Experimenting with waveguides using the LiteVNA

[Mechatrommer]

The developers included harmonics to extend the frequency range so customers could play with it.

have you seen how crappy is the harmonic mode (6-9?GHz) when doing just a simple SOL calibration? iirc its very crappy.

[rf-messkopf]

I did play a bit with some free simulators a while back in response to someone's question on layouts.   The free simulators are very limited and I spent more time trying to work around those limitations than reaching my goal.   So for now, solder and PCB materials are my tools of choice for trying out the basic concepts.  No doubt this will become a limiting factor later on but fine for the basics.       

There is openEMS, which is a 3D FDTD EM solver, which is perfectly capable of simulating waveguide structures. It does not have a a GUI interface though. So you will need an external CAD program to draw conductor structures, and use a third party tool for 3D result visualization, e.g. Paraview. Meshing and the simulation itself is controlled by an Octave script.

As an example, here is a video on an openEMS simulation of a waveguide magic tee, which demonstrates field visualization and the extraction of S-parameters:
.

There was a very interesting talk by Thorsten Liebig, the openEMS developer, at FODSEM 2019:
.

The openEMS website hosted at the University of Duisburg-Essen is currently down because there has been a large-scale hacking attack on the university's systems, see here for some news coverage if you read German.

[joeqsmith]

have you seen how crappy is the harmonic mode (6-9?GHz) when doing just a simple SOL calibration? iirc its very crappy.

Sure and I have posted data for it.  This thread is already extremely large and I imagine overwhelming so I have included a link to save you some time searching for it.       Good enough for that first simple experiment. 

[joeqsmith]

There is openEMS, ...

It's been a few years since I looked for free software.   I assume you have used it and will have a look.   Which tools did you use with it?

[joeqsmith]

After damaging the covers with the hacksaw, I made two new ones and soldered them in place rather than using the O-rings.   I then squared off the edges.  Next I ran a 2D swept of one part.   

[joeqsmith]

I made up a horn for the second part using PCB material. 

[joeqsmith]

I epoxied the horn to the transition and then soldered the whole mess together.    It's not pretty, so I added some primer to give it that professional look.   A setup like this from Pasternak would cost over $1000.   Here's the real deal: 

https://www.pasternack.com/wr-90-waveguide-standard-gain-horn-antenna-20-dbi-sma-pewan090-20sm-p.aspx

The horn was mounted to the stage and swept.   I actually ran the two antennas back-to-back within minutes of each other.   I stored the transition without the horn to memory and overlaid them to provide an easy comparison of the effects.   

[joeqsmith]

Adding a horn to the second transition. 

[rf-messkopf]

It's been a few years since I looked for free software.   I assume you have used it and will have a look.   Which tools did you use with it?

So far I've been using openEMS only through QucsStudio, which automatically generates the 3D model and simulation script, and runs it. However, this limits you to microstrip structures, i.e., planar copper over a solid ground plane. But it has the benefit that the EM simulation is integrated in a circuit simulator, so you can add lumped elements and other structures very easily. I've been using it to tweak microstrip filters, bias tees and couplers, which works very well. Up to now this has been covering all my needs.

To simulate waveguide structures and devices, or antennas, with openEMS you would need a 3D cad like FreeCAD, and Octave to write and run the simulation scripts. If you only want frequency domain data like S-parameters, you could use Octave's own plotting functions, or dump the S-parameter data to a file, e.g. in Touchstone format, by setting up a script to do that. If you want animated field visualizations it is my understanding that you need something like ParaView, as has been shown in the videos.

Some time ago I've been looking at openEMS's documentation and examples. It is my impression that if you know some Matlab/Octave, the scripting part is not rocket science. You have to gain some understanding of FDTD simulations though, in order to set up things like the mesh and the excitation (frequency range, decay energy, etc.), as well as material properties, to get reliable results in finite time. But that is already the case for microstrip with QucsStudio, even though there are far less degrees of freedom.

If I were to venture into this area, I would have to learn some 3D CAD first. To get started with openEMS it would probably be a a good idea to get some of the examples on the homepage working and then tweak them. Too bad it is down at the moment, there is only the Github stuff available. But there is certainly a steep learning curve involved, as with many things related to electronics, and there is never enough time in life (I'm currently trying to get my head wrapped around Verilog and Lattice FPGAs). 

