Problem with modeling of HFSS (Correlation between HFSS, SIwave and Measurement)

I am trying to model HFSS in order to get good agreement with the measurement results.

As we know, SIwave has its own method of forcing the boundary condition to maintain the TEM mode of the pcb, and it saves a lot of times and technical issues by giving a highly accurate simulation results for pcb structure.

For HFSS, it is no more easy as we need to figure out how to correctly set the lumped port, boundary conditions, error of tolerance, etc.

As in the attachment, I have tried hard to try to get results for a microstrip line in HFSS as close to SIwave as possible.
In SIwave, by forcing the solution frequency to 20GHz, the mesh gets finer, and we could get significant dip at each resonant frequencies.
But in HFSS, no matter how I changed the boundary conditions (lambda/4 or 10 times the dielectric thickness), or even make the lumped port as wide as possible, deembed the lumped port, set the error of tolerance for interpolating sweep to as low as 0.1%, the result is still a smooth wave as if there's reflection going on.

How should I correct my HFSS model in order to have a good agreement with SIwave result?

Thanks in advance.

- - - Updated - - -

I have no idea how should I set my boundary conditions to maintain TEM or even quasi-TEM of my microstrip line

Hi nelsonys,

Your "out" port is completely overlapping object SIG001_2, which is copper. What are you trying to accomplish with this setup?

Ports in HFSS excite fields on and between conductors. Having a conductor completely overlapping the port should produce a non-sensical result.

You also don't have a conductor touching the outer side of the port, so how are you planning to excite a TEM mode?

Hi PlanarMetamaterials,

Thank you so much for letting me know my stupid mistakes as I have no idea on how to create a good lumped port.
For this case, it is a round via and I have no idea how could I create a sheet without overlapping the conductor.

Forgive me as I am an absolute newbie. For SIwave, I just have to point at the signal conductor and the ground plane to create a lumped port.
Would you mind to explain more on the theory of not overlapping, and the correct way to define lumped port for this case?

Thanks a lot for your great help!

- - - Updated - - -

and appreciate that if you wouldn't mind to explain a bit about the some background theories on how to excite a TEM mode using lumped port sheet in HFSS.

Hi nelsonys,

Firstly, it doesn't appear you need lumped ports. I think you could get away with wave ports, which will automatically select the correct impedance. But that is jus a minor issue -- if you have the right impedance, lumped ports are the same.

Ports in HFSS excite various modes, which don't necessarily have to be TEM. The modes that are excited are determined by the conductor geometries on the port.

Take for example a coaxial cable, which supports a TEM mode (and is the mode which I believe you're trying to excite). The properties of this mode depend on the geometry of the setup (inner and outer conductor radii), as well as the filling material. If you only had the inner conductor, the TEM mode would not be excited, as a single conductor on its own does not support a TEM mode. The port needs to know that there's an outer conductor in order to excite this TEM mode, therefore, you need to have a conductor on the outside edge of the port. The inside of the coaxial cable between the inner and outer conductor (typically teflon, but could be anything) also helps determine the properties of the TEM mode, but importantly, it is the region in which the power flows through the cable. The port needs to know that this is a dielectric -- so putting a conductor in front of it makes the port think that it's looking at a giant metallic slab.

So, in the end, you want a port which completely covers the inner conductor, and at least touches (but could also cover) an outer conductor. There needs to be a dielectric (air in your case, I think), between them.

Hi PlanarMetamaterials,

Thank you very much for your insights!

So for this case, which if I need to create lumped port across Signal via and Ground via, do you think it is possible to excite the proper mode?

I've tried hard to comprehend your insights and had a check on those ports.
I think the way I created the port is, it is on the solder resist layer instead of buried inside PCB dielectric. I think this might be one of the cause of not having results like SIwave did.

And one more problem is, Vias are all with facets, I am not able to create a nice rectangular sheet right between them...

Kindly advise.

I'm not really sure what's going on with your "ground vias". This excitation method doesn't seem very practical.

How are you feeding this device? If it's with a coaxial cable, the usual method is to connect the pin to the microstrip trace, and the shield to the microstrip ground.

I suppose some strange connector would be able to implement the four separate ground paths that you have at each end, but then your best best is to model this connector in HFSS as well, and feed it with a regular coax mode.

Thank you Sir for your advice.

You mean for such an uneven distribution of dielectric and metals, it is not possible to force the lumped port to excite TEM right?

Yes you are right Sir. Those holes are supposed for SMA connector for S-parameter measurement use. Let me try if could import a connector and create a lumped port across it.

Ah, yes, one of those face mount types -- I should have recognized that. Yes, creating the model and exciting the coax mode is the way I would do it.

Good Luck