Request for Assistance with Feko Simulation Errors: ERROR 33052 & ERROR 39518

NoYa
NoYa Altair Community Member
edited October 28 in 質問と回答 (Q&A)

I am encountering issues while running simulations for dipole antennas at 100 MHz using Feko, and I would appreciate your assistance in resolving them.

1. Basic Cylinder Dipole Antenna
The 100 MHz calculation for this antenna is performed using MLFMM (Multilevel Fast Multipole Method) to reduce computation time. However, the following error is displayed:

ERROR 33052: Internal Feko error.
This error prevents the simulation from completing, and I have checked the output file cylinder_dipole_100MHz.out for additional details, but I have been unable to identify the cause.
2. Full-Sheath Dipole Antenna
For the full-sheath dipole antenna, I receive the following error:

ERROR 39518: An edge port may not be defined on the surface of a dielectric body.
I tried setting the surface of the edge port as PEC (Perfect Electric Conductor) on the dielectric boundary, but this did not resolve the issue. I would appreciate any guidance on how to proceed.
Could you please provide insights or recommendations to address these issues? The relevant output files are available if needed.

Best regards,
NoYa

Answers

  • Torben Voigt
    Torben Voigt Altair Community Member
    edited October 21

    Hi NoYa,

    Without knowing the models I can only make assumptions. The MLFMM error might be due to lacking memory. Maybe the mesh is much finer than 1/10 of the wavelength? For MLFMM this will increase the memory usage.

    For the edge port make sure that the region on both sides of the faces are the same delektric medium (or free space). Else you will have a "dielectric boundary" where edge ports are not allowed.

    Best regards,
    Torben

  • NoYa
    NoYa Altair Community Member
    edited October 21

    Hi NoYa,

    Without knowing the models I can only make assumptions. The MLFMM error might be due to lacking memory. Maybe the mesh is much finer than 1/10 of the wavelength? For MLFMM this will increase the memory usage.

    For the edge port make sure that the region on both sides of the faces are the same delektric medium (or free space). Else you will have a "dielectric boundary" where edge ports are not allowed.

    Best regards,
    Torben

    Hi Torben,

    Apologies for not attaching the model earlier. Here it is. My environment has 32GB of memory, and I suspect that the issue might be due to the mesh being finer than 1/10 of the wavelength, which is 3 meters at 100 MHz, leading to potential memory shortages. Do you think I should consider adjusting the mesh size to address this?

    Regarding the full-sheath dipole antenna, I have set up the regions of the faces as shown in the attached file. Could you please review these configurations and let me know if there are any adjustments you would recommend?

    Thank you for your assistance.

    Best regards,
    NoYa

     

  • Torben Voigt
    Torben Voigt Altair Community Member
    edited October 21

    Hi NoYa,

    I'll have a look at your two models tomorrow. Unfortunately, I can't make it today. But what I can say for now: MLFMM is suitable for electrically large models (> 4 lambda). MLFMM is completely unsuitable for electrically small models. More tomorrow.

    Best regards,
    Torben

  • NoYa
    NoYa Altair Community Member
    edited October 21

    Hi NoYa,

    I'll have a look at your two models tomorrow. Unfortunately, I can't make it today. But what I can say for now: MLFMM is suitable for electrically large models (> 4 lambda). MLFMM is completely unsuitable for electrically small models. More tomorrow.

    Best regards,
    Torben

    Hi Torben,

    Thank you for your prompt response. I appreciate the clarification that MLFMM is suitable for electrically large models (> 4 lambda). It helps me understand why it might not be appropriate for my current setup. I'll look forward to your further insights tomorrow.

    Best regards,
    NoYa

     

  • Torben Voigt
    Torben Voigt Altair Community Member
    edited October 22

    Hi NoYa,

    1. Basic Cylinder Dipole Antenna

    There is a 30 m high concrete tube and a 30 m high water cylinder with a diameter of 50 m in the model. This sounds at first as if MLFMM could be used well (model size a few wavelengths), but there is a problem here: the concrete cylinder is largely inside the water cylinder and is meshed according to the wavelength in the water.

    image

    This leads to a very large number of meshing elements that are significantly smaller than lambda/10 of the free-space wavelength. And this then leads to a significant increase in the memory required by MLFMM.

    With enough available RAM, you can of course simulate the model without any problems (65 GByte with 16 processes, ~8 minutes), but I don't suppose you have that much.

    I have attached the files of the finished simulation here.

    The question is: Should it really be a large water cylinder or would you rather use a dielectric halfspace? MLFMM would then no longer be possible, but the memory requirement with MoM should still not be too high. However, I would have to check whether there might be problems with the very high permittivity and conductivity of the water (epsr 72, sigma 4.8).

    2. Full-Sheath Dipole Antenna

    I had already given the answer here: Edge ports must not be placed on dielectric boundaries (= different dielectric regions on both sides of the surfaces). The inside of your dipole is PEC, surrounded by Purewater.

    image

    So you only have to change the inside of the dipole to Purewater. Model is also attached.

    Hope this helps!

     

    Best regards,
    Torben

  • NoYa
    NoYa Altair Community Member
    edited October 23

    Hi NoYa,

    1. Basic Cylinder Dipole Antenna

    There is a 30 m high concrete tube and a 30 m high water cylinder with a diameter of 50 m in the model. This sounds at first as if MLFMM could be used well (model size a few wavelengths), but there is a problem here: the concrete cylinder is largely inside the water cylinder and is meshed according to the wavelength in the water.

    image

    This leads to a very large number of meshing elements that are significantly smaller than lambda/10 of the free-space wavelength. And this then leads to a significant increase in the memory required by MLFMM.

