Menu

Discussion Forum

New Discussion
Filter

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Meshing of thin large metal sheet

Posted May 11, 2015, 6:45 a.m. EDT RF & Microwave Engineering, Mesh 3 Replies

Paul Antkowiak
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Hi,
I am using comsol to simulate a cubical hf-shielding chamber.
It consists of 6 aluminum panels and has an edge lengt of about 2.5m.
Since the cube is not seemless in reality, I left gaps in between the panels.
the panels are 2.5x2.5m wide and 5mm thick. My problem now is that the whole object is pretty large, while the panels are very thin. since I have to simulate eddy currents in the walls, the meshing has to be fine enough for the wall thickness but since everything else is much larger than the thickness of the panels I need a quite coarse mesh everywhere else so I can run the simulation without having to wait too long.
Do you have any suggestions how to find a good compromise?
I thought about using a swept mesh for each of the panels (triangular on the large areas and it gets quadrangular on the thin side planes) but then I could't find out how to make a transition to the tetrahedral mesh of the space around the chamber.
3 Replies Last Post May 11, 2015, 2:11 p.m. EDT
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago May 11, 2015, 7:22 a.m. EDT
Paul,

you can convert the quadrilateral elements to triangles, check the available functions in the meshing.

The other question is whether you really need to resolve the eddy currents. If the skin depth is much smaller than the 5 mm plate thickness you might consider to use impedance BCs at the plate surfaces and exclude the plate volumes from the model.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Paul, you can convert the quadrilateral elements to triangles, check the available functions in the meshing. The other question is whether you really need to resolve the eddy currents. If the skin depth is much smaller than the 5 mm plate thickness you might consider to use impedance BCs at the plate surfaces and exclude the plate volumes from the model. Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Paul Antkowiak
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago May 11, 2015, 7:45 a.m. EDT
Hello Edgar,
thanks for your reply!
Unfortunately this model is supposed to become more and more adapted to a situation with high frequency interfering fields (100Khz-5Ghz) but low frequency magnetic field pulses aswell. Also the wall thickness might get smaller later on. Therefore I have to resolve the eddy currents because one important question will be how quick they will decay.
Hello Edgar, thanks for your reply! Unfortunately this model is supposed to become more and more adapted to a situation with high frequency interfering fields (100Khz-5Ghz) but low frequency magnetic field pulses aswell. Also the wall thickness might get smaller later on. Therefore I have to resolve the eddy currents because one important question will be how quick they will decay.
Robert Koslover Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago May 11, 2015, 2:11 p.m. EDT
Some comments: Accurately modeling small imperfections in RF shielded chambers via FE is not very straightforward, in part for the reasons that you have already noted. Leakage at relatively high frequencies is dominated by gaps, cracks, oxidized/corroded (quite possibly hidden) joints, holes, etc. If you get the geometries and properties of those features wrong, (which is usually the case, since those holes/gaps are generally unplanned/accidental) then your model results will be wrong too! If the chamber is not loaded with lossy items/materials, the leaked fields will ring about inside it (if they are above cavity mode cutoffs) and there will be strong dependencies upon small features. The complexity is even worse if your chamber isn't empty. (And an empty chamber is mostly a useless chamber!) So frankly, you are unlikely to model it correctly. Also, if the shielding is fairly good (which is the usual design goal), then leaked fields will be quite weak (which is as they should be). However, this introduces yet another problem: numerical noise. Comsol Multiphysics is much better at computing accurately the stronger fields in just about any problem than the much weaker ones. If what you are looking for is, for example, 60 dB (or worse, say 90 dB) down from the numbers you are working with on the strong-field side, you may run into all sorts of nonsensical values in your outputs, due to numerical noise. This is especially problematic in time domain models. Leakage at lower frequencies, such as directly through the sealed/solid/unbroken material walls is easier to model, but at the same time, seldom of much interest, unless your chamber's objective is blocking signals such as 60 Hz background fields or similar low-frequency noise. So... what to do? Stick to modeling the things you can test and verify and for which the geometries and materials are well defined. Things like specific sizes of gaps and holes in specific materials. You can also do some models of individual cases of larger problems, paying close attention to the kinds of issues noted above, so that you don't produce a bunch of computed curves that have virtually nothing to do with reality. In other words, be very careful and don't get too ambitious or careless with these kinds of models. Good luck.
Some comments: Accurately modeling small imperfections in RF shielded chambers via FE is not very straightforward, in part for the reasons that you have already noted. Leakage at relatively high frequencies is dominated by gaps, cracks, oxidized/corroded (quite possibly hidden) joints, holes, etc. If you get the geometries and properties of those features wrong, (which is usually the case, since those holes/gaps are generally unplanned/accidental) then your model results will be wrong too! If the chamber is not loaded with lossy items/materials, the leaked fields will ring about inside it (if they are above cavity mode cutoffs) and there will be strong dependencies upon small features. The complexity is even worse if your chamber isn't empty. (And an empty chamber is mostly a useless chamber!) So frankly, you are unlikely to model it correctly. Also, if the shielding is fairly good (which is the usual design goal), then leaked fields will be quite weak (which is as they should be). However, this introduces yet another problem: numerical noise. Comsol Multiphysics is much better at computing accurately the stronger fields in just about any problem than the much weaker ones. If what you are looking for is, for example, 60 dB (or worse, say 90 dB) down from the numbers you are working with on the strong-field side, you may run into all sorts of nonsensical values in your outputs, due to numerical noise. This is especially problematic in time domain models. Leakage at lower frequencies, such as directly through the sealed/solid/unbroken material walls is easier to model, but at the same time, seldom of much interest, unless your chamber's objective is blocking signals such as 60 Hz background fields or similar low-frequency noise. So... what to do? Stick to modeling the things you can test and verify and for which the geometries and materials are well defined. Things like specific sizes of gaps and holes in specific materials. You can also do some models of individual cases of larger problems, paying close attention to the kinds of issues noted above, so that you don't produce a bunch of computed curves that have virtually nothing to do with reality. In other words, be very careful and don't get too ambitious or careless with these kinds of models. Good luck.

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.

Suggested Content