Coaxial excitaion + PML yielded divergence

[gadiLhv]

I'll start with the useless setup, then explain where it actually came from.

1. Where it Ended

Observe this minimal setup, of a short, short-circuited coaxial cable

It's excited by a "classical" openEMS excitation (with an analytical function).

Originally run with this mesh:

Now let's look at the energy decay:

Seems unnatural, right? That's because it isn't. Observe the E-field animation from this simulation.

Coax_Mesh_1mm.mp4

Now here's what I gathered so far:
This is a transverse resonance (to the width of the coax) that occurs at a very high frequency. Claude claimed to have calculated ~40 GHz. I haven't confirmed myself, sorry.

The PML, for whatever reason I am trying to gather, is amplifying the transverse mode.
Here is the mode, BTW.

Here is the script to recreate this:

coax_short_pml.py

2. Where it Started

Now I didn't start from this. I started from a patch simulation. Here is a similar example script:

coax_fed_patch.py

And there, the field values didn't explode uncontrolably, probably because the structure radiated. But maybe it would get to that, eventually

3. What's the solution?

Well, this is the simpler part. For some reason, I have'nt figured out yet, a refined mesh solves this. Probably because the more "accurate" port structure doesn't allow this mode to propagate.

Here, now the patch converges!

And the short coax.

3. Discussion

I really don't know at this point what is the exact trait of the PML that causes this to happen. I need to dive back into the PML formulations for that.

Any thoughts to share?

Now you know, yet another peril of time-domain simulation.

[thliebig]

PML is known to be inherently not stable. If you hit it with fields it does not like, this is what happens. See also the FAQ section in the docs. Not much we can do about it except use an appropriate mesh and keep fringe fields away from it.

[gadiLhv]

What's the mechanism that causes a mode with such a high cutoff to even propagate though, @thliebig ? Is it because it is a very short t-line?

[thliebig]

This very coarse mesh produces a very weird looking waveguide... not sure what happens there