Passivity assessment and enforcement utilizing eigenpairs information

Abstract Rational models can be a cause of unstable time domain simulations if they are non-passive. One commonly applied method for ensuring model passivity is to combine a passivity assessment step with a passivity enforcement step in an iterative loop where the model's residue matrices are updated in each pass. This paper shows a new variant of such scheme that is computationally more efficient than an existing one. The advantage is achieved by transferring eigenpairs information between the two steps, rather than frequency samples where passivity violations exist. This leads to fewer inactive constraints in the constrained least squares problem associated with the passivity enforcement step, and thereby faster solving. The new approach is combined with the residue perturbation method known as RP-NNLS for maximum performance. The resulting procedure is demonstrated for the modeling of components with many terminals, a white-box transformer impedance matrix, grounding mat admittance matrix, and a black-box transformer model obtained via frequency sweep measurements.