Experimenting with waveguides using the LiteVNA

[joeqsmith]

Interesting question from viewer Wave64.  Sadly they removed it.   Turns out what they really were asking was why the offset short (guessing the short as well) is a dot rather than an arc.  I'm sure most viewer's who follow the channel know that I have been using ideal coefficients but I suspect that in this case, the poster did not understand what that meant.  Here are a couple of articles on the subject that may help other understand why my setups always show dots after cal rather than arcs.   


https://edadocs.software.keysight.com/kkbopen/when-we-measure-the-open-or-short-reflection-on-a-vna-why-is-the-result-in-arcs-but-not-a-dot-on-a-smith-chart-as-most-people-expect-620142095.html

https://coppermountaintech.com/introduction-to-the-metrology-of-vna-measurement/

[coppercone2]

I love the units provided for the calibration standards, + something  ^-42 henrys

[joeqsmith]

I pulled the specs for the 11664-20021 WR90 shim from here:

https://testequipment.center/Product_Documents/Agilent-P11644A-Specifications-6D081.pdf

With a nominal thickness of 9.78, the delay works out the 32.63ps.   3/8 standard would be 14.67mm 48.95ps.  The Maury 3/8 standard is 14.48mm 48.32ps.   So, a bit of an offset.  If I wanted to tighten things up, I would start by bolting/clamping the parts together.  Then measure the standards depth.  With so many other sources of errors, I see no reason to chase that rabbit.

[coppercone2]

I was actually trying to determine a torque limit for the 3d printed resin parts. Maybe your parts can handle it, but flashlight inspection is basically demanding that I get conductive gaskets of some sort for the interface.

I finally played around with VNA calibration, using the sliding load.. .but of course the antenna is right at the end of the calibration kit.

I need a saturn printer so I can get into lower frequencies without fooling around with glue seams and such, its hard enough as it is to paint these things IMO. Or just settle for shorter horns. Plated loop antennas too maybe.


hopefully a little bit more misering will get me out of this impasse

[joeqsmith]

You could make a metal plate that would wrap around the flange, "U" shape.  Spread out the force. 

I have a few PLA parts that I never did anything with.   It's certainly possible to measure the flange flatness and attempt some sort of flashlight leakage test.  With you working with different waveguides, I don't think it would be very helpful but I'm willing to have a look if you like.   

[coppercone2]

Yeah all I did is just hold a flash light on the other side while its pressed in. I think I need to lap the paint.

The first horn I smoothed out the supports via sand paper and optics plate, the second horn I just used as it was pulled from the work place on the elegoo mars printer, I assume its pretty flat but I did not check, but they dont pass the flashlight test.

Do not look forward to trying to sand that paint because the aperture is in the middle and the paint on the edge is IMO of dubious integrity. Its been cured for a while so I can try a gently wet sand.. the problem I had last time was the sand paper I had curled up and it was pissing me off

I can check the flatness if I buy some of those stupid ass 3.6V AA cells for the dial indicator I have (like xeno cells), because that one fits on the granite block. I got the granite 4 years ago and I aint do shit with it yet lol, 'federal tools' are about as fun to work with as the IRS


I wonder how many attempts its gonna take to get the proper wave guide coupler lol, this whole endevor is pure wrangling. When I installed a garden hose faucet into a brick and cinderblock wall behind a tree I pressed out of the way, which was maybe a 1 hour job, and stretch that into 2 years, its the same feeling. The tree was just bearly small enough to push out of the way if you really really tried hard. (freeze proof faucet that is long). brittle plastic and copper = masonry plumbing 100% similarity

[joeqsmith]

I was thinking to run it on the profilometer (optical).  I have a granite table at home that I use to setup my engines.  My V blocks and table are just big enough to set a crankshaft up on it.  If you just want a ballpark idea of flatness, this would be a very simple check.  I could take a few photos in the dark using the flashlight with  the flange laying on that table while I'm at it.   Maybe run one that wasn't coated and one that was.   

[szoftveres]

Interesting question from viewer Wave64.  Sadly they removed it.   Turns out what they really were asking was why the offset short (guessing the short as well) is a dot rather than an arc.  I'm sure most viewer's who follow the channel know that I have been using ideal coefficients but I suspect that in this case, the poster did not understand what that meant.  Here are a couple of articles on the subject that may help other understand why my setups always show dots after cal rather than arcs.   

I removed it, it was a pointless conversation; the above links didn't address my comment.

A 1/4 wave offset short is a 1/4 wave impedance transformer, transforming a short to open, but only on a single frequency. In the waveguide world, the only way you can present a high-Z termination to a VNA is by transforming a short at a single frequency with a 1/4 wave impedance transformer. Waveguide cal algorithms know about this and compensate for this, and when you re-measure your offset short (or as a matter of fact, any "stub") you are supposed to see an arc, as with any 1/4 wave stub. Except, when your calibration algorithm wrongly assumes that an offset short is a 1/4 wave offset short on every frequency - in that case you get the dot when you re-attach it after calibration.

