An Adaptive Sampling Technique for Passivity Characterization and Enforcement of Large Interconnect Macromodels

This paper deals with the characterization and enforcement of passivity for linear lumped interconnect macromodels. An adaptive accuracy-controlled frequency sampling process is employed to identify a set of frequency bands where the macromodel is locally passive. These results are employed as a preliminary step, enabling the fast computation of imaginary eigenvalues of the Hamiltonian matrix associated to the macromodel. Then, iterative perturbation is employed to remove these eigenvalues from the imaginary axis and to achieve global passivity. The resulting scheme is highly optimized for macromodels having large dynamic order and with a sparse structure. Significant speedup factors up to two orders of magnitude are achieved with respect to a standard implementation of the same passivity compensation scheme based on a full eigensolver