VNA for cable characterization

[joeqsmith]

... Both were cal'ed using the NanoVNA cal kit.

Let's try repeating this experiment also using my supplied cal kit using the sorted Mini-Circuits ANNE load.  Both VNAs using the idea model.   All connections were properly torqued for each measurement.   Both VNAs were swept from 300kHz to 9.0GHz for each test.    The LiteVNA used was the latest hardware revision: 64-0.3.1, using the supplied cables.   Both VNAs were allowed to warmup a half hour. 

With the same standards, math, and torque, we have isolated some of the differences.   There is still the cables, connectors, VNA, DUT placement, other....   

This first test is showing the RG400 cable.   Green trace is the LiteVNA,  White is the Agilent.   

[joeqsmith]

The TEFLEX is a really old cable.   The connector repeats but the cable is poor.  Move it, things change.   

[joeqsmith]

I have two of the Pasternak RG58 cables that are roughly the same.  They were purchased at the same time, same connectors.  I think I swapped them when I ran the PNA last time which is why you see the difference.   

Green (Lite)/Gold (Agilent) is the cable 1

Yellow (Lite)/White (Agilent) is cable 2

Note just how different the two SMAs behave. 

[joeqsmith]

And finally, the homemade 70mm 50-25-50 Beatty airline.   Green is the LiteVNA, White the Agilent. 

Even with the Beatty standard which doesn't flex, there is a fair difference between the two setups.   If we used my better standards with proper coefficients, we could remove some of that ripple and get a better absolute measurement with the Agilent.   Still, the standards I have are not really good enough for what you are asking.   It's that hunting for perfect coax article I linked for you.     

[joeqsmith]

I wanted to give you some idea on how the movement of the test cables matters at the levels you are asking about.  I repeated the test after four ties were wrapped around the TEFLEX cable in an attempt to stabilize it.  This is also why I suggested when you use the LiteVNA to test your cables, you move them around a bit and see what effects that has.   

White is the Agilent, Green is the LiteVNA. 


https://www.youtube.com/shorts/2iLCt-u6M2M

CMT showing their TDR feature.   At roughly 7:30, they change the scaling to display ohms.

[pdenisowski]

I wanted to give you some idea on how the movement of the test cables matters at the levels you are asking about.  I repeated the test after four ties were wrapped around the TEFLEX cable in an attempt to stabilize it.  This is also why I suggested when you use the LiteVNA to test your cables, you move them around a bit and see what effects that has.   

Yes, phase (and amplitude) stability of cables is something that is often overlooked.  In my experience, the difference between cheap cables and expensive cables is often primarily the difference in phase stability. 

When you're constantly connecting, disconnecting, bending, moving, etc. cables (like VNA test cables), this becomes even more important.

Or you could just buy a big roll of tape* 


* This is a joke - taping cables to the bench is not a substitute for good quality cables

[joeqsmith]

At 3:00 in, notice how they move the cable to check the stability.  Their cable is a much higher grade than my TEFLEX example.   I had seen Brian's video showing the stability of the cables they supply but wasn't able to locate it.   



Another one I suggest OP watch on accuracy. 



Gore's demonstration of their test fixture:

[joeqsmith]

Or you could just buy a big roll of tape* 

* This is a joke - taping cables to the bench is not a substitute for good quality cables

Those are four coffee bag ties.  I like them because A) they are thicker and much more heavy duty than garbage bag ties, B) they are free.   

In my experience, the difference between cheap cables and expensive cables is often primarily the difference in phase stability.

From that original paper, they talk about the original Tektronix cable not only having in-house developed connectors but also the stiffness of the cable.  The RG400 cable I show is very stiff compared with the TEFLEX which is reflected in the data.   

[joeqsmith]

You can get high flex cables with decent specs but you do pay for them.  How much?? Looking at Digikey to get some idea on cost from various suppliers that would be used with a VNA.  They carry many of the name brands. 