You can fix your calibration algorithm, by de-embedding, or adding transmission line sections (i.e. compensating for phase), depending on how far you are in frequency (and which way) from the 1/4 wave of your offset short.

I hope this helps.

[coppercone2]

I was thinking to run it on the profilometer (optical).  I have a granite table at home that I use to setup my engines.  My V blocks and table are just big enough to set a crankshaft up on it.  If you just want a ballpark idea of flatness, this would be a very simple check.  I could take a few photos in the dark using the flashlight with  the flange laying on that table while I'm at it.   Maybe run one that wasn't coated and one that was.

well no pressure because our printers are different, I don't plan on getting into extruders any time soon, its not useful data for me other then curiosity. I believe you will have different figures because of

1) build plate heater, I believe you have significant thermal differential in the structure as its being built in a extruder printer
2) thermal contraction of extruded plastic on a local level
3) possibly moisture related stuff

I think there is very little heating occurring in the resin process, its essentially fluid cooled the whole way.


But, I guess we both are painting it. Maybe that is a common ground between the two manufacturing methods.And uh it might be interesting to know what the plating is doing.

I will order some xenos for the DI

[joeqsmith]

OK then, just a simple edge of uncoated flange on horn #4.  Stacked coated horn #1 on top of it and used your flashlight suggestion.  Shown with lab lights off and camera pointing right at the joint where the two flanges meet.  No clamps or bolts, just gravity. 

Also note that horn #1 is right off the printer.  No prep work.

[coppercone2]

Hmm nice quality your paint job must be alot better then mine. I bought some 3.6V cells for the depth indicator, just need to buy some mounting arm for the dial indicator now, should be able to get nice measurements with those two.

[coppercone2]

I have a arm, that is 3 sections, 1/4 and 3/8.

I managed to put a post on the granite plate that has a living hinge and a 12-24 screw thread on it.

So I need some way to attach the 3/8 inch shaft to the 12-24 screw and then a dove tail thing to the 1/4 shaft that fits on the digimax 1040 dovetail, which seems to reset to 0.0001 off sometimes, but thats good enough.

So I think I just need 1 swivel thing and one other thing. At least i quantified the problem now.

[coppercone2]

ok I got two starrett snugs that should be able to attach the fowler TI to an arm and then the arm to the stand, I just need to make a adapter that is 12-24 thread to 1/4 inch smooth shaft out of brass that I can stick into the stand, I will braze that together I guess.

been putting off figuring out how to do that for like 4 years lol, those snugs are not cheap

At one point I almost decided to soften and thread the indicator arm so it screws into the 12-24 thread, but I think snugs are the proper way to do it lol

[coppercone2]

also do you know what I should expect from a test indicator, the one I have will return to 0.0001 after I move it, sometimes, but usually to 0.0000.

Is it something I can fix by unscrewing the arm and say cleaning it? I put a drop of silicone oil on there.

[joeqsmith]

All of my general purpose dial indicators have a 0.001 scale.   I have mostly Starrett.  One from Flowler that I keep just for setting up my cam shafts.   I have one of these G models for example but MUCH older...
https://www.starrett.com/metrology/product-detail/711GCSZ

I do have a set of connector gages and test standards.  These are rated for 0.00025.  If I want to measure something tighter, I use one of the profilometers at work. Showing a few bits from a carbon seal from one of my bikes.

[coppercone2]

thats really nice. I figured out how to deal with my setup, the off center drilled 12-24ish hole can easily be enlarged to 1/4 -20 . I threaded a brass rod to 1/4 - 20 on the end so I can stick that in there and put the snugs on it. should have done this 4 years ago

Its fortunate this does not need to be a strong joint, because that thread I managed to make is some real smelly bs, but now the big cast iron ? arm thingy has a brass 1/4 inch rod sticking out of it for dial indicator arm attachment

[joeqsmith]

Interesting question from viewer Wave64.  Sadly they removed it.   Turns out what they really were asking was why the offset short (guessing the short as well) is a dot rather than an arc.  I'm sure most viewer's who follow the channel know that I have been using ideal coefficients but I suspect that in this case, the poster did not understand what that meant.  Here are a couple of articles on the subject that may help other understand why my setups always show dots after cal rather than arcs.   

I removed it, it was a pointless conversation; the above links didn't address my comment.

A 1/4 wave offset short is a 1/4 wave impedance transformer, transforming a short to open, but only on a single frequency. In the waveguide world, the only way you can present a high-Z termination to a VNA is by transforming a short at a single frequency with a 1/4 wave impedance transformer. Waveguide cal algorithms know about this and compensate for this, and when you re-measure your offset short (or as a matter of fact, any "stub") you are supposed to see an arc, as with any 1/4 wave stub. Except, when your calibration algorithm wrongly assumes that an offset short is a 1/4 wave offset short on every frequency - in that case you get the dot when you re-attach it after calibration.

You can fix your calibration algorithm, by de-embedding, or adding transmission line sections (i.e. compensating for phase), depending on how far you are in frequency (and which way) from the 1/4 wave of your offset short.

