Multilevel Thresholding Method for a Sparse Representation of Reduced Matrix in CBFM

Abstract

The characteristic basis function method (CBFM) is a direct solver which can effectively reduce the memory usage and CPU time requirement. The reduced matrix in CBFM is a compressed form of original impedance matrix in method of moment (MoM) and used for describing the mutual coupling between one block and another. Unfortunately, more memories are still consumed when forming the reduced matrix as the electrical size of target increases. A multilevel thresholding method is proposed to get the sparse representation of the reduced matrix to further eliminate the redundancy of the reduced matrix. The accuracy and efficiency are verified by numerical results.