Maury Microwave offers the SV-185-FF-48 for $5,678/ea.   
https://store.maurymw.com/cdn/shop/files/2Z-002.pdf

Looking at Pasternak (I think Fairview Microwave makes their cables for them with custom blue anodize),  you get an idea of how the connector options effect the price:
PE3VNA1801-24 SMA Male to SMA Male, 24", $1130/ea
PE3VNA2603-24 3.5mm Male to 3.5mm Male,  24",  $2020/ea
PE3VNA4004-24 2.4mm Male to 2.4mm Male, 24",  $2730/ea

Looking at Keysight, their PN# 85131E 3.5mm cable is $5342.   

LiteVNA with cables and standards is now $150ish.  That sorted load cost me under $100 and time to sort them. 

[joeqsmith]

This was the webinar I was thinking of. 
https://coppermountaintech.com/webinar/time-domain-measurement-and-time-domain-gating/

There is another that OP may be interested in on measurement uncertainties. 
https://coppermountaintech.com/webinar/effects-of-random-factors-on-the-metrology-of-vna-measurement/

[joeqsmith]

Same setup but measuring a section of RG59.  I don't have a good SMA to 75 ohm female BNC.   There is an SMB/F-BNC/M to BNC/F-BNC/F to test cable.     

[joeqsmith]

Same RG59 section/adapters,  using the crossover point.  Measured terminator at 54.0 ohms.   Sqrt (98.67 * 54) = 72.99 ohms.

[joeqsmith]

I wanted to try and measure the cable using the RCL meter.   Test cable is 6'2".  RG59 is 22.1pF per foot. Or roughly 130.1pF.   Measured 126.2pF, so we are in the ballpark.   Installing an BNC to SMA to SHORT, I measured 0.806uH.    So 79.9 ohms.   As with my 50 ohm cable measurements, readings with the RLC meter are again roughly 5 ohms higher compared with using the VNA.

***
Measuring a 220pF cap, RLC meter reads 221.7pF.   A 100nH measures 0.105uH. 
 

[Pinörkel]

Thank you for the many interesting measurements and reading suggestions. I already started digging into that. The more I read, the more I get the impression that one needs to get at least a 15000€ VNA bundled with the guy who build it, just to tell you that if you did not spend 10 times more on a calibration kit and another fortune on adapters, you could as well have tried just eyeballing your measurement using scaleless calipers. (Owning the company that does the local weather forecast might also help.) 

Nonetheless, my replacement liteVNA64 should be on the way. For now, I only managed to get some very cheap BNC-F to SMA-F and SMA-M adapters to go with it. The supplier of the liteVNA had only those and I wanted to have something to evaluate what types of adapters I would need and which work best. Strangely there seem to be only two types of SMA-adapters on the market: very cheap china sub-3€-stuff and ridiculously expensive adapters that cost well over 50€ up to infinity, nothing in between. At first, I wanted to get some of this nice looking compact type to directly screw onto the liteVNA ports. Then I noticed that the center pin would rotate every time they were screwed on or off and that can not be good for the gold plating of the liteVNA port contacts.

The following two from Fairview seem to be of at least acceptable quality (they have a datasheet).
https://www.fairviewmicrowave.com/sma-male-bnc-female-adapter-sm4721-p.aspx
https://www.fairviewmicrowave.com/sma-female-bnc-female-adapter-sm4709-p.aspx

At the moment, I am not sure what the best type of connection would be - directly to the VNA ports or with the short supplied SMA cables in between. As far as I see it, the liteVNA can only be calibrated right to its ports when using the supplied calibration kit and minimizing the amount of required adapters sounds like a good idea.

[joeqsmith]

As shown, some errors can be reduced just by using good practices.   

As far as I see it, the liteVNA can only be calibrated right to its ports when using the supplied calibration kit and minimizing the amount of required adapters sounds like a good idea.

  Of course you can calibrate with what ever you want between the VNA and the standards.  Obviously the calibration can not overcome a bad interface.  Software can't correct for a bad cable!!   

For all of the data I have presented here, I have used the cables supplied with the LiteVNA.  I am using a female to female adapter at the end of the cable. I am always calibrating at the end of the adapter. I tend to leave these parts attached to the VNA to avoid the wear. 

