Seeking Help with 3-Segment Stripline/MSL Stepped Impedance Matching Transformer Simulation

[Sarajevo67]

Hello,

I hope this message finds you well. I'm reaching out for some assistance with a Stripline/MSL simulation I've been working on, and I was wondering if someone could lend me a hand.

I'm currently trying to simulate a 3-segment Stripline/MSL stepped impedance matching transformer, but I'm facing some challenges. Despite going through OpenEMS tutorials and examples and spending a lot of time trying to adapt them for my specific needs, I haven't been able to find a solution that fits my requirements. I must admit that I'm still relatively new to OpenEMS, so I apologize in advance if I've missed some basics along the way.

I wanted to share with you the basic requirements and parameters for this project:

• I'm considering both Stripline and MSL designs, but I might prefer Stripline if it leads to significantly faster simulations.
• The transformer has 3 segments with different widths (W1, W2, W3) and lengths (L1, L2, L3) starting from port(1).
• The overall length of the transformer is the sum of L1+L2+L3.
• To achieve wideband matching, I have antenna measurement parameters in the form of [Freq. R+iX] pairs at 64 different frequencies.

EDIT/01-08-2023/ This is rough sketch:

I have two specific questions that I'm hoping you can help me with:

1. How can I construct and mesh the 3-segment SL/MSL? The examples I've come across mainly deal with port attached lines and offset measurement ports, so I'm unsure how to adapt them for my case.
2. Is there a way to present the known antenna measurements to port(2)? If not possible, could you guide me on how to measure (on port(2)) the propagated reference excite impedance (U/I) from port(1)?

I truly appreciate any guidance or insights you can offer. I understand that my questions might seem basic, but I'm eager to learn and progress with this project.

Thank you for taking the time to read this message.

Best regards,

DD

[evanfoss]

Hi DD, Could you give a rough sketch of this? What I am picturing is not too different from the example I did here. http://www.repo.hu/cgi-bin/pool.cgi?project=pcb-rnd&cmd=show&node=s_param Best, Evan

…

On Tue, Jul 25, 2023 at 8:34 PM Sarajevo67 ***@***.***> wrote: Hello, I hope this message finds you well. I'm reaching out for some assistance with a project I've been working on, and I was wondering if you could lend me a hand. I'm currently trying to simulate a 3-segment Stripline/MSL stepped impedance matching transformer, but I'm facing some challenges. Despite going through OpenEMS tutorials and examples and spending a lot of time trying to adapt them for my specific needs, I haven't been able to find a solution that fits my requirements. I must admit that I'm still relatively new to OpenEMS, so I apologize in advance if I've missed some basics along the way. I wanted to share with you the basic requirements and parameters for this project: - I'm considering both Stripline and MSL designs, but I might prefer Stripline if it leads to significantly faster simulations. - The transformer has 3 segments with different widths (W1, W2, W3) and lengths (L1, L2, L3) starting from port(1). - The overall length of the transformer is the sum of L1+L2+L3. - To achieve wideband matching, I have antenna measurement parameters in the form of [Freq. R+iX] pairs at 64 different frequencies. I have two specific questions that I'm hoping you can help me with: 1. How can I construct and mesh the 3-segment SL/MSL? The examples I've come across mainly deal with port attached lines and offset measurement ports, so I'm unsure how to adapt them for my case. 2. Is there a way to present the known antenna measurements to port(2)? If not possible, could you guide me on how to measure (on port(2)) the propagated reference excite impedance (U/I) from port(1)? I truly appreciate any guidance or insights you can offer. I understand that my questions might seem basic, but I'm eager to learn and progress with this project. Thank you for taking the time to read this message. Best regards, DD — Reply to this email directly, view it on GitHub <#100>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AATDEXVOJOINSJPRAC3WAUTXSBQ2JANCNFSM6AAAAAA2X2CFTQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- https://github.com/evanfoss

