Error with slightly modified Simple_Patch_Antenna script

[DatKnight]

I've had a go at trying to simulate a Patch Antenna in openEMS, starting by just modifying the existing Simple_Patch_Antenna.m script. The only changes I've made are to the patch and substrate dimensions and dielectric constant:

But when I run the script I get the following error:

Thing is if I inspect the port variable:

So I've got no clue why it throws an error as I've not messed around with the port dimensions, and the port should still be on the PCB?

[biergaizi]

The only changes I've made are to the patch and substrate dimensions

This is the red flag. Many simulation setups can be surprisingly sensitive to dimensions, and there are many traps for new users: wrong meshing, wrong port locations, ports or components misaligned to meshes, coordinates too close to absorbing boundary conditions, violation of the 1/3-2/3 rule...

In this case, the number of cells on the Z axis is probably too low, just barely enough to finish the original simulation:

substrate.cells = 4;

Once the substrate height is slightly increased, the mesh became unable to correctly cover the port on the Z axis, so the port stops working.

To fix that, use:

substrate.cells = 10;

[DatKnight]

That worked! Thank you so much!

Is there anywhere I can find more about these rules of thumb/traps/quirks with the software? There isn't much I can find on the wiki that is in that level of detail, am I mostly just limited to trial and error and asking questions on here?

[thliebig]

Yeah, having a full list of all this necessary knowledge would be nice, but I guess it will be never complete. But I want to point out that it not as much about "this software" as for the FDTD method and EM-modeling ;)

[DatKnight]

Are there any good resources for learning more about that?
I've learned some at uni but never went too in detail about the specifics of EM modelling or the FDTD method

[biergaizi]

https://docs.openems.de/more.html

[LubomirJagos42]

maybe this is not best advice but since I start playing with openEMS I had to walk through all that caveats and so and most frustrating is building up simulation from scripting, like but my opinion to make simulation build up from ground without knowing API is putting there some layer of unknown and after that there are other layers of unknow so I created GUI for FreeCAD to generate these octave or python scripts similar way how HFSS used to define simulation and put there lot of user input checks to generate at least runnable simulation you can look here:
https://github.com/LubomirJagos42/FreeCAD-OpenEMS-Export
here are tested examples:
https://github.com/LubomirJagos42/FreeCAD-OpenEMS-Export-Plugin-Documentation

if you want to walk through this trial and error way faster in words to put more effort into building it visualy and at least to have clue how I solved some caveats for simulations this will help you got oriented in openEMS faster

definitely what you should do is ALWAYS check your simulation in AppCSXCAD and think what it's doing, this is just to provide some way from top to bottom like some guide how you can implement your simulation, let say inspiration

when you master scripting way you can always do your model in scripting, hope it helps

[biergaizi]

ALWAYS check your simulation in AppCSXCAD and think what it's doing

I'd like to add that it's also often a good idea to enable field dump so one can visually examine what the simulation is doing in ParaView as a 3D animation - which is extremely valuable for debugging malfunctioning simulations.

[DatKnight]

I've changed the simulation again to move the feed point to the edge of the patch and I'm getting the same error :l

Been having a look at trying to see what might be wrong, but for some reason the port isn't showing up in AppCSXCAD?

It should be turned on.

There's also some errors that pop up in the console when I try to rotate the view - I've been making do by just looking at the xz/xy/zy planes as it lags it the hell out when I rotate with the mouse.

I've googled it and it seems to be a generic issue, I'm not sure if it's something anyone here as encountered before?

@biergaizi I've tried having a look at adding a dump box for Et fields and viewing those:

But I just get an empty box with nothing showing up when I select E-Field and run the animation, this also happens with the original script with unmodified parameters, even though the simulation has no problems.

I'm really confused here, will going and doing my homework to understand the maths help me use the software and stop so much of this trial and error? I feel like I should be able to get something simple like a Patch antenna working without too much hassle.

[biergaizi]

My CPU isn't that out of date, I imagine a 7 year old CPU presumably shouldn't have any issues with OpenGL.

Thanks for posting detailed spec. I believe the root cause is: nigels-com/glew#172 In short, OpenGL GLEW library is currently incompatible with Wayland because it wants to use X11's GLX, but is unable to find it. So far there's no fix in the stable release. As you can see under that issue, almost everything that uses GLEW are affected, with links to ~20 projects.

