3

I'm modelling a 50 ohm coaxial cable in HFSS that is excited with a waveport and terminated with a 50 ohm impedance boundary.

The Z0 of the wave port is 50 ohms (refined through tweaking dimensions of the Teflon and pin size around standard SMA 50 ohm cable designs).

The problem is that S11 mag in dB (at 4 GHz) is only -3.26 dB. I'd expect that to be much lower given the 50 ohm impedance boundary.

This result doesn't change when changing the following:

  • Changing the length of the cable (S11 mag doesn't change, but phase does, as expected)

  • Changing port characteristic impedance - Zpi, Zpv,Zvi, Zwave

  • Changing PEC to copper Using curvilinear coordinates for curved surfaces Deembedding (same on and off)

  • Normalizing port to 50 ohms (no change, as expected due to Z0 being very close to 50 already)

  • Mesh density (smallest tried is Lambda refinement of 0.25)

  • Doing a full 3D analysis (not utilizing symmetry)

I feel I'm missing something obvious..

enter image description here

Transistor
  • 175,532
  • 13
  • 190
  • 404
Ian
  • 131
  • 4
  • 1
    I've realized that the "impedance" boundary is designed for sheet impedances, defined as ohm/square.. So I tried with a radiation boundary and got -70dB S11 mag.. But I feel like that's cheating, however it does demonstrate what I hoped to demonstrate – Ian Oct 25 '18 at 21:11
  • 1
    If you have worked out the answer to your question, please post an answer for the benefit of future readers. – The Photon Oct 26 '18 at 05:49
  • I appreciate your suggestion, but unfortunately I wouldn't consider the approach I mentioned above a solution. It's a way of achieving the lack of reflection at that interface, which would be characteristic of a matched resistive load, but it's not the same as modelling such a load at that interface, which would have frequency- and port characteristic impedance- dependent variability. Others have suggested the only way forward is to model with parallel distributed resistive loads to achieve 50 Ohms. That may be the true solution, but I've not yet implemented that to check. Will report back – Ian Oct 29 '18 at 15:45
  • It sounds as if you are trying to use the wave port as a model for some actual load you have attached. The wave port, when configured correctly, must have a low return loss. Whatever it is you have at the termination of the coax must be included in the model. One approach could be to model the coax to planar transition (if that’s what you have) and add the complex impedance in post processing of the s-parameters. – electroGeek Jan 14 '22 at 06:38

0 Answers0