AnsweredAssumed Answered

How to match momentum simulation with schematic simulation result?

Question asked by Riju on Jan 31, 2019
Latest reply on Feb 6, 2019 by volker_muehlhaus

Like • Show 0 Likes0
Comment • 7

Hello,

I am designing a simple open stub filter which suppresses 2nd and 3rd harmonic frequency and gives a required value of Zin at the operating frequency (or first harmonic) using two quarter wave open stubs designed at 2nd and 3rd harmonic frequency respectively. The schematic has 8 parameters, which I can TUNE to get the desired response.

The problem arises after doing moemtum simulation, the result is almost matching at the lower frequency band but not at the higher frequencies.....dips are shifted and Zin at fundamental freq also slightly detuned.

I have tried to match the momentum result with the schematic response using OPTIMIZATION, but it can't achieve what I want and importantly, it takes huge time.

What is the process to match the momentum simulation results with the schematic exactly in less number of steps?

No one else has this question

Outcomes

volker_muehlhaus Feb 4, 2019 1:05 AM

Are you using your Momentum results in schematic with symbol + emModel?

If you look at emModel settings, there is one tab for interpolation. You can use interpolation for the stub length, so that small changes in stub length are evaluated by interpolation instead of starting another EM run. THis can speed up your EM-based optimization.

Good luck!

Volker

Actions

Riju @ volker_muehlhaus on Feb 4, 2019 7:22 AM

I first simulated and optimized the schematic and generated layout from that optimized schematic. I also tried to optimize the momentum result with the goals being that final optimized schematic result. (as shown in the picture below)....but the momentum optimization is taking long time and it doesn't match with the desired goal after hundreds of iteration.

Is there any "faster" way to match the momentum result with the optimized schematic result?

Actions

volker_muehlhaus @ Riju on Feb 4, 2019 7:39 AM

Consider emModel interpolation which I recommended above. If you open the emModel settings, you will find the interpolation tab. Use meaningful values.

Actions

Riju @ volker_muehlhaus on Feb 5, 2019 11:36 AM

I have run for 19 hrs and only 76 iterations are finished...very very slow! Even after 76 iterations, the result is not matching with the opmised schematic

Actions

volker_muehlhaus @ Riju on Feb 6, 2019 12:07 AM

Have you tried my suggested solution? I try to avoid direct EM optimization and prefer more efficient methods like pertubation technique, but your trivial case should work fine. It seems that your model isn't meshed efficiently.

Don't expect full agreement between circuit simulation and EM. We do EM because it is more accurate (if done properly) and includes coupling effects that are missing in model-based circuit simulation (nodal connection only).

Good luck!

Actions

Riju @ volker_muehlhaus on Feb 6, 2019 12:27 AM

Yes, after enabling the interpretation option with default interpretation range it became much slower.

I have enabled the edge mesh and transmission line mesh.

Could you please elaborate on the perturbation process that you use?

Actions

volker_muehlhaus @ Riju on Feb 6, 2019 12:50 AM

I don't know what you did with emModel interpolation - for my EM work that uses interpolation, it becomes much faster because many EM simulations can be skipped and instead interpolated data is used. However, it is mandatory to use good meaningful values for the "Interpolation Delta".

For pertubation technique, you might search for work by Dan Swanson or Jim Rautio. "Port tuning" is another keyword.

Applied to this case, you would set up a multi-port model where you cut of the very last segment of the open ended stub, and place a port there. Then, use the multi-port emModel in schematic and connect an open ended stub circuit model. You then have 95% of the model EM-simulated, and can tune the residual (very short) open ended stub length delta_l at circuit simulation speed. Then, iteratively include the resulting length delta_l (plus or minus) into the EM part. For the line between the stub, same idea: cut out a small length and place ports, then attach an MLIN in schematic so that you can fine tune the +/- length at circuit simulation speed.

It's difficult to describe, but there is a lot of literature on that topic from the last 10 years or so. This technique has become quite popular with filter designers.

Actions

7 Replies

volker_muehlhaus Feb 4, 2019 1:05 AM
Mark Correct
Correct Answer
Are you using your Momentum results in schematic with symbol + emModel?
If you look at emModel settings, there is one tab for interpolation. You can use interpolation for the stub length, so that small changes in stub length are evaluated by interpolation instead of starting another EM run. THis can speed up your EM-based optimization.

Good luck!
Volker
Like • Show 1 Like1
Actions
- Riju @ volker_muehlhaus on Feb 4, 2019 7:22 AM
  Mark Correct
  Correct Answer
  I first simulated and optimized the schematic and generated layout from that optimized schematic. I also tried to optimize the momentum result with the goals being that final optimized schematic result. (as shown in the picture below)....but the momentum optimization is taking long time and it doesn't match with the desired goal after hundreds of iteration.
  
  Is there any "faster" way to match the momentum result with the optimized schematic result?
  Like • Show 0 Likes0
  Actions
  - volker_muehlhaus @ Riju on Feb 4, 2019 7:39 AM
    Mark Correct
    Correct Answer
    Consider emModel interpolation which I recommended above. If you open the emModel settings, you will find the interpolation tab. Use meaningful values.
    Like • Show 0 Likes0
    Actions
    - Riju @ volker_muehlhaus on Feb 5, 2019 11:36 AM
      Mark Correct
      Correct Answer
      I have run for 19 hrs and only 76 iterations are finished...very very slow! Even after 76 iterations, the result is not matching with the opmised schematic
      Like • Show 0 Likes0
      Actions
      - volker_muehlhaus @ Riju on Feb 6, 2019 12:07 AM
        Mark Correct
        Correct Answer
        Have you tried my suggested solution? I try to avoid direct EM optimization and prefer more efficient methods like pertubation technique, but your trivial case should work fine. It seems that your model isn't meshed efficiently.
        
        Don't expect full agreement between circuit simulation and EM. We do EM because it is more accurate (if done properly) and includes coupling effects that are missing in model-based circuit simulation (nodal connection only).
        
        Good luck!
        Like • Show 0 Likes0
        Actions
        Riju @ volker_muehlhaus on Feb 6, 2019 12:27 AM
        Mark Correct
        Correct Answer
        Yes, after enabling the interpretation option with default interpretation range it became much slower.
        
        I have enabled the edge mesh and transmission line mesh.
        
        Could you please elaborate on the perturbation process that you use?
        Like • Show 0 Likes0
        Actions
        volker_muehlhaus @ Riju on Feb 6, 2019 12:50 AM
        Mark Correct
        Correct Answer
        I don't know what you did with emModel interpolation - for my EM work that uses interpolation, it becomes much faster because many EM simulations can be skipped and instead interpolated data is used. However, it is mandatory to use good meaningful values for the "Interpolation Delta".
        
        For pertubation technique, you might search for work by Dan Swanson or Jim Rautio. "Port tuning" is another keyword.
        
        Applied to this case, you would set up a multi-port model where you cut of the very last segment of the open ended stub, and place a port there. Then, use the multi-port emModel in schematic and connect an open ended stub circuit model. You then have 95% of the model EM-simulated, and can tune the residual (very short) open ended stub length delta_l at circuit simulation speed. Then, iteratively include the resulting length delta_l (plus or minus) into the EM part. For the line between the stub, same idea: cut out a small length and place ports, then attach an MLIN in schematic so that you can fine tune the +/- length at circuit simulation speed.
        
        It's difficult to describe, but there is a lot of literature on that topic from the last 10 years or so. This technique has become quite popular with filter designers.
        Like • Show 0 Likes0
        Actions

How to match momentum simulation with schematic simulation result?

Outcomes

Related Content

Recommended Content