General Block Structure-Preserving Reduced Order Modeling of Interconnect Circuits

Ning Mi,  Sheldon Tan,  Boyuan Yan


In this paper, we propose a generalized block structure-preserving reduced order interconnect macromodeling method (BSPRIM). Our approach extends structure-preserving model order reduction (MOR) method SPRIM into more general block forms. We first show how SPRIM-like structure-preserving MOR method can be extended to deal with admittance RLC circuit matrices and show the 2q moments are still matched and symmetry is preserved. We also show that $2q$ moment matching can't be achieved when the RLC circuits are driven by both current and voltage sources. We also improve SPRIM by introducing the re-orthonormalization process on the partitioned projection matrix. Then we present our BSPRIM method to deal with more circuit partitions and show the general block structure preserving MOR for circuits formulated in impedance and admittance forms. The reduced models by the proposed BSPRIM will still match the $2q$ moments and preserve the circuit structure properties like symmetry as SPRIM does. Experimental results demonstrate effectiveness of the proposed methods and it outperforms SPRIM in terms of accuracy with more partitions.