How to fix "Error 32536: Losses are too large when using dielectric bodies with the MLFMM"？

Yufeng Cai

While running the model, I am getting  errors as "Error 32536: Losses are too large when using dielectric bodies with the MLFMM". But when I change the dielectric medium from  conductivity=1000 to loss tangent=0.003, the error disappeared. Could you please tell me what does it mean ? And how to clear this error?

Torben Voigt

Yufeng Cai said:

Considering the confidentiality, my model may not be able to share in the altair community. I have tried to set the region as free space, then set metallic on its surface, and finally run successfully with the MLFMM solver. Do you think this is right?

Hi @Yufeng Cai ,

In fact, this idea also came to me, but only after I wrote. I think this is the right approach. With a conductivity of 1000 S/m, the material actually has more metallic than dielectric properties. Accordingly, the currents should also be on the surface, which makes the approach plausible.

Best regards,
Torben

Torben Voigt

Hi @Yufeng Cai ,

a conductivity of 1000 S/m corresponds to very high dielectric losses, while a value of tand = 0.003 corresponds to rather low dielectric losses.

Example: If you convert conductivity to loss tangent and vice versa (can be done using an appilication macro in CADFEKO) at 1 GHz:

sigma = 1000 S/m -> tand = 4493.8
tand = 0.003 -> sigma = 0.00066759

 

Best regards & a Happy New Year,
Torben

Yufeng Cai

Torben Voigt_20420 said:

Hi @Yufeng Cai ,

a conductivity of 1000 S/m corresponds to very high dielectric losses, while a value of tand = 0.003 corresponds to rather low dielectric losses.

Example: If you convert conductivity to loss tangent and vice versa (can be done using an appilication macro in CADFEKO) at 1 GHz:

sigma = 1000 S/m -> tand = 4493.8
tand = 0.003 -> sigma = 0.00066759

 

Best regards & a Happy New Year,
Torben

Thank you for your reply! In my simulation, the parameter of this dielectric is ε_r=4, and the conductivity is 1000 S/m. How can I define this material in FEKO and which solver should I use?

Torben Voigt

Yufeng Cai said:

Thank you for your reply! In my simulation, the parameter of this dielectric is ε_r=4, and the conductivity is 1000 S/m. How can I define this material in FEKO and which solver should I use?

Hi @Yufeng Cai ,

As you know, MLFMM is recommended for electrically large models (>4 lambda). Maybe you could simulate your model with standard MoM?

Best regards,
Torben

Yufeng Cai

Torben Voigt_20420 said:

Hi @Yufeng Cai ,

As you know, MLFMM is recommended for electrically large models (>4 lambda). Maybe you could simulate your model with standard MoM?

Best regards,
Torben

But my model is electrically large, it's hard for me to simulate with MoM. Thanks for your replying！

 

 

Torben Voigt

Yufeng Cai said:

But my model is electrically large, it's hard for me to simulate with MoM. Thanks for your replying！

 

 

Hi @Yufeng Cai ,

I first thought of the "Dielectric surface impedance approximation", which was implemented especially for dielectrics with particularly high losses. But it seems that the conductivity of 1000 S/m is much too low even for that approach...

The only other thing you could try is MoM with higher order basis functions. This is another method besides MLFMM to reduce memory requirements for electrically large models. I'm not too optimistic, but it's worth a try.

Would it be possible to attach your model here?

Best regards,
Torben

Yufeng Cai

Torben Voigt_20420 said:

Hi @Yufeng Cai ,

I first thought of the "Dielectric surface impedance approximation", which was implemented especially for dielectrics with particularly high losses. But it seems that the conductivity of 1000 S/m is much too low even for that approach...

The only other thing you could try is MoM with higher order basis functions. This is another method besides MLFMM to reduce memory requirements for electrically large models. I'm not too optimistic, but it's worth a try.

Would it be possible to attach your model here?

Best regards,
Torben

Considering the confidentiality, my model may not be able to share in the altair community. I have tried to set the region as free space, then set metallic on its surface, and finally run successfully with the MLFMM solver. Do you think this is right?

Torben Voigt

Yufeng Cai said:

Considering the confidentiality, my model may not be able to share in the altair community. I have tried to set the region as free space, then set metallic on its surface, and finally run successfully with the MLFMM solver. Do you think this is right?

Hi @Yufeng Cai ,

In fact, this idea also came to me, but only after I wrote. I think this is the right approach. With a conductivity of 1000 S/m, the material actually has more metallic than dielectric properties. Accordingly, the currents should also be on the surface, which makes the approach plausible.

Best regards,
Torben

Yufeng Cai

Torben Voigt_20420 said:

Hi @Yufeng Cai ,

In fact, this idea also came to me, but only after I wrote. I think this is the right approach. With a conductivity of 1000 S/m, the material actually has more metallic than dielectric properties. Accordingly, the currents should also be on the surface, which makes the approach plausible.

Best regards,
Torben

Thanks for your answer！