Information flow in digital metal-oxide semiconductor circuits

Static algorithms designed to symbolically describe an electrical network and derive subnetworks of a given maximal size are presented. These algorithms further provide the solution order of the subnetworks. This network splitting allows for strong local feedback that would otherwise cause convergence speed problems. Signal-flow graphs that describe the influence of each element in the circuit on its neighbors serve as the basis for these algorithms. Cycles in the signal-flow graph represent feedback in an electrical network. By removing the cycles, one can naturally determine (by leveling) a favorable solution order for the differential equations and small-equation systems. The work presented here also provides the basis for designing the parallelization of a differential equation solver.<>