VNA for cable characterization

[Pinörkel]

Hello,
I am new to RF measurements and would like to perform some high precision cable impedance measurements to characterize some short(0.5m to 2m) 50Ω BNC cables of different types in the MHz and low GHz range.

I read that cable impedance measurements of the required accuracy can be done with a TDR sampler like the Tektronix 7S12 with a respective pulse generator and sampling head. A setup like this should be able to measure the impedance along the cable with an accuracy of +/- 0.1Ω or +/- 1 mρ and a time resolution in the sub nano second range. It should also be possible to do frequency dependent impedance measurements with a suitable milli-ohms-meter like the R&S URV5 with a suitable insertion unit. Unfortunately, both vintage setups are extremely difficult to acquire and I do not have the required bench space and operating knowledge for a suitable sampling oscilloscope. So, I tried to dig into the topic in order to maybe find some kind of measurement setup that could somehow fulfill my needs and stumbled onto vector network analyzers.

It seems that high quality VNAs, like most kind of test equipment, can get ridiculously expensive, but there are also affordable solutions, like the liteVNA 64, which might or might not be sufficient, for what I want to do. However, until now I could not figure out what measurement precision can be expected from a device like the liteVNA 64 with respect to cable impedance and TDR measurements. This is where I hope I can get some answers here.

I already found the very feature packed Solver64 software from forum member joeqsmith, who has put tremendous amounts of work into providing countless features like TDR measurements for the inexpensive VNAs. I read that the performance of these devices depends heavily on the usage of high quality calibration standards, which might easily cost more than ten times the price of a budget VNA. So my question would be: does anyone here know what TDR measurement performance in terms of impedance deviation can be achieved with a liteVNA 64 or similar device, combined with some affordable calibrations standards? Or am I mistaken and a VNA is not the right tool for this task? Any help in this matter would be appreciated.

[jwet]

I saw that you got no responses so I'll give some guidance to get a conversation started perhaps.  I'm not a renowned expert at all but there are some on this board.

I think a standard VNA, cheap or fancy is not a great high precision tool for your application.  The problem is that since VNA's depend on reflections and the reflections will be very small (loads close to 50 ohms) , relative errors will be large.  The dynamic range of a decent VNA is 80+ db, cheap VNA's are about 70 dB.  Flatness of a good VNA is .05 db which will turn into fraction of an ohm.

There are RF impedance measurement devices that use an V/I technique.  One is the HP4396B that I'm familiar with.  It looks and acts like a VNA but in impedance measurement mode, it uses the I/V technique.  HP has an impedance measurement handbook that is pretty good and goes through their solutions and limits of accuracy.  Good standards are expensive and will be required to get accurate absolute results.

Good Luck.

[Pinörkel]

Thank you for the kind and helpful reply. So, it might be as I already feared: my best bet is to lurk on ebay and wait a few years for a chance to get some vintage TDR equipment. Unfortunately, this stuff is much harder to get here ein Germany than in the US. In any case, I will have a look at the HP impedance handbook you mentioned. Until now my main literature source about VNAs was The VNA Applications Handbook from Bonaguide and Jarvis.

I saw that you got no responses so I'll give some guidance to get a conversation started perhaps.  I'm not a renowned expert at all but there are some on this board.

I think a standard VNA, cheap or fancy is not a great high precision tool for your application.  The problem is that since VNA's depend on reflections and the reflections will be very small (loads close to 50 ohms) , relative errors will be large.  The dynamic range of a decent VNA is 80+ db, cheap VNA's are about 70 dB.  Flatness of a good VNA is .05 db which will turn into fraction of an ohm.

There are RF impedance measurement devices that use an V/I technique.  One is the HP4396B that I'm familiar with.  It looks and acts like a VNA but in impedance measurement mode, it uses the I/V technique.  HP has an impedance measurement handbook that is pretty good and goes through their solutions and limits of accuracy.  Good standards are expensive and will be required to get accurate absolute results.

Good Luck.

[joeqsmith]

You never state your requirements,  what your goals are, or the reasons behind them.   Reading your post, I get the impression that you feel that 50 vs 50.1 is a big deal for your application, like the quote below taken from the linked article.  I want to understand why? 

