A new accuracy-controlable method for compressing the MoM impedance matrix of an electromagnetic problem based on a basis change plus a truncation with a threshold is presented here. The idea is to switch to singular basis functions, which are defined on relatively large subdomains of the object, representing perfectly the far field of the subdomain and yielding compression for well-separated blocks. Although the construction of the matrix is not computationally efficient compared with other matrix compression methods, it is still quite interesting from a theoretical point of view. Furthermore, it clearly explains certain characteristics of some of those methods, particularly the errors in the solution of those based in pseudoglobal functions.

Peer Reviewed