-----BEGIN PGP PUBLIC KEY BLOCK----- Version: GnuPG v2 mQENBFYy4RYBCAC183JomLtbdAlcKiaPDoVHq52LDmVmH75aiEc69m7YxDt54/ai VtYCAobbGVIyn3Hlz3uhF6LnPl/6Lm1VdnCfpwu3KQhCO6ds10ow2C30X4ohCqOd hCVg5C+ILmQkEffFrFODy3ji+PYTF4pADvHCWsTMv0hf0llwFOJsBCK6cl02IffE JPqy4PjM1nZ9HpzT84JBaG/4OGvTZ8SQ2yFUl265jagvygPTf88H1xpZHH1r8dB1 stjUHLmPH8AOyDgKxFchgGeDc3p/vJtgDDIXAFfDXG0NSRovLmtaQdGxe47Zf/go bXiEM7YL2WqQe5zfEA919JxkEwlDKYniOSVzABEBAAG0N0V2YW4gRm9zcyAoVGhp cyBpcyBteSBwdWJsaWMga2V5LikgPGV2YW5mb3NzQGdtYWlsLmNvbT6JATkEEwEC ACMFAlYy4RYCGwMHCwkIBwMCAQYVCAIJCgsEFgIDAQIeAQIXgAAKCRCIpQTcE8nN bbBaCACAm8pU5lG1ev2Fsw68Axtcl57SJrYieqX96c3YuYH9JpqMqJRnd9nDKw9X tQuvuH7tUk0VbOaDqReOYJVI/4c5wb9AaOFp6K2DUcupq6XhgXpvz3HzoPwjAdIj XuQzdRUx5+innTJrSkGuBYW/CZ2zqEx4xfLlq4rO0hoTUMR8QVp2cCrkw6BT0m86 APIw/ZnjoxM8IEzr7MxfRIg3qpzrZk28rmhx+k78Jyk61UhwcCPGIm/pjUopTwYJ 3YBdRB2cYD2aN7A1JVf5cRmSQYooHBGpH0kYvomGk97PKqypVuJ7OpG9xM58wUcC qUVt9hKlePLzP8csYjt8onqI7qIIuQENBFYy4RYBCADlH8spG3WkCx62vB5mr5Z0 SCDd/RcyA4A5y5EOj5KurQkrSWpgi9Ho1yKruMJ6blQR2qkc66KqH9pnXDm/ZI1M K/wdW3ngETxBmXoozzFMT89aEWIVR5/PFodWK1elekE9iJxACuR98Zg2QttTD3x8 A9w8VEyMLOXcDTrPFpHegMKswFBg5iuMulAdXAoGejWTI3n+qKFpabHm2Lfs6wjk 5rjucpTdeFK6UeWF1xAvNxXibuu5BlGwv53930qIXRwO/Gn2Rh5DXWxKU2fEIme/ xgQQmIsDeUoWbfybdjw/x7Q0LW4mINiLDQcGHHRQKFIxbAJCT3USPLGh5xwE9/Er ABEBAAGJAR8EGAECAAkFAlYy4RYCGwwACgkQiKUE3BPJzW0uYAf9Hf30n8tM3mR2 Zo6ESE0ivgdgjaJtAWrBUx7JzAzPjBnBOlNnu5Y9lVEqetvUPH6e3PvaHYUuaUU8 0HwxuKBW9nUprgV6uIu1DZmlcp+SxpbuCy7RDpNocRLNWWFMaYYzznmTgfnTgD4D gCq8Mf1mcfrluTkOAo+QNqbMfl1GISClopRqxVuAo59ewgMnFujwgd8w12BwWl24 CzqOs5HqcUslePj+LzcjSNgVCklYwKl+0dsb/fctMOCtHodwqm2CBJ+zydvNmYkD fxda/J91Z1xrah5ec++FL0L4vs+jCiIWJeupJFKlr1hCMZiiGH7W554loK5l4jv3 EY347EidAw== =Ta4p -----END PGP PUBLIC KEY BLOCK-----

[Sarajevo67]

Hi Evan,

Thank you for your response, it helps to clarify some ambiguities I had.

