RL-GO for a Gradient-Index (GRIN) Lens
Hi Torben,
In the attached I am trying to model a Gradient-Index (GRIN) lens using RL-GO ray tracer as the MoM-MLFMM surface equivalence solver can't model this lens due to memory constraints. The diameter of the lens is greater than 20 lambda which should be ideal for ray methods. The lens is defined using many volumetric regions with a unique dielectric constant. The lens is illuminated by a spherical wave expansion source. In order to validate the accuracy of the RL-GO solver the permittivity of the lens was set to be equal to free space and near field data was computed on a planar grid.
TThe RL-GO near field data must match the FEKO simulation with no lens case, but there is a huge discrepancy in the RL-GO result. I have attached both the .sph file and cadfeko model for your reference. I tried adaptive ray launching accuracy option high but there is nno significant improvement in the accuracy of the solver.
Can you have a look at the model and let me know if the GO solver can handle these type of lens.
Answers
-
Hi @Shan25 ,
Feko loves warnings, so sometimes users just don't take note of them anymore. However, there are warnings that should definitely be taken seriously. In your case, the lens is too close to the spherical modes source.
"WARNING 40147: Near fields for spherical modes are computed in the cut-off region beta*R < N, reduce spherical wave order N or increase distance R"
A Spherical Modes source is a far field source, which means that no geometry or even near field request is allowed within a certain radius. The size of this radius is calculated like this:
R_far = N*lambda/(2*pi) ; N=Number of modes
In the .out file you will also find this radius specified.
A good tip to reduce this radius is to place the origin of the far field request for the calculation of the spherical modes as close as possible to the center of the antenna to be calculated.
Example:
0 -
Torben Voigt_20420 said:
Hi @Shan25 ,
Feko loves warnings, so sometimes users just don't take note of them anymore. However, there are warnings that should definitely be taken seriously. In your case, the lens is too close to the spherical modes source.
"WARNING 40147: Near fields for spherical modes are computed in the cut-off region beta*R < N, reduce spherical wave order N or increase distance R"
A Spherical Modes source is a far field source, which means that no geometry or even near field request is allowed within a certain radius. The size of this radius is calculated like this:
R_far = N*lambda/(2*pi) ; N=Number of modes
In the .out file you will also find this radius specified.
A good tip to reduce this radius is to place the origin of the far field request for the calculation of the spherical modes as close as possible to the center of the antenna to be calculated.
Example:
__Hi Torben. What about far field source, located near vertex of paraboloid in two reflector configuration. In the my model (see figure below), i have small square hole in paraboloid (big reflector) for feeder (hole size=feeder aperture size). I've got a warning
"WARNING 3718: Far field condition not met for near field computation of a far field point source"
__I honestly ignored it. Can it lead to incorrect results? Or using FF source more safe, than using spherical mode source?
__Thanks in advance.
0 -
Hi @Shan25
With Far Field Sources it is the same as with Spherical Modes Sources, except that the forbidden radius is calculated differently: R_far = 2*lambda*D/(pi^2) ; D=Directivity
Again, you can find the radius in the .out file.
Therefore, near field sources are suitable as equivalent sources in the vicinity of geometry.
Best regards,
Torben
0 -
Hi Torben,
We reduced the number of modes in the spherical wave source by shifting the center of the coordinate system. The problem with the near field pattern still exist for the unity lens obtained using RLGO. Is there a accuracy issue with the RLGO? does it work only for a simple homogeneous lens?
0 -
Hi @Shan25 ,
Can you attach the latest model? Of course the accuracy of asymptotic solvers like RL-GO or PO depends on many factors. Unlike full wave solvers like MoM or MLFMM, where you can always expect 100% accuracy, there are of course different assumptions that are made for asymptotic solvers. However, your example here is very good for RL-GO and should also show good results. I would like to have a look at the last example (incl. .sph file).
Best regards,
Torben
0 -
Torben Voigt_20420 said:
Hi @Shan25 ,
Can you attach the latest model? Of course the accuracy of asymptotic solvers like RL-GO or PO depends on many factors. Unlike full wave solvers like MoM or MLFMM, where you can always expect 100% accuracy, there are of course different assumptions that are made for asymptotic solvers. However, your example here is very good for RL-GO and should also show good results. I would like to have a look at the last example (incl. .sph file).
Best regards,
Torben
Hi Torben,
I agree to your point that the accuracy of the geometrical optics and physical optics methods is dependent on scattering size with respect to the wavelength. You can use the model and the .sph file that I attached to the original post. I also tried the RLGO dielectric lens example described in the FEKO manual. The RLGO works for this case, but the PO solver with full ray tracing has a significant accuracy issue in the both near and far- field regions. I have attached the files for your reference. I strongly feel that RLGO is not accurate when you have a stack of dielectric layers.
0 -
Shan25_22239 said:
Hi Torben,
I agree to your point that the accuracy of the geometrical optics and physical optics methods is dependent on scattering size with respect to the wavelength. You can use the model and the .sph file that I attached to the original post. I also tried the RLGO dielectric lens example described in the FEKO manual. The RLGO works for this case, but the PO solver with full ray tracing has a significant accuracy issue in the both near and far- field regions. I have attached the files for your reference. I strongly feel that RLGO is not accurate when you have a stack of dielectric layers.
I am aware that the RLGO does not include the diffracted fields from the dielectric wedges, so I don't expect to be accurate in the near-field region, but it should match the exact method in the far-field
0 -
Shan25_22239 said:
Hi Torben,
I agree to your point that the accuracy of the geometrical optics and physical optics methods is dependent on scattering size with respect to the wavelength. You can use the model and the .sph file that I attached to the original post. I also tried the RLGO dielectric lens example described in the FEKO manual. The RLGO works for this case, but the PO solver with full ray tracing has a significant accuracy issue in the both near and far- field regions. I have attached the files for your reference. I strongly feel that RLGO is not accurate when you have a stack of dielectric layers.
Hi @Shan25 ,
You mentioned that you reduced the number of modes in the .sph file. Does Feko still complain with
"WARNING 40147: Near fields for spherical modes are computed in the cut-off region beta*R < N, reduce spherical wave order N or increase distance R"
?
If yes, the reason for the inaccuracy still that the source isn't outside of the far field radius
If the warning doesn't show up anymore, did you check if the reduced number of modes still represents the radiated field well enough (eg. by comparing full antenna model with .sph source in freespace)?
PO is very unsuitable for this kind of models, because it is not able to calculate the transmission through dielectrics correctly. Only the reflection is calculated correctly.
0
Altair Employee