Discussion Forum

I encourage you to attach your model to your post, so that others may open it to see for themselves what may be the source of your difficulty.

Hi, apologies, I tried to attach the file, but when I edited my post the attachment was removed. Here it is again.

Charles,

Several notes regarding your model:

1. Scattering Boundary Condition is not appropriate BC in this case.

2. You don’t need to include metal domain in the model. So, remove domain 1 from emw interface. Then Default PEC condition at the air/metal interface would represent lossless metal waveguide wall. If you want include metal losses, replace default PEC by Impedance Boundary Condition and make Copper material defined on the boundary (instead of default Domain in the Geometric entity level)

3. Set Wave excitation at Port 1 “On” (currently, you have Wave excitation at both ports “Off”).

4. Numeric port BC requires a Boundary Mode Analysis study type. It should appear before the Frequency Domain study step. Use one Boundary Mode Analysis node per Numeric port and assign each to the appropriate port. So, replace your “Step 1: Mode Analysis” by the following three
“Step 1: Boundary Mode Analysis” for Port 1
“Step 2: Boundary Mode Analysis 2” for Port 2
“Step 3: Frequency Domain”

This procedure is described in the following model:
www.comsol.com/model/download/178207/models.rf.waveguide_adapter.pdf



Regards,
Sergei

Dear Sergei,

Thanks for your reply. A few responses of my own. First of all, I have in fact implemented already mostly what you have described in another version of my model (not posted). Unfortunately, there is one fundamental problem with this method--it relies on a constant cross-section, and in fact, the Boundary Mode Analysis solves for the mode of the boundary, not the mode of the whole waveguide, as I understand it. Certainly, since it is a 2D solver, it couldn't possibly find a mode for an axially dependent geometry. Therefore, this method won't be valid for modeling a helix, which as I said is my primary interest in first modeling a cylinder. Additionally, it doesn't seem to accurately solve the modal analysis, based on a textbook analysis of a cylinder, there should be a donut-shaped mode above the second cutoff frequency. Comsol does not find a different mode until above the 3rd or 4th cutoff frequency. Again, if I did not know these modes in advance, then I would have no confidence (based on this performance) that it found the first few modes.

Re 1: There should be nothing wrong with an SBC here, if any field makes it outside of the waveguide, I want it to dissipate, exactly what an SBC does.

Re 2: While I don't "need" to include metal, I want to, because I will have to for future models (though I will likely abandon COMSOL for open source software anyways). Just because I don't "need" to include metal does not mean that the physics should be any different if I do include metal. I wish to model a geometry where the waveguide must be represented as a structure and not just as a boundary condition.

Re 3: Using the MODE ANALYSIS study node, I tried all different configurations of port boundary conditions--including having SBC's instead of ports. This comment is specific to the boundary mode analysis problem, which as stated above is not so useful here.

Re 4: See initial paragraph.

Thanks for your response, Sergei, but unfortunately it doesn't answer my primary question: is COMSOL even capable of solving for a propagation mode of a truly 3D geometry? It claims to be able to, but it seems to depend on essentially treating it as a 2D problem, rendering the solutions pretty untrustworthy.

Best,
Charlie