Here is general example of stepped impedance transformer:

My dimensions are different (widths 0.530, 2.5 and 3.6 mm , lengths 12, 31 and 31 mm height 0.635 mm, for simplified air dielectric), but it does not change a much.
I think that I can solve transmission lines 1/3 2/3 meshing, thanks to @VolkerMuehlhaus advice, but I still need clarification for these questions:

1. How to "present" range of impedances to the port(2)? I have complex 'Zload' antenna measurements vector, for range of frequencies and most of tutorials and examples (including your), handle just 'Feed_R' as lumped port resistance (50.0 Ohms in most cases).

2. Where to put and how to use 'MeasPlaneShift' and 'FeedShift' in this case? This question is mostly related to the integrated SL and MSL ports and their implementation in metric sensitive applications like impedance transformers.

Thank you,

DD

[VolkerMuehlhaus]

could you guide me on how to measure (on port(2)) the propagated reference excite impedance (U/I) from port(1)?

You can calculate impedance Zin measured into the port like this:

port1.CalcPort(sim_path, f, ref_impedance = 50)
Zin = port1.uf_tot / port1.if_tot

That said, I would not use pure EM simulation to optimize your circuit. If the step-in-width is moderate, circuit models in circuit simulation will be reasonably accurate, and you can optimize this so much faster compared to EM. I would try that first, and use EM mainly for verification of the optimized layout.

Regarding your frequency dependent port termination: The Matlab environment for openEMS provides CTB (circuit toolbox) so that you can combine EM and circuit simulation. This way, you can use standard 2-port simulation in 50 Ohm environment, and then connect your actual load afterwards in circuit simulation. I haven't used that workflow in openEMS using CTB, but you might want to explore that possible solution.

https://github.com/thliebig/CTB/tree/5ac29367ebb7e30e78dd3db13a21464756855599

Good luck!
Volker

[VolkerMuehlhaus]

Stepped impedances are "stepped", just in theory, aren't they?

If this comment refers to simulation port impedance: it is good practice in EM simulation to simulate using 50 Ohm ports, and then use that 2-port data in circuit simulation where you connect other external impedances. There is no need to have mixed impedances directly in the EM tool, you can always apply any termination later, with no loss in accuracy.

But if you prefer to run your transformer directly in EM with no subsequent circuit simulation, you can use the equation shown above to get input impedance into the antenna port, and compare that to your complex frequency dependent target value. Or did I misunderstand your requirements?

[Sarajevo67]

EDIT: I posted this before reading your second reply. Let me prepare more detailed setup.

I want pure EM verification of SL/MSL stepped impedance concept, by comparing analytical results with simulations.

I hope that is not wrong approach?

[VolkerMuehlhaus]

Ok, then I would suggest to simply calculate in openEMS the input impedance into the antenna port, with the other port terminated into 50 Ohm. This can then be compared to your antenna impedance in Re/Im plots, or you can calculate the mismatch between openEMS impedance at antenna port and your actual antenna impedance.

This calculation of input impedance from FDTD voltage and current is independent of the port impedance used in your model! This means there is no need (and no advantage) to terminate the antenna ports with your complex antenna impedance in openEMS.

port1.CalcPort(sim_path, f, ref_impedance = 50)
Zin = port1.uf_tot / port1.if_tot

[Sarajevo67]

it is good practice in EM simulation to simulate using 50 Ohm ports, and then use that 2-port data in circuit simulation where you connect other external impedances. There is no need to have mixed impedances directly in the EM tool, you can always apply any termination later, with no loss in accuracy.

Does this mean, that I can run openEMS simulation of TL with 50 Ohm ref impedance ports on both sides and latter, in postprocessing, just plug antenna impedances on port(2) and postprocess it using simulation data?

