Some problem with simulation by SONNET IE3D and HFSS

[xd2186]

Hi every guys.
I am performing a simulation optimization for a LTCC filter,But very different result between IE3D and HFSS&SONNET, SONNET and HFSSresults were almost identical.Pictures are as follows.

Why does this happen? who can help me? Thanks in advance.:-x

 

[volker@muehlhaus]

Why does this happen?

From my experience with LTCC EM simulation, results should agree much closer. There is some difference in your models.

 

[xd2186]

All model was created by DXF output of IE3D, maybe some wrong with HFSS&SONNET setting?

 

[volker@muehlhaus]

All model was created by DXF output of IE3D, maybe some wrong with HFSS&SONNET setting?

DXF gives you the polygons in the x-y plane, but there can be mistakes in the stackup, or in the modelling of conductor thickness, or boundary condition, or mesh. Sonnet and HFSS results are also very different if you look at the upper stopband. Can you upload the Sonnet model?

 

[xd2186]

DXF gives you the polygons in the x-y plane, but there can be mistakes in the stackup, or in the modelling of conductor thickness, or boundary condition, or mesh. Sonnet and HFSS results are also very different if you look at the upper stopband. Can you upload the Sonnet model?

That is Sonnet model and HFSS model.thanks

 

[BigBoss]

The Result of IE3D is "too good"..Others seem more realistic..

 

[volker@muehlhaus]

In Sonnet you have defined 3µm conductor thickness for loss, but used thin sheet (zero thickness) for the actual EM model meshing.
You have defined 45µm dielectric, is that between the metal conductors? Or do the 3µm metals penetrate into the dielectric, for an effective 45µm - 3µm = 42mm distance in the capacitors?

 

[volker@muehlhaus]

Your Sonnet model is essentially OK. You should use thick metal modelling and a smaller cell size, that has some effect.
I have used Sonnet a lot for LTCC work and trust results - if the layout and stackup are accurate.




Updated models:
View attachment bpf2070_5_volker.zip

 

[xd2186]

The Result of IE3D is "too good"..Others seem more realistic..

I used IE3D for some microwave planar circuits such as microstrip filters, it is very accuracy. Why is IE3D inefficient for 3D circuits(LTCC)?

 

[xd2186]

In Sonnet you have defined 3µm conductor thickness for loss, but used thin sheet (zero thickness) for the actual EM model meshing.
You have defined 45µm dielectric, is that between the metal conductors? Or do the 3µm metals penetrate into the dielectric, for an effective 45µm - 3µm = 42mm distance in the capacitors?

I have considerd it.The ceramic layer thickness is 42.5um(greentape is dupont 951 50um),conductor is 3um.In my experience,just defined 45um.

 

[xd2186]

Your Sonnet model is essentially OK. You should use thick metal modelling and a smaller cell size, that has some effect.
I have used Sonnet a lot for LTCC work and trust results - if the layout and stackup are accurate.

View attachment 148908
View attachment 148909

Updated models:
View attachment 148910

For your suggestion,sonnet setting is alright, I try to find my IE3D incorrect setting. Thanks for your help, it is so kind of you.

 

[xd2186]

I have compared results of IE3D and sonnet, This is example of IE3D, I have exported dxf by IE3D and created model in sonnet.
There are some different between them yet.

 

[volker@muehlhaus]

I don't know how good and accurate that IE3D model is. Here is an LTCC appnote that my colleague created long ago with Sonnet. That includes measured vs. simulated.

View attachment MWEE_JuneJuly03.pdf

However, in your last Sonnet model you have some poor connections (round via in square cutout).



Also, don't forget that Sonnet analyzes in a metal box (=ground). You might want to keep some more distance to the side walls. Some of your lines are rather close and parallel to the box side walls.

 

[xd2186]

It is my fault.I'm so careless.For your suggestion, I modified the sonnet model.Result difference between sonnet and IE3D is small now.


I will continue to find mistakes in IE3D model.

Thanks you very much!

 

[volker@muehlhaus]

One area that you can check: the mesh in overlapping metal polygons. Only if the mesh is fine enough or aligned, the correct charges can be placed on the mesh and capacitance is accurate.

Sonnet does that mesh alignment on adjacent layers automatically. For IE3D, you can visually check the mesh: it should have mesh boundaries where the capacitor metals overlap. This allows to place charge in the large metal plates with the exact size of the overlapping smaller polygon.

 

[xd2186]

One area that you can check: the mesh in overlapping metal polygons. Only if the mesh is fine enough or aligned, the correct charges can be placed on the mesh and capacitance is accurate.

Sonnet does that mesh alignment on adjacent layers automatically. For IE3D, you can visually check the mesh: it should have mesh boundaries where the capacitor metals overlap. This allows to place charge in the large metal plates with the exact size of the overlapping smaller polygon.

The problem hasn't been solved, thank you all the same!

 

[xd2186]

One area that you can check: the mesh in overlapping metal polygons. Only if the mesh is fine enough or aligned, the correct charges can be placed on the mesh and capacitance is accurate.

Sonnet does that mesh alignment on adjacent layers automatically. For IE3D, you can visually check the mesh: it should have mesh boundaries where the capacitor metals overlap. This allows to place charge in the large metal plates with the exact size of the overlapping smaller polygon.

I have compared difference of current distribution between IE3D and sonnet, but it look like simular. I have no idear:bang:

Current display parameter2GHz 0deg.

 

[volker@muehlhaus]

For the capacitors, we want to see charge (not currents). Sonnet can plot charge, for IE3D I don't know.

The problem that I mentioned is seen when capacitor plates have different size. For this case with equal sizes, there should be no mesh alignment issue anyway.

The IE3D currents look a bit coarse to me.

 

[xd2186]

For the capacitors, we want to see charge (not currents). Sonnet can plot charge, for IE3D I don't know.

The problem that I mentioned is seen when capacitor plates have different size. For this case with equal sizes, there should be no mesh alignment issue anyway.

The IE3D currents look a bit coarse to me.

I can't find charge display module in IE3D, be insteaded by vector current.I have plot charge response by sonnet,I don't know if it is correct.

 

[xd2186]

For the capacitors, we want to see charge (not currents). Sonnet can plot charge, for IE3D I don't know.

The problem that I mentioned is seen when capacitor plates have different size. For this case with equal sizes, there should be no mesh alignment issue anyway.

The IE3D currents look a bit coarse to me.

SONNET optimization ability is unsatisfactory ,only one mothed to optimize.How can I improve it .
It seem like AWR and IE3D ability to optimize obvious much better than sonnet.