For now, try running AppCSXCAD under XWayland instead.

env WAYLAND_DISPLAY= /path/to/application/you/want/to/run

This forces WAYLAND_DISPLAY to be an empty variable.

[DatKnight]

My CPU isn't that out of date, I imagine a 7 year old CPU presumably shouldn't have any issues with OpenGL.

Thanks for posting detailed spec. I believe the root cause is: nigels-com/glew#172 In short, OpenGL GLEW library is currently incompatible with Wayland because it wants to use X11's GLX, but is unable to find it. So far there's no fix in the stable release. As you can see under that issue, almost everything that uses GLEW are affected, with links to ~20 projects.

For now, try running AppCSXCAD under XWayland instead.

env WAYLAND_DISPLAY= /path/to/application/you/want/to/run

This forces WAYLAND_DISPLAY to be an empty variable.

That works great thanks! How did you go about finding the solution may I ask?

[DatKnight]

ok, thanks, I corrected your simulation:

* commented smith chart reflection due there is some error and because of that it's not continuing so to get you other graph I disabled that

* added post_processing at beginning if set to 1 it's not doing simulation, skipping it and continue to graphs

* E, H probes made planar, not 3D, I set their thickness to 0, also I still don't know how to get slices in Paraview from 3D, but you want planar E, H probe, at least this way you got nice animation in paraview when set as @biergaizi mentioned in previous comment

your design is on good way, S11 is bad, but it's showing there is some resonance :D

patch_antenna_DatKnight.txt

Thanks so much! Never occurred to me to just turn off the Smith Chart first and see if that works lol, I'm still curious why it doesn't work though.
I've got to figure out how to add some matching to the simulation now, hopefully that doesn't cause more issues.
I'll give your FreeCAD addon a try, I want to be able to get to a point where I know what I'm doing scripting wise, but if this eases the learning curve then that's great!

Thanks so much everyone!

[biergaizi]

@LubomirJagos42: E, H probes made planar, not 3D, I set their thickness to 0, also I still don't know how to get slices in Paraview from 3D,

To get a 2D slice from a 3D model, first open and visualize the model as usual. Then click Filter menu, select Search. Type Slice. Click the matched Slice option. In the menu bar at the left, click Z normal, and click Apply. The result is a slice across the Z dimension. To change the location of the slice, change the 3rd coordinate of the Origin property. Don't forget to click Apply. The only problem I've found is that the dimension is in SI (I think) so the coordinate values are often extremely small and difficult to type - if it's messed up, click Reset to undo the unapplied change.

See ParaView documentation for details: https://docs.paraview.org/en/latest/UsersGuid/filteringData.html#slice

That works great thanks! How did you go about finding the solution may I ask?

Using search engines and making educated guesses from experience. I just searched the error message on the Web, and clicked several links. One link brought me to the GLEW bug tracker, the reporter mentioned Wayland, so I became suspicious and made an educated guess that it's probably the culprit. Then, I asked your system specs. You said Fedora, so my suspicious was almost confirmed. I then installed a fresh Fedora on a virtual machine and reproduced the problem. I've now submitted a patch to the VTK upstream, so the root cause will be fixed by VTK team soon.

[georgy-moshkin]

I'm really confused here, will going and doing my homework to understand the maths help me use the software and stop so much of this trial and error? I feel like I should be able to get something simple like a Patch antenna working without too much hassle.

Hey, let me show of what is possible:
https://www.youtube.com/watch?v=v8uxhbu36Ug
While I have not released any free examples yet, I have planned two informative openEMS tutorials related to patch antennas. I haven't encountered any strange port errors, probably because I rigorously implement meshing code and spent a lot of time to obtain results similar to few research papers. In some cases, I put 10 mesh lines on a feeding line between "1/3 rule" lines to get correct results, otherwise it is +-30 Ohm compared to correct results. This was really a breakthrough for me, because I've got the same results as in CST or HFSS graphs from one paper. Also, I use "algorithm 3" for recursive mesh smooth. During my experimentation, most problems was caused by smooth mesh routines together with detectedges routines, when there are two very closely spaced meshing lines forced algorithm to create unnecessarily dense meshing lines. Thus, in most cases I put all meshing lines manually, and then run smoothmesh with "algorithm 3" for nice smooth mesh generation. Also, for learning I think that coarse mesh lines are preferable, because it saves a lot of time. So you can try more ideas faster. For real antenna design I would use dense mesh.