Radiation boundary of HFSS

source: http://www.edaboard.com/thread44722.html

Questions on the radiation boundary of HFSS

I am using HFSS to simulation a patch antenna with a lumped port. I know the rule of setting up the dimension of radiation boundary to be more than lamda/4. However, I found out that with different boundary sizes, the simulation results are different. What is the problem? How can I get consistent results?


Answers
To evaluate the size of radiation boundary with the simulation accuracy, you may keep on increasing the radation boundary size (until the HFSS breakdown), until the simulated results are converged, hereby you can confirm the simulation is accurate enough. We have to introduce radiation boundary to truncate the infinite space into one confined simulation space because it is FEM based. Normal, the nearest distance between your simulation strucutre and the boundary of radiation box should be lamda/4, where the lamba is the longest wavelength of your simulation frequency band.

In simulation of antanna structure specially p@tch antenna analysis converge slowly to the answer per pass.

so first, change pass number in analysis setup to about 20 pass. (It's very important)

second, if the result is not converge te desired solution enlarge radiation box.

third, if this to operation is not enough use PML boundary to achive the best result.

Thank you so much on your great help. I changed it to 20 times, now the result converged, it is great. However, I have another two questions:
1) what is PML?
2) If I kept increasing the boundary size, the resonant frequency will not change, but I can get a better performance (s11 is smaller). So what kind of performance index can I claim to my antenna?

Perfectly Matched Layer (PML) is a fictitious boundary condition.
it's a kind of abserbing boundary condition with lossy & unisotropic meterial.
its parameter is calculated so that it has no reflection theoretically.
In HFSS if you wanna near radiation box to the structure smaller than lambda/4 or you want to have more accurate results you can use this kind of boundary condition. so first select external box faces and next go to HFSS menu, then Boundaries, and then PML Setup Wizard to set its parameters.

to have a better response with less passes and a boundary at lamg/4 ( because greater boundaries causes much more time for simulation) , you can use mesh refining. you should refine the meshes on the structure and the meshes on the boundary. like that, your meshes will remain the same, no matter how far the boundary is. each pass will take more time (cause the number of meshes are greater) but it will converge very soon.

However, could you tell me how to refine the mesh? Thanks,

You should always seed the radiation boundaries (~lambda/10) before simulation.

in order to reduce the problem size you can use symmetry E and H planes.

you have the mesh refining option in hfss. you could choose to refine it on a surface or inside a structure. as "savedadogs" said, lambda/10 is ok. but i usully seed the meshs on the structure and lower plate of my radiation box.
and by using pml at near lambda/7 or so, you'll save a lot of time in simulation.