Jump to content

Torben Voigt

Members
  • Content Count

    119
  • Joined

  • Last visited

  • Days Won

    2

Posts posted by Torben Voigt


  1. Hi bouvy,

     

    I thhink that the Directivity in that case should not be available in POSTFEKO at all, but on the other hand

    WARNING  3976: Directivity cannot be computed for far field calculations involving real ground with losses, gain will be computed instead

    makes it quite clear.


  2. Hi bouvy,

     

    Coming back to your original assumption that the Directivity should be ~3 dB larger than for an isotropic radiator. Could you please clarify why you expect this? And please remember: Directivity will not be calculated when using infinite dielectric ground. Only Gain will be calculated.

     

    Apart from the results, FEKO 7 is quite outdated. If possible you should install the current version of FEKO (2018.1).

    bouvy likes this

  3. Hi bouvy,

     

    The attached model gives a maximum Directivty of 2.49 dBi. Please note that for the Reflection Coefficient Approximation a minimum distance of lambda/10 should be between ground and segments/elements. That's why warnings are given:

     

    WARNING   667: A segment is close to the ground

    WARNING   668: A triangle is close to the ground

     

    I compared with Exact Sommerfel Integrals where the maximum Directivity is slightly higher (3.19 dBi). I assume that the lower the frequency, the worse the results, because the assumption of segments/elements being > lambda/10 away from the ground not fulfilled.

     

    Due to the small distance geometry and ground (1 cm) the Reflection Coefficient Approximation should only be used for frequencies > 30 GHz. I would recommend using Exact Sommerfeld Integrals instead.

    image.png.829250cf3fee056097072e2727a5a4a6.png

     

    Please also note, that for dielectric halfspaces the loss in the dielectric is unknown and therefore only Gain will be computed (Directivity = Gain).

    bouvy likes this

  4. Hi FieldForcer,

    1. Did you get any warnings (I think there should be)?
    2. In general I would recommend to use continuous (interplated) range. FEKO will then catch all the resonces along the frequency range, whereas with discrete frequency points all the information betwwen the calcilated points is neglected.
    3. Of course 0 Hz is impossible. Our recommendation for fmin and fmax for time analysis is
      fmin = 1/sd and fmax = (Nt/sd)/2
      with sd the signal duration and Nt the number of samples.

    There is a video on Youtube about FEKO's time analysis (beginning at ~32:00). It's from 2014, but the basic rules still count: https://www.youtube.com/watch?v=YAruJMWRalI&t=8s

     


  5. Hi FieldForcer,

    I was referring to the length od the edges of the mesh on a surface. At 60 Hz and standard mesh settings, a sphere with 50 mm diameter will be meshed similar to this:

    image.thumb.png.90910dfdb0452aed3a84ffff20bceed5.pngimage.png.510ee2783d72b56f0c4275b9a18f4670.png

     

    The edges on the surfaces should be around 1 cm in that case which would in principle allow you to request nar field points 1.5 cm from the surface.

     

    Please note the mesh size of tetrahedra also depends on the dielectric properties of the medium.


  6. Hi FieldForcer,

     

    Sorry for being imprecise: With "properties are freespace" I meant a dielectric with properties of freespace (eps_r = 1, tand = 0). I was referring to

    On 7/24/2018 at 4:17 AM, FieldForcer said:

    first model has two coils and second model has two coils with a sphere. Properties of sphere are free space. Ideally I should get same field patterns, but I get different results. 

    I assumed that you tried to solve the sphere with SEP, so I recommended VEP.

     

    Regarding the Singular field errors, which version of FEKO are you using? In case it's the current version 2018.0.2 I would have to look into it. Is it posiible to attach the model?


  7. Hi Kartik,

     

    FEM modal ports must be assigned to dielectric boundary faces, contrary to MoM waveguide ports, which are assigned to PEC faces.

     

    image.thumb.png.1d14a59aee17e71abc0434e45c4f8782.png

    Also, there is a PEC face blocking the wave transmission:

    image.png.1fb3ede68821a3068f247596dee8b815.png

    Setting also this face to dielectric boundary should help.

    Furthermore, the top part of the horn is isolator, shouldn't it be Air instead?

    image.png.66cacbebfe20b25e40257128fc54c275.png

    Last but not least, FEM modal ports between two dielectric regions are not supported (that's what the warning is referring to). I would rather cut the outer air box like this:

    image.png.e27b23597764011603b3e71e71eff2e3.png

    The model is attached: Anechoic_2-1_alt.cfx

     


  8. Hi Pavlam,

     

    One imoprtant addition: The calculation the Rx / Tx coefficients from the power flow through near field is only well-defined if the size of the near field requests is N * size of unitcell (N is an arbitrary interger value). In your case the size could be dy*dz or dy*2dz,...


  9. Hi hykr,

     

    One way would be using the "Project" function in CADFEKO (see attached video). However, the length of edges will change if they aren't orthongonal to the direction of curvature, It will depend on the degree of curvature, how much the length will change. For complex models there may be better solutions.

     

    Regarding the best solver, this of course depends strongly on complexity and size of yor model. In case there are both small/complex parts and large parts, it may be feasible to decompose the mode, e.g. by replacing small/complex parts by equvalent near fields.

     

    If you could attach an example model, I may be able to help more.

    hykr likes this

  10. Hi Pavlam,

     

    The reason for the runtime issue is still under investigation. It could be worked around by setting up the unit cell differently:

    image.png.6172c731c3723a5133030ca21e26ed0a.png

    Note that your mesh setting of lambda/5 does reduce the memory requirement, but will most likely not lead to trustworthy results. The unit cell has a size of 1.6 x lambda and thus will demand some memory with a mesh size of lambda/10 (>20 GByte). I see from your *.out file that you have 64 GByte installed, so that should not be a problem.

     

    BUT: Due to the size of the unit cell there are grating lobes for each frequency (Grating Lobes). They prevent an accurate calculation of the transmission / reflection coefficients. A simple workaround here is shown here:

    image.png.2ad49e14550b3aeaf371f392a220cc4d.png

     

    Note, P_incident is calculated by 2nd model, with no geometry present, just calculating the power going through NF front.

    I will add the files here once the simulation has finished.

×
×
  • Create New...