A new technique for the efficient solution of singular circuits

Singular circuits are those that have a continuum of solutions (or no solution) and are characterized by rank deficiency in the (linearized) circuit matrix. Such circuits often arise in practice-examples are filters with poles at zero, chains of transmission gates that are off, and circuits that rely on charge storage and transfer. In this paper, a technique for the efficient solution of such circuits is presented. The method is based on solving for the minimum-least-squares solution of the singular system. Unlike traditional methods for least squares solution, the new approach exploits the sparsity of the circuit matrix, making it practical for large industrial circuits. The method is best for applications with relatively small singular subspaces, as is the case in most circuits. Applications to industrial designs testify to the efficacy of the new technique; an example in which more than a week of design time would have been saved is presented