[SOLVED] Simulation of Submicron Passive Components

[BigBoss]

Generally I use ADS's Momentum in the simulations of deep-submicron passive components. But again generally it does not converge and spends many hours with huge memory requirements even working on UNIX or PC.
Sometimes (indeed often ) results are terrible regarding to measurements.

Is there anybody have some expriences deep-submicron passive planar components that are layed-out highly ohmic silicon substrate ?

What can he/she offer to acquire better results ?

( Sorry if any topics were opened before )
If it's necessary , I can add DXF or GDS2 file.

Good Luck..

 

[wave-maniac]

Edited.

 

[BigBoss]

The dimensions are 1.8mmx1.5mm. The fact that is a impedance matching circut and a LC Balun that work at 2GHz.
There are many vias that have 20um diameter to connect metallic layers which are separated by 0.5um .
Substrate 300um height on Si material.

Problem is meshing takes many time and memory. Obviously momentum
does not start to simultae because memory limits are over while meshing.

I'm searching another solution.
Good Luck

 

[wave-maniac]

Edited.

 

[loucy]

In terms of meshing, I think Momentum is the best among the commercial 2.5D MoM codes.

You should first precompute the mesh and see if it hangs. I doubt it is a memory problem. Please upload the layout for us to learn.

 

[BigBoss]

I have attached gds2 file and momentum layout definitions.
Please check and put your comments.
GL

 

[wave-maniac]

Edited.

 

[wave-maniac]

Edited.

 

[BigBoss]

But this structure is not a simple inductor . It consists of spiral inductor and a capacitor that is connected by very small vias. Therefore , sthe structure could be tought as 3 dielectric layers and 2 metallic layers. I have simulated this one and obtained nonsense S parameter response like S11>1 etc.
I think there is a trick that I couldn't find but How..
GL

 

[wave-maniac]

Edited.

 

[BigBoss]

I have attached this as project.zap file.

 

[loucy]

Is this related to the problem of MoM codes at low frequency?

 

[loucy]

Is there a setting in Momentum for selecting Loop-star basis, which is more suitable for low frequency.

I would also try the RF mode of momentum.

 

[BigBoss]

Strangely, I select linear frequency sweep 1.7GHz-2.2GHz but momentum sweeps different frequency points such as 0GHz, 3.6GHz,6.8Ghz etc. And then simultes the structure and returns back results and some errors.

I really do not understand.

Because the same structure could be simulated many times in version 2002 both UNIX and PC version.

When I used RF mode results are more realistic.

Either there is a bug or setup problem.

Anyway , I will try it in Sonnet to find a solution.

Thanks a lot.

 

[wave-maniac]

Edited.

 

[loucy]

I think the RF-mode use some basis which are similar to the so called "loop-start" basis. It performs better if the cell size is much much smaller than the wavelength. For the usual rooftop basis, the moment matrix is ill conditioned. It is not easy to get an accurate solution.