Articles
EM Software Adds Non-Radiating Network Analysis Functionality To Its Software Suite
October 23, 2007
Stellenbosch, South Africa -- EM Software announced that version 5.3 of its FEKO software suite is now available. FEKO
is a comprehensive computational electromagnetics code (CEM code) used widely in telecommunications, automobile, space, and defense industries. FEKO offers several solution techniques (MoM, MLFMM, PO, GO, UTD and FEM) enabling the solution of a broad spectrum of EM problems (e.g. 3-D antenna design, antenna placement on electrically large
structures, microstrip-antennas, microstrip-circuits, EMC, biomedical, and scattering). Highlights of Suite 5.3
- General non-radiating network analysis. The combined analysis of electromagnetics (e.g. antennas) with linear circuits are now possible. Computation can be reduced greatly by breaking large problems into smaller element blocks (e.g. microstrip circuits). Network elements can be represented by S-, Z- or Y-parameter data blocks.
- Geometric optics for large dielectric objects. Electrically very large dielectric structures (>20?) can now be analysed with a ray-launching scheme, rather than with the MoM, FEM, MLFMM or dielectric PO, delivering a more accurate and better scalable solution.
- Complete redesign and rewrite of OPTFEKO. OPTFEKO is now fully integrated into FEKO GUI components allowing variable parameters in CADFEKO models to be optimised. New optimisation methods include particle swarm optimisation (PSO) and genetic algorithms (GA).
Kernel Extensions
- Waveguide port excitations with dielectrics. The existing waveguide port exciation method was extended to allow the inclusion of dielectric objects in the model, e.g. dielectric blocks in a waveguide filter.
- Lumped complex impedance for the FEM region of models. This is a "load-dual" of the line current excitation for the FEM region that is also available in FEKO.
- Polygonal plate UTD now supports three or more plates sharing a common edge. Polygonal plates could, e.g., now form a corner reflector.
Hardware Support and Kernel Optimisation
- Improved MLFMM parallelisation. Various communication protocols between process have been optimised even further, but most significantly, the computation of the SPAI preconditioner has been greatly improved with the implementation of "ghost" boxes and load balancing.
- Fast near-field calculations with the MLFMM. The traditional MLFMM implementation is used to compute currents on the structure, but MLFMM principles are now also applied to the computation of field points, rather than using classical integration routines.
- Improved memory management on 32-bit Windows systems with new DLL loading algorithms. The existing limit where approximately 1.4 GByte memory could be allocated has been increased to 1.8 Gbyte.
SOURCE: EM Software & Systems - S.A. (Pty) Ltd.