Please, show me, how can I do that? My knowledge is limited to the extent that best I know is how to calculate parallel RLC in lumped element, to get valid single frequency Zload. I already prepared loop through antenna parameters, to put them in single frequency simulations. Very time consuming... LOL

Thank you dr Muehlhaus,

DD

[VolkerMuehlhaus]

Does this mean, that I can run openEMS simulation of TL with 50 Ohm ref impedance ports on both sides and latter, in postprocessing, just plug antenna impedances on port(2) and postprocess it using simulation data?

Yes, absolutely, that works perfectly fine in theory, or using external tools such as QUCS: get the S2P data for your transformer (simulated using 50 ohm ports), then use S1P data for your antenna and cascade them.

We only need to find the implementation that works best in your workflow, to create that circuit simulation from openEMS resuts. In my day job, I am working in EM support for commercial simulation tools, so I am not experienced what is the easiest solution using open source tools.

If you are using openEMS with Python, have a look at scikit-rf, that looks like a very capable RF circuit simulation toolset. You could put that as postprocessing into your openEMS Python script, to attach the antenna impedance (S1P data) at port 2 of your openEMS transformer data.

https://scikit-rf.readthedocs.io/en/latest/tutorials/Connecting_Networks.html

Best regards
Volker

[gadiLhv]

Hey there @Sarajevo67

Can you elaborate a bit about why you are building this three stage transformer? Is your load connected width of the right most piece?

openEMS is a 3D engine. For my matching networks\transformers, I always prefer to start off with a circuit solver. QUCS is open source, free, and perfect for starting to sniff around how to do this matching.

Also, you can easily connect your measured load and see the results there.

After having rough results in QUCS (or other tools), you can model the transformer in 3D, export the SnP file and see how well you did in the circuit solver.

Does that make sense to you?

[gadiLhv]

[VolkerMuehlhaus]

Your model looks good overall, but I'm afraid that your 1D lumped port (wire!) will cause too much port discontinuity and degrade accuracy.

I have drawn a sketch to show the ingredients of a microstrip port. The termination resistor is optional, only used if you port starts/ends inside the simulation box, at some distance from the boundary. If your port starts/ends at the boundary using PML, the port is specified without termination resistor and PML boundary will do the termination.

~~

If you are still struggling with creation of Touchstone S2P file of your transformer for use in QUCS , you can have a look at example Patch_Antenna_Phased_Array.m found in matlab\Tutorials. There the simulated 3-port S-parameter results are exported using write_touchstone(). That function also works for 2-port data.

[Sarajevo67]

So TL dimensions, within measurement planes, are only relevant dimensions?
Just to clarify termination resistor, in case when start is within box, do its difference from characteristic impedance of the MSL, causes any reflections that disturb measurements?

Also, I took a chance and tried to forge S1P and S2P ABCD matrices within Octave script. Please don't laugh too much :-) , if I did it completely wrong. Here is my attempt:
EDIT 10.08.2023.
It turns out that I made a mistake in the construction of the termination matrix:

for i=1:numel(f)
b(1,1,i) = 1.0;
b(1,2,i) = Zload(i);
b(2,1,i) = 0.0;
b(2,2,i) = 1.0;
endfor

Here is validated code:

%% load antenna measurements %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
load 'antennax128_1090_2360.out'
f = NecMat(:,1)'; % get frequencies as row vector
Zload = complex(NecMat(:,2),NecMat(:,3));
%% forge 1 port antenna ABCD matrix
b = [];
for i=1:numel(f)
    b(1,1,i) = 1.0;
    b(1,2,i) = 0.0;
    b(2,1,i) = 1.0 / Zload(i);
    b(2,2,i) = 1.0;
endfor
RefImp = 50;

I hope that this is the way. Very few sources describe how to get ABCD matrix from S1P.

