Non uniform cross section transmission line

[faraday98]

Hi everyone,

I am trying to model a transmission line that is periodically shunted with an RC load along the transmission line's length. Currently I am trying to do this with HFSS, but not having any success. The application is an integrated circuit design (~CMOS), so most of the dimensions (for every object) is in the um range. The goal frequency is about 25 GHz.

HFSS behaves as if the load is never present. I presume this is because the cross-section where a load is present is different from the cross-section at the wave port. It seems HFSS, in order to calculate the characteristic impedance and the effective index, assumes that the configuration at the wave port is the whole model. Is this correct? If so, is there any method to obtain such information for transmission lines that are not uniform? It would be best to have the characteristic impedance/index of the whole design, but if that is not possible, at least one that varies with position would also be helpful.

I have also tried exporting the RLCG parameters, and then calculating the characteristic impedance. But this was also not successful, as the RLCG parameters also behave as no load is present. I end up getting the same impedance as from HFSS' solution data.

Aside, does the computation of the S-parameters have any limitations with regards to the cross section seen by the waveport? Clearly, that wouldn't make much sense for a 3D solver, but I just want to be sure.

I have tried with both driver terminal and driven modal.

I am also open to using other tools, if they are known to solve this kind of problem.

Thanks.

 

[PlanarMetamaterials]

Hi faraday98,

HFSS should be able to handle this properly, I've done it often myself. If you're using waveports, ensure that the correct mode is being excited at both ports - this may answer your question "does the computation of the S-parameters have any limitations with regards to the cross section seen by the waveport?"

All of your assumptions seem to be correct. What are you using for capacitors?

 

[faraday98]

Hi,

I have tried two different methods for the RC simulation. One method was to make up fictitious material. Knowing the volume of the material, the conductivity (for R) and the permittivity (for C) was calculated to match the R and C of the load. In another attempt, the lumped RLC boundary condition was assigned to a plane, which was placed horizontally touching two vias.

From since my last post, I am thinking that the characteristic impedance (Z0) doesn't make much sense for this structure. This is because, in all definitions I have encountered, the Z0 is only meaningful if the impedance (V/I) is the same for the whole line. Since, here I have loading that is periodic, I don't think Z0 is a proper metric of anything. Rather, I should focus on trying to obtain Zin and Zout and design matching based on that.

Concerning the waveport, I am unsure whether I should include the load in it. I presume this is the correct way, as the mode would change as it propagates and encounters the first shunted load, etc. Or should I include the load in the waveport as well?

Thanks

 

[PlanarMetamaterials]

Periodic structures like this can be a bit tricky. The Z0 refers to the characteristic impedance of what we would call the "host" transmission line - that is, the (CPW? microstrip?) line that exists in the absence of loading. Once periodic inclusions are taken into account, the overall impedance is given by the Bloch impedance. However, if you are exciting the host transmission line, you still should be using the host transmission line's Z0. In this case, you would not want the first loading capacitor to be near the wave port.

When I do this type of simulation, I generally leave a quarter wavelength (or more) of unloaded transmission line between the loaded sections and the waveports. This is so that you only excite and receive the TL mode at the waveports. De-embedding the ports then becomes important for phase measurements.

Hope this helps.