[coppercone2]

Thats a nice horn. I wonder if you can lay a wire in the seam to make that joint stronger.

I will tell you though that using PCB material is not a bad idea because if you try this with thin copper sheet, which you can form nicely, it becomes very fragile, and you would need to like put fiberglass on it or something to make it any bit decent.

For working the copper, if you are reusing it, you can throw it on a BBQ for a while then dunk it in water (clean it up), then use a rolling pin to flatten it out very nicely.

And I have experimented with using pen electroplating of solder joints like that and it works like crap.

[coppercone2]

Also if you make your own FR4 PCB, I think you you can do this to get perfect right angles inside

1) make a FR4 segment of the correct shape for a horn side.
2) when you glue the copper to the FR4, make it bigger then it needs to be, so free copper hangs off the edge.
3) Put it together so there is a full copper seam on the inside and the outside looks like a pirogi/dumpling
4) flood the inside with solder
5) use a long square edge file to cut a square on the inside


It will be mechanically correct then

[joeqsmith]

I have played with both FreeCAD, and Octave some.  I tried to use FreeCAD to create a model for some free FEA software.  They showed a lot of very nice drawings with it but I soon realized these were drawn with a professional CAD package and then imported into FreeCAD.  It turned into a big waste of time.   That's been at least 10 years since I last looked at it and I am sure it has improved.

My last time playing with Octave, I wanted to work on some large data sets and set out to build it for 64-bits.   I dare say that was a few weeks effort trying to locate everything needed to rebuild it.   In the end, I did get it working but there were a lot of bugs.   I used that same version I think with your scripts.   Again, guessing they have moved beyond the 32-bit world by now and have most of the problems solved.

Watched the first video.  He never shows the 20 weeks of effort that went into getting to that point.      He could have made a 10 part series on this one component, walking you through all of the painful steps.   I would watch it!

A few of us were recently posting about the Lattice tools and licensing.   I've used their very old PAL software and their newer software for small CPLDs.  I don't remember having any problems with it.

[joeqsmith]

Thats a nice horn. I wonder if you can lay a wire in the seam to make that joint stronger.

I will tell you though that using PCB material is not a bad idea because if you try this with thin copper sheet, which you can form nicely, it becomes very fragile, and you would need to like put fiberglass on it or something to make it any bit decent.

Wire could possibly improve the strength.  I could also fibreglass the entire structure which would really improve its strength.   Adding a cover to the front would also help.    Triangles are good!   Double sided would have also helped as we could tack both sides. 

For working the copper, if you are reusing it, you can throw it on a BBQ for a while then dunk it in water (clean it up), then use a rolling pin to flatten it out very nicely.

The copper sheet I show is very thin and I will typically grab what ever I have on the desk in front of me to flatten it out.   

Also if you make your own FR4 PCB, I think you you can do this to get perfect right angles inside
...

For angles, I just block up the parts and solder them.   I typically tack solder the parts at first and then run my beads once things are together.   In this case I'm not too concerned.   All I am doing now is just a toe dip to test the waters....   Eventually when I start working with WR90 I will have some idea what the performance should be.  I doubt we could buy a production horn and transition as poor as what I have shown. 

I think we could make for example a slotted line with everything (2 transitions, slotted waveguide)  in one part without too much trouble but at some point, it would be much nicer to have a set of building blocks to mix and match.   

[coppercone2]

for soldering FR4 together I remember there is a document somewhere that describes the construction of the chassis for radios and he recommends that you actually do not use a right angle but like a 91 degree angle because it contracts when it cools down. That can improve quality of those things also, to have angle offset blocks around for rapid prototyping with PCB

[joeqsmith]

Interesting.  I've always just used 90 deg blocks and called it good.    I have used it to make cases for some of my projects where I want something fast and am not concerned about the looks of it.  For example, opening video shows a couple of cheap RF detector boards built into a PCB case for a simple Watt meter:

https://youtu.be/JHsgbu7brYI

From the following site

In most radars, the feed horn is covered with a window of polystyrene fiberglass to prevent moisture and dirt from entering the open end of the waveguide.

 
https://www.radartutorial.eu/03.linetheory/Horn.en.html

I could certainly add a sheet of PCB to one of the horns and compare results.   I like their 3D plot.  I noticed that my software doesn't calculate anything when running 2D mode.  Just an oversight.   

