Interpreting S-parameters in Circuit (w/Link to HFSS Design) vs. HFSS

  • Last Post 23 January 2019
Julia posted this 09 January 2019


I am having trouble interpreting the S-parameters in Circuit Designer and HFSS. I am looking at the S21 between a tuned loop antenna (for f=2.4GHz) and two dipoles (Front Dipole and Back Dipole). I have attached the reported values for 15 different HFSS design variations. I am using circuit designer to tune the antenna (the schematic image is included below). The S-parameters computed in HFSS seem more realistic, but the HFSS design alone does not have any of the matching components for the loop antenna so I am concerned these numbers are not accurate for the situation I am trying to simulate. I would anticipate the circuit S21 to be more accurate (and higher) since the loop antenna is tuned for 2.4GHz, but it is much lower. 

How should I be interpreting these S-parameter values in Circuit Designer vs HFSS? What is the "correct" value to report if I am looking at the S21 between the tuned loop antenna and each dipole? 

Thank you!





Order By: Standard | Newest | Votes
Peter Serano posted this 10 January 2019

The S21 between each of the dipoles should be the same in HFSS and Designer since in Designer you're just adding a 50 ohm port to the N-port block in the circuit schematic.

You won't be able to see the real S-Parameters of any tuned coil in HFSS since the tuning takes place in the circuit schematic. When you solve the circuit and push the excitations back into HFSS, the fields in HFSS will be post-processed to take into account the lumped element tuning from the circuit schematic - but there is no way to see these 'tuned' S-Parameters in HFSS.

The only way to get S-Parameters to match in both HFSS and Designer is to have a circuit port tied directly to the N-port block with a short circuit - like you have with the two 'WiFi_Dipole_Back/Front_Feed' ports...

Peter Serano - Senior Application Engineer - High Frequency Products

Julia posted this 10 January 2019

Thanks, Peter! 

So there is no way to see the S21 between the dipole and the tuned coil even in the circuit schematic? What exactly am I seeing if I look at the S21 between the "WiFi" port and the "Back_Dipole" port in the circuit? 

Peter Serano posted this 10 January 2019

You can see the S21 between the dipole and the tuned coil in the circuit schematic by looking at the S21 between the ports 'WiFi' and 'Back_Dipole'. - You cannot see this data in HFSS however since the ports 'WiFi' and 'Back_Dipole' only exist in the circuit realm.

You were asking about comparing S-Parameters between HFSS and Designer, correct? The only S-Parameters that will ever match between the two realms in this case is the dipoles - and that is only because you are directly connecting the HFSS port and the Designer port with a short circuit in your Designer model.

Peter Serano - Senior Application Engineer - High Frequency Products

Julia posted this 10 January 2019

I'm curious about the difference between the S21 values for the tuned coil and dipoles (the S-parameters from the circuit) and the S21 values for the coil that is not tuned and the dipoles (the S-parameters from HFSS). 

I would anticipate the S21 values between a dipole and a coil, both resonant at 2.4GHz, to be higher (less negative) than the S21 values between a dipole and a loop that is not tuned for 2.4GHz. However, the S21 values are higher in HFSS, which would be for the case of a non-resonant coil, than in the circuit. I'm just confused why this might be. 

Peter Serano posted this 11 January 2019

It's hard to say what's going on with the non-tuned coil - in fact I wouldn't even call that non-tuned loop in HFSS a 'coil' - remember when you calculate S-Parameters between any two ports, you are assuming there is a 50 ohm termination on all other ports in the system. So the S21 between the dipole and any of the ports on the loop in HFSS is the result of energy coupling in some non-obvious way that is not necessarily meaningful and/or useful. In this case it could be possible that an arc section of the de-tuned loop is acting like another dipole!

If you'd like to explore what's going on, you can try going into the edit sources dialog in HFSS (Fields -> Edit Sources) and set a 1W excitation for one of the dipoles. Then plot J_Surf_Mag on the surface of the loop coil and animate it vs. phase... You'll be able to see how the energy flows around the ports on the loop coil over the full wave of the excitation. - Keep in mind again that this isn't necessarily meaningful to any real-world function, since in the real world, those ports on the loop coil are not terminated with 50 ohms, they are terminated with capacitors and other lumped element circuits defined in the circuit schematic. - Then you can go back to the circuit schematic, set a 1W sinusoidal power source to drive your dipole port, re-run the linear network analysis, and push the excitations into HFSS and compare how the current flows for the tuned coil system...


Peter Serano - Senior Application Engineer - High Frequency Products

Julia posted this 14 January 2019

That is a good check and I will try to do that to see if the energy is flowing as predicted! To better approximate the physical system should I instead try to include the lumped elements I currently have in the circuit in the HFSS Design? It makes sense that the S21 values in HFSS, as I have it now, are not really meaningful.

Are the S21 values in the circuit a better metric for the path loss between the port in the coil and the dipole(s)? I am mainly concerned because the values I am getting in the circuit seem very low compared to what I have measured in practice. The free-space path loss with the loop and the dipole separated a few meters is typically 20-30 dB lower in the circuit than predicted. 

Thanks again for the help! 

Peter Serano posted this 17 January 2019

I would not bother with redefining any ports in HFSS with LumpedRLC Boundaries. There's no way to model any combinations of lumped elements (like your coil's matching network) with a single LumpedRLC boundary - you'd have to physically model additional ports for each individual lumped element. You'd probably also find that your tuning is slightly off from the circuit model and you'd have to completely re-run the heavy FEM computation whenever changing any of the lumped element values.

In general, in any hybrid Circuit/FEM models, you can safety ignore any S-Parameter data in HFSS. - So the S21 values in the circuit are the *only* metric in this particular model for the path loss between the tuned coil and the dipoles. - Have you checked the S11 of each of the dipoles to ensure they are also tuned properly? - Choosing an analytically derived lambda/2 length for the overall dipole length will get you close but it may not be perfectly tuned. - For the S21 values to be accurate in this case, you should have at least a -15dB match at your center frequency for the S11 of both dipoles and your tuned coil's driving port in the circuit schematic. 

Peter Serano - Senior Application Engineer - High Frequency Products

Julia posted this 21 January 2019

Okay, I figured the S-parameters in HFSS were not very meaningful. I just wanted to make sure since they seemed more realistic than the s-parameters in the circuit. The S11 of each of the dipoles is about -14.8 dB and for the port on the coil with the excitation it is -20.5 dB. 

Are these sufficient enough to have accurate S21 values between the ports? 

Peter Serano posted this 23 January 2019

Yes, that should be sufficient. When I used to build RF coils, -15dB was the minimum S11 acceptable for each coil channel - and there isn't much improvement in performance beyond a -20dB match...

Peter Serano - Senior Application Engineer - High Frequency Products