Microwave Integrated Circuit Parameter Extraction Of A Fast Integral Equation Methods

The method of moment(MoM) is often effectively used in the extraction of passive components in modeling integrated circuits. This extraction, however, involves a dense system of linear equations. Using direct solution method is prohibitive for large problems. This thesis presents IES3, a fast integral equation solver for three-dimention problem. The approach takes the advatage of the smoothness of the integral equation kernel and can compress via the singular value decomposition(SVD). Furtrermore, it does not depend on the analytic form of the Green's function,so it may be applied to arbitrarily complex media. A dense matrix arising from MoM can be stored in sparsity after compressed. Those direct iterative slovers such as CG and GMRES are not applicable. The thesis reconstructs the two slovers based on the matrix compressing.Some parameters in IES3 have a final effect on compressing efficiency and computational accuracy. The thesis makes an investigation on these parameters and gives advice on appropriate choice. The preprocessing time can be considerable for the large problem. A new conception called sub-matrix threshold is introduced in the thesis and the modified algorithm can obtain much faster preprocessing time with a small loss of compression.