Abstract
—Time-domain circuit analysis and simulation is commonly performed through intrinsically serial methods, such as those implemented in the popular software package SPICE. However, the practical utilization of these methods is limited to circuits of a certain size, as for larger problems the simulation time becomes prohibitive. Parallelization of such serial routines has been proposed as the main alternative to accelerate the analysis and overcome the problem, even if the speed-up that can be achieved by this strategy is bounded, according to the Amdahl's law. Currently, there is a lack of intrinsically parallel methods which would allow to detach the efficacy of the solution from the problem size. In this article, we develop a new theoretical approach to circuit analysis from an intrinsically parallel point of view. We propose a fixed-point method and determine its convergence condition according to the theory of asynchronous iterations. We also perform a series of tests that show that our method is faster in most cases than those based on the traditional intrinsically serial approach. In particular, we obtain an empirical proof that our approach is independent of the problem size, offering great opportunities for scalability.