Experimenting with waveguides using the LiteVNA

[coppercone2]

do you think its better if I don't have the right calibration standard connectors to
1) remove the adapter that was allowing me to connect to the standard and put a new adapter (substitution)
2) add a extra adapter to connect the DUT

So
1) subtracts something known and replaces it with a unknown
2) adds more unknowns that are well behaved (high quality coaxial adapters), but does not subtract anything that was calibrated for (longer)

I have a hunch that #2 is better, even if your measurement chain is longer

[coppercone2]

I did not get my test indicator adapter yet, but I did manage to do this

I made a bootleg profilometer, IDK how good this is, but I took a comparator stand, and I zero it to 0.000 mm (this one is in metric, the federal digital one) with the horn underneath with the indicator pushing down on the waveguide flange (half way between the apature and the long side edge).

The comparator stands were pulled from garbage and fitted with ebay stuff, but the test I did is I took a old gauge block (possibly east german origin) that is in OK shape, then cleaned it, did a lapping motion a few times on a pink granite block, and then cleaned again with alcohol lint free wipe. I then put it in the middle of the comparator, zero it, then I shift the gauge block around left right, up down, and at 45 degree angles, I did not get any motion past 0.001mm (I think the stand was shifting though, maybe old grease, because it was unrelated to the block motion, it just ticked up). Need to play with /regrease (been 4 years since I greased it) a little maybe. Like it tick up to 0.001mm, but then when I shift the block around, it stay at 0.001mm, it was not following the blocks motion (like sliding around a wedge). Maybe heat related too, kinda need a cold ass zombie to work in mechanical metrology rather then warm human).

I need like a curved gauge block that gets maximally flat so I can mechanically 'chop' my measurement to detect if the stand is shifting/comparator is drifting (digital, old) or if the bottom plate is uneven. Thought it was totally broke at first too, because the measurement head of the indicator was loose, so it was slightly turning/flexing the first time I tried, leading to totally unstable numbers. ! I don't think I should be worrying about 0.001mm though that is 4/10000, probobly can drive myself crazy trying to resolve that accurately


Before measurement I 'lapped' the horn on a known flat steel piece without any abrasive, just to maybe squish unstable blobs.


Then I shifted the horn around and watched the maximum deflection.

I got more deflection on the edges outside of the holes (maybe the indicator spring is flexing it).

Maximum Deflection Results:
near rectangular inner apature
1) silver paint : maximum ~0.1mm
2) nickel paint : maximum ~0.25mm

Outside perimiter including screw hole areas
1) silver paint Maximum : ~0.25 mm
2)nickel paint : maximum : ~0.45 mm

So the nickel paint is quite uneven, and the silver paint is not perfect but better. Seems natural for the paint to be screwed up near a hole, unless you paint then drill its gonna have surface tension related problems there because of an edge formed.


I should print a bare horn and leave a unpainted test object for paint/coating studies when I make things. I am not sure if it was aligned to begin with (the build plate might be misaligned). But still interesting.

I expect the test indicator (lever arm style) on granite block to be much more applicable to this then misusing a comparator.

Maybe I can record a video doing it slowly and then make an actual 2d plot on graph paper with a grid matrix

[xrunner]

Sounds like an edited ChatGPT screed with advertising links for toaster ovens added. 

[joeqsmith]

I thought my new friend szoftveres would have chimed in by now.  If the point they were making was that I shouldn't be so lack in my measurements, I doubt many viewers would disagree.  Then again, many have been upset over my use of a simple torque wrench.    

Here's my last bit of data from yesterday.  To get us closer to what was shown during that last video, I changed the sweep range to 9.8 to 10.02GHz and calibrated the LiteVNA with my extender.   I then measured the Narda reflection standard using the ideal model.  I repeated the measurement with the added delay and loss set to zero (like I have been showing).   In both cases, the IFBW was set to 1kHz.   

While I can't use the old Agilent as a reference, it does provide an idea of the effects between the two models. 

***
Forgot to add the Smith chart.  We can't have that.

***
AppCAD doesn't seem to allow for Touchstone files with different ranges, points...  METAS doesn't offer a Smith chart viewer but it does have a cartesian.   You can see how the PNA (at a lower frequency) compares with the LiteVNA ideal and corrected models.   

[coppercone2]

I would not call it upset over a torque wrench, its just that some people hammer the topic like crazy and then you get scared because you look at the original price of the hardware, and it just seems scary to deviate. But IMO torque is a touchy topic, even with car wheels. Like that old story about union workers dropping their torque wrenches to shut down a line, it makes it seem very important that even with how... loose managers can get with specifications and quality for the bottom line quota, torque can still get to them. Very easy to imagine if the specification was less important, it would be easy for a manager to bypass the 'torque wrench problem' as irrelevant because they are sure its good enough and no one will complain anyway if its 'off by a little'. 

