Simulation does not produce meaningful results

Hi,

I am having issues with running a simulation on a model that im trying to recreate. Im trying to recreate the model in the following article. You can click the link to reach the article or you can simply download it from the attachment. I have created the model and im getting results that are not meaningful. I've only created topology A from the article. I've gone through the tutorials but i couldnt fix the issue no matter how much i tried. Can you help me fix please?

Link to the article

https://onlinelibrary.wiley.com/doi/abs/10.1002/mop.33959

1st attempt.cfx

Micro Optical Tech Letters - 2023 - Can - Dual‐band sub‐6‐GHz frequency filtering with optical.pdf

Find more posts tagged with

English

Feko

Accepted answers

All comments

Torben Voigt

Hi Ali,

I will have a look.

Best regards,
Torben Voigt

Torben Voigt

Hi Ali,

You accidentally entered MHz instead of GHz I also recommend FEM for such closed waveguides, as the simulation is significantly faster and saves more memory than with MoM, see attached model.

Note, I didn't check correctness of the dimaensions in your model, also didn't check if mesh needs to be refined.

Best regards,
Torben

test_FEM.cfx

Torben Voigt

At first glance, it looks as if the dipoles are not centered in the z-direction. Is that correct?

And I believe that the overlapping dipoles should have different lengths (L1 and L2), right?

Ali Emir Uğurluoğlu

Hi Ali,

You accidentally entered MHz instead of GHz I also recommend FEM for such closed waveguides, as the simulation is significantly faster and saves more memory than with MoM, see attached model.

Note, I didn't check correctness of the dimaensions in your model, also didn't check if mesh needs to be refined.

Best regards,
Torben

Hi Torben,

i think you are running an old version of feko. Because i get some kind of incompatibility error while trying to open the file on the attachment. Thanks for the effort though.

Best regards,

Ali

Torben Voigt

Hi Torben,

i think you are running an old version of feko. Because i get some kind of incompatibility error while trying to open the file on the attachment. Thanks for the effort though.

Best regards,

Ali

Hi Ali,

I'm using the latest release, which is Feko 2024.0.1 (Feko 2024.1 is about to be released in the next days).

Best regards,
Torben

Torben Voigt

Hi Ali,

Once you have installed Feko 2024 and updated to 2024.0.1, you should use the new attached model test_FEM_newly_created.cfx. I recreated it based on the paper and the results look promising:

Best regards,
Torben

test_FEM_newly_created.cfx

Ali Emir Uğurluoğlu

Hi Ali,

Once you have installed Feko 2024 and updated to 2024.0.1, you should use the new attached model test_FEM_newly_created.cfx. I recreated it based on the paper and the results look promising:

Best regards,
Torben

Hi Torben,

It appears that as a student ı can only access the 2023 version of feko. I wont be able to open the project you sent me. Would you please tell me the changes you ve made on my project?

Sincerely

Ali

Ali Emir Uğurluoğlu

At first glance, it looks as if the dipoles are not centered in the z-direction. Is that correct?

And I believe that the overlapping dipoles should have different lengths (L1 and L2), right?

I didnt realize that the dipoles were not centered in the z axis. Also the length of L1 and L2 are the same i think.

Torben Voigt

Hi Ali,

I just re-build the model again from scratch in Feko 2023.1.2. Please check, if length and width of the dipoles is correct. For me, the information in the paper is somehow unclear.

Best regards,
Torben

2023_FEM_newly_created.cfx

Ali Emir Uğurluoğlu

Hi Ali,

I just re-build the model again from scratch in Feko 2023.1.2. Please check, if length and width of the dipoles is correct. For me, the information in the paper is somehow unclear.

Best regards,
Torben

Hi Torben,

Your model works very well. Thank you for your time and work. I ve learned a lot from your model. I tried to mimic your model to get the same results but even so, i failed. Can you show me what i am missing? Thanks in advance.

Sincerely,

Ali

FailedModel.cfx

Torben Voigt

Hi Ali,

The first thing I notice is that there is a StandardConfiguration and a parameter configuration. Since you are actually only interested in S21, the calculation of the StandardConfiguration is superfluous and costs unnecessary time.
Pay attention to this notice:
FEM region is completely enclosed by PBC/PEC/modal port boundary surfaces
This means that you can use the "Direct Solver" for FEM, which can mean a significant acceleration.

You can also deactivate the coupling between MoM and FEM.
The medium of your sample is accidentally defined as Air, but should be PVC.
You have defined the dipoles as a volume (thickness 35 µm). In this case you should ideally also use Air (and FEM!) for the inner medium.
It is very important to pay attention to whether the outer surfaces of the dipoles are defined as "PEC" or "default"! With "default", they become dielectric boundaries as soon as you set the inner surface to Air.
If I were you, I would simply define the dipoles as PEC rectangles (as in my model 2023_FEM_newly_created.cfx). The thickness of 35 µm is thick enough that no losses due to skin effects are to be expected. You get a simpler model and faster results with PEC rectangles
I would recommend meshing the dipoles nice and fine. They are the most important elements here. In your case, it is best to set a local mesh size to the regions of the dipoles.
To avoid having to simulate an unnecessarily large number of frequencies, I recommend "Continuous (interpolated) range" instead of "Linearly spaced discrete points". This way, Feko only calculates as many frequencies as are needed to accurately map the entire frequency range.

