ADS - Momentum mesh problem - taking >>1.5hr to finish

[dynaren]

arc resolution ads mesh

The structure is a ring-like microstrip structure with a 90-degree tight coupled curves whose spacing is 0.23mm.

Substrate:
1st layer - the above ring-like structure, 35um thickness, perfect conductor
2nd layer - rogers 5880 with 2 vias that are 2D distributed model,
3rd layer - infinite ground plane with an opening inside.

Mesh settings:
10cells/wavelength in 2nd and 3rd layer(no edge mesh), 20 cells/wavelength in 1st layer (automatic edge mesh)@ 2.5GHz and the arc resolution is 45 degree.

With the above settings, the mesh generation seems >>1.5 hrs.

I opened up the mesh report. There are multiple lines saying that

"Arc refacetting rejected. Refacetted arc violates original connectivity.
Arc: centre 0.000255241, -0.000100787 radius 1047.04 (in MIL)
Arc refacetting rejected. Refacetted arc violates original connectivity.
Arc: centre 0.000255241, -0.000100787 radius 1047.04 (in MIL)
Point of conflict: -436.393, 951.476 (in MIL)....."


With only 8 cells/wavelength for global mesh only took less than 30 seconds to finish.

Is there any ways to reduce the mesh generation time and still keep the simulation accuracy?

Thank you for your time and attention.

 

[Tucco]

ads edge mesh

Hi,
Maybe you can try to make a "defeaturing" of your geometry. If you have circular geometry defined by perfect circles, try to make "convert to polygon" in layout.
Or make a new layout to have less number of segments of all curves.

A another solution is that you have a grid spacing with a too high definition, that is some parts don't touch really the others and Momentum can genere conflict with these problems.

 

[kspalla]

2009 ads momentum problems

Yes by using "union" you can merge layout in a layer. This will reduce the computation time.
just select all and edit---> merge---> union.

 

[dynaren]

momentum mesh

Thank you for your replies.
This structure does being defined by perfect circles and has been united in CST. I export the structure to ADS using dxf.

"Convert to polygon" works perfect!

Thanks again...this saved me a lot of time!

Added after 5 hours 2 minutes:

Btw, I found another way to get quick mesh when the structure is very complicated.
Use File->generate artnetwork in ADS to generate another layout. Only need to replace ports and that's it....20 - 30 cells/wavelength mesh in different layers took just a couple of secs to finish....

 

[si14]

mesh ads layout

If you have circular geometry defined by perfect circles, try to make "convert to polygon" in layout.

Or make a new layout to have less number of segments of all curves.

Yes by using "union" you can merge layout in a layer. This will reduce the computation time.

This structure does being defined by perfect circles and has been united in CST. I export the structure to ADS using dxf.
"Convert to polygon" works perfect!
Use File->generate artnetwork in ADS to generate another layout. Only need to replace ports and that's it....20 - 30 cells/wavelength mesh in different layers took just a couple of secs to finish....

Hi Tucco, kspalla and dynaren:
I have a simple but heavy structure in ADS. The shapes are some u-strip transmission lines and some circles. I have simulated that before, but wanted to test your approaches.

I tested all the mentioned methods, but the number of meshes didn't change.

So basically I think these methods are good for structures imported from other simulators.

Do you happen to use these methods to reduce the simulation time of a structure that sketched in ADS.

Thanks.

 

[Tucco]

momentum meshing 2009

Hi,

In this case, to reduce the mesh generation time, a solution is to increase the arc factor in mesh setup. That is, with 30° Mom create less triangle & edge mesh than with 5 °, for example.

The accuracy of simulation remains good up to a critical value of the arc, function of the curve radius of your geometry. The acceptibility is also function of the process definition. Often, this approximation generate less error that the effective accuracy of the simulator it self !

Compromise is the thing which a designer has to make better !!

 

[si14]

momentum mesher

Hi Tucco:

As you said, one obvious way is to increase the mesh sizes in arcs and edges, therefore reducing the simulation time. But I was talking about the previous mentioned methods in this topic. Have you worked with them? I tried all of them, but no chance in decreasing the simulation time.

Thanks.