Composing Scalable Solver for Simulation of Electronic Circuits in SPICE

The most efficient solvers use composite procedures that adaptively rearrange computation algorithms to maximize simulation performance. A similar concept can be integrated into a process of electronic circuit analysis, where the combination of different algorithms allows scalability of simulation performance. In this paper, we propose a new adaptive internal solver based on Biconjugate gradient stabilized method for the iterative solution of nonsymmetric linear systems supplemented with incomplete LU factorization as an efficient replacement for the direct solver implemented in program Spice for solving large-scale circuits. We describe basic concepts of a simulation of electronic circuits with nonlinear time dependent devices and present implementation examples of the proposed methods. Optimal setting of the method and its application in program Spice is shown in comparison to other modern iterative solvers for nonsymmetric linear systems.