I may try and calibrate the Agilent and see what difference it makes.   Another test would be to try and characterize the standards I use with the LiteVNA.   I've seen people post about doing this but I have never used anything besides the ideal model with them. 

***
One thing to consider if you do decide to not use a cable between the VNA and your load, you may end up putting undue mechanical stress on the connector and damaging the VNA.  You also have the added wear of the connector.  I had seen some of the hams mount SMA to UHF connectors on theirs.  They would use panel mount types and bolt the entire assembly down to remove the majority of stress on the SMAs.

[joeqsmith]

Searching for that Tektronix cable, appears there are others that have looked at it as well.  In this first link, they were trying to use some version of my software with the cable connected between the two ports.  I'm not sure why. 

https://physik.co-i60.com/2022/01/new-addition-to-the-lab-nanovna-v2-plus4/

https://groups.io/g/TekScopes/topic/my_tdr_evaluation_of_the/31377831

[Pinörkel]

OK, time for some feedback. My liteVNA64 finally arrived yesterday and I had some time for trying to fiddle out how to make it work.
Getting the Solver64 software to be able to communicate with the device via any COM port was a four hour nightmare. Manually installing the correct CDC driver was easy, but some parts of the NI-VISA or Labview components that are responsible for serial communication repeatedly refused to install right on all three machines, I tried this on. I finally got it working by beating the heck out of some stubborn DLLs.

Then I tried to find a suitable measurement setup for the liteVNA, the supplied calibration kit and the four el-cheapo BNC to SMA adapters. Good thing, I ordered those. I put the supplied short SMA cables on the liteVNA and tried to connect everything else there. The calibration was performed with the supplied SMA-F to SMA-F adapter between the SMA cables and the cheap calibration standards. For cable test measurements, I left the adapter in place and connected an SMA-M to BNC-F adapter to attach the BNC cable as the DUT. At it's end, I used a BNC-F to SMA-F adapter to connect the calibration kit load as a 50Ω termination.
Measurement setup:


Then I played with calibrations for different start and stop frequencies, with connecting different cables, repeatability and SMA connectors. Insight number one: Some of my equipment and my RF measurement skills are crap. Improving the latter is the key to identify how to improve the first.

For testing, I used three self made RG58-CU cables, made from the same cable roll using Telegärtner BNC connectors. Two of the cables are of similar length of about little more than a meter and the third one is about 1.5 meters. I measured all three cables in the ranges of 50k-100M, 50k-3G and 50k-6.3G. I am still not sure, which ranges would be the best to use here. At least the measurements were somewhat consistent.
50k-100M:

50k-3G:

50k-6.3G:


Could the large spikes at the beginning and the end of the traces be an indicator of the really bad quality of my SMA to BNC adapters or is this just an expected effect of a BNC connection?

Regarding the repeatability, I redid some measurements after disconnecting and reattaching the DUT cable from everything. I also repeated some tests with just a redone calibration. I discovered that I had to tighten the SMA connectors with quite a bit of force by hand, to get some repeatability. Just disconnecting and reconnecting the cable sometimes lead to impedance differences of multiple ohms with the whole impedance curve shifted up or down(see image). I am still not sure what caused this. If, for example I had not tightened the double SMA-F adapter at the end of the SMA cable sufficiently or it came loose because of reattaching the DUT, I would have expected to see some kind of local impedance spike, but not a complete up- or down shift of the whole curve. Similarly for the experiments with three redone calibrations, I had some measurements with the same effect while other measurements were very close to each other. Then again, there might have been SMA connection issues on some calibrations.
Reconnection repeatability:

Recalibration repeatability:


If I am correct, this maybe hints at using a 0.5Nm torque wrench to tighten the brass SMA connectors. Someone told me, it should be quite difficult to over-tighten an SMA connector by screwing it on by hand without any tools. So maybe I can get around investing over 350€ in a good quality SMA torque wrench by just tightening the connectors very hard without using any force amplifying tools?

I will do some further testing in the next few days.

