Discussion Forum

I very rarely used the temw mode so I had forgotten there was no selection for an imperfect conductor.

You will be able to obtain some insight using the perfect conductor boundary condition. However you should consider doing a frequency-dependent calculation which contains complementary information. In principle it would be possible to use those results to perform an inverse Fourier transform to get the time response.

Actually, I need to see the transient behavior of EM waves over time to see how they propagate. A frequency-dependent solution might not be suitable for my case study.

David suggests to perform a sufficiently broad and fine frequency scan and do a an inverse Fourier transform to obtain the time dependency. You will need to make yourself familiar with Fourier transform theory. The window function which is the Fourier transform of the time dependent excitation is key. It works, I have done it in acoustics models. But it is not exactly plug and play.

Cheers Edgar

In principle, you can also model the walls (to perhaps a couple of skin depths?) as fully meshed domains within a time domain model, since the temw formalism will allow you to specify a finite conductivity of a domain. This may or may not be practical computationally in your model. If you do take this approach, be sure to define the mesh of the lossy wall volumes carefully, so as to properly model the field gradients in the vicinity of the skin depth, yet not run out of memory. This will be easier if you can do a 2D model instead, or make any other use of symmetry.

Is there any example or article about this principle? Actually, I a little bit confused about your explanations.

See the example attached. This example is not optimized for any purpose in particular. A single cycle of RF is launched by a post inside a square cavity, and then it bounces around in the cavity. There is one lossy domain (note that sigma = 1.0 S/m in the rectangle on the right-hand side). This is only a 2D model. I included a probe to show how the energy in the cavity varies with time (see the Results section, to plot that). All that said, I encourage you to follow the suggestions made by David Greve and Edgar Kaiser above, if you can! They know what they are talking about.

Very nice example Robert!

Cheers Edgar

The finite-conductivity approach suggested by Robert will probably be most accurate for a limited range of frequencies, where the skin depth is of the order of the mesh size. So if the exciting waveform is nearly a delta function (with a wide spread in the frequency domain) this would not be the best approach. On the other hand a windowed sinusoid would work very well with the finite-conductivity wall.

If OP chooses to mesh the wall it will be necessary to be very judicious with the approach used. Probably a swept mesh is the way to go.

If the holes are large enough (so that propagation outside is dominant) it may not even be necessary to model dissipation within the box.

Thanks Robert. Actually I defined four walls made of copper (not lossy) and put a air gap on the surface. EM wave should be spreads out from this gap but it doesn't work. I attached the file. I appreciate it if you take a look of it.

The gap is leaking radiation, it is just not very well visible in the linear plot. I often plot log10(normE) to visualize fields over a large range.

I made a little animation that shows the effect.