Model order reduction via eigen decomposition analysis

In this paper, we have proposed a novel model order reduction technique via rational transfer function fitting and eigenmode analysis considering residues. We define a constant as a key in the sorting algorithm as one of correlations in order to sort the order of eigenvalues. It is demonstrated that the accuracy via eigenmode analysis considering residues is improved. The proposed algorithm is a general method to match pole values with frequency domain poles for linear RC and RLC systems. Calculation of pole eigenvalues and eigen vectors can be done with more sophisticated analysis with the same level or smaller cost in the proposed algorithms in comparison to passive reduced order interconnect macromodeling algorithm (PRIMA). The experimental results show that our algorithm reduces up to 90% errors compared to the existing model order reduction algorithm, such as PRIMA, in wide frequency environment with the same number of poles in comparison.