[joeqsmith]

I try and hammer every topic I cover, within reason (until I grow bored).     

I've known softheads that would sabotage their code.  The idea was they would set themselves up to be a hero and save the day when they solved the problems they created.   The end game of course is to make yourself look valuable.  Typical managers are too weak to pick up on it.         

[coppercone2]

I do have a lock in amplifier that seems appropriate for some kind of laser system. Those graphs are cool

[coppercone2]

ah thats nice I got the surface plate with a test indicator working. The silver one looks to be like 0.003 flat across most of the surface on the granite block. I just need to put a drop of glue on the indicator and paint over the crack and I get some pics. Very cool tool.

I wonder if I can measure any RF connectors with that setup. I have the special micrometer for APC-7 connectors, but I wonder if there is other stuff to measure.

The rusty v-blocks I cleaned up show that they are flat to 0.0005 inches

The mini ground jewelry anvil is also flat to like 0.0004 (almost mint).

[joeqsmith]

Still playing with the hybrid T.   

[joeqsmith]

Szoftveres, as I demonstrated in the video and so far with these tests, I have not been securing or properly aligning the waveguides.  Basically, like my use of the ideal model, I saw no reason.   But, I thought I would repeat the test, taking the time to bolt and torque them.   This should at least give you an idea of how the Lite and my algorithm compares with the Agilent PNA.   

The PNA's IFBW was changed from 35kHz to 5kHz.  The number of data points for both VNAs was increase from 201 to 801.  The frequency on both VNA was set between 8.5 and 9.0GHz.


PNA_torque_gravity:  Showing the previous data I tool with the PNA where I laid the Narda reference standard on the transition compared with the proper setup. 

PNAvsLite_OffsetShort:  From your posts, you seem focused on the 1/4 offset short standard I made.   This graph compares that standard being run on my Agilent PNA compared with the Lite using my extender.   As you can see, taking the time to properly set up the measurements has tightened things up.   I'm still not clear about your comments.   I could be doing something wrong as you suggest making bad assumptions and what not but again, keep in mind that the data off the Agilent PNA is their built-in algorithms and coefficients.  That is nothing to do with me.   

PNAvsLite_slidingref:  The Narda sliding reference attached to the Agilent and Lite.   

The LiteVNA does have a bit of noise but as I mentioned, I do have the latest hardware which should improve it somewhat.  Then it's further crippled with my poor man's extender.   Still, keeping in mind everything total (standards, transition, cables, extender, LiteVNA64), maybe $700.  Not bad when you consider the price of a decent VNA.

I have been making changes to the LiteVNAs software as a result of playing with these Waveguides.  I was tired of manually calculating the frequencies depending on the setup and added a few other improvements.   Once I am done playing around with the WGs, I'll upload the latest version. 

[joeqsmith]

Using the same VNA settings with the LiteVNA64 latest 3.1 hardware.  The same firmware was installed.   Plot showing the data normalized to my old Agilent PNA. 

While the one spur still shows up at the same location, the other two large ones have been cleaned up.  With phase being the big concern for Szoftveres, consider it's basically within a degree.   Of course, if I use the ideal model the phase will certainly read 0 as I have shown many times.   

The algorithms that are in question came from various  application notes from HP, Agilent & Keysight.  I'm sure we are all aware how there are mistakes, missing information, and how they evolve.  I could also have made mistakes in my math.   Point being that it should come of no surprise that the LiteVNA behaves much like the Agilent as the algorithms should be very close.

Maybe Brian at Copper Mountain Tech will send us one of their VNAs so we have something to compare the Lite with at higher frequencies....    He's not visited here in some time.  Hope all is well.

[Yannick.R]

If you have another transco transfert switch, you could build a full two port system with your extender.
Your transco switch is rated up to 18GHz, so you could use it in x band.

A first one, allow to switch the
Output of the coupler - 50 ohms terminations or downconverter.
High frequency port 2 - downconverter or 50 ohms terminations

With this one you have S11, S21

A second like you use at lower frequency to switch port 1 port 2
And you will have S22, S12

Then you can play with other calibrations (TRL or LRL) which are nice because you don’t care if the reflect is a dot or a curve. This is not useful. The only information used for solving is that the reflect is on the edge of the smith chart.

[joeqsmith]

If you have another transco transfert switch, you could build a full two port system with your extender.
...
Then you can play with other calibrations (TRL or LRL) which are nice because you don’t care if the reflect is a dot or a curve. This is not useful. The only information used for solving is that the reflect is on the edge of the smith chart.

