Adaptive Cross Approximation (ACA) integration into existing MoM codes


Authors:
Juan M. Rius with J. M. Tamayo, A. Heldring, J. Parrón and E. Ubeda

The authors are (or have been) with AntennaLab,
Dept. of Signal Theory and Communications,
Universitat Politecnica de Catalunya, Barcelona, Spain.
e-mail: rius@tsc.upc.edu

ACAsolver:
ACAsolver is a fast iterative solver for compressible linear systems, with multilevel Adaptive Cross Approximation (ACA) matrix compression. ACA is based on the fact that interaction between far-field spatial subdomains has a reduced number of degrees of freedom. Although the solver has been developed for Electromagnetic Integral Equation problems discretized by Method of Moments (MoM), it can be applied to any physics or engineering linear system of equations with a compressible matrix that contains rank-deficient pseudo-submatrices.

We believe that software integration between University research groups is the key for successful analysis of very complex antenna configurations, that cannot be handled yet by commercial Computational Electromagnetics (CEM) packages. The integration effort started in Antennas Center of Excellence (ACE) network can be the seed of a future Linux-like open library of CEM engines and solvers.

Adaptive Cross Approximation (ACA):
Among many matrix compression algorithms that have been used by the CEM community to accelerate the solution of MoM linear systems, we have chosen ACA because it is -to our knowledge- the simplest to integrate with existing MoM codes.

Selected key references:

Downloads:
ACAsolver 1.0 with MoM-2D EFIE-TM example (MATLAB)
ACAsolver 1.0 with MoM-3D example (MATLAB)
ACAsolver 1.0 user guide

Planned next release:
Pyhton-scipy version of ACAsolver 1.0.

License terms:
ACAsolver is free-source software licensed under the GNU General Public License
ACAsolver license


ACAsolver block diagram