Lossy substrate and thin layers in CST 5.0

[Supaswing]

imaginary wave impedance

Hi all,
I am trying (hard) to simulate planar inductor on silicon in MWS. After long I stick with the two following main problems:

Port definition : the substrate has a resitivity of about 1ohm.cm and I can have trouble to couple the energy correctly in the conductors most of the port energy being dissipated in the substrate. I tried to separate the ports from the lossy substrate bya piece of stripline but then I have a lot of reflections on the lossless/lossy interface and I have no idea how to match them.

Mesh problem : we use several thin layers, dielectric and magnetic and I think have to rely on the PBA even so I cant imagine how the reuslts would looks like
I have port memory exeption when I try to mesh the thin layers manually. By the way, did somebody models thin layer only with one cell, can he tells me the accuray he got?


after countless hours I still can't consider having consitent results even though I sometimes go correct results.

David

 

[asdfaaa]

1. Port definition is not related to the substrate (Silicon here) properties, about the resistivity of the silicon substrate, you should assign it in material properties.
2. For very thin planar metal, there is no need to manually assign the mesh, because normally the default meshing schemes will use TST meshing, which allows in each cell, have thin metal components (maximum number of different materials in each cell is two, if more than two, it will generate errors.) For more details, you may check with the help files about TST.

Regards,

 

[Supaswing]

Thank you for your fast reply but I think I didn't formulate my question in a good way,

The thin layers are in SiO2 not metal, and the magnetic layers, while conductive should be treated as anisotropic material, I want to mesh them because of the strong gradients in the Z directions, but doesnt seems feasible as MWS doesnt seems to bear very small cells compared to the size of the device.
The second point is comes from the fact that the port covers a wide portion of the lossy substrate, exciting more the substrate than the conductor and messing with qTEM modes that should appear in the conductor region.

Regards,
David

 

[asdfaaa]

Thank you for your fast reply but I think I didn't formulate my question in a good way,

The thin layers are in SiO2 not metal, and the magnetic layers, while conductive should be treated as anisotropic material, I want to mesh them because of the strong gradients in the Z directions, but doesnt seems feasible as MWS doesnt seems to bear very small cells compared to the size of the device.
The second point is comes from the fact that the port covers a wide portion of the lossy substrate, exciting more the substrate than the conductor and messing with qTEM modes that should appear in the conductor region.

Regards,
David

1. As my understanding both SiO2 and Si are isotropic material, Si is lossy, and have 1000 Simems/M bulk conductivity. SiO2 is better dielectric material than Si. All those properties you can assign it in materials properties. I can not understand what do you mean for "the strong gradients in the Z directions".
2. For a very thin layer, you can also manually assign mesh properties, you need use larger aspect ratio first, I cant remember the default aspect ratio, but it will work.
3. Even for lossy substrate, your waveport still can properly excite TEM or Q-TEM mode as you want if the waveport is correctly setup. Only difference is the lossless case, wavenumber is pure imaginary, lossy case, wavenumber is complex number, I dont think it can affect the waveport excitation.

Regards,

 

[loucy]

FDTD is not the right choice for simulating the spiral inductor. Both problems you described (dispersion in port mode, and fine geometrical features) are tough to handle within FDTD. It is unlikely that the MWS has any "deep secret" to help you. Go for some MoM tools.

 

[neoflash]

Try HFSS or IE3D on spiral inductor simulation. Normally, they gave much better results.

Hi all,
I am trying (hard) to simulate planar inductor on silicon in MWS. After long I stick with the two following main problems:

Port definition : the substrate has a resitivity of about 1ohm.cm and I can have trouble to couple the energy correctly in the conductors most of the port energy being dissipated in the substrate. I tried to separate the ports from the lossy substrate bya piece of stripline but then I have a lot of reflections on the lossless/lossy interface and I have no idea how to match them.

Mesh problem : we use several thin layers, dielectric and magnetic and I think have to rely on the PBA even so I cant imagine how the reuslts would looks like
I have port memory exeption when I try to mesh the thin layers manually. By the way, did somebody models thin layer only with one cell, can he tells me the accuray he got?


after countless hours I still can't consider having consitent results even though I sometimes go correct results.

David

 

[Supaswing]

Let's come back on this issue, ok?
I agree with Loucy and Neofalsh, a MoM software would be the best choice, even though I tried the evaulation of EM3DS and we could obtain any results from it for our case, due to the strange, non desactivable, too close, shadow lines creations (projections from the edges of the conductor).
But my boss want me to get results with what we have in the company (MWS and Maxwell3D).
I made some new tests by removing the coil and remplacing it by a stripline,
In is this case the electric field under the conductor correspond to the reality.
but under the coil there is almost no electric field, about the same than 2mm fro m the coil
To help the dipsersion at the port in the lossy case substrate, the support of MWS told me to try surrounding the port with ideal material with same wavelength than Silicon??
I tried to replace the Silicon by PEC in the port vicinity, no big change.
In these two case I suspect the reflection is way too high as S12 is always lower than S11 even at low freq.

BTW How do you match lossy to lossless material, it doesnt seem possible to me.

In the coil case I have errors of about 100-200% on all major charatecrics (resonance, inductance, effective resistance).
Does somebody succeed on a similar case or should I run away??

 

[asdfaaa]

Hi, we have done one simulation for planar inductor on silicon-dioxide, I didnt encounter the problems you mention. I would like to take a look of your simulation model.

Regards,

 

[Supaswing]

Yes I send you here the file of one of my air core inductor. It was made with MWS 5.1, the mesh is not so heavy but characteristic of the others.
This inductor should have a inductance of about 0.45uH a Qmax of 8.5 and a frequency resonance of 30MHz.
The filling material has a permittivity of 3.8, the ground is made here around the inductor.
We told me too that the only difference between lossy and lossless case should be a imaginary wave impedance instead of a real wave impedance, can you develop as for me you can't compare, since you have currents in the substrate?

 

[asdfaaa]

HI, I checked and ran the simulation, some issues need to be addressed here:
1, CST wave port dont have capability to treat lossy case because of lack of 2-D eigenmode solving in waveport. Even your case is lossy substrate, but this can not be considered in the wave port excitation. More details you can find it in CST help file.
2, the planar inductor design, your metal layer is thicker than the SiO2 thickness, and the conductivity of silicon is 160s/m. First, check the wawveport excitation on the waveport plane, you can find the most part of energy is still trapped in Si, not SiO2, and because of the conductivity of the Si you choosed, the current is shorted by the Si, that is the reason you can not get any E of current. Possible solution is to increase the thickness of your SiO2 layer, or use high-resistivity Si substrate.

For more discussions, you may give me your email through PM, we can discuss in email.

Regards,