You could connect a transfer relay directly to your LiteVNA to measure the four S-parameters.  My software supports that today.   You could play at lower frequencies using the NanoVNA and work with Touchstone files along with your favorite math tools.  Once you have the math worked out, maybe include it into one of the open source programs.   I don't believe that the open source programs support driving an external relay but you could certainly add that. 

Depending on your goals, the LiteVNA may work out for you.  Keep in mind that all of these low cost VNAs have their problems.  Agree, they are fun to play with. 

Maybe post what you are doing and the group can try and help you out. 

[Yannick.R]

Yes, I have bought one transfer switch on ebay and I wait the delivery. And I have already installed your software solver32. And I have still to play with an ftdi device and to understand how to pilot the switch with your software and a ftdi. Then I will try to use others calibrations than SOLT.

But I was just suggesting that you could mix your ideas (the use of a switch and the use of an expender) and with your extender and 2 switchs, you can manage to do a full two port x-band vna with the litevna.
You have already a transfert switch that is rated for 18GHz operation according to your manual. If you have another, you have all you need (and cable and time)

[joeqsmith]

Yes, I have bought one transfer switch on ebay and I wait the delivery. And I have already installed your software solver32. And I have still to play with an ftdi device and to understand how to pilot the switch with your software and a ftdi. Then I will try to use others calibrations than SOLT.

But I was just suggesting that you could mix your ideas (the use of a switch and the use of an expender) and with your extender and 2 switchs, you can manage to do a full two port x-band vna with the litevna.
You have already a transfert switch that is rated for 18GHz operation according to your manual. If you have another, you have all you need (and cable and time)

That's a good start, keeping in mind that used parts may no longer meet the manufacture's specs.  It's one reason I suggested you start out at lower frequencies.   Connecting the relay directly as I mentioned, should help simplify your setup.  While you're waiting on your relay,  you may want to download the source for one of the open sourced programs and get the tools setup so you can build it.  Then you can make what ever changes are needed for your experiments.   

Keep us updated on your progress.

***
To automate switching between S11&S21, you may want to look at the old HP33311B relays.

[Yannick.R]

Yes this is better option than playing with a full transfer switch and match: “ The HP33311B has
4 RF ports with one internal 50ohms termination designed for applications requiring a transfer switch or a cross switching element.”

One step after the other but I will keep an eye on this reference.

[joeqsmith]

Yes this is better option than playing with a full transfer switch and match: “ The HP33311B has
4 RF ports with one internal 50ohms termination designed for applications requiring a transfer switch or a cross switching element.”

One step after the other but I will keep an eye on this reference.

Datasheet may be found here:   https://www.testequipmenthq.com/datasheets/Agilent-33311B-Datasheet.pdf

If your only goal is to play with the math though, you may want to simulate everything in software.  This is typically the route I take when adding a new complex feature to my software.   If you're using a low cost VNA,  buying used parts, using poor quality cables and connectors, not using proper assembly techniques and then trying to calibrate it with homemade standards, lots can go wrong.    Sometimes it's better to decouple from the hardware.   

Someone was posting about their attempts to connect a transfer relay + ? to their low cost VNA.  They had repurposed an HP box and it was far more fancy looking than anything I have shown.   Still, the performance was so bad, I used it in a video as an example of what not to do. I tried to get them to simplify their setup to try and help track down their problems but I don't think they ever made it that far.   

The best one I saw was from a person posting how the four S-parameters could be measured using two low cost VNAs connected together with a few Tees (and some software).      Their reason was to avoid the cost of a transfer relay. 

[coppercone2]

I don't think anyone should discourage use of proper test equipment, just realize it is not a substitute for know how and skill. But I don't blame anyone for wanting to work with a chassis, I don't think its some right of passage that you graduate from bare PCB puzzle.

But I will say it again, learning the equipment is hard, and just working with 'boxes' requires alot of learning anyway.

It's your money... money =/= skill. Don't get mad because you think you can spend your way out of a problem.. most people know this by the time they get their first job at a electronics company, that money does not solve problems, but people do. Its less annoying to spend money but no less difficult.

The problem I think is that you need alot of different things, and people are getting upset because they buy a few expensive things, but just don't have enough money to build the test system. It does not magically do more, its just nicer. Its plumbing, programming and electronics all in one.. I hope people realize that its not easy and cool their tempers lol

[joeqsmith]

I don't think anyone should discourage use of proper test equipment, just realize it is not a substitute for know how and skill. But I don't blame anyone for wanting to work with a chassis, I don't think its some right of passage that you graduate from bare PCB puzzle.