Perhaps now maybe the “50 Ohm” cable idea makes some sense and you are now a “50 ohm” systems zealot. You now strive for “perfect 50 ohms” in all your cabling, connections and devices. You have become so unreasonable that you insist that all systems be EXACTLY 50 ohms.   Well now you are in trouble.

https://www.dsinstruments.com/support/understanding-characteristic-impedance-vswr-reflection-coefficient/

[Pinörkel]

Thank you for the link and for pointing out that I did not provide enough information for the question to be answered. I tried not to overload the question with irrelevant information, to prevent readers from thinking: Too long didn't read. Thankfully, the situation is not as bad as in your quote. I have mainly two goals with this:

The first and less critical goal is that I want to be able to verify the quality of BNC cables I buy or make. In the past, I have bought cables from sources I thought I could trust, only to discover later, that the connectors were of terrible electrical quality or the cables had abysmal frequency dependent return losses. Also some cables may go bad with time unnoticed, e.g. foam dielectric coax cables which have been bend too much. This made me believe several times that certain measurement gear was bad, when in fact it was only a matter of bad cables. So being able to ballpark my cables would be nice to find out, which cables to throw away and which not to use for precision measurements.


The second goal is that I have repaired some vintage calibration gear. For a correct calibration of the gear and the devices that will be calibrated with it later on, there is the requirement of using a precision 50Ω +/- 1% cable with a length of 36 inches. These cables are then paired with the calibration device in order to later be able to make precise statements about the calibration accuracy of devices calibrated with this combination. Unfortunately, the original cables used for this e.g. the Tektronix 012-0482-00 are near to being unobtanium nowadays and I know from people that worked with these cables on a daily basis back in the days, that they had a tendency to go out of spec after a while. So, original vintage cables cannot be trusted without testing and a way is needed to validate them.

Due to the bad availability of original cables I tried to get information on the precise specifications of the cables to see, if maybe there was a way of making those using parts that are available today. According to my current findings, the validation of a cable should at least include verifying that the impedance stays within 50Ω +/- 1% below 1GHz and that the VSWR below 3GHz is less than 1.3. A few years ago, Dennis Tillman did an in-depth TDR evaluation of an original cable (link) to reverse engineer its specifications using a Tektronix 7S12. This evaluation suggests that an impedance measurement accuracy of at least +/- 0.1Ω or +/- 1 mρ is possible and also allows a precise look at what is going on in the used BNC plugs (given a high quality 50Ω termination is used). So, 0.1Ω will certainly be no big deal for my application as long as the 1% spec is met, but I had the impression that this resolution may be the minimum required to see IF the 1% spec is met.

Based on this, I managed to identify several combinations of available parts(BNC plugs and cables of different types) that could be suitable for creating respective cables. However, the the quality of the cables not only depends on the parts, but also on the manufacturing process. The latter has some potential to be screwed up by me. So, just buying several combinations of parts and making different types of cables would be useless without a way to measure and compare the cables. At the moment, I do not have sufficient gear or knowledge to reliably take the required measurements. So, I am now trying to find out what I need to know to take the required measurements and if there is a way to acquire the respective measurement gear. After all, just owning measurement gear usually does not mean that you know how to use it correctly. So, time for some research. At the end, if possible at all, it would be nice if I could come up with a recipe, that can reliably be used by others to create calibration cables of this type, optimally for a reasonable price.

You never state your requirements,  what your goals are, or the reasons behind them.   Reading your post, I get the impression that you feel that 50 vs 50.1 is a big deal for your application, like the quote below taken from the linked article.  I want to understand why? 

Perhaps now maybe the “50 Ohm” cable idea makes some sense and you are now a “50 ohm” systems zealot. You now strive for “perfect 50 ohms” in all your cabling, connections and devices. You have become so unreasonable that you insist that all systems be EXACTLY 50 ohms.   Well now you are in trouble.

https://www.dsinstruments.com/support/understanding-characteristic-impedance-vswr-reflection-coefficient/

[pdenisowski]

There are many different ways of measuring the impedance of cable, but if I had to choose a methodology that's a good compromise between accuracy, simplicity, and cost, I would use a VNA and the methodology I describe in this video:

[gf]

In order to preserve the leveled output of the SG503 generator, wouldn't it be sufficient to verify that the cable's |S21| is flat enough (say +/- 0.1dB) over the generator's 250kHz...250MHz range?

