[SOLVED] Simulation Vs Measure for best results

[guytoub]

Hello,

how do you manage to configure ADS correctly in order to have simulations that correspond to your measurements?
What is the degree of mesh during your simulations that give you good results?
I would like to design an antenna and have it manufactured by a company but I would like a repeatability of 95%.
For the moment my results are not convincing and I do not know which parameters to use.
Subtract values provided or simulation parameters.
Thank you for your help .

 

[volker@muehlhaus]

Your question is very generic, and there are many possible mistakes in using EM tools. Use a mesh that is fine enough where it matters, and make sure your ports (including ground ref) are defined properly.

In recent ADS releases, there are different emSetup templates for different applications. Settings for a 5 GHz PCB are obviously different from a 200 GHz on-chip antenna simulation.

It is a good idea to look at current densities from simulation, this might show if mesh is fine enough.

You might also want to check if Keysight offer Momentum trainings these days (possibly online?)

 

[guytoub]

Your question is very generic, and there are many possible mistakes in using EM tools. Use a mesh that is fine enough where it matters, and make sure your ports (including ground ref) are defined properly.

In recent ADS releases, there are different emSetup templates for different applications. Settings for a 5 GHz PCB are obviously different from a 200 GHz on-chip antenna simulation.

It is a good idea to look at current densities from simulation, this might show if mesh is fine enough.

You might also want to check if Keysight offer Momentum trainings these days (possibly online?)

i had this problem with antenna array at 10GHz using RO4003 Substrate

 

[volker@muehlhaus]

RO4003 has tight tolerances, and you should see very good agreement if your Momentum setup is correct.

Try finer mesh
Visually check current density
Make sure you have defined thick metal model (expand into dielectric above)
If you have solder resist on your board, make sure to include that in Momentum

 

[BigBoss]

Take the Manufacturing Tolerances into account.
A good design is a insensitive one.So, sensitivity has to be minimized during manufacturing,Therefore there are some golden rules such as..
-not to use very thin lines
-not to use very close coupled lines
-avoiding sharp/acute corners

etc.In order to verify the sensitivity of your design against manufacturing, you'd better do Monte Carlo Analysis ( very time consuming ).

 

[guytoub]

Take the Manufacturing Tolerances into account.
A good design is a insensitive one.So, sensitivity has to be minimized during manufacturing,Therefore there are some golden rules such as..
-not to use very thin lines
-not to use very close coupled lines
-avoiding sharp/acute corners

etc.In order to verify the sensitivity of your design against manufacturing, you'd better do Monte Carlo Analysis ( very time consuming ).

did monte carlo also works with antenna design?

--- Updated Apr 30, 2021 ---

RO4003 has tight tolerances, and you should see very good agreement if your Momentum setup is correct.

Try finer mesh
Visually check current density
Make sure you have defined thick metal model (expand into dielectric above)
If you have solder resist on your board, make sure to include that in Momentum

thanks volker .
they told me that for a planar antenna i could mesh with Lambda G/10 grid

 

[volker@muehlhaus]

they told me that for a planar antenna i could mesh with Lambda G/10 grid

Who is "they"?

lambda/10 is very coarse, do not expect good results with that setting. Even the default lambda/20 is too coarse in several cases. For accurate results, fine enough mesh density is really important.

 

[guytoub]

Who is "they"?

lambda/10 is very coarse, do not expect good results with that setting. Even the default lambda/20 is too coarse in several cases. For accurate results, fine enough mesh density is really important.

you've right , i tried Lamba/40 and i've got good result . i've got something like 100 Mhz beetween the simulation and the measure. I think the connector has something to do with it because I don't simulate it.