DocumentCode
2768950
Title
Application of the tree-cotree splitting for improving matrix conditioning in the full-wave finite-element analysis of high-speed circuits
Author
Lee, Shih-Hao ; Jin, Jian-Ming
Author_Institution
Dept. of Electr. & Comput. Eng., Univ. of Illinois at Urbana-Champaign, Urbana, IL
fYear
2008
fDate
5-11 July 2008
Firstpage
1
Lastpage
4
Abstract
In recent years the development of high-speed circuits leads to a demand in accurate and reliable full-wave electromagnetic analysis from high frequencies all the way to DC, and thus brings a challenge to current full-wave techniques. A full-wave solver yields solutions with deteriorated accuracy or even breakdowns at low frequencies due to ill-conditioned system matrices. The same situation also happens when very small elements (compared to the wavelength) exist in the finite-element discretization, which is similar to a low-frequency case [1]. This is actually very common in circuit problems since there are very dense meshes both inside and outside conductors for modeling the conductor loss and the field singularity, respectively. In this case an element may not only be very small, but also have a high aspect ratio, which make a system matrix very ill-conditioned even at not too low frequencies.
Keywords
finite element analysis; high-speed integrated circuits; trees (mathematics); waveguides; finite-element discretization; full-wave electromagnetic analysis; full-wave finite-element analysis; high-speed circuits; matrix conditioning; solution space projection; tree-cotree splitting; Application software; Circuit analysis; Computational electromagnetics; Conductors; Electric breakdown; Electromagnetic analysis; Finite element methods; Frequency; Maxwell equations; Space power stations;
fLanguage
English
Publisher
ieee
Conference_Titel
Antennas and Propagation Society International Symposium, 2008. AP-S 2008. IEEE
Conference_Location
San Diego, CA
Print_ISBN
978-1-4244-2041-4
Electronic_ISBN
978-1-4244-2042-1
Type
conf
DOI
10.1109/APS.2008.4619419
Filename
4619419
Link To Document