***
double negative.

I found a sheet of fibreglass that is just about the right size for the horn.   I'll glue it up. 

[joeqsmith]

Two horns with fancy paint.   Plots showing effect of adding the cover.   There may be some difference masked by the Lite's noise. 

[joeqsmith]

Previous test was two horns pointed at one another then inserting the sheet of fibreglass between them.   I glued on the cover to the second horn.   I then swept both horns.  Blue trace is horn with no cover.  Patterns are sightly different due to errors in construction.   

Next I ran a 3D sweep on the horn without the cover.

[joeqsmith]

https://www.electronics-notes.com/articles/antennas-propagation/rf-feeders-transmission-lines/waveguide-junctions.php

To look at the operation of the Magic T waveguide junction, take the example of when a signal is applied into the "E plane" arm. It will divide into two out of phase components as it passes into the leg consisting of the "a" and "b" arms. However no signal will enter the "EH plane" arm as a result of the fact that a zero potential exists there - this occurs because of the conditions needed to create the signals in the "a" and "b" arms. In this way, when a signal is applied to the H plane arm, no signal appears at the "E plane" arm and the two signals appearing at the "a" and "b" arms are 180° out of in phase with each other.

***
This site appears to have it correct:
https://www.everythingrf.com/community/what-is-a-magic-tee

[joeqsmith]

MagicT_InpE_Out1_2:  LiteVNA's Port1 attached to the magic-T's E-Plane.  H-Plane port is terminated.   Port2 is then connected to each of T's co-linear ports.  We can see the 180deg phase shift as expected. 

MagicT_InpH_Out1_2: LiteVNA's Port1 attached to the magic-T's H-Plane.  E-Plane port is terminated.  Port2 is then connected to each of T's co-linear ports.  We can see the phase remains constant.   

https://www.microwaves101.com/encyclopedias/magic-tees

[joeqsmith]

MagicT_InpCoLin_OutH_E_CoLin:  LiteVNA's Port1 attached to one of the magic-T's  co-linear ports.  Port2 is then connected to the H-plane (yellow), the E-plane (violet) and finally the second co-linear port(red).   

There's more leakage than I would expect but no surprises.   Keep in mind LiteVNAs poor performance in this range and no attempt was made to calibrate the system.

MagicT_InpH_OutE:  LiteVNA's Port1 attached to the magic-T's H-plane.  Port2 is then connected to the E-plane.   Unused ports are terminated. 

[joeqsmith]

Another good ham link.   
http://www.w1ghz.org/waveguide/Waveguide.htm

[joeqsmith]

I want to attempt to construct a waveguide with some hobby brass next.   The plan is to again include the transitions.  It's a lot of thermal mass and I am thinking maybe silver solder.   

[coppercone2]

welcome to distortion land if you try to braze corner joints

If you don't bend a corner piece for a lap joint you will have a ugly time IMO when its getting red hot

but I never tried with thin brass. It might work. The amount of fixturing people use for like stable work is kinda ugly (never got that far, its just tooo mucch). Also ventilate well because thats alot of brass surface area to be heating, it might fume.

[joeqsmith]

I would use a jewelers torch head.  If you saw that home made current shunt I constructed for the UT61E DMM, this is how I silver soldered the brass contacts to the shunt.   Those were very small parts compared to this.

The brass I am using is 1mm thick from K&S (see link).  You can find much cheaper sources than the one I linked.   I cut off a section and tried to bend and solder to it.  Even with a small iron, it wasn't a problem to get it hot enough to flow.  So maybe I can get away with using my larger iron. It bends fairly easy and is soft enough that there are no signs of any cracks.    Thinking two L's rather than a U channel.

https://www.ebay.com/itm/225111789751

[joeqsmith]

I have not bent the ends or solder anything except the connectors.  The radius is about as tight as I dare make it.   I laid some material to close off the ends and held it together with my hand.    Yellow trace is with the waveguide apart, Violet with it together and Red with a thru applied.   The waveguide has a bit of loss but hopeful this will improve once I actually assemble it.