CST - combline filter

Hello everyone, this is my first post here
I have recently started to work with CST and designed few 3dB couplers, FM band. There my experiences were good for results were in accordance with measurements.
But I have hard time trying to simulate combline filter, 4 resonators, UHF band. Simulations take long time and results seem not accurate. I tried to put density control rectangle near critical parts in structure, results are better but I am not sure if they are accurate enough.
Being beginner in using this software I ask for any advice from exp. users. What numbers should I put in Mesh properties? How much should I refine at PEC edges for this resonant structure?
Thanks.

Hi dule !
It's normal that you need long time for solving in CST( based on FEM code ). Without several processors and lot of memory you can not simulate and optimize so big structures in CST .
For structures like combline filter try with specialized software like as WASP-NET ( http://www.mig-germany.com).

You can also try to use eigenmode solver (see modal analysis in documentation) in CST-MWS. This solver must be more useful for filters and other resonant structures.
Best regards,
Kit-the-great

I simulated by CST combline filters.
And I had many problem's such devics and solve them only after much addition mesh cells.
I think that in CST is not very good automeshing. Adaptive meshing also dont help.
Many problem your need to solve by hands and head .
For large devices, with not very complicated structure, CST works good.
But if you have large difference between elements dimensions and your device also very resonance structure, you must make mesh refinement by hands and you have huge size mesh cells.
In HFSS the mesh's refinement is the better And for resonance devices I had more good results.

Eigenmode in CST is good only for not very complicated structures such as one, two cavity filters. But for more complicated structures you need to increase mesh very much. So you will have very long execute's time and not very good results.

Just a word from me. When using EM simulators we usualy assume ideal materials, conductors with infinite conductivity and losless dielectrics. This is not an issue with MoM and FEM software, it effects the accuracy for the insertion loss but not the simulation time. In FDTD software it efects the simulation time as the energy of the pulse is bouncing through the structure for a long time making simulations long and/or inacurate. Introducing realistic losses might help solving the problem as the energy of the pulse is dumped much quicker. Of course, with cavities there is no help, they are high-Q structures for real!
FlyHigh

I know that it is a little late to add something to this thread. but I'm also building a filter in cst 2006 and I also have problems with simulationtime and accuracy. I can't tune the filter in CST since each calculation last 24h or more. I have heard that some people uses MATlab togehther with cst for optimization. Does anyone know how one can make the tuning with Matlab?
I am going to try some software from Comsol and they have MoM and FEM solver. My question is: is Comsol a good program to design a combline filter with?

Hi,
which CST Solver do you use? With 2006B they implemented the FD solver based on tetrahedral mesh and well as a fast MOR solver for resonant structures. For Resonant structures, these solves can be much faster then the the TD Solver. Which version of CST do you use?
F.

I use CST 2006B

Hi loke,
do you use the FD TET solver? If you like you can send me your design and I have a quick look. Calcualtion time for 24h seems rather long for me. Usually the calcaultion time is much shorter then this (as long as the mode fits into the memory). Using 2006B you should have no trouble with this because CST added 64 bit support for all solvers.
F.

Hi,
I did have similar problems with long simulation times in mws while
high Q structures were considered. However I realised that significant
reduction in simulation time was achieved by carefully going through
the possible options within each solver (transient or eigenmode).
For example, for high-Q structures, as you may know, it could be faster
to simulate from DC (0 Hz) up to a desired frequency(f_upper) rather
than to simulate between two frequencies, with the lower one close to
f_upper (i.e. Fourier between time-frequency).
As a beginner then, I found valuable information relevant to the above
matter by reading the 'Advanced Topics' booklet. However, each project
requires perhaps an individual approach.
I eventually relied my phd research on MWS, with actual measurements
(DRAs) being in close proximity with MWS results (|error|<2-3%).
Interesting...

Dear
for the comlbine filter design use a frequency domain solver like HFSS , the mesh strategy is much flexible and famous designer and company use this tools for filter design.
I design a duplexer at uhf using " my mesh strategy" I mean try to use dummy object to force the solver engine to compute fast and to converge rapidly also take into account the edges in the filter construction were the field is more interest.
Also keep in mind to use wall symmetry this may reduce computation time.
Also u can use " the divide and conquere " principle and the optimize ur filter in the schematic level and not in the EM level.
Hope this help

Hello All,
Has any one used Vector Fields CONCERTO 6.0 for these (Combline Filter) applications...
CONCERTO uses the state of the art modeling techniques such as finite element (FEM/FEA), finite difference time domain (FDTD), and Moment method (MoM), the code is well suited to a wide range of low and high frequency applications.
Vector Fields CONCERTO offers a complete range of two dimensional (2D) and three dimensional (3D) electromagnetic design, modeling, analysis and simulation software.
http://www.vectorfields.com/content/view/80/97/

Hi loke !
1st try to make the filter from a vacum, I think this is what you'll need in order to solve the filter correctly.
For that reason you have to make the BackGround material to PEC.
About the fact you have a 2006B version of MWS, use the FD (Frequency Domain) solver with/without the tetrameshing, as RF-Simulator mentioned.
Good luck,
Itzik

Pls upload the HFSS combline filter project files.

What I do to simulate large structures is that I first use smaller meshes for faster simulations to have a rough idea. After that I simulate with more meshes to have accurate result. From my experience, the finer the mesh, the more accurate the result. There is slight differences between simulated and measured results since the materials substrate, teflon etc. is not well defined.
If you have material file you can put in CST you will get restults closer to measurement results.