TY - GEN
T1 - Novel high precision UFFT methodology for fast analysis of 3D Planar circuits embedded in shielded layered media
AU - Rautio, Brian J.
AU - Okhmatovski, Vladimir I.
AU - Lee, Jay K.
N1 - Publisher Copyright:
© 2014 IEEE.
PY - 2014/11/12
Y1 - 2014/11/12
N2 - Unified-FFT (UFFT) is a novel 3D-Planar Method of Moments (MoM) solver of the Electric Field Integral Equation (EFIE). UFFT is the resultant algorithm of combining the FFT-enhanced Matrix Solve Operations (MSO) currently utilized in Sonnet Suites with FFT-enhanced Matrix Fill Operations (MFO). It has been shown, in conjunction with iterative MSO based on GMRES, to solve matrix vector products scaling with O(NlogN) operations and O(N) memory for planar, single-plane geometries (B. J. Rautio, V. I. Okhmatovski, J.K. Lee, IMS 2013). Previous UFFT implementations have achieved O(NlogN) scaling through acceleration of iterative matrix-vector products within GMRES by separating near and far interactions and calculating far interactions implicitly as in PFFT. In this work, the same uniform grid used with MFO is extended for use with MSO, negating the need to treat near and far interactions independently and allowing the entire matrix to be stored implicitly with no significant loss of precision. The resulting algorithm, UFFT-Grid Totalizing (UFFT-GT), achieves O(NlogN) operations and O(N) memory scaling with minimal loss in precision vs. full matrix inverse beyond numerical noise.
AB - Unified-FFT (UFFT) is a novel 3D-Planar Method of Moments (MoM) solver of the Electric Field Integral Equation (EFIE). UFFT is the resultant algorithm of combining the FFT-enhanced Matrix Solve Operations (MSO) currently utilized in Sonnet Suites with FFT-enhanced Matrix Fill Operations (MFO). It has been shown, in conjunction with iterative MSO based on GMRES, to solve matrix vector products scaling with O(NlogN) operations and O(N) memory for planar, single-plane geometries (B. J. Rautio, V. I. Okhmatovski, J.K. Lee, IMS 2013). Previous UFFT implementations have achieved O(NlogN) scaling through acceleration of iterative matrix-vector products within GMRES by separating near and far interactions and calculating far interactions implicitly as in PFFT. In this work, the same uniform grid used with MFO is extended for use with MSO, negating the need to treat near and far interactions independently and allowing the entire matrix to be stored implicitly with no significant loss of precision. The resulting algorithm, UFFT-Grid Totalizing (UFFT-GT), achieves O(NlogN) operations and O(N) memory scaling with minimal loss in precision vs. full matrix inverse beyond numerical noise.
UR - http://www.scopus.com/inward/record.url?scp=84916205227&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84916205227&partnerID=8YFLogxK
U2 - 10.1109/USNC-URSI.2014.6955454
DO - 10.1109/USNC-URSI.2014.6955454
M3 - Conference contribution
AN - SCOPUS:84916205227
T3 - 2014 USNC-URSI Radio Science Meeting (Joint with AP-S Symposium), USNC-URSI 2014 - Proceedings
SP - 72
BT - 2014 USNC-URSI Radio Science Meeting (Joint with AP-S Symposium), USNC-URSI 2014 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2014 USNC-URSI Radio Science Meeting (Joint with AP-S Symposium), USNC-URSI 2014
Y2 - 6 July 2014 through 11 July 2014
ER -