At the end of two pass processing I forged S2P ABCD matrix:

%% forge two port ABCD matrix %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
a = [];
a(1,1,:) = ((1+s11).*(1-s22) + s12.*s21) ./ (2*s21);
a(1,2,:) = RefImp .* ((1+s11).*(1+s22) - s12.*s21) ./ (2*s21);
a(2,2,:) = ((1-s11).*(1+s22) + s12.*s21) ./ (2*s21);
a(2,1,:) = 1./RefImp .* ((1-s11).*(1-s22) - s12.*s21) ./ (2*s21);

multiplied them, extracted new S11 and calculated impedances:
Edit: CTB functions doing more elegant work here.

%% multiply a and b ABCD matricies to concatenate them %%%%%%%%%%%%%%%%
for i=1:numel(f)
    c(:,:,i) = a(:,:,i) * b(:,:,i);
endfor
%% calculate impedance %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Out3DS = a2s(c, RefImp);
OutZ   = s2z(Out3DS, RefImp);
Zin    = squeeze(OutZ(1,1,:));

I compared results, with my optimizations and found some similarities. My guess is, if my attempt is not waste of time, that, for more accurate results, I need to move measurement plane.

I will definitely export Touchstone, in search for validation on QUCS, but I was eager to try within Octave also.

Best Regards,

DD

[VolkerMuehlhaus]

I haven't checked all your ABCD calculations, it's too long since I used that at university. Sidenote: you can also find functions for ABCD to S and vice versa in openEMS Matlab circuit toolbox (CTB).

So TL dimensions, within measurement planes, are only relevant dimensions?

Yes, what matters is the length between the measurement planes. You can switch on voltage and current sensor display in CSXCAD to double check where the measurement plane is. In my picture the MeasPlaneShift is equal to the port length between start and stop. If port length is longer than MeasPlaneShift, that will act as additional length of line (included in results in that case).

Just to clarify termination resistor, in case when start is within box, do its difference from characteristic impedance of the MSL, causes any reflections that disturb measurements?

The termination resistor should have the same value as the port impedance. It doesn't need to agree with line impedance: you will might get reflections in "raw" S-parameters from EM because of the mismatched terminations, but everything works perfectly accurate as soon as you put your S-parameters into circuit simulation (or ABCD calculation) and terminate them with the line impedance.

So let me repeat: the reference impedance used in EM simulation has an effect if you look at the "raw" results directly from EM, but doesn't matter when using the data in a circuit environment. S-parameters can be converted to any other reference impedance. If you place your EM data into a circuit with different source and load impedance, that is what determines reflections, regardless of the port impedance used in EM simulation.

[Sarajevo67]

I haven't checked all your ABCD calculations, it's too long since I used that at university. Sidenote: you can also find functions for ABCD to S and vice versa in openEMS Matlab circuit toolbox (CTB).

Yes, I found it, but now I'm not sure, which S1P ABCD matrix create function to choose, for shunt or series impedance? As I see, the one I coded is for series impedance.

[gadiLhv]

There are plenty of tools for touchstone generation. Please find attached example. When you have the matrices, do yourself a favor and concatenate them with a circuit solver (e.g. QUCS). Cheers

…

On Sun, Aug 6, 2023 at 3:41 PM Sarajevo67 ***@***.***> wrote: Thank you, I'm doing just MSL (in openEMS), following these steps. Helix is simulated in NEC, because openEMS "hates" thick wire... LOL Next, I have to find Octave tool for construction and cascading S2P and S1P nets, otherwise I have to use Python. As soon, as I make script for meshing and simulation, I will post it here. Thanks again and Best Regards, DD — Reply to this email directly, view it on GitHub <#100 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA52FR6S772MII6GOYI47OTXT6GGVANCNFSM6AAAAAA2X2CFTQ> . You are receiving this because you commented.Message ID: ***@***.*** com>

[Sarajevo67]

When you have the matrices, do yourself a favor and concatenate them with a
circuit solver (e.g. QUCS).

I will take your advice.

Thanks,

DD

[Sarajevo67]

Here is simulation from QUCS:

It is quite close, to my analytical simulation. I found some issues related to TL impedances, but it is for new discussion.
Thanks a lot,

DD

[gadiLhv]

Good to hear

Nice work!