Characteristic Mode Analysis of a coaxial-probe fed microstrip patch antenna on a finite PEC-backed lossy dielectric

AGirlHasNoName

Hi,

I want to analyze the modal excitation of a coaxial probe-fed rectangular microstrip patch antenna on a finite circular-shaped PEC-backed TMM-10i substrate (please, see the attached .cfx file). However, every time I try to simulate the structure, I am getting the same error message saying that the characteristic modes were all non-radiating.

Could you please guide me on how to successfully use the TCM solver in FEKO?

Additionally, I had to make the dielectric substrate lossless (loss tangent = 0). Is there a way to use the TCM solver for lossy dielectrics?

Thank you in advance. I am eagerly looking forward to suggestions.

 

Mel

It appears that you are using Feko 2019.0. This error does not occur in later versions, i.e. 2019.3. But it is strongly recommended to upgrade to Feko 2021. Unless you attached a different model here?

But you are correct in that lossy media are unfortunately not yet allowed.

Note that usually a finer mesh is required to accurately model the higher order modes.

Not sure if it was intentional, but if you only want to do characteristic modes, you do not need the StandardConfiguration:

 

 

AGirlHasNoName

Mel,

Thank you for your earlier response.

Since, I don't have the license for feko-2021, I re-modeled the structure using feko-2020.1. This time, I incorporated the following changes:

1. Increased the number of frequency points from 100 to 500 (between the start and the end frequencies).

2. Created a finer mesh.

3. Requested more number of modes (over 30).

 

However, none of the modes were computed over the entire frequency range and it asked me to request more number of modes. The problem was not resolved even after requesting 100 modes! How can I solve this issue? I shall be looking forward to your suggestions.

Thank you in advance.

Mel

The number of modes requested refers to the number of modes computed at each frequency. Feko tracks the modes in that the label (mode number) for a given mode is maintained across all frequencies. However, only the most significant modes for each frequency are computed, which can change with frequency. As the modal significance decreases, the mode may no longer be computed and would appear to abruptly stop and a new mode would take its place. This is why a sufficient number of modes should be requested in order to track the modes from the lowest frequency.
However, the higher the number of modes requested, the smaller the required mesh size will be to accurately model these modes.

With such fine meshing required, it is recommended to use double precision since the MoM matrix becomes increasingly singular with finer meshing. In some cases the default dielectric method (Surface Equivalence) could suffer from "non-radiating" modes. ("Observations on Computational Outcomes for the Characteristic Modes of Dielectric Objects", H.Alroughani, J.L.T.Ethier and D.A.McNamara)

You could change the dielectric method to Volume Equivalence:

And use double precision for further stability:

I have run the model with VEP and a mesh size of about lambda/13 in the dielectric (lam0/40), over 101 frequencies, and the first few modes seem ok - mode 3 "stops" but again, this could be due to meshing. Unfortunately finer meshing causes a high computational load for the VEP compared to SEP.