[tggzzz]

If I am correct, this maybe hints at using a 0.5Nm torque wrench to tighten the brass SMA connectors. Someone told me, it should be quite difficult to over-tighten an SMA connector by screwing it on by hand without any tools. So maybe I can get around investing over 350€ in a good quality SMA torque wrench by just tightening the connectors very hard without using any force amplifying tools?

if you only want to do One or two, make your own; it is easy enough.

Get an open end ring spanner where the open end fits the SMA. Measure the distance to the ring end, and calculate the force needed for the specified torque (or torques).

Get a spring scale with an appropriate FSD, and use it to pull on the spanner's ring end at 90degrees.



After using that a few times, you will have.sufficient "muscle memory" that the spring scale isn't so necessary.

[joeqsmith]

... My liteVNA64 finally arrived yesterday and I had some time for trying to fiddle out how to make it work.  Getting the Solver64 software to be able to communicate with the device via any COM port was a four hour nightmare. ....  I finally got it working by beating the heck out of some stubborn DLLs.

Having used LabView for several years at various companies and different PCs and OSs,  getting it to run hasn't really been a problem.   I've ran into people downloading random files from NI somehow expecting them to work.  Once in a while people have had a different VISA installed and run into problems.  I don't know what "beating the heck out of some stubborn DLLs" would mean but sounds like you were able to sort it out on your own.  That IMO puts you far ahead of most people who contact me about it.  So good job!!   There is a newer version that I have not uploaded that would automatically replot the data when you change the TDR memory and attributes.  It also changes the average function from a LIFO to a running average.  If you want it, let me know and I will upload it.   

Insight number one: Some of my equipment and my RF measurement skills are crap. Improving the latter is the key to identify how to improve the first.

Welcome to the club. 

I measured all three cables in the ranges of 50k-100M, 50k-3G and 50k-6.3G. I am still not sure, which ranges would be the best to use here. At least the measurements were somewhat consistent.

Higher frequency = increased the length resolution.   More data points = longer length that can be measured.   (you can see this in your graphs, 6GHz is clearly higher resolution).   At 50k, the LiteVNA has some issues but should be alright.  Looking at at the data I provided for you, I used 300k to 9GHz, 401 data points, 4kHz IFBW.  I suggest that you use my settings as a starting point.  You should at least be able to replicate what I have shown you. 

*** Bonus to you for not telling me how 10 data points increases the length resolution    hams 
 

Could the large spikes at the beginning and the end of the traces be an indicator of the really bad quality of my SMA to BNC adapters or is this just an expected effect of a BNC connection?

These are your connections. Every discontinuity will cause a disturbance in the impedance. A 20GHz VNA would do a better job showing what is going on there but with the Lite, we can at least get some idea which are better.   Yes, BNC connectors are poor.  Guessing that is why Tektronix made a custom one.

If I am correct, this maybe hints at using a 0.5Nm torque wrench to tighten the brass SMA connectors.

I couldn't tell you how many times people have written me about how pointless torque wrenches are and how they damage connectors.  Oh well.

So maybe I can get around investing over 350€ in a good quality SMA torque wrench by just tightening the connectors very hard without using any force amplifying tools?

Those 3D printed brakeover wrenches we made worked well after calibration.   The plastic wears fast and you would need to continually check it but they don't cost a lot to print.  Used is also an option.  You can always check it for calibration and adjust if needed.   

[pdenisowski]

If I am correct, this maybe hints at using a 0.5Nm torque wrench to tighten the brass SMA connectors.

I couldn't tell you how many times people have written me about how pointless torque wrenches are and how they damage connectors.  Oh well.

Oh, you absolutely can damage connectors with torque wrenches if you use them incorrectly.   But used properly, a torque wrench will help avoid many common types of damage.

So maybe I can get around investing over 350€ in a good quality SMA torque wrench by just tightening the connectors very hard without using any force amplifying tools?

Those 3D printed brakeover wrenches we made worked well after calibration.   The plastic wears fast and you would need to continually check it but they don't cost a lot to print.  Used is also an option.  You can always check it for calibration and adjust if needed.   

