Discussion Forum

I just took a brief look. Hmm. I would attack this problem somewhat differently, but I'll offer you a couple really quick & easy suggestions for now: (1) you have a symmetry in the x=0 plane, right? So I suggest you take advantage of it to cut your problem space in half there (with a PMC boundary condition on the symmetry plane). (2) Also, go to the Settings section, Discretization, and replace Quadratic by Linear. This will enormously reduce your memory requirements and will also increase the speed of the computation. I suggest you do those things first, then let us all know how much that helps and if you still want to discuss other possible issues in your model! Added: I just noticed that you've got all your elements driven together, as if they were acting in unity, as one single port. But there is insufficient symmetry to cause them to truly act identically, so it seems to me you are attempting to enforce an inappropriate boundary condition on those ports? If I were preparing the model, I would specify a separate port for each element. This change certainly can't hurt, and it just might help. Also, I think you might want to use a finer mesh in/on the antenna elements themselves. But... one step at a time, I suppose.

Thank you for your valuable feedback.

I'll put into practice your tips. Should I have any other issue, I'll let you know.

Good afternoon,

After switching my discretization to linear and having implemented a PMC as suggested, I managed to reduce the computation cost to less than 10 seconds, and I'm really thankful for that.

There's just a couple of things.

Firstly, my far field pattern was kind of distroyed, as I no longer have a directional pattern, but a ball with spikes, I don't know up to what point is that important.

Secondly, I tried to separate the excitations in different ports and it gave me errors of "must be placed between two different PECs", so I'm keeping them in the same port at the moment.

Despite the previous points, I still regard the results obtained with the electric field values as acceptable in the multislice section, idk.

Pablo

1. In regard to the ports, I suggest you define the metal-material properties in the vicinity of your ports to be PECs. This should fix your port problem (this is a proven solution, by the way, for that issue.) You can generally join such surfaces or domains to non-PEC materials of your choice, without issues.
2. One reason a radiation pattern may be spikey is that you don't have enough points in your far field plot. But in your case, also:
3. Your cell sizes in the radiated field region are far too large. Don't make any cell bigger than roughly lambda/6. Your cell sizes in the radiation volume are ~ lambda??! If so, that will surely generate nonsensical results for computed radiation patterns. Consider modeling a smaller test volume and use an appropriate max mesh size. Only after you get that to work well, and you get all the bugs out of it, then consider increasing your model size.
4. One more thing. Your far-field aperture integration computation is also specified incorrectly. Now that you have the PMC symmetry plane present, you must tell the far field computing algorithm that there is a symmetry plane present, and what kind of symmetry to apply. In the Model Builder, go to: Electromagnetic Waves, Frequency Domain, then click on Far-Field Domain 1, then click on Far Field Calculation 1. Now, in Settings, expand the Far-Field Calculation item, and click the checkbox "Symmetry in the x=0 plane". The symmetry type: should show "Symmetry in E (PMC)". (If it shows something else, change it to that one.)

Ok, I made it finally with a lambda/3 size after five hours

Lambda/6 overflows my memory.

I think that's all for now.

Thank you so much!

Pablo