design of cylindrical cavity+hfss
First of all, i would like to thank every EDA Board member who replayed to my inquiry specially the following members:
Jian ,
maxwelian ,
akedar ,
navuho ,
sengar_vijay
Here, i attached the 8-slots magnetron cavity that I'm trying to solve and get its resonance frequencies by using the HFSS eigenmode solver.
Again, the problem that I'm facing right now is that i already get the solution from the HFSS and i get another solution (using another approach) from a Matlab Code
and both solutions doesn't match !
I feel that the correct solution is the Matlab one. It gave me resonance frequencies starting from 9.5 GHz up to 17.7 GHz. The HFSS simulation gave me results at the same band and very close too. However, both results must be exactly the same and completely tie down.
Also, I attached a very simple cylindrical cavity that i made to check the output results from the HFSS in order to start trusting HFSS. The correct results for this cavity is calculated by the very basic microwave laws which means that they must be correct but the HFSS also gives different values which is strange!
If this problem is done, ill be so close to finish my research and get the masters degree and I'm sure that you can guess how worry i feel now.
I tried to upload the HFSS project file on the reply but I'm not sure it will succeed or not.
Anyway, if any body is eager to help please send me back your e-mail address here or on my e-mail address which is:
m_goher@yahoo.com
Thanks is advance
Regards,
Goher
Hi Goher,
the problem might be the "faceted" representation of HFSS. Any curved geometry object will be represented by a segmented mesh. This introduces a geometry error. The real life geometry and the mesh geometry vary slightly. I guess this is the reason why you do not get the "perfect" results even for analytical examples.
The problem is: any mesh adaptation works on the basic mesh. The mesh adaptation will improve the field sampling but the geometry is fixed to the initial mesh! All programs which use a tetrahedral mesh have this kind of problem. I believe HFSS has some kind off analytical correction for this known problem which might reduce the error somewhat but from you post it seems that this does not give you perfect results. I have not used the eigenmode solver of HFSS this often and I do not know how good this correction works. Maybe someone else can comment on this? However, you certainly should get results within 1% or 2% of the analytical results.
The only program I know which improves the field AND the geometry when using a denser mesh is the eigenmode solver of CST. (If anyone knows other programs as well -> comments welcome). Using a denser mesh (manual or automatic mesh adaptation) you should ge be able to get results with an error of less then 0.1%.
The Eigenmode solver in CST uses a rectangular mesh but they take into account sub cell information using this PBA technique. For each denser mesh field and geometry are newly sampled.
Can you post your geometry and and the cylinder here as a SAT file?
F.
I have to agree with RFSimulator. CST has a much better eigensolver in my opinion. I have used it to obtain the eigenmodes of very complex structures. Though the simulation takes a lot of time, I am very happy with it.
Dear Goher,
I have been studying on the waveguide and cavity of a microwave oven for my master thesis.
I do not know any comments about your problem but here attached my simualtin which we can connect both of them and create all sections of a microwave oven.
By the way İ want to ask you one question:
In my thesis i need to simulate many modes at the same time especially TE 564 is the solution.
How can I tell HFSS to produce especially mode 564?
Your asswer will be very helpful for me, i have seen that u are using many modes in your analysis.
Regards
Hi !
I checked quickly your projects.
1. The main mistake is that you try to calculate full model despite of the natural symmetry plains. In that case the number of
mesh elements is unnecessary large and the results could be really strange because of spliting the dipole (or higher) resonances on two polarizations.
The Frequencies of these two modes may be different since the mesh produced by HFSS is nonsymmetrical.
So you need to split your model with XZ and YZ plains and define proper boundary conditions (E or H).
Moreover you may split it more once by the half length if you defined full cavity in Z direction.
2. You don't need to set PerfectE condition to outer boundary, HFSS assumes it by default
3. Don't use metal (copper) elements in your design if you can avoid it. Just subtract it from the main body and than define lossy (conductivity) boundaries.
4. You try to mesh the skinlayer which is totally useless at freq. about 10 GHz
The depth is about microns or order 10^-5 comparing to the cavity size.
5. Don't try to calculate many modes (20) at one time. It's a lot. Split your analyses on few bands and calculate it one by one with about 5 modes max at once.
6. Control the mesh of your objects ! You need to define proper Model resolution and SurfaceAproximation before you start calculation. Also it will be a good idea to
add more elements (do mesh refinement) to the critical part of your cavity
7. Don't use lower order mesh (better to add more memories :)
8. After the calculation check the convergence (<0.1 % is good) , EM fields in the cavity and actual mesh. The Fields should be smooth, without unphysical behavior.
Hi navuho,
I would agree with your settings. How do you get around the geometry error. I certainly agree that this error is not this big for most structures. But in order to get extreme accurate results you have to take into account the faceted mesh.
Do you do some surface refinement of the mesh to keep the error small? Would be nice to have an idea how much surface refinement you need in order to get the error bellow a certain threshold (0.5%, 0.1%, 0.05% ..). I guess this refinement factor will change with cylindric and spherical shapes.
F
Yes, you are completely right. I talked many times with HFSS support about the possibilty to draw the faceted surface (Ansoft remove such feature since HFSS v9)
But they changed ideology and completely moved the face control to the mesher.
Now I use 3d-party programs to draw the model.
Actually I don't trust to HFSS mesher because it produces unregular mesh with many redundant cells.
I prefer to draw segmented surface "ab initio" and refine (iterate) volume mesh only.
Generally the 3D object is the fugure of curvature rotation around the Z-axis. The good start point is to adjust the volume of your segmented model with analitical one.
To do that I wrote simle program in mathcad to estimate the error and find the optimum relation between the number of curvature segments and the steps of rotation around the axis.
The type of curvature (convex, concave) approximation (inscribed, circumscribed or median) is also important and I wrote couple macroses as well to realize it.
Finally I do import geometry to HFSS using sat-format.
Concerning CST, from my expirience the CST eigenmode solver is realy strange.
I mean it is converged fast to a sertain level (~1 %) but if you need to get more precise it almost impossible, the convergency becames so slow and irregular...
The HFSS is much better from this point (of course you have to spent time to prepare the outer surface). Also CST starts the eigenmode calculation always
from the lowerest mode which sometimes is unacceptable. Both CST solvers (AKS and JD) don't support subgridding and impedance boundaries.
So... my current choice for accurate cavity modeling is HFSS :)
Hi navuho,
Your preparation of the HFSS model sounds pretty advanced !
Concerning the CST simulation. I belive in the 2006B version can can specify a "target" frequncy. The Eigenmode calculation will start only above this frequncy. Lower modes are not calculated. I guess this is what you are looking for?
Intersting to hear that you design converges only slowly to a higher accuray. Would you mind to share some of your designs (analytical shapes). If I find some time I could to do some studies.
F.