There are cheap Chinese torque wrenches available from Amazon for about 30 USD.  No, they are not metrology grade, but they are definitely better than using a non-torque wrench  

Or you could buy quality like these (just posted today, in fact )

[joeqsmith]

Or you could buy quality like these (just posted today, in fact )

Not as fancy as yours, video showing different printed plastic wrenches with my home made calibrator.  Hard to beat the price. 

Oh, you absolutely can damage connectors with torque wrenches if you use them incorrectly.   But used properly, a torque wrench will help avoid many common types of damage.

Yes, you can damage them as I demonstrate here.   

[joeqsmith]

CDI's website:
http://www.cditorque.com/

Pasternak's wrenches:
https://www.pasternack.com/click-type-rf-torque-wrenches-category.aspx

[vk6zgo]

Oh, you absolutely can damage connectors with torque wrenches if you use them incorrectly.   But used properly, a torque wrench will help avoid many common types of damage.

You can damage cylinder head studs with a torque wrench if the signalling device fails.
Been there, done that, got the T-shirt.

[Pinörkel]

I 3D-printed a flex-style indicator-needle-type like torque wrench(the thinnest one with the least amount of torque) just to compare the resulting SMA nut tightening with my hand tightening. The results were surprisingly close. Based on that, I also had the impression that it should be quite difficult to over-tighten an SMA connection by only using hand-tightening. I think, I will get an adjustable ~40€ class SMA torque wrench, once I have some more expensive connectors and stick to hand-tightening for now.

Apart from that, I did additional experiments regarding my poor repeatability and played around with the sweep settings. With the settings proposed by @joeqsmith (300k to 9GHz, 401 data points, 4kHz IFBW), I got a kind of repeatable offset of ~1.5Ω when compared to my initial 50k to 6.3G settings, which I cannot really explain. If I repeat the calibration with either of the two settings, I get kind of repeatable results with small deviations, but just loading a calibration of the other sweep parameters type (and not even moving anything else) will shift the whole curve up or down starting with the peak produced by the first adapter.

Result differences between sweep parameter types:


Next, I had a closer look at the SMA-M to BNC-F adapter between the SMA cable and my DUT. Swapping it with the second adapter I have of that type, also produced a nearly 2Ω offset. Additionally, I discovered that with one of the adapters, applying even a little bit of lateral tension on the connection will shift the curve around. The image shows a measurement of the two adapters in green and red and the white curve is the result of applying a small amount of tension on the connection. I would take from this that both adapters have an inaccurate impedance, bad coupling, and one of them is even mechanically defective and will not be used anymore. The respective adapter also had a kind of mushy end stop when screwing it on.

Result differences between different adapters:


Looking at available SMA adapters at Mouser, I noticed that nearly only the ones of Rosenberger had datasheets with VSWR specifications and torque recommendations for their products resulting in over 40€ per piece(e.g. parts 476-32S151-K00L5 and 476-32K151-K00L5). So, those might only be suitable to buy in small numbers for calibration work or high precision measurements. Then again, I would need some adapters for daily less critical work to protect the expensive ones from unnecessary wear. Maybe the ones of mini-Circuits could be an option, because they at least specify VSWR (parts: 139-SF-BF50+ and 139-SMF-BF50+)

I also found this description on how to use several Rosenberger parts as an affordable calibration kit, suitable for sub 5GHz VNAs: https://hhft.de/sma-cal-kit

[joeqsmith]

I 3D-printed a flex-style indicator-needle-type like torque wrench(the thinnest one with the least amount of torque) just to compare the resulting SMA nut tightening with my hand tightening. The results were surprisingly close.

Thinnest doesn't tell us much.  Materials will play into it.  Even knowing that, it still offers nothing.  You need to measure the torque then provide that detail.   

Based on that, I also had the impression that it should be quite difficult to over-tighten an SMA connection by only using hand-tightening.

I wouldn't think you could over-tighten by hand, but you could certainly under-tighten.  Want to know, then measure.

I think, I will get an adjustable ~40€ class SMA torque wrench, once I have some more expensive connectors and stick to hand-tightening for now.

