Any FEKO user's who've done BIG geometry?

Hi All,

I am trying to simulate in FEKO, at 2-18GHz in 2GHz steps, an 8foot diameter ground plane which is not flat (so I cannot use UTD-unifrom theory of diffraction for flat electrically large pec faces). I am using a monopole and line port at the center and trying to also use MLFMM.

I have alot of resources on my workstation but it keeps running out when I run at frequencies about 8GHz. I do know that others HAVE been able to simulate this up to 18GHz.

When I ran with symmettry it told me that there were problems using the line port at the center, I should define the port symmetrically(?), so i'm not sure how to fix that.

I think symmetry (electric) may be the only way to save enough resources to run this problem?

Any help fixing this symmetry would be great, thank you!

~Nick

How many triangles is in your model? Why simulation depends on frequency ou are using - problems only over 8 GHz? - Do you have different meshes for different frequencies?

At 8GHz I have ~254,000 triangles. My mesh does depend on my frequency as it is a portion of the wavelength. This means the mesh triangles get smaller and smaller as the frequency gets larger. I am trying to compensate for this by cutting the number of triangles down using symmetry.

In the Help Centre on the FEKO website there's some tips on how to reduce the memory for the FMM

Try to use also mesh from 8 GHz for higher frequencies as well. Yes, you will have some accuracy leak, but probably it can be 10%-15% and you might accept it?

You might also probably use non-uniform mesh - depending on the currents distribution on the surface - if they are concentrated mainly in some active region you might probably reduce number of triangles in the model.

I found a better solution: Under the MLFMM tabs I set the box size to 0.2(wavelengths) and used the Sparse Approximate Inverse (SPAI) preconditioner to reduce memory requirements. I was able to run up to 14GHz so far, though these remain very slow and large problems.