[joeqsmith]

In order to preserve the leveled output of the SG503 generator, wouldn't it be sufficient to verify that the cable's |S21| is flat enough (say +/- 0.1dB) over the generator's 250kHz...250MHz range?

In OP long post, they fail to mention anything about the actual test equipment,  only the need for a Tektronix 012-0482-00 cable.  Was the SG503 Tektronix's only use case for it?  Sounds like a higher performance cable may not be better for this application.  They setup the generator for that one specific cable, what ever it is.   

It does appear you can buy them new.  For a standard,  $100/ea seems like a no brainer. 
https://www.testequipmentdepot.com/tektronix-012-0482-00-bnc-male-to-male-precision-1-coaxial-cable-50-ohm-36-in.html

Because OP mentions the LiteVNA and I had mine out, ran a quick test using the following cables.  All cables are fitted with a 50ohm BNC and SMA.  All connectors are different.  All cables are within a foot of each other.  Using my BK RLC meter to measure each cables capacitance and inductance to derive the impedance.


TEFLEX
60.1

Pasternak
RG58C/U
56.5

MIL-C-17G THERMAX
RGS-400
56.8

All read high using this method with my meter.   Using the LiteVNA w/ TDR, calibrated using the V2Plus standards with Mini-circuits ANNE for load (sorted for RL), and the ideal model.  Sweep range set to 6GHz.    Should give you some idea of what to expect.  Maybe using good standards that have been characterized and using those coefficients you could improve your measurements enough to meet your goal.  Of course, that may cost more than just buying a new cable. 

If my only goal was to test cables I have,  I would physically move them around while measuring them to make sure they are stable, but would consider the LiteVNA even with poor standards and ideal model,  good enough.

[Pinörkel]

@gf:
According to my current knowledge, regarding the special case of the SG503, the answer to your question is a clear "maybe". The output flatness seems to be one of the main design features of the 012-0482-00 cable. This is achieved by using a precision cable with custom made 50Ω BNC connectors with a reasonable return loss. In addition to that, a cable with a solid center conductor and a solid dielectric is used, most likely to increase the robustness against bending and kinking. According to the manual, deviating too much from the intended cable length of 36 inch can cause amplitude variations of up to 4%, but getting the cable length right is a trivial task.

Lets assume it would be sufficient to make a cable from high quality BNC connectors and any low loss high quality 50Ω cable (e.g. RG233 or H155), that has a flat enough $$|S21|$$. As I understand it, for this to meet the original specs, the required flatness in dB has to be derived from the allowed impedance deviation of 1%.

I read through some of the stuff you linked and some additional documentation from Keysight.
With $$Z_{t}$$ being the transmission impedance of the cable and $$Z_{0}$$ being the system impedance which equals 50Ω, I found the following equation to compute $$Z_{t}$$ from $$|S21|$$.
$$ Z_{t} =  Z_{0} \frac{2(1-|S21|)}{|S21|} $$
Since S21 seems to be a complex number, I assume its magnitude has to be used here, which should be a linear value.
To get to and from the logarithmic dB values you mentioned, the following two terms can be used according to my sources:
$$|S21|_{dB}= 20\cdot log_{10}(|S21|)$$
$$|S21|=10^{\frac{|S21|_{dB}}{20}}$$
I will math-fumble with that next weekend.


@joeqsmith:
Thank you very much for going through the trouble to make some example measurements of different cables. From what I can see there, the liteVNA, combined with your very nice solver64 software should be more than enough for my basic cable quality testing application. It looks like it will be able to give me a good idea of the cable impedance and also the return loss, to evaluate and compare my bench cables. This capability alone will most likely make me buy a liteVNA. Although it will not produce high accuracy measurements, it will be enough for that case and has a very compact shape and competitive price.

As for the 012-0482-00 cable: the SG503 was probably the most common application of this cable, although it appears in several other Tektronix calibration gear service manuals. The SG503 was optimized to perform best with a cable of this type an then each individual generator is matched to a specific 012-0482-00 cable during calibration.

The thing with new 012-0482-00 cables is that all the sources I could find up to this date only deliver to locations in the US or Canada, but not to Europe. Also, some sources who list the cable as "not in stock, but can be ordered" will call ridiculous prices (>500€) for something that can most likely be made from parts between 30€ and 50€.

