ADS Layout - tune parameters

[jdstavares]

Hi everyone!

I'm trying to tune some parameters after layout, but whenever I try to do it, this error comes along: "Warning detected by hpeesofsim during HB analysis `HB1'. COMPONENT `X1.em_data': Simulator uses frequency higher than frequency range for model. The S-parameters are extrapolated. †"

In frequeny plan layout I have 2.45 GHz and 0 GHz and in the schematic I have both the HB and LSSP pointed to 2.45 GHz. So not sure what is happening.

 

[RealAEL]

What do you have set for the Order in the HB simulation? Your EM simulation frequency range must include all harmonics specified and not just the fundamental frequency.

 

[jdstavares]

Oh got it!
But, I tried now using order 1, to consider only the fundamental freq. and still can't perform the tuning, it claims that I need to simulate the cell first, which I did...

Error detected by hpeesofsim during HB analysis 'HB1'. Fatal error occurred while evaluating model MomCmpt',∈stanceMomCmpt′,∈stanceX1.em_data'. †

 

[BigBoss]

RE-simulate the EM structure one more time, I think the database is corrupted.
Use 7 harmonics or more in HB simulations to get more accurate results. Fundamental frequency only will give you a bit erroneous result. Therefore, simulate your EM structure up to the last harmonic frequency to obtain accurate and consistent result.
For instance, you work on 2.45 GHz and 7th harmonic becomes around 18 GHz and use DC to 18 GHz in Momentum set-up. (Microwave mode) and then update the EM database.

Update: Do you try to tune an EM structure without parametric dimension? It's not possible.
You have to insert associated dimensions into cell definition as parameter so you can slide this variable during EM simulation. It's a bit difficult but no way to tune an EM structure.

 

[jdstavares]

I did that, but now I'm having huge difference between the results in the schematic and the results over the layout. From the schematic I had efficiency of 30%, and now the efficiency is basically 0. Moreover, the output is also pretty bad compared with the schematic. Is this normal?

I believe that some differences may happen, but not as huge as this.

 

[BigBoss]

I did that, but now I'm having huge difference between the results in the schematic and the results over the layout. From the schematic I had efficiency of 30%, and now the efficiency is basically 0. Moreover, the output is also pretty bad compared with the schematic. Is this normal?

I believe that some differences may happen, but not as huge as this.

This is a different story. If there is a huge difference between schematic and layout version of the circuit, there is absolutely an error while transferring the schematic onto layout or EM setup.
Either substrate or simulation set-up itself might not be appropriately organized.

 

[jdstavares]

This is a different story. If there is a huge difference between schematic and layout version of the circuit, there is absolutely an error while transferring the schematic onto layout or EM setup.
Either substrate or simulation set-up itself might not be appropriately organized.

Ok... and how can I find that error?

 

[D0D]

Have you solved the problem? I have the exact same warning as you！

 

[jdstavares]

Have you solved the problem? I have the exact same warning as you！

Sorry can you specify which problem, because I posted more problems (in this same thread)

 

[rd_dinesh]

Oh got it!
But, I tried now using order 1, to consider only the fundamental freq. and still can't perform the tuning, it claims that I need to simulate the cell first, which I did...

Hi, How did you solve this error? Even I got the same error and unable to resolve the issue

Error detected by hpeesofsim during HB analysis 'HB1'. Fatal error occurred while evaluating model MomCmpt',∈stanceMomCmpt′,∈stanceX1.em_data'. †

View attachment 179612

 

[rd_dinesh]

How to resolve this error while tuning my layout?

 

[volker@muehlhaus]

The solution is what the message says: before using manual tuning, you need to run a parameter sweep over the entire range, so that the emModel is filled with pre-computed results from the sweep. Only then, with existing EM data for all parameter values, you can use real time tuning.

 

[rd_dinesh]

The solution is what the message says: before using manual tuning, you need to run a parameter sweep over the entire range, so that the emModel is filled with pre-computed results from the sweep. Only then, with existing EM data for all parameter values, you can use real time tuning.

should I sweep my parameters in layout?(is there any way to do so?)

or

should I sweep my parameter values in em cosimulation schematic?

 

[volker@muehlhaus]

It means that instead of tuning the emModel in schematic, you should first run a parameter sweep of you emModel in schematic. This will take time, but it fills the emModel with data for different parameter values. You need to set sweep range and step to the values that you use later in real-time tuning.

I haven't used real-time tuning for emModel myself. But all information is given in the ADS dialog that you showed!

--- Updated May 4, 2023 ---

PS: If you look at your emModel settings, there is also an option to use interpolation, so that tiny differences in value are computed from existing emModel data instead of running yet another EM simulation. But use that with care!

 

[rd_dinesh]

It means that instead of tuning the emModel in schematic, you should first run a parameter sweep of you emModel in schematic. This will take time, but it fills the emModel with data for different parameter values. You need to set sweep range and step to the values that you use later in real-time tuning.

I haven't used real-time tuning for emModel myself. But all information is given in the ADS dialog that you showed!

--- Updated May 4, 2023 ---

PS: If you look at your emModel settings, there is also an option to use interpolation, so that tiny differences in value are computed from existing emModel data instead of running yet another EM simulation. But use that with care!

I have almost 20+ parameters, how can I sweep all these variables in em cosimulation schematic.
By the way this will also ruin my existing data display file. I have to rewrite all my trace equations.

My peers suggest to do optimization.
But my doubt is how will optimization work when tuning was unable to.
Finally is there a good way to make my em model complete for real time tuning before I use it in my cosimulation schematic?
Thanks for your quick response.

 

[volker@muehlhaus]

I have almost 20+ parameters

From 20 years experience as an EM simulation expert: direct EM-based tuning or optimization is the wrong approach with so many parameters. It will take just too long.

Finally is there a good way to make my em model complete for real time tuning before I use it in my cosimulation schematic?

You already asked that, and the answer is above. There is no magic solution. For your case, you better forget real time tuning. Optimization is indeed better, if you can start from values that are close to target, because then you don't need to pre-calculate all possible parameter combinations.

My advice is to invest a few weeks thinking about more efficient workflows, where you don't need to EM-tune over 20 parameters. Brute force EM is not a practical solution.

Best regards
Volker (Keysight certified expert EDA)

 

[rd_dinesh]

From 20 years experience as an EM simulation expert: direct EM-based tuning or optimization is the wrong approach with so many parameters. It will take just too long.



You already asked that, and the answer is above. There is no magic solution. For your case, you better forget real time tuning. Optimization is indeed better, if you can start from values that are close to target, because then you don't need to pre-calculate all possible parameter combinations.

My advice is to invest a few weeks thinking about more efficient workflows, where you don't need to EM-tune over 20 parameters. Brute force EM is not a practical solution.

Best regards
Volker (Keysight certified expert EDA)

Thanks sir, may be I need to think about it.