Weird behavior in PML

[SengerM]

I am trying to simulate the propagation of a wave along a PCB track. This is a preview of what I want to do:

Everything seems to proceed well while the wave is still "far away" from the end of the track:

Then, this happens:

What could be the reason for this? (Before this, I run the same simulation but with a straight track (i.e. without the zig zag) and everything seem to have worked perfectly.)

CSX_model.zip
source.zip

[VolkerMuehlhaus]

Is this using your modified version of openEMS? It seems that it doesn't run on my installation.
Your XML can't be opened because it links to external files, with filenames hard coded in your local file system. Overall, you make it really hard to support you :-/

Just a general thought: how about convergence? Does your residual energy increase towards the end of your simulation?

[thliebig]

I guess you are using a PML on the right side to terminate the line? A PML8 is 8 cells thick, thus I assume your bend is too close to it. PML's are not unconditionally stable and really do not like reactive near-fields. If you make your line longer and thus the bend further away everything should be stable... Otherwise you try a lumped port on the right just the same as on the left?

[SengerM]

Thanks for the explanation, really appreciate it. Indeed I have a PML_8 on the 6 walls, and the track (and the rest of the PCB) goes straight into the PML in the right side.

If you make your line longer and thus the bend further away everything should be stable

You mean the 3D parts to extend further away from the mesh, like this? When you say "the bend", do you mean the edge of the 3D parts?

[VolkerMuehlhaus]

Try to make the straight line segment longer on the right side, where it hits the PML.

[SengerM]

I made a few tests and this is what I found for this specific case:

1. thielbig was right in his answer, the problem seems to be related to instability in the PML.
2. Making the PCB extending beyond the PML did not solve the issue.
3. ✅ Making the PCB longer than the PML and removing the vias as shown in the image below solved the issue.
4. Removing the vias and not extending the PCB longer, I did not try.

As a beginner, what I conclude is that in the PML it is better not to have sharp things.

[VolkerMuehlhaus]

3. removing the vias as shown in the image below solved the issue

So you had vias in those outer 8 cells where the PML8 is created?

[SengerM]

Yes, I had. I also have in the other simulations that are working, maybe they are slightly shifted such that they work. It is hard to tell how they intersect the mesh lines because I don't know how to pan the view in AppCSXCAD after having zoomed in.