I hope this helps.

With your comment about it being a pointless discussion and the links I provided not addressing your comment, I was surprised to see you post again.   I'm actually glad you did.  I pondered a day to see if I missed something in your most recent post.  Sorry but it did not help me understand your point.   

As I stated, pretty much all of data I have presented for these low cost VNAs has used the ideal model.  So the SOL and SSL will always converge to a dot.  You understood the two articles and suggest this doesn't address your comment.   We both seem to understand that the open should be an arc but you seem to think it has something to do with the algorithm and not my use of ideal coefficients.   

Start with why you feel I need to fix my calibration algorithm and why you feel that using the ideal model does not explain what I am showing. 

[szoftveres]

Start with why you feel I need to fix my calibration algorithm and why you feel that using the ideal model does not explain what I am showing.

I think that you wrongly assume that the same rules apply to both rectangular waveguide- and e.g. coaxial line calibrations, which is a honest mistake.

In the SMA coax calibration world, it's perfectly normal to ignore the parasitics of your cal standards and use "ideal" coefficients, as you correctly stated. An SMA cal standard, having let's say 1mm of transmission line before the actual termination (50 ohm resistor, open, or short) means an insignificant phase shift at let's say in the sub 5GHz world.
If your center frequency (I'll refer to it as 'f' from now on) is at 2GHz, the added phase shift at 2f caused by the parasitic 1mm transmission line will increase, but will still be negligible.

In the waveguide world however, (as I said above) the only way you can present an infinite impedance (equivalent to and "open" standard in the SMA world) is by transforming a short, with a 1/4 wave impedance transformer, which you achieve with a 1/4 wave offset short.

The phase shift of a 1/4 wave transmission line around f is not just a parasitic - it's predictable, and is extremely abrupt, so much so, that at 2f, your 1/4 wave offset short standard will look like a short again.

And I think here's where you wrongly assume that phase delays around f in a 1/4 wave offset short are just parasitics, and therefore can be ignored.

The phase delay of a section of transmission line that has a length that is 1/4 wave is extremely significant, and by calibrating it out (i.e. assuming it's just parasitics) you essentially take the ability away from your VNA to measure phase meaningfully.

The calibration algorithm for waveguides therefore have to compensate for the phase shifts when using an offset short.

I hope this helps, an I can assure you that your articles were interesting and useful, I just wanted to point out a concern (independent from your articles) with your method of calibration which is worth looking into.

[joeqsmith]

I'm still lost so bear with me.   Lets start very basic.   If we look at the standard definitions for the WR90 calibration kit (X11644A), the difference between 1/4 offset short is the delay.  Note the parasitics are all zero.   If we compare this with Solver's ideal standards we can see the delays are all 0 and it's using a 50 ohm reference.

[joeqsmith]

Attached using the same standards shown during the video you referenced, to calibrate my old Agilent VNA, using only the internal software. *** I would normally run this VNA remote using my software but because the algorithm is what you are questioning, we needed to remove it from the equation.   During calibration I selected the X11644A (X band waveguide) standards and measured S11 only.   After performing the calibration , the offset load was attached and measured.   

Keep in mind that this VNA is limited to 9GHz.  I suspect this is what we both expect.

***
Sorry that the Agilent's print screen makes it difficult to see the data so I took a photo as well to better show the results.

***
Add details.

[joeqsmith]

The PNA can export Touchstone files and using third party tools like AppCad and METAS, we can view this data.   Attached showing the measured offset short. 

Up to this point, I suspect we are both on the same page. 

[joeqsmith]

As I mentioned, my old Agilent is limited to 9GHz so I calibrated the LiteVNA using the same standards with the ideal model over the same range.   A few things to mention.  For the most part, I keep the Lite's IFBW set to 4KHz.  This is the maximum.   This is how I ran it during that video.   The Agilent was set to 35kHz. 

Like before, I just laid the standards on top of the transition.  Basically, I want to replicate what I had shown with the exception of changing the range to 8.5-9GHz so we can compare the results.   

Now, we can see the effects of having the ideal models delay set to 0.   The magnitude is very similar but the phase is not correct.  It's always zero, which when we plot the data on the Smith chart, we see a dot rather than the arc.   

[joeqsmith]

Assuming you have no questions, let's add some delay to the model. 

As I mentioned, I default to the 4kHz IFBW.  The Lite running standalone defaults to 2kHz.  Typically I am willing to take a hit but I have also repeated the calibration using a 1kHz IFBW to show the improvements.   No guide pins and we can see error from my very poor technique (not properly attaching the standard to the transistion).  Still its good enough for what we are looking at. 

***
add details.

[joeqsmith]

How does the offset short measurement compare between the LiteVNA and the old Agilent.  Considering the setup, it's alright.

[joeqsmith]

During the video I had shown the unterminated transition.  It's not a great reference considering all the clutter in the lab right now. 

Also shown is the Short, Offset Short and Load measured with both the Agilent and Lite. 

Hopefully by walking you through the basics of what I had shown during the video, we can now sort out where the confusion is.