How to choose appropriate mesh size for my EM simulation?

[rd_dinesh]

Mesh size is a trade off between accuracy and simulation time.
So, how do I choose appropriate mesh size for my design?

 

[navet]

All depend on your need and the applications you are targeting ? On which software are you working on ?

I would say in general, in the early stages of developement you will go with a basical mesh configuration and as far as your design is becoming mature (at the end) you will use fine mesh to secure the results.

You should always keep in mind where are located the critical areas of you design, for exemple if your are dealing with microstrip RF coupler or filter coupling effect are part of the desing so you have to be sure that the mesh always fit the complexity of your desing. Some areas can be more sensible to the mesh than others.

 

[BigBoss]

Roughly speaking
You generally start with coarse meshing then you make it a bit finer then observe the difference.
If the difference is small so that means the main convergence has been obtained.
Iterative method is also valid here as well as in every RF problem.

 

[rd_dinesh]

All depend on your need and the applications you are targeting ? On which software are you working on ?

I would say in general, in the early stages of developement you will go with a basical mesh configuration and as far as your design is becoming mature (at the end) you will use fine mesh to secure the results.

You should always keep in mind where are located the critical areas of you design, for exemple if your are dealing with microstrip RF coupler or filter coupling effect are part of the desing so you have to be sure that the mesh always fit the complexity of your desing. Some areas can be more sensible to the mesh than others.

Hello,
I'm designing matching circuit for PA in Keysight ADS.

--- Updated May 17, 2023 ---

Roughly speaking
You generally start with coarse meshing then you make it a bit finer then observe the difference.
If the difference is small so that means the main convergence has been obtained.
Iterative method is also valid here as well as in every RF problem.

Thanks for the idea.

 

[volker@muehlhaus]

matching circuit for PA in Keysight ADS.

Momentum mesh algorithm is really good for such cases. You can use default setting and enabled Edge Mesh, that's it. You can then try more mesh density and compare, but I don't expect relevant differences.

But there is one detail where I have seen many PA macthing network simulation fail: port with at the transistor terminals. Make sure your port representing the transistor terminal has the same width as the transistor lead. In this case, it is wrong to just place the port on the edge because then the entire edge is the port.

See appnote here:

Dr. Mühlhaus Consulting & Software GmbH » Edge/Area pins in Momentum EM simulation

Your RF EDA solution partner since 1993

muehlhaus.com

 

[rd_dinesh]

Momentum mesh algorithm is really good for such cases. You can use default setting and enabled Edge Mesh, that's it. You can then try more mesh density and compare, but I don't expect relevant differences.

But there is one detail where I have seen many PA macthing network simulation fail: port with at the transistor terminals. Make sure your port representing the transistor terminal has the same width as the transistor lead. In this case, it is wrong to just place the port on the edge because then the entire edge is the port.

See appnote here:

Dr. Mühlhaus Consulting & Software GmbH » Edge/Area pins in Momentum EM simulation

Your RF EDA solution partner since 1993

muehlhaus.com

View attachment 182845

Thanks for this.
One question, should I place this edge pin, exactly on the edge or somewhere inside?
or should I construct a separate pad for these terminals and place the edge pin there?
How about using a rectangular port instead?
Should we do the same thing with the capacitors and other components also?

 

[volker@muehlhaus]

As described in the appnote: This topic is most important for the transistor terminals connected to the much wider matching network: large parasitic L from step in width, at a low impedance point.
For the other components it is not so sensitive, so you can use standard port placement there.

Regarding your other questions: it's all answered in the appnote.

 

[biff44]

try a mesh size, and look at the results.
then make it smaller mesh size. run it again, and do the results look the same? if so the original mesh size was small enough. if not, you need to keep running simulations until lowering the mesh size no longer makes the results vary wildly.