I recorded the steps I did, so you can see how I apply the changes to your model.

Best regards,
Torben

07-08-2024_08-13-03.mp4

RepairedModel.cfx

Torben Voigt

Hi Ali,

I see now that the results in the "RepairedModel" are still significantly different from mine. I'll have to check what you did differently.

Torben Voigt

Ok, now I see that you used different lengths for the dipoles. I used "L1" (29.4 mm) and "L2" from the paper, you used 20 mm. Also for the widths of the dipoles I used "w1" (2 mm )and "w2" (2 mm) from the paper, you used 1.5 mm.

Thus, the models are different.

Ali Emir Uğurluoğlu

Hi Torben,

My model has been corrected for the most part thanks to your efforts. But one small issue is that i see a ripple other than the 2 resonant frequencies. Any idea why this happens? Your model with the Rectangular dipoles had it as well but it was relatively smaller. I sketched the same model in CST studio and the results dont show any ripples. Thanks in advance.

Sincerely,

Ali Emir Uğurluoğlu

FailedFixedModel.cfx

Torben Voigt

Hi Torben,

My model has been corrected for the most part thanks to your efforts. But one small issue is that i see a ripple other than the 2 resonant frequencies. Any idea why this happens? Your model with the Rectangular dipoles had it as well but it was relatively smaller. I sketched the same model in CST studio and the results dont show any ripples. Thanks in advance.

Sincerely,

Ali Emir Uğurluoğlu

Hi Ali,

This sometimes occurs when using "continuous (interpolated) range". It is possible that an improved mesh will fix this, but you would have to try it out. I would start by meshing the dipoles a little more finely. If you were to use "linearly spaced discrete points", you would probably not have this problem, but you would probably have to simulate several thousand frequencies...

Torben Voigt

Try this:

It works for me in Feko 2024.0.1

Ali Emir Uğurluoğlu

Try this:

It works for me in Feko 2024.0.1

I made the change and it didn't work at all. Do you have any other suggestions?

Torben Voigt

Here is my model and results. Not sure why your model behaves different, but most likely your dipoles (cuboids) require much finer meshing. As I said, I recommend rectangles.

Best regards,
Torben

test_FEM_newly_created_2.zip

Ali Emir Uğurluoğlu

I should do it with rectangles i guess. cant get rid of the ripples even testing with many frequencies. Thanks for the help Torben, you're doing god's work!

Sincerely,

Ali Emir Uğurluoğlu

Torben Voigt

I should do it with rectangles i guess. cant get rid of the ripples even testing with many frequencies. Thanks for the help Torben, you're doing god's work!

Sincerely,

Ali Emir Uğurluoğlu

Hi Ali,

For the sake of interest, I will also try it with 35 µm thick cuboids.

Best regards,
Torben

Torben Voigt

Hi Ali,

Here is my model with physical thickness. There is no ripple in the result. I think with such thin regions you have to measure accordingly fine. In my case, the dipoles are meshed with 35 µm * 20. You can investigate whether you can enlarge the mesh again until the ripples reappear. Fortunately, such a fine mesh in FEM does not immediately mean that memory and runtime explode (in contrast to MoM).

You probably know that one should always perform a Mesh Convergence Stufy to check whether the mesh is fine enough. Of course, in 99% of cases you don't feel like it and don't do it...:) If you don't want to do this manually, you can use AMRFEKO. I would recommend setting the threshold to Low. Here is a video about AMRFEKO: https://www.youtube.com/watch?v=pmKj63tsXVQ&t=597s

Best regards,
Torben

test_FEM_newly_created_physical_thickness.zip

Ali Emir Uğurluoğlu

AMRFEKO didnt work, though the ripple was smaller. The interesting issue is that i dont get any ripples at topology B where the same sided dipoles are not parallel but 45 degrees apart. You changed the meshing of dipole faces to 35e-3* 20. you mean by 35 micrometer *20 , its 35e-3 right?

Torben Voigt

AMRFEKO didnt work, though the ripple was smaller. The interesting issue is that i dont get any ripples at topology B where the same sided dipoles are not parallel but 45 degrees apart. You changed the meshing of dipole faces to 35e-3* 20. you mean by 35 micrometer *20 , its 35e-3 right?

As I said, ripples sometimes occur when continuous frequency range is used. In many cases this can be remedied by using a finer mesh (here 20* thickness of cuboids). You have to bear in mind that with your dipoles with a physical thickness of only 35 µm, there is an extremely small distance between the top and bottom of the dipoles. This can lead to the need for such a fine mesh. Apparently, interesting phenomena occur there at a certain frequency. I chose the value more or less at random, a larger mesh might also work.

With rectangles instead of cuboids, we wouldn't even be having this whole discussion

Best regards,
Torben

Quick Links

All Categories
Recent Discussions
Activity
Unanswered
日本語 (Japanese)
한국어(Korean)
Groups