Barke
Abstract

Several different methods can be used for the simulation of analog circuit with uncertain parameters. The uncertainties can have different root causes, for example manufacturing tolerances, aging or ambient temperature. Common simulation methods like the Corner Case or Monte Carlo method have the disadvantage that they do not include all characteristics caused by all the possible parameter combinations. Methods based on range arithmetics can be used as an alternative. They guarantee the inclusion of all characteristics. The equation systems that need to be solved are in general implicit and nonlinear. In previous works an algorithm called EPD method was introduced. However, the size of the circuits and the size of parameter uncertainties which can be solved by this method is limited. This is due to the overapproximation caused by the affine arithmetic and the properties of the EPD method itself. One approach to increase the convergence range is the utilization of splitting and merging. The parameter space is divided, calculations are done with divided parameters and the results are merged afterwards. To not slow down the simuation by this approach it is reasonable to perfom the split only if a solution is not possible otherwise. Two critierions are presented to perform such an automatic split decision. Furthermore, different splitting and merging stratgies are compared in terms of runtime, overapproximation and required number of splits. The results show that the circuit size and the maximum size of parameter uncertainties can be increased by this method. By chosing the right strategies runtime can be traded for overapproximation. The optimal strategies and the gain in convergence range depent on the circuit which is simulated.

Organisation(s)
Institute of Microelectronic Systems
Type
Doctoral thesis
No. of pages
117
Publication date
2018
Publication status
Published
Electronic version(s)
https://doi.org/10.15488/3663 (Access: Open)