Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Huge difference between results from schematic simulation and momentum co-simulation

Status
Not open for further replies.

dangan1993

Junior Member level 3
Joined
May 29, 2020
Messages
28
Helped
0
Reputation
0
Reaction score
0
Trophy points
1
Activity points
214
Hi everyone,
I am trying to design a dual-band Class E power amplifier at 2.4/3.5 GHz using only transmission lines for input matching and output matching networks. In the first step, I designed and simulated the circuit on ADS schematic using microstrip lines as in Figure 1

Fig1microstrip_simulation.PNG

I extracted the results, the performance at both frequency bands was as expected (about 5.3 W and over 70% PAE).

After that, I had created layout look-like components for input matching network and output matching networks individually, then append them to discrete elements (VDC, transistor and two terminations) and perform EM co-simulation.

To be carefully, I just used a EM model of input matching network, the output matching network is still in the microstrip version (like Fig 2)

Fig2.layout_cosimulation.PNG

The performance is significantly decreased (about 0.5 W at 2.4 GHz and 1.7 W at 3.5 GHz). I had checked input impedance of the network, it changed a lot, and this huge mismatch leads to power degradation.

Then, I tried to optimize the EM input matching network, the performance can recover to acceptable value, but the layout is changed significantly, compared to that of microstrip schematic.
I am so worry about that phenomenon :thumbsdown:, and whether I had made mistakes in momentum co-simulation or is this normal in high-frequency circuit design?
Note that, the similarity happened to the output matching network.
Thank you for your time.
 

The crossover schematic model accuracy might not be super accurate for such very different widths, but should be acceptable. Most likely this is a mistake in your EM modelling. Can you upload an archived workspace? Would be nice if that is reduced to this relevant part, and does not include too much unrelated stuff.
 
The crossover schematic model accuracy might not be super accurate for such very different widths, but should be acceptable. Most likely this is a mistake in your EM modelling. Can you upload an archived workspace? Would be nice if that is reduced to this relevant part, and does not include too much unrelated stuff.

Thank you for your reply.
The link below contains my workplace with two files Dualband_ClassE and TLY_Dualband_ClassE, respectively.
Can you help me identify the problem?
https://drive.google.com/file/d/1A7tJltP9RyqPaabYKYXhsIRDxBwcst9N/view?usp=sharing
 

Crossover looks like extreme.. Correct your matching circuit.
 

Crossover looks like extreme.. Correct your matching circuit.

Hi BigBoss,
Thank you for your conment. But how can I correct it. I think it is common to include a cross at intersection of lines to connect them?
 

What ADS version is that? ADS 2009? I imported your ADS2009 workspace to ADS 2020. I couldn't see any obvious mistakes in EM modelling, but of course didn't see your original ADS 2009 settings.

- - - Updated - - -

But as already mentioned, the crossovers are rather extreme and you should not expect the nodal circuit models (equation based) to be as accurate as EM simulation. The MCROSS adds relevant extra length to each branch for such extremely different widths.

I will not spend more time on your model because you can't read my (possibly modified) workspace back into your very old ADS version anyway.

mcross.jpg

mcross2.jpg
 
Last edited:

What ADS version is that? ADS 2009? I imported your ADS2009 workspace to ADS 2020. I coulnd't see any obvious mistakes in EM modelling, but of course didn't see your original ADS 2009 settings.

- - - Updated - - -

But as already mentioned, the crossovers are rather extreme and you should not expect the nodal circuit models (equation based) to be as accurate as EM simulation.

View attachment 159631

View attachment 159632

Hi volker,
Unfortunately, I am using ADS 2009, and I am not sure whether the files can be imported correctly in newer version. But like you said, I can accept the much worse result of EM simulation as a normal thing?
 

You haven't shown what the difference is, but yes: EM should be more accurate here.

What I would recommend: do a careful investigation of the matching network sections (only that) and compare circuit model S-parameters to EM simulation results.
 
Last edited:

You haven't shown what the difference is, but yes: EM should be more accurate here.

What I would recommend: do a careful investigation of the matching network sections (only that) and compare circuit model S-parameters to EM simulation results.

Hi volker, I already checked the S-parameters of each individual section, and extracted their input impedance, EM model give different impedance, compared to that in circuit model, and this is reason leading to power degradation in my circuit.
In my opinion, I expect the EM simulation result is similar (little bit changed) to the circuit simulation result, and then the expected performance can be easily recovered under an EM optimization scheme.
But the huge power decrease makes me curious whether either it is a common phenomenon in microwave circuit design or it is an error from my simulation settings.
 

Hi volker, I already checked the S-parameters of each individual section, and extracted their input impedance, EM model give different impedance, compared to that in circuit model, and this is reason leading to power degradation in my circuit.

You didn't share your results for the matching network, so I tested the input matching network. Indeed, results are different. I don't see a mistake in modelling, but maybe someone else does.

Parameters between schematic view and layout do agree.

inputmatch_schematic.jpg

inputmatch_layout.jpg

testbench.jpg

testbench_result.jpg


substrate.jpg
 

Electromagnetic Planar ( other 3D ones too ) Structures have a "range of validity" and models are based on this validity.
This looks like a simple SPICE model.It has also a validity region and cannot be used out of the range..
 

Hi volker, you tested S11 and S22 in the same layout input matching network. Can you explain why the results are different?
 

Hi volker,
Sorry for my late response.
The below is reflection coefficients that I did for circuit model and layout model, respectively. You can see, the result is different to each other that make my circuit not to work well in EM simulation. It seems to have a frequency shift. It is normal?

S_Parameter.PNG

results.PNG
 

I believe that you see large differences with the said "extreme" microstrip geometry. Unfortunately I can't locate a comparison of both results in your threads.

Like Volker, I would trust the EM simulation in the first place. There's however a prerequisite, does the port connection geometry in your simulation correspond to the real hardware, e.g. a wide trace with a "point" connected 50 ohm port. If not, the EM simulation may be equally useless.
 

Hi volker, you tested S11 and S22 in the same layout input matching network. Can you explain why the results are different?

I compared emModel view and schematic model view, as shown by the red arrows.

My only explanation is what was written above: the crossovers are rather extreme and you should not expect the nodal circuit models (equation based) to be as accurate as EM simulation. The electrical model behind the MCROSS schematic model is rather simple, see ADS help for that model. EM can be more accurate, as it solves the actual geometry.

And indeed, what FvM mentioned is also important if you have large step-in-width between feedline and transistor:
https://muehlhaus.com/support/ads-application-notes/edge-area-pins
In your EM model, the port width is the same as the line width.
 
Last edited:
  • Like
Reactions: FvM

    FvM

    Points: 2
    Helpful Answer Positive Rating
What I forgot to mention: To access the different views from the same schematic symbol, I used the "Set View for Simulation" functionality. This is a new functionality in ADS 2011 or later.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top