Retrofitting some fast methods into existing MoM codes

When retrofitting fast methods into an existing MoM code, one might think that N/sup 2/ matrix elements must be generated limiting the speed increases possible. Recent work on sparse MoM codes for PC applications has reduced LU matrix factorization time to significantly less than standard matrix fill time when the number of unknowns, N, approaches 200000. This paper reports on results using methods that reconstruct the MoM matrix from a small sample of its elements. A compressed form is then created which may then be directly changed into other compressed forms. Results suggest that when problem sizes approach 500000 unknowns, matrix fill can be reduced from 100 to 10 hours on a modern PC, which is then competitive with fast solutions by sparse LU factorization or by iteration.