Is There A Resolution Limit For HFSS?

Hi all, I'm simulating a simple coupled line microstrip resonator.

Basically, I just want to characterize a series of channels from 200GHz to 300GHz by changing the length L of the resonator. The step size preferred is about 1 micron (L changes in a geometric sequence) from 150 micron to 250 micron.

However, it looks some of the resonance peaks returned by HFSS are OVERLAPPED or DISORDERED (f_res should change 'monotonically' with L), although the general trend looks very good. I really want to get a series of more or less evenly-spaced channels over that frequency range.

Can anyone tell me if my step size have reached the resolution limit of HFSS? Should I change the "model resolution" settings? How can I get rid of those overlaps and disorders?

The resonator frequency is sensitive to the mesh. For accurate center frequency you need a good mesh.

FEM requires adaptive meshing, and you set the stop criteria for meshing. Maybe your mesh frequency or mesh criteria (what delta S convergence level is good enough) need to be changed.

Thank you SO MUCH, sir. May I ask you a few more questions?

I have manually set mesh's maximum length to 1/50 wavelength on my conductors (microstrips), but for delta S, I'm still using the default value 0.02. Maybe I should change delta S to be something like 0.01 or even smaller?

And...I'm actually using fast sweep and a solution frequency of 300 GHz now. Usually people set the solution frequency at the center of the frequency range (~250GHz in my case), but if I do so, I'll have no resonance peaks near ~250GHz (250 GHz and its neighbors are skipped!). Do you think the result is reliable if I set my fast sweep's solution frequency to 300GHz?

I really appreciate it!

You should have a plenty of time and set the frequency sweep to discrete. I never trust any other sweep.

I second this. Especially under resonant conditions, fast sweeps may give inaccurate results.

Thanks! Maybe I should give discrete sweep a chance, lol

I need to cascade the S matrices of all the channels so I want the simulation as precise as possible. If there are overlaps or disordered channels, the cascade will be a mess!

Thanks a lot! I hope discrete sweep can give me some better outcomes.

If you have a resonator, your mesh refinement should be done at/near the resonance frequency.

Thanks for your response! I still have a question about the mesh size. Say if I want the simulation to resolve a 1 micron difference in resonator's length, does that mean my mesh size on the resonator should be less than 1 micron? I mean I can specify the maximum edge length in the mesh operation section, right?

No. Your 1/50 wavelength is already really fine.

But you need to make sure that mesh refinement is done at/near the resonance frequency, and not at some frequency where your resonanator is reflecting most energy.