If you like to guess, by all means, fingers work fine.   You mentioned early on you have one of the books I read.  Have you taken the time to read it?  It's a pretty good introduction into how to use a VNA, and is filled with tips.  Sadly, I have not found a way to absorb the materials outside of taking the time to read them.

Apart from that, I did additional experiments regarding my poor repeatability and played around with the sweep settings. With the settings proposed by @joeqsmith (300k to 9GHz, 401 data points, 4kHz IFBW), I got a kind of repeatable offset of ~1.5Ω when compared to my initial 50k to 6.3G settings, which I cannot really explain. If I repeat the calibration with either of the two settings, I get kind of repeatable results with small deviations, but just loading a calibration of the other sweep parameters type (and not even moving anything else) will shift the whole curve up or down starting with the peak produced by the first adapter.

As I mentioned, the upper frequency effects the resolution of the distance measured.  I doubt you would measure a 1GHz clock with a 1MHz scope.  You may find the slow scope greatly attenuates the signals amplitude.  Its similar when you try and measure very small distances without a high enough frequency.  You may find your 60ohm connector measures 50ohm as you lower the upper frequency.  You saw this effect in the very first test data you uploaded.  At that point you were unclear what was causing the spikes but I explained it was the connectors and that the LiteVNA is not good enough to measure them because it is such a low frequency device.  This is explained in that book you have. 

The lower frequency will get you a better DC measurement.  50k vs 300k shouldn't make too much difference in the context of these tests.  As I mentioned, I used 300k because that is the lower limit of my old Agilent and I wanted to provide you with some comparisons between the two.   Also as I mentioned, the LiteVNA may start to have some problems at 50KHz. Should be alright but again, I don't think you are gaining much for this particular test by running it this low.

Next, I had a closer look at the SMA-M to BNC-F adapter between the SMA cable and my DUT. Swapping it with the second adapter I have of that type, also produced a nearly 2Ω offset. Additionally, I discovered that with one of the adapters, applying even a little bit of lateral tension on the connection will shift the curve around. The image shows a measurement of the two adapters in green and red and the white curve is the result of applying a small amount of tension on the connection. I would take from this that both adapters have an inaccurate impedance, bad coupling, and one of them is even mechanically defective and will not be used anymore. The respective adapter also had a kind of mushy end stop when screwing it on.

    Welcome to the club.  I didn't tiewrap that TEFLEX cable to make it look all nice and neat and take up less space.  I wrapped it to try and mechanically stabilize it to get a repeatable measurement and to show YOU how important these details are.  I am not surprised that different connectors would be more stable.  Compound that with hand tightening....  based on your accuracy goals,  you have some work ahead of you.   

Looking at available SMA adapters at Mouser, I noticed that nearly only the ones of Rosenberger had datasheets with VSWR specifications and torque recommendations for their products resulting in over 40€ per piece(e.g. parts 476-32S151-K00L5 and 476-32K151-K00L5). So, those might only be suitable to buy in small numbers for calibration work or high precision measurements. Then again, I would need some adapters for daily less critical work to protect the expensive ones from unnecessary wear. Maybe the ones of mini-Circuits could be an option, because they at least specify VSWR (parts: 139-SF-BF50+ and 139-SMF-BF50+)

Well, how about we say maybe the connectors will help.  I think there are other variables you could tie down before introducing your 3.5mm precision adapters into the mix. 

I calibrated the LiteVNA64 (latest revision) using three different settings, changing both the start and stop frequencies.  The SMA connector measured a peak of 55 ohms with the two 9GHz tests and 53.5 with 6.3GHz.  Matching your 1.5Ω difference.  Again, maybe it is 57 ohms.  We really can't tell without having a better VNA.  Consider that home made Beatty standard.  It's 70mm!!  We can just see it settle.  The LiteVNA64 with harmonics can still provide some small amount of data.  Why not use it?

The coax of course is MUCH longer than the connectors.  So now the upper frequency has less of an effect on the impedance and we should get the same values if there were no other variables.   The fact yours changed so much tells me there are other errors with your setup.