How to make the array simulation faster?
I am simulating 25x25 dipole array in feko and I notice that this simulation takes more than a day to simulate and feko eventually hangs up. How can I hasten the process? Each of the element in the row is excited with a different amplitude and phase, this repeats with other 24 rows.
Edit: I have added the model file
Answers
-
Hi Niharika,
that sounds strange. I made a test with an 25 x 25 dipole array and runtime was 14 seconds on my laptop with Feko 2023.1. You may create an Altair support ticket and share your model there.
Kind regards
Christoph
0 -
Christoph Mäurer said:
Hi Niharika,
that sounds strange. I made a test with an 25 x 25 dipole array and runtime was 14 seconds on my laptop with Feko 2023.1. You may create an Altair support ticket and share your model there.
Kind regards
Christoph
in my simulation, the dipole array is not uniformly excited. Can that cause such issues?
0 -
No, that should not be the reason. Also in my test example I defined a non-uniform excitation, using the "generate antenna array" in the application macros.
0 -
Christoph Mäurer said:
No, that should not be the reason. Also in my test example I defined a non-uniform excitation, using the "generate antenna array" in the application macros.
I even tried the MLFMM solver, but seems like that does not work and throws an error.
0 -
Hi Niharika,
MLFMM should work for such an array, too. And for most large arrays it is even the better solver option. Please share the model (and error message) with Feko support to check what is going wrong here.
Kind Regards
Christoph
0 -
Christoph Mäurer said:
Hi Niharika,
MLFMM should work for such an array, too. And for most large arrays it is even the better solver option. Please share the model (and error message) with Feko support to check what is going wrong here.
Kind Regards
Christoph
This is the model. I have not used Mlfmm for this version of the model
0 -
I think your problem is related to failing convergence of the adaptive frequency interpolation. I recommend to use discrete frequency steps.
Additionally I would remove the S-Parameterconfiguration (because this request is already in the Standard Configuration), choose double precision and at this high frequency a finer resolution for the far field pattern request is necessary. I got results with MoM and with MLFMM (but I tested only for one frequency)
The pattern of the result looks really nice!
1 -
Christoph Mäurer said:
I think your problem is related to failing convergence of the adaptive frequency interpolation. I recommend to use discrete frequency steps.
Additionally I would remove the S-Parameterconfiguration (because this request is already in the Standard Configuration), choose double precision and at this high frequency a finer resolution for the far field pattern request is necessary. I got results with MoM and with MLFMM (but I tested only for one frequency)
The pattern of the result looks really nice!
A quick follow up question, did you perhaps change the excitation for the array? I followed the steps you asked me to do and I get a different pattern for the excitation I have but it really did help in faster simulation.
0 -
Maybe we have different frequencies. I created the pattern at 50 GHz0
Altair Employee