non-symmetric S-parameter in HFSS

Did anybody ever simulate symmetric structure (e.g. CPW, microstrip) in HFSS 9.2 ? It is really ridiculous, the S11 is not the same as S22. Any comment on this?

The reply may not be relevant to the above question directly but I felt I needed to thell that in CST-MWS 5 , I simulated a microstrip and it is clear that S11=S22 and S12=S21 ( symmetrical structure ).

-Arun

wlcsp said:

Did anybody ever simulate symmetric structure (e.g. CPW, microstrip) in HFSS 9.2 ? It is really ridiculous, the S11 is not the same as S22. Any comment on this?

Click to expand...

S11 and S22 will not be equal unless the mesh is also symmetric. Since HFSS uses adaptive meshing, the mesh will almost certainly not be symmetric. FDTD codes uses a cartesian gridding; by its very nature this gridding is symmetric.

However, S11 and S22 should be very close. If they are not close, I suspect you have an error in your setup.

If your model is a device , like power divider, the S11 will not equate to S22.

Wiley said:

S11 and S22 will not be equal unless the mesh is also symmetric. Since HFSS uses adaptive meshing, the mesh will almost certainly not be symmetric. FDTD codes uses a cartesian gridding; by its very nature this gridding is symmetric.

However, S11 and S22 should be very close. If they are not close, I suspect you have an error in your setup.

Click to expand...

I agree with you, I also thought about it. My result shows less than 1 dB difference. However, I was wondering where the HFSS start to generate the mesh. Is there any trick to make the mesh symmetric?

wlcsp said:

Wiley said:

S11 and S22 will not be equal unless the mesh is also symmetric. Since HFSS uses adaptive meshing, the mesh will almost certainly not be symmetric. FDTD codes uses a cartesian gridding; by its very nature this gridding is symmetric.

However, S11 and S22 should be very close. If they are not close, I suspect you have an error in your setup.

Click to expand...

I agree with you, I also thought about it. My result shows less than 1 dB difference. However, I was wondering where the HFSS start to generate the mesh. Is there any trick to make the mesh symmetric?

Click to expand...

I see no way to make it. But the difference between parameters, that must be equal, indicates the value of simulation error. If You have the difference about 0.1 dB - this indicates that the simulation error not less than 0.1 dB. In this case I'm usually replace both parameters by its simple average. Usually this procedure slightly decrease the error and restore the structure symmetry.

Best regards,
Kit

Kit-the-great said:

I see no way to make it. But the difference between parameters, that must be equal, indicates the value of simulation error. If You have the difference about 0.1 dB - this indicates that the simulation error not less than 0.1 dB. In this case I'm usually replace both parameters by its simple average. Usually this procedure slightly decrease the error and restore the structure symmetry.

Best regards,
Kit

Click to expand...

It's a good idea kit. Did you ever compare it with measured data?

even the mesh hfss generated is not symetric. s11 should be very close to s22

try increasing mesh density..I mean seeding the mesh.

Hi,

HFSS allows you to control the accuracy of your solution.
Please do the following:
1) Increase port solution accuracy from 2% to 0.1%
2) Increase solution accuracy from 2% to 1%

In order to speed up the simulation on "simple" problems (CWG,MS,Diff Pairs), do the following:
3) Increase adaptation per pass from 20% to 30%
4) Set seeding on the traces with a minimum mesh of width of the trace.
5) Use Wave Ports in Driven Terminal instead of Lumped ports and use De-embeding. For example instead of simulating a trace that is Lambda long, simulate only 1/4 Lambda and deembed another 3/4 Lambda.

Regards,
Itai Frenkel