Summary
International Symposium on Antennas and Propagation
2011
Session Number:FrP1
Session:
Number:FrP1-49
Optimization of Block Size for CBFM in MoM
Keisuke Konno, Qiang Chen, Kunio Sawaya, Toshihiro Sezai,
pp.-
Publication Date:2011/10/25
Online ISSN:2188-5079
Summary:
Method of Moments(MoM) is known as one of the powerful techniques for numerical analysis of antennas and scatterers[1]. With a direct solver like Gauss-Jordan method, CPU time for the MoM is O(N3 ) where N is number of unknowns. Therefore, the direct solver can not be applied to the MoM when N is too large. Previously, iterative methods like conjugate gradient (CG) method has been proposed for reduction of CPU time [2]. CPU time for each iteration is O(N2 ) and total CPU time becomes smaller than O(N3 ) when number of iteration steps is smaller than N. However, iterative methods are not effective for ill-conditioned problems because number of iterations for the problems is proportional to N[3]. Therefore, total CPU time required for anlysis of ill-conditioned problems is still O(N3 ) even when iterative methods are used. CBFM (Characteristic Basis Function Method) is also known as one of the powerful techniques for analysis of large-scale problems[4]. Since the CBFM does not include iterative procedure like the CG method, CPU time required for the CBFM can be reduced for the ill-conditioned problem. So far, it has been found that CPU time required for the CBFM depends on number of blocks M. However, relation between number of blocks M and number of segments N, which gives minimum CPU time, has not been investigated. In this paper, optimum number of blocks M is derived theoretically as a function of N. The numerical simulation shows that minimum CPU time for the CBFM is realized by the optimum M.