Multi-Reflection Simulations in Dual-Foci System

artgru

I aim to perform simulations of both far-field and near-field with multiple reflections for a system featuring two focal points. How should I select the solver method for each element of the system across different frequencies ranging from tens to hundreds of GHz, in order to capture reflections in the system at the output of the last element and also to inspect the reflection at each of these elements? I understand that this is an extensive project and likely dependent on the computational power of my computer. However, it would be beneficial to have a procedure for selecting methods, starting from the simplest with low accuracy to the most accurate one.

Torben Voigt

Hi Artur,

(I expected that you would create a support ticket, but this is also ok I guess)

Could you attach the model you're thinking of? Sounds like electrically large (several wavelengths), so I would suggest MLFMM.

On the other hand, you write "frequencies ranging from tens to hundreds of GHz", which makes MLFMM difficult again, as the mesh at the highest frequency is much too small for the low frequencies. You should then split the model so that fmax/fmin is not larger than ~3 (e.g. 10 GHz-30 GHz, 30 GHz-90 GHz, 90 GHz-270 GHz, ...).

For reflectors PO or LE-PO could also be a very good choice.

It would be really good if you had an example model.

Best regards,
Torben

artgru

Hi Torben,

Although in this case, I took a random (6MHz) sample, and I am attaching the diagram to this message. In subsequent setups, let's call them advanced, there may be more optical reflecting elements, as I learned from the team I work with. In short, I chose an example of reflection from CADFEKO. Next, I rotated one reflector by 45 degrees, copied it, and made a mirrored reflection of the second reflector onto the first one. I didn't know that assigning the LE-PO method to the reflector would result in a lack of reflection on the second reflector. As I later found out, for reflection to occur, the method for the first reflector should be like the MLFMM antenna structure, and the last reflector should have LE-PO. Now the question is whether this means that when I add many other reflector s as part of the optical path, all the way to the last one, they should all have MLFMM and the last one LE-PO. Here I wonder what the selection procedure looks like.

I will also note that in the extended version in the future, I plan to replace the reflectors from parabolic to ellipsoidal, and should I also change the solver method. Let's say ultimately they will be in the GHz range, but I will attach this sample project here. I will add that it does not have perfectly preserved angles and distance relationships. It is just an example on which I am trying to practice and develop, reaching advanced systems that we will design in the team. Would the mesh selection matter because in many cases, we wouldn't need to apply super accuracy?

If it comes to any other examples, I will send them later...

If it helps, the machine for this has an NVIDIA RTX2080Ti card and 64GB of RAM.

Best regards,

Artur 

Torben Voigt

Thanks Artur,

I will have a look asap. You're right, with PO there will be only one reflection (more are possible, but extremely time consuming), so if reflector 1 is solved with PO or LE-PO none of the reflected field will reach reflector 2. This is also part of today's training session, by the way. 

For further questions in the future, I would really recommend creating support tickets on Altair One.

Besr regards,
Torben

Torben Voigt

Hi Artur,

The horn antenna in your model is 222 m large... Should this be mm maybe?

Can you please check and rebuild if needed?

Best regards,
Torben

artgru

Torben,

This is just an example I wanted to check and learn how it works, disregarding distances. Later on, the distances will be on the scale of a laboratory table. I think that once I have specific examples with the correct distances and absorb the knowledge from the recent days of lectures. My future inquiries will be more precise. 

I just wanted to know if there is a simple way or a specific methodology to indicate, without greater accuracy in the final result of simulated fields, that such a solver method at such frequencies and distances.

It turns out to be a bigger challenge because mesh density or is also important, not just that.

At the moment, time doesn't concern me, even if it requires 3 days of calculations, as long as there is a simple methodology for which solver methods to use, that's fine at the beginning. 

I'm glad you've shown how to get in touch with you (and your team) in case of further challenges on the road to total success.

Best regards,

Artur

Torben Voigt

Hi Artur,

Without further investigation today (no time) I would recommend MLFMM. If you want to reduce computational requirements you may use PO for the last reflector. Also you could replace the horn by an equivalent spherical modes source (like I did today in the training).

With the spherical modes source you could then potentially use RL-GO for the whole setup.

Best regards,
Torben