[joeqsmith]

I'm not sure what a good set of BNC connectors and cables would cost.   The SMAs were shown towards the start of my data, not the BNCs.   Then you have the crimp tool.  Once you have all that, you still have to verify it meets what ever your requirements are which maybe you could rent the equipment?   Get all of that sorted and you still need to get the cable working with what ever equipment you have (which I am still not sure of). 

If the end goal is just to have a leveled output, maybe there are better options.   Maybe just a good splitter, detector, PC and some software to control your RF generator or attenuator?  I played with some of those dirt cheap log amp eval boards which seemed to be fairly accurate at the 50MHz I was playing with them.   Still, tough to verify.  It may be time to think of a good used Watt meter and send it off for calibration?     

[tggzzz]

... only the need for a Tektronix 012-0482-00 cable.  Was the SG503 Tektronix's only use case for it? 

Calibrating Tek 1502 TDRs, both checking the impedance and adjusting for length.

The 1502 (not A B, 1503 or any combination thereof!) has a 140ps system risetime, so you can resolve impedance changes 1-3cm apart. Lovely devices; I have been known to cart them around hamfests to check the quality of the cables (BTW, don't but one without seeing it working; it requires a functional NiCd battery before it will turn on)

If the OP has a suitably fast scope, they could make their own TDR. Use one of Leo Bodnar's 50ps step generators, or even a few 74LVC1G* gates can push 2.5V steps into 50ohms with 250-300ps risetime. Need to be rather careful about the layout, though.

[joeqsmith]

There are many different ways of measuring the impedance of cable, but if I had to choose a methodology that's a good compromise between accuracy, simplicity, and cost, I would use a VNA and the methodology I describe in this video:

Attempt to use this technique with the Pasternak  RG58C/U cable.  This cable is 3' or 0.9144 meters long.  Using their 75/length in meters for fstop, call it 75MHz.    They talk about start not critical.  100kHz or lower?  I set it to 300k in case I decided to try and replicate with my Agilent (limited to 300k).   We are down into the muck, so set the IFBW to 1.3kHz.  I had to enable the magnification for the Smith chart so we could see something.  Then had to add two more gain ranges. 

Again, using low grade standards and ideal model.   Yellow showing the load, red is cable + load.   I am not sure why they care about the stop frequency rather than maybe just wanting the first cross over?  Maybe more data points?  If we use the second cross over of 50.5,  sqrt (50.5 * 50) - 50.25.   The first cross over is at 50.1 giving us 50.05.   

Guessing better standards and VNA would help but I think we are just asking too much from the low cost setup.   

[joeqsmith]

Also note the amount of drift at the low frequency end.  This was after about an hour of warmup.   

[joeqsmith]

After several more hours of running.  This particular LiteVNA64 is the latest hardware revision.

Still, for the $120 we paid for these, they are excellent. 

[pdenisowski]

Still, for the $120 we paid for these, they are excellent.

I have several "hobbyist" VNAs and they are indeed an excellent value for the money. 

I also think that the availability of these hobbyist VNAs is good for the "professional" VNA industry as well.  Historically, it was hard to get hands-on experience using a VNA since they were priced well above what a hobbyist could afford.  The Nano, etc. VNAs provide a great opportunity to both build up VNA skills in the "community" as well as perform basic (or "undemanding") measurements.

The cable impedance methodology I posted is something I've only ever done with a "professional" VNA, so it's perfectly possible it might not work well (enough) with a "hobbyist" VNA. 

Incidentally, Alan (w2aew) did a video on this same methodology using a NanoVNA

[joeqsmith]

Skimmed it enough to notice how he takes the easy route and used 75.   Would have liked to have seen him try it at 50.   That said, the original NanoVNA may do a better job than the LiteVNA in this case.   

I assumed OP wasn't interested in professional systems, but with you having both, maybe they would like to see how a  couple of cables would compare using your professional vs low cost VNAs/standards.   Because they are asking about a 50ohm cable, maybe stick with that. 

[pdenisowski]

maybe they would like to see how a  couple of cables would compare using your professional vs low cost VNAs/standards.   Because they are asking about a 50ohm cable, maybe stick with that.

Good idea.  I'll do it both ways in the next day or so and post the results.

[joeqsmith]

If you have a standard (air line) with a known impedance to use as a reference, it would be really interesting to see how close the various setups are to the published values.   

