Co-simulation setup in HFSS

[vdbhope]

Hi, I am trying to setup cosimulation in HFSS to tune cavity filters. What should be the port impedance of the lumped ports assigned between resonator and the ground? Using impdance like 50 or 100ohms will detune the filter. Will this have any effect on the tuning when running circuit simulation?

 

[PlanarMetamaterials]

Welcome, vdbhope,

For a cosimulation, you should be using a driven terminal setup [unless you have a valid reason not to, such as direct excitation of a waveguide mode]. In this case, it shouldn't matter what you assign the terminal impedances to be, the external circuit will specify the actual input/output impedances. The internal terminal values are, I believe, what are used in the absence of external circuitry.

 

[vdbhope]

Welcome, vdbhope,

For a cosimulation, you should be using a driven terminal setup [unless you have a valid reason not to, such as direct excitation of a waveguide mode]. In this case, it shouldn't matter what you assign the terminal impedances to be, the external circuit will specify the actual input/output impedances. The internal terminal values are, I believe, what are used in the absence of external circuitry.

Thank you for the reply! I have been using driven modal setup since I am exciting a coax cavity using an SMA connector. Do you think I can use driven terminal setup here? I am trying to follow the procedure explained in this paper - https://ieeexplore.ieee.org/document/967097.
The author assigns lumped ports in between the resonator and the top ground plane.

 

[PlanarMetamaterials]

For the single-mode coax, either driven modal or driven terminal should work. It sounds like the rest of your setup is correct -- I would stick with 50 Ohms, since this is the impedance of the SMA connector -- but it shouldn't matter for the purposes of cosimulation. It will matter if you are just looking at the (scattering) results of the HFSS simulation directly.

 

[vdbhope]

For the single-mode coax, either driven modal or driven terminal should work. It sounds like the rest of your setup is correct -- I would stick with 50 Ohms, since this is the impedance of the SMA connector -- but it shouldn't matter for the purposes of cosimulation. It will matter if you are just looking at the (scattering) results of the HFSS simulation directly.

Okay then! I am going to try and tune the filter using the setup. One question though, why driven terminal for cosimulation and why not modal?

 

[PlanarMetamaterials]

One question though, why driven terminal for cosimulation and why not modal?

Generally in a circuit model, the various wires, excitations, and other components are represented as being either in series or shunt with various conductors (or "terminals").

Once you extend the modal concept to multiple modes, it doesn't make much intuitive sense to place components, for example an inductor, in shunt between two modes. There are valid reasons for doing so, but the vast majority of the time someone is looking at a circuit, they will assume that various connections are describing a physical layout.

 

[vdbhope]

I have started co-simulation. But the solution is not optimizing to the goal values. I am using the ratio S11/jS21 as the optimization function. I am not sure what the issue here is. Is it the range of values of capacitances in the co-simulation? Can anyone share some references that guide on step by step co-simulation setup.

 

[PlanarMetamaterials]

I regularly do cosimulations, I think the HFSS help documents were sufficient for me. Sorry, I'm not sure if I know of any other references.

I've noticed the generic optimizer in HFSS can be pretty bad. My solution is generally to change the type of optimization. I prefer the Legacy quasi-Newton Gradient for most cases. Be sure to set up the advanced options so that it doesn't end prematurely.

 

[vdbhope]

I regularly do cosimulations, I think the HFSS help documents were sufficient for me. Sorry, I'm not sure if I know of any other references.

I've noticed the generic optimizer in HFSS can be pretty bad. My solution is generally to change the type of optimization. I prefer the Legacy quasi-Newton Gradient for most cases. Be sure to set up the advanced options so that it doesn't end prematurely.

Yeah I thought that I would need to change the type. Will definitely try that. Have you done co-simulation to tune filters? If yes, what do you use to model the coupling between cavities? Shorted transmission lines or series capacitance? What should be the length of the transmission line in that case? I am having a hard time understanding which one to use and why. For now, I am using shorted transmission line of beta*l<pi/2 at center frequency.

 

[PlanarMetamaterials]

Yeah I thought that I would need to change the type. Will definitely try that. Have you done co-simulation to tune filters? If yes, what do you use to model the coupling between cavities? Shorted transmission lines or series capacitance? What should be the length of the transmission line in that case? I am having a hard time understanding which one to use and why. For now, I am using shorted transmission line of beta*l<pi/2 at center frequency.

No, sorry, I haven't worked with filters.

Good luck with your simulation.