Very bad results of HFSS for an X-band Waveguide Filter

[Matthew_Dawson]

x-band filters

We have analyzed, simulated and built an X-band waveguide filter. we performed a simple circuit analysis of the filter and an analysis using HFSS and later with CST microwave studio. The discrepancy between measured and circuit analysis results is small however simulation results show significant frequency. I have double checked everything yet this problem was not resolved. I'd be grateful to hear your suggestions. Here I attached simulation files and results.

 

[ansonjimli]

mesh seeding,hfss

you should mesh more densely to get an accurate results. HFSS is good for this waveguide structure.

 

[burton]

cst studio mesh

HFSS is better for high Q filter, but mesh seeding may be used, especially for the coupling section with evanscent waveguide.

 

[Matthew_Dawson]

microwave filter with waveguide hfss

Thank you for your replies.
The Q-factor of the filter is 20. There is 1.8% frequency shift in the response of the filter. this problem was not resolved even when I used very dense mesh. (1.5GB of RAM was used for an x-band filter)
I ran the simulation with CST microwave studio, as well and used very dense mesh. It took 6 hours to do the simulation. However, results of CST Microwave studio is similar to HFSS and exhibits a frequency shift of 1.8% which results in considerable error due to narrow band behavior of the filter.
I'd appreciate to hear your comments.
Matthew

 

[ansonjimli]

results in hfss

as burton said , the coupling part is very important and lmain for the load effect .you could use dummy project to improve the dense of this part.

 

[chenqihit]

Hi, Mattew. I am also doing a project designing a waveguide filter in HFSS and my result shows that the passband frequency shifts from the required frequency. It is similiar to your problem. So did you figure out a way to fix this problem?

 

[PeterMartin]

HFSS and CST/MWStudio are equally good tools for simulating WG filters.
MWS has a choice of several solvers. Their FD solver is good for filters , their transient solver is not so good , it is more suited to less resonant but broadband structures like antennas and couplers, but it is still usable and useful as final check .

It sounds like you are doing a check anyway between HFSS and MWS and are getting close agreement but in reality you are just over 1.5% off in your centre frequency.

For a filter that's not too bad. They can be very sensitive to machining tolerances, or imperfections in soldered joints etc, -which is why tuning screws are often needed. Thermal expansion can be another problem for the same reason. You might want to calculate what machining inaccuracy will produce the error you are measuring.

If you are slightly too high in frequency adding a tuning screw in the cavity should fix the problem.