how to simplify the substrate in ADS momentum?

[wireless man]

ads momentum slm

Dear All,
Now I am building a transformer in ADS momentum based on 0.13 um RFCMOS process.
The substrate file (.slm) should have 8 metal layers, between every two metal layers, there are 4 dielectric layers. That means there are almost 40 layers in the substrate file (.slm) totally.
However, the ADS Mom cann't afford this substrate calculation, I cann't get the result even after two days running.
What the matter? Is the substrate file tooooo complex? Should I do some simplifications and how to do? Thanks all.

 

[kspalla]

Yes
It is complex as I am doing 18 layer board design with struggle.
The thing is we need a fast machine with huge RAM memory.

Added after 4 hours 24 minutes:

There is shortcut if you say that these 4 dielectric layers are same.
May I know what are these layers?
If possible share the piece of design.

 

[psmon]

If the dielectric layer is much thinner compared to the others, you can ignore the layer and increase the thickness to the dominant layer.

 

[iamdrew]

Are you using all ~40 dielectric layers and 8 metal layers, or can you simplify it? For instance, if you have a ground plane on the 4th metal layer, then you would only need to include half of your layers.

Or, as was previously stated, if the material is the same or similar, you should be able to "merge" them into a single dielectric layer. Although, to me this sounds like it maybe should be solved in a 3D environment.

Good Luck!

 

[wangyufn]

"8 metal layers, between every two metal layers, there are 4 dielectric layers" , looks like TSMC 0.13um RFCMOS. I did not work on such a complex substrate. However, I know some friends used this technology a few years ago and they met a similar problem. At that time their solution is to use some formula to combine the 4 dielectric layers into an effective one. I am not sure about the accuracy, however, you can try it.

Another thing I want to suggest that to check the step that causeing the problem, substrate calculating, meshing, loading matrix or solving? I guess it is loading matrix that killed your time. Maybe you need to review the mesh summary to estimate the needed memory size.

Some tricks mentioned by others are very useful, e.g. bottom metal ground plane.