🎉Community Raffle - Win $25

An exclusive raffle opportunity for active members like you! Complete your profile, answer questions and get your first accepted badge to enter the raffle.

Join and Win

Near-Field Planar Characterization Simulation of a Rectangular Horn Antenna with a probe - FEKO

Eros · 2025-03-11T11:52:46+00:00

There was an error rendering this rich post.

Eros

Hello everyone,

I am currently facing difficulties while performing a simulation of a typical near-field planar characterization. Particularly, I am considering the case of a rectangular horn antenna using an Open-Head WG as the probe. Here's a detailed description of my setup and the issues I'm encountering:

I have created 3D models of both the horn antenna and the probe.
I developed a macro that requests the SParameters to consider S21 or S12 as the open-circuit voltage measured in an anechoic chamber.
Through the macro, the simulation process involves:
1. Running the simulation for a given probe position to get the SParameters.
2. Moving the probe to the next measurement point.
3. Re-running the simulation to obtain the updated SParameters.
This process enables me to perform a scan over a planar grid (as for a standard uniform half a wavelength field sampling) getting as many .s2p files as the points of the grid.

High Computation Time:
- The simulation is computationally expensive because it generates a new mesh for each acquisition point.
- This occurs despite using a GPU and attempting to optimize the process with the MLFMM (Multilevel Fast Multipole Method).
Noisy Results:
- The results obtained within reasonable computation times (<3 minutes per point) are noisy.
- Levels at -30 dB compared to the peak of S21 or S12 are highly oscillatory and asymmetric concerning the center of the scanning domain.

Is there a way to reduce computation time and improve the accuracy of the results? Any suggestions or alternative approaches would be greatly appreciated.

Thank you in advance for your help!

Product/Topic Name : FEKO, CADFEKO

Find more posts tagged with

Feko

How To

Troubleshooting

Performance

Accepted answers

All comments

Torben Voigt

Hi Eros,

Could you attach your base model, so I can have a look? Maybe I find things to improve.

Best regards,
Torben

Eros

Hi Torben.

Yes I can. Thank you very much for your time and interest.

Kindly,

Eros

try3.cfx

Eros

https://community.altair.com/discussion/comment/191089#Comment_191089

Hi Torben.

Yes I can. Thank you very much for your time and interest.

Kindly,

Eros

try3.cfx

Torben Voigt

Hi Eros,

The frequency is not defined in your model. What is it?

Best regards,
Torben

Torben Voigt

Ah, found the variable.. Sorry!

Torben Voigt

Hi Eros,

Since the model is larger than 3 x lambda you shold definitely switch to MLFMM. ALso I don't see a reason for double precision (typically only used for very low frequencies).

GPU is not recommended if you have more than 1 CPU core (which you probably have).

On my machine with 6 cores the simulation takes around 10 seconds.

try3_alt.cfx

Best regards,
Torben

Eros

https://community.altair.com/discussion/comment/191118#Comment_191118

Dear Torben,

Thank you very much for your help and the information you provided - it’s been really helpful.

However, I had already attempted this setting initially, and the computational time issue arose when I switched from “single-point acquisition” to “multi-point acquisition”, moving the probe across different sampling points.

What I observed is that when the probe is repositioned, all the precomputed matrices become ineffective for calculating the new S-parameters. Moreover, "measuring" along the Y direction, the values are strongly asymmetrical with respect to the value at the center (Y=0) and oscillatory around -30 dB compared to the peaks of S21 and S12. I have used the MLFMM and a mesh of lambda/20.

To provide more clarity, I have attached an image showing the values of the S12 in dB.

Thank you again for your support.

Best,

Eros

9.jpg

Torben Voigt

Hi Eros,

Yes, when moving some part of the geometry it results in a completely new model and the whole simulation will be performed again. That't the same for MoM and MLFMM.

As far as the asymmetry is concerned, please note that we are talking about values in the -30 dB range, i.e. extremely small numbers. If a mesh part is moved, it will be re-meshed, and the resulting mesh may not always be exactly the same. If the mesh does not remain identical, you will always see slight asymmetry in the results. I don't remember what it was like in Feko 2022, but maybe you can lock the mesh?

Best regards,
Torben

Eros

https://community.altair.com/discussion/comment/191136#Comment_191136

Hi Torben,

Thank you again for your support. I have actually resolved the issue by replacing the horn with an aperture source.

Best regards,

Eros