Muilti-region systematic reduced field testing (RFT) for matrix thinning

The RFT method is a spatial domain method yielding a good approximation for a sparse matrix, by economizing on the representation of far field interactions within the MoM matrix. The subdomain basis and testing functions are quite arbitrary, as opposed, e.g., to the wavelet transformation. The basic two-region RFT principle is expanded to include a number of regions. A new systematic derivation of the sparse matrix is also given. In the multi-region RFT, for each subdomain basis function, we divide the scatterer surface into several regions, viz. a near field, intermediate field and a far field region. Interactions within the near field region (and other rough regions of the surface) are modeled in the customary MoM manner, whereby a testing function is assigned to each 0.1/spl lambda/ cell, generating a dense diagonal region (and a number of other dense columns) in the matrix. In the intermediate and far field regions, however, it is recognized that the 0.1/spl lambda/ samples are highly correlated because the far field has fewer degrees of freedom, as it varies slower and has lower spectral content. Therefore the 0.1/spl lambda/ sampling rate is excessively fast in the far field regions and tends to generate information which is virtually redundant.