MWS Eigenmode Solver Question

[jpsmith123]

Hello,

I'm wondering what happens when the eigenmode solver apparently fails to converge upon a solution...does it just keep on going forever trying (until it is manually terminated), or does it at some point give up on its own?

Regards,
Joe

 

[jpsmith123]

Just in case anyone's interested...

Apparently the solver just keeps on going, ad nauseum.

I wanted to experiment with periodic boundary conditions, so I decided to see if the solver could calculate the first few propagating modes of a partially dielectric filled circular waveguide, and then also find the phase and group velocity.

Maybe, being that I'm not a MWS expert, I did something wrong.

Anyway, after watching it get nowhere for three days, I decided to pull the plug.

Regards,
Joe

 

[asdfaaa]

Yes, this is common situation, I am also struggling with eigen slover.
Please check attached file, wish it is helpful.

Best Regards,

 

[asdfaaa]

Hello,

I'm wondering what happens when the eigenmode solver apparently fails to converge upon a solution...does it just keep on going forever trying (until it is manually terminated), or does it at some point give up on its own?

Regards,
Joe

It does NOT fail to converge upon a solution, but converge very very slowly. Because your model is high Q value, the energy drop down very slowly.
Two comment:
1, try to use AR filter as I attached.
2, give your dielectric material loss tangent. In most case, it will not generate new resonant frequency, but will shift original resonant frequency a bit (still acceptable).

Best Regards,

 

[navuho]

2jpsmith123


MWS Eigenmode Solver grew up from well known MAFIA and it's a kind of standart for such type of calculations. So, 100% that you are wrong somewhere.
I just cheked simple geometry as you described and all is OK.

 

[navuho]

asdfaaa

Because your model is high Q value, the energy drop down very slowly...
1, try to use AR filter...
I think you mixed up with TD solution. The EigenMode solver always works
with high Q models (moreover low Q requires more time to calc. )
AR filter is also applicable only in TD.

 

[asdfaaa]

asdfaaa

Because your model is high Q value, the energy drop down very slowly...
1, try to use AR filter...
I think you mixed up with TD solution. The EigenMode solver always works
with high Q models (moreover low Q requires more time to calc. )
AR filter is also applicable only in TD.

Thank you for your rectification. Most of my time I use HFSS eigen solver to do simulation, yes, most of time the simulation results make me crazy.
Comment about eigensolver of HFSS and CST is welcome.

Best Regards,

 

[jpsmith123]

Hello Navuho,

Thanks for your reply.

Your model runs fast and converges nicely. (One difference is that I was using a lossy ceramic and running the "JD" solver for lossy materials. That seems to slow things down dramatically, although I don't think that was the problem per se).

Anyway, I'm going to continue to experiment to see what's going on and try to understand it.

BTW, I was surprised that, of the first twenty modes for your structure, I didn't see any TM01 or TM01 like modes.

Regards,
Joe

 

[navuho]

I was surprised that, of the first twenty modes for your structure, I didn't see any TM01 or TM01 like modes
Of course, because I defined different symetry planes (electric and magnetic )

I was using a lossy ceramic and running the "JD" solver
Aaa... "JD" is a new solver that CST just imlemented in a last version of MWS.
1. Sometimes "JD" is converged very slowly (from my experience)
2. Sometimes "JD" misses a eigen roots (frequencies) of geometry.
So, personally I'm not trust to this method. Be careful to use it !

In opposite, "AKS" is very stable, well-known method with assured convergence.
I recomend to use it in most cases, if not only you are definitely need "JD" solver.

 

[jpsmith123]

Hello Navuho,

I didn't realize you defined symmetry planes...yes making the yz plane an "electric" plane would certainly kill TM01.

Regards,
Joe

 

[jpsmith123]

Just in case anyone's interested, I finally got the simulation to work (finding the phase and group velocity of a partially dielectric filled circular waveguide) and the results seem to be in good agreement with published results...

Regards,
Joe