Passivity-preserving model order reduction of linear time-varying macromodels

Abstract

We study model order reduction (MOR) of continuous-time linear time-varying (LTV) systems. Examples include circuit or interconnect models found in VLSI marco-modeling. Specifically, a time-varying version of positive-real balanced truncation (PRBT), called LTV-PRBT, is proposed, which preserves the passivity of LTV systems for stable global simulation. Implementation details are discussed together with a brief outline of a discrete-time counterpart of LTV-PRBT. Dynamically changing state dimension is allowed for accurate modeling at the lowest possible order. Numerical examples then verify the effectiveness of the proposed approach over existing LTV MOR methods.