For these experiments, outside of using the old Agilent to verify some of my results, the LiteVNA has been fine.

But I will say it again, learning the equipment is hard, and just working with 'boxes' requires alot of learning anyway.

True, my Agilent is a bit cumbersome to drive but IMO they kept it similar enough to their previous VNAs that someone familiar should be able to run it out of the box without a manual (for the basics).   

It's your money... money =/= skill. Don't get mad because you think you can spend your way out of a problem.. most people know this by the time they get their first job at a electronics company, that money does not solve problems, but people do. Its less annoying to spend money but no less difficult.

Making money seems more difficult than hanging onto it.   

I bet a few companies in the US bought their way out of recent supply chain problems.  Hopefully we learned a few lessons.  Doubt it.

The problem I think is that you need alot of different things, and people are getting upset because they buy a few expensive things, but just don't have enough money to build the test system. It does not magically do more, its just nicer. Its plumbing, programming and electronics all in one.. I hope people realize that its not easy and cool their tempers lol

Many I correspond with seem more frustrated than anything.  They typically have a ham license, know everything there is to know about RF and then spend their $50 on a VNA in order to measure VSWR in the most complex way possible.   I wonder how many of the pocket VNAs never made it out of the box because the owner realized that they were going to have to put in some major effort in to use it or that they really didn't have the skills they thought they had and that their ham license is not quite the same as having an education and working in the industry.   

Your last sentence sums up why I enjoy electronics. You will never learn it all or even a small percentage and there are so many different aspects to it.

[Yannick.R]

  I wonder how many of the pocket VNAs never made it out of the box

These are wonderful toy. (And quite good for something to play with). Even not for doing something very useful. Just for the fun.
I am very impressed by what it is possible to do with these very cheap (vs professional equipment) devices.

[joeqsmith]

These are wonderful toy. (And quite good for something to play with). Even not for doing something very useful. Just for the fun.
I am very impressed by what it is possible to do with these very cheap (vs professional equipment) devices.

Can't agree more.    I wonder if they are being used in any of the university classes yet.

[Yannick.R]

Don’t Know in USA. In France, most of the universities are public and money must be spend with very stricts regulations. They can’t buy on aliexpress for example. So never heard of that and probably not. But I’m agree that could be a very good tool for the freshman for the price of a basic multimeter.

[joeqsmith]

Interesting.  Maybe there is a work around where you have the students buy their own.   Seems like that these were made for education.  Beginner damages it (assuming they will) they get a new one.  I'm guessing the access to higher end equipment is limited in the class.

I was on groups.io  (https://groups.io/g/liteVNA/topics) where someone had asked about TRL and WG.  I did post a link to this thread.  With you having a similar interest, you may want to join that group and see if the two of you could work together on it. 

For the experiments I am running, most have been relative measurements.  At most I may normalize the data.  Coppercone2 stirred the pot when they attempted to measure the SWR of their own horn.  (pg7)      I had not planned to make any absolute measurements with the printed horns but plans change.    The three methods I used during the video yielded three different answers.  Not surprised that using a VNA with unknown cheap coaxial standards to cal it, and then we introduce the cheap transition's error.   I'm surprised we got as close to the slotted line as we did once the higher quality transition was used.   

To further reduce the errors, we needed to account for the transition.   And while it was suggested that my algorithms have an error and I was making wrong assumptions,  I'm guessing from the OPs lack of response,  that's been put to rest.  Using the Narda and Muray standards  to verify the the homemade standards, I believe they are enough for these experiments.   

While my old PNA has limited support TRL, it doesn't support waveguides.  They may have offered it with later firmware but I have the most recent version my hardware supports installed.  I have attached a paper that discusses TRL, using WR90 as an example. 

****
I thought I should double check that statement and it does support it.  See attached photos.  Which had I paid attention, I actually showed it previously:
https://www.eevblog.com/forum/rf-microwave/experimenting-with-waveguides-using-the-litevna/msg4801919/#msg4801919
 

[coppercone2]

do you think time domain could be interesting? I have the beadless air line to try that with, its meant for time domain (TDR like?) calibration.

I wonder if that could tell anything about the interface I was so curious about.

I feel like time domain vs bolt torque would be interesting on the horn coupler interface. But I still need to repaint the horn after sealing the chipped waveguide with putty.

[coppercone2]

I bit the bullet and bought 2x wr62 adapters, that are within the cal range of the equipment with favorable connectors. 

Better to just do this right then fool around guessing about the 2ghz above cal range when I see things go south

Just the horns will be less long. I am comfortable to say the horns I made are probobly doing something and might work for something basic but the measurement values are trolls