    With enough available RAM, you can of course simulate the model without any problems (65 GByte with 16 processes, ~8 minutes), but I don't suppose you have that much.

    I have attached the files of the finished simulation here.

    The question is: Should it really be a large water cylinder or would you rather use a dielectric halfspace? MLFMM would then no longer be possible, but the memory requirement with MoM should still not be too high. However, I would have to check whether there might be problems with the very high permittivity and conductivity of the water (epsr 72, sigma 4.8).

    2. Full-Sheath Dipole Antenna

    I had already given the answer here: Edge ports must not be placed on dielectric boundaries (= different dielectric regions on both sides of the surfaces). The inside of your dipole is PEC, surrounded by Purewater.

    image

    So you only have to change the inside of the dipole to Purewater. Model is also attached.

    Hope this helps!

     

    Best regards,
    Torben

    Hi Torben,

    Based on the previous suggestions, it was advised to set the inside of the dipole to purewater to avoid problems with edge port placement on dielectric boundaries. While I understand that edge ports should not be placed on boundaries between different dielectric regions, I am still unclear why changing the inside of the dipole to purewater would allow the edge port to function properly. My concern is that this configuration would result in the port facing a dielectric surface, which typically leads to issues.

    Could you please explain why this specific setup (setting the inside of the dipole to purewater) allows the edge port to work without errors? Additionally, if there are any specific conditions or additional settings that need to be adjusted to ensure the edge port connects correctly to the PEC region, I would appreciate any further clarification on this matter.

    Additionally, I have tried to create a model based on the one you previously provided, where instead of using a cylindrical water column, I used a plane to represent the seawater. This was to address the memory limitations I encountered with the cylindrical model, as I aim to calculate the propagation constants in both the air layer and underwater. However, I am still facing difficulties with the edge port setup, as it continues to indicate that the port is placed on the surface of a dielectric material, leading to errors.

    I apologize for bringing up this issue again, but could you kindly provide further guidance on how to properly configure the edge port for this model, especially in this plane-based setup?


    Best regards,
    NoYa

  • Torben Voigt
    Torben Voigt Altair Community Member
    edited October 23

    Hi NoYa,

    the condition for an edge port to work is that the same material is defined on both sides of the surfaces. In your model cylinder_Fullsheath-dipole.cfx the inside of the dipole is PEC and it is surrounded by a Purewater region (which is then surrounded by a PVC region, but that doesn't matter).

    image

    The edge port therefore sees PEC on one side (inside the dipole) and Purewater on the other.

    For the edge port to work, the outside and inside of the dipole must be defined as Purewater.

    Note: Could it be that you have mixed up the regions somehow? I've been wondering the whole time why the inner region is Purewater and the outer PVC. That's a strange dipole ;) But I honestly wouldn't understand it the other way around either. A dipole wrapped in PVC makes sense, but why the additional Purewater region if there is a large Seawater region anyway...?

    But hey, it's up to you :)

    Best regards,
    Torben

  • NoYa
    NoYa Altair Community Member
    edited October 25

    Hi NoYa,

    the condition for an edge port to work is that the same material is defined on both sides of the surfaces. In your model cylinder_Fullsheath-dipole.cfx the inside of the dipole is PEC and it is surrounded by a Purewater region (which is then surrounded by a PVC region, but that doesn't matter).

    image

    The edge port therefore sees PEC on one side (inside the dipole) and Purewater on the other.

    For the edge port to work, the outside and inside of the dipole must be defined as Purewater.

    Note: Could it be that you have mixed up the regions somehow? I've been wondering the whole time why the inner region is Purewater and the outer PVC. That's a strange dipole ;) But I honestly wouldn't understand it the other way around either. A dipole wrapped in PVC makes sense, but why the additional Purewater region if there is a large Seawater region anyway...?

    But hey, it's up to you :)

    Best regards,
    Torben

    Hi Torben,

    I was able to understand the content of the area that you previously explained. Thank you very much for your assistance.

    The attached model consists of a dipole made of a cylinder and a wire. I am trying to calculate the S-parameters at 100 MHz, but when including both the wire and the cylinder, the simulation gets stuck at 82% and does not proceed any further. This time, I defined the seawater region using a plane.

    I have increased the memory capacity to 64GB and am using 16 processes, but it still stops at the screen showing 82%. Could you please help me understand why this is happening?

    Best regards,
    NoYa

  • Torben Voigt
    Torben Voigt Altair Community Member
    edited October 28

    Hi NoYa,

    I am currently working with your models. The solver is obviously struggling with the extremely high values of seawater for a dielectric (epsr 70, sigma 4). The WARNING 3604: Too many interpolation points for the substrate message must be taken seriously. I don't believe the simulation ius actually stuck, but at least it takes very long.

    Unfortunately, I don't know of a suitable workaround at the moment. When using the cylindrical seawater region again instead of the Sommerfeld halfspace, the memory requirement is much too high for your computer.

    I will now try to find out whether the ACA solver can help us and also whether the newly implemented CBFM (Characteristic Basis Function Method) is helpful here. However, both remain to be seen.

    image

    image

    Note, in any case you should use symmetry where possible. In your models you can use an electrical symmetry plane at x=0.

    Best regards,
    Torben

  • Torben Voigt
    Torben Voigt Altair Community Member
    edited October 28

    Hi NoYa,

    You are lucky! If you use only 4 (or less) CPU cores to calculate the cylindrical seawater region with MLFMM, the memory requirement can be reduced enough to fit into your 64 GBytes. You can find the files in the attachment.

    Note, with standard MoM it will require ~355 GByte.

    Best regards,
    Torben