Approximation of Large-scale System using Equivalent-Norm Frequency selected Projection

In the design of integrated circuits, model order reduction is a dynamic tool to decrease the complexity of the large-scale systems arising in the verification of chips. The development of an accurate and compact model of electronics and on-chip interconnects in the design is central to the time cost and storage space requirement of integrated circuits (IC). The current IC design tools use 3-D extraction to develop resistor, inductor, and capacitor-based state-space representation, which results in a very large-scale system representation. In this work, a reduced equivalent-norm frequency selected projection method is proposed, which optimally selects expansion points for the approximation of a large-scale on-chip system and solves it as an eigenvalue problem. The reduce-order results are accurate across a selected range of frequencies and perform better than the existing Arnoldi-based methods for model order reduction. In addition, the proposed approach assures stability and passivity, with simulation comparison showing better numerical accuracy compared to existing methods.