Question about multiple port simulation on octave interface

[LeoGalloGomes]

Hi, it's me again. I'm trying to understand how to turn on and off ports in a simulation loop to simulate asymmetrical circuits. Is it enough to change the excitation status and run openEMS or should I do something else?

I tried analyzing the directional coupler tutorial, but found it confusing.

Thanks again :)

[JTSvejda]

Hello, maybe I can help. You essentially need to create a loop to activate the ports sequentially. Each port element has an 'excite' parameter, which is essentially a boolean to toggle the excitation on and off:

for active_port_n = 1:N_ports
    for port_n = 1:N_ports
        % This loop creates your ports, where just one port is active
        [CSX, port{port_n}] = AddLumpedPort(CSX, <<prio>>, port_n, <<R>>, <<start>>, <<stop>>, <<dir>>, port_n==active_port_n)
    end
    % For each iteration of the active_port_n loop you run the simulation
end

This method is utilized within the set of these two tutorial files: matlab/Tutorials/Patch_Antenna_Phased_Array.m, matlab/Tutorials/Patch_Antenna_Array.m.

I hope you find this helpful :)

[LeoGalloGomes]

Hi @JTSvejda !
Thank you for your response, and yes, that was very helpful. Each time there's a new active port the "main" simulation, Patch_Antenna_Array.m gets called, I got it now.

Thank you so much!

[VolkerMuehlhaus]

I tried analyzing the directional coupler tutorial, but found it confusing.

The basic structure of multi-port simulation in that example is to put the single excitation model into a function, and then call that function for each port, one after another. So essentially you build there entire model for each port excitation! Results for each excitation go into a separate result directory.

What did not work at least for my Python models: It was not possible to re-configure the port excitations for an existing CSX, and re-run with that modified ports configuration. In my models I really had to re-build everything all over again for each different excitation, that is why I encapsulated model setup & simulation into a function, with port excitation as a variable. The coupler example for Matlab does the same thing.

One detail in that model that you might find confusing: instead of evaluating S-parameters directly, the model calculates Y-parameters from from port voltage & current, and then converts the full Y-matrix to S-matrix.
I know that method from my commercial FDTD tool. If you evaluate FDTD S-parameters directly, sometimes results are not exactly passive and not exactly reciprocal, due to all these small precision errors like truncating the time signal below a certain energy threshold. By using [Y] to [S] conversion some of these errors average out, and you get a consistent [S] matrix.
My best guess is that the author of the couple example saw some differences in forward and reverse parameters which should be identicaly, and solved it by using [Y] to [S] evaluation.

[VolkerMuehlhaus]

In the meantime I looked at your model visually in AppCSXCAD, to see what you designed and visually check the mesh.

Having an electrically small model is fine, if the overall mesh count is reasonable and the mesh size is not smaller than necessary. I also created RFIC models like yours, and they simulate for a while (some models for hours), but they finally converge.

The PEC boundaries are ok, the energy should exit the model at the ports and there is no radiation here.

I think you should do the test with limited timesteps and check the port voltages. In my commercial FDTD solver I once had constant energy when the port was shorted or unconnected by mistake, I don't recall the exact details. So look at the port voltage and check if the excitation pulse looks "normal" with gaussian shape.

[VolkerMuehlhaus]

@LeoGalloGomes When I run your model, the solver reports energy=0.0 from the beginning, any stays there. Zero energy. I think there is something wrong at the port excitation already, possibly open or short. I will look into your model again now.

[VolkerMuehlhaus]

@LeoGalloGomes I changed your ports from 2D sheets to 3D objects, and excited port 3 for testing. This runs fine and converges. Results for s33 = (port{3}.uf.ref) ./ (port{3}.uf.inc) also look fine.

The only code change were in the port defintion:

port_width = pad_length/8;
[CSX, port{1}] = AddLumpedPort( CSX, 500, 1, 100, [-pad_width/2 pad_length/2-port_width/2 Metal5.zmin], [-(pad_width/2+pad_gap) pad_length/2+port_width/2 Metal5.zmax], [1 0 0], false);
[CSX, port{2}] = AddLumpedPort( CSX, 500, 2, 100, [pad_width/2 pad_length/2-port_width/2 Metal5.zmin], [(pad_width/2+pad_gap) pad_length/2+port_width/2 Metal5.zmax], [1 0 0], false);
[CSX, port{3}] = AddLumpedPort( CSX, 500, 3, 50, [-3.25 geometry.ymax Metal5.zmin], [3.25 (geometry.ymax-1) Metal1.zmax], [0 0 1], true);

[LeoGalloGomes]

Ah, I think I got it, I probably forgot to add a mesh.y point on the ports. That's why the energy was stuck at 0, the pad was outside the grid, and thus no energy was being input into the structure... silly mistake. Thank you so much for your help on that, @VolkerMuehlhaus !

However, if I may press the matter a bit further, how could I excite the 100-ohms ports 2 and 3 to get the transmission response (and also S11 seen from the pad itself)? That's my main problem now, and something that's blocking me from advancing on other fronts as well (as my chipathon contribution uses a lot of S-CPWs, and ustrip/S-CPW transitions).

[VolkerMuehlhaus]

@LeoGalloGomes For the ports 1 and 2 which are in parallel, just excite them simultaneously and process them as already prepared in your model. That part looks fine to me.

I only started from port 3 to keep it simple, and get some results for checking.

[lowpon]

Hi all,
When using the AddCurvePort or AddLumpedPort function with a string instead of true or false, as it is in directional_coupler example or Patch_Antenna_Array tutorial, I get the following warning :
"deprecated: a string as excite option is no longer supported and will be removed in the future, please use true or false"
Thus, my question is : How to get ports excited one by one in a loop without a parameter here ?

[VolkerMuehlhaus]

edit: Sorry, I dind't notice that this is about Octave, and posted a Python example. Removed!