wavegiude analysis that is more than 10 wavelengths long

[sviodo]

combining modes 1 and 2

i need to analyse waveguide that is that is more than 10 wavelengths long
i couldn't cope with the problem in hfss (problem was too big), maybe you could recommend me appropriate tool?

With regards,

 

[yuyu]

Dear Sviodo,
I think that you can try to use CST Microwave Studio.
This SW is a 3D e.m. simulator, FDTD method based.
If the problem is too big, it is possible to slice the model in several blocks each e.m. analyzed and described by its own generalized S matrix.
Then it is possible to combine the S paramiters by a circuital simulator taking in account higher order modes also if they are evanescent modes at the interfaces between two blocks.
For this purpose I suggest the software CST Design Studio.

 

[eirp]

Hi, sviodo!

Please describe more the problem if possible. If you have straight waveguide with arbitrary cross section, it can be made by your own..
If you're familiar with with Matlab programing and solving PDE's, you can solve Helmholz the equation for transverse eigenvalues and eigenfunctions. If you compute enough number of modes (very fast), it's possible to build S matrix.
Similar method is used in MWS (eigenmode solver). Look into manual how it works. This solver is appropriate and efficient for this task.
Regards,
Eirp

 

[sviodo]

thx for your advice, i develope the program in matlab but i need to check design before it'll be produced, seems that i can use mws

Regards,

 

[loucy]

You have discontinuities all over these 10 wavelength? Why can't you use divide and conquer.

 

[sviodo]

you suggest to analyse sth like 20 divided parts and combine results?
i do belive that this could be solution in case of 2-3 divided parts but does not apply to my case...

plz correct if i misundersood you

 

[filterman]

Using FTDT (CST,QW3D,Fidelity,etc) is O.K. , but if problem can be analized in mode-matching software(mician's wizard or Wasp-net) then your results will be faster and more accurate-you might end up with a producable non tunned device.

 

[eirp]

Dear Sviodo,

please can you post a simple sketch of your problem?
Maybe we can learn something new and interesting!
Eirp

 

[loucy]

Why you can do 2 to 3 divided parts, but not 20 divided parts?

I think "divide and conquer" is a simplified case of mode matching. Suppose one can ignore the higher order mode coupling between these parts, then each part is a simple two port. Higher order modes should be included only when the discontinuities is close (1 lamda_g or 1/2 lamda_g ? there is no rule) to the dividing plane.

 

[mike_bihan]

why not use 2-D simulator

why not use 2-D simulator?

 

[sviodo]

you sure that you can use 2d simulator to analyse for example magic T
or E junction ?

 

[Kit-the-great]

You may try to use MuWave Wizard for fast calculation, or divide the structure to simple parts and analyze each by Markuvitz's Waveguide handbook formulas. In any cases, please, try to post the some skethes of Your problem for our better understanding of it.

With responce,
kit-the-great

 

[mike_bihan]

sorry

I confuse that with cpw

 

[lindaknoll2002]

I have used hfss to simulate waveguide structutres with more than 20 wavelength long and did not have any problem. The only thing i remembered is that in the old version of HFSS it does not allow the ratio between longest and shorted objects be more than 4000. So make sure no object is too small.

Linda.