[Pinörkel]

I'm not sure what a good set of BNC connectors and cables would cost.  The SMAs were shown towards the start of my data, not the BNCs.   Then you have the crimp tool.  Once you have all that, you still have to verify it meets what ever your requirements are which maybe you could rent the equipment?

A pair of good quality BNC crimp connectors starts around 16€ and the cable will add another 4€ to 10€, not including shipping. Very high quality connectors can cost double the price. I have an acceptable crimping tool available, but it does only work for some RG58 and RG223 and H155 cable connectors. I may have the chance to access some expensive equipment at my university, but that would be complicated to get and of course require that I can prove that I know how to use it. You wouldn't let someone drive your Ferrari if you weren't sure they were up to the task? So even if I could lend an expensive VNA or TDR test device, I would not do it before having learned how to operate it properly.

In another post you mentioned that the original NanoVNA may do a better job than the LiteVNA for the cable testing case. Would that be because the NanoVNA has a lower noise floor in its fundamental based range below 300MHz? Because, otherwise the liteVNA seems to be much better spec'ed: fundamentals up to 6GHz, more sampling points, etc.


@tggzzz: Yes, the tube based 1502s seem to be nice tools for TDR applications. There is a common hack to fake the presence of the battery and make them run on wall power. I already have one of those on my watch list, but did not find one in acceptable condition yet. Unfortunately, I do not have any gear qualifying as suitable for high frequencies. My fastest self-repaired vintage scope has only lousy 50 MHz. Once I have the necessary bench space available that will change. I already made some experiments with stacked Schmitt triggers for a pulse generator in the past, but my gear was not up to the task to validate the results.

@pdenisowski: Thank you for your helpful input and the offer to maybe take some hobbyist vs pro VNA measurements. Those would be quite interesting.

[pdenisowski]

Okay, so I measured an approximately 1.5 meter long SMA cable using both a R&S ZNL and my (ancient) NanoVNA.  Both were cal'ed using the NanoVNA cal kit.

As expected, using the measurement procedure I talk about in the video on a 50 ohm cable leads to a "dot" in the middle of the Smith Chart in both cases.  I zoomed in very close on the ZNL and got a "first crossing" impedance with a real part of 50.287 ohms.  Plugging into the equation give a cable impedance of 50.143 ohms. 

I had trouble figuring out how to zoom in on the NanoVNA, either on the device itself or using NanoVNA Saver - mouse scroll wheel doesn't work.  Without sufficient zoom, I can't accurately measure the first crossing of the resistive axis.

I'll look into this some more, just FYI for now.

Edit:  I was a little sloppy with the marker placement, etc. - just comparing procedures at this point

[G0HZU]

Don't forget that the Zo of 'real' coaxial cable will tend to rise (and become increasingly complex) at lower frequencies due to the resistive losses in the copper.

Therefore, I'm not sure how valid the test is at 1.7MHz assuming you want to (accurately) know Zo up at VHF or UHF. There are other ways to measure it using a VNA or an impedance analyser. Also you can explore how much the Zo increases at lower frequencies for a given cable type.

Skinny cables tend to be lossy so the Zo climbs rapidly at lower frequencies. Try repeating your VNA test using a skinny cable like RG174 to see what I mean.

[tggzzz]

@tggzzz: Yes, the tube based 1502s seem to be nice tools for TDR applications. There is a common hack to fake the presence of the battery and make them run on wall power. I already have one of those on my watch list, but did not find one in acceptable condition yet. Unfortunately, I do not have any gear qualifying as suitable for high frequencies. My fastest self-repaired vintage scope has only lousy 50 MHz. Once I have the necessary bench space available that will change. I already made some experiments with stacked Schmitt triggers for a pulse generator in the past, but my gear was not up to the task to validate the results.

I find the battery hacks are unreliable.

If you want to run them from mains without a battery, the simplest way is to have a 12V 1.5A PSU, and use 4mm banana sockets to connect to the battery terminals down in the case. Make sure you get the polarity right! Alternatively replace the transformer with a small 240/110Vac->12Vdc converter, and attach that to the battery terminals.

You can use NiMH cells, but the 1502 is very rough on NiCd cells, and NiMH are less robust and their capacity is often "questionable".

[tggzzz]

I had trouble figuring out how to zoom in on the NanoVNA, ...

Check whether your NanoVNA has a "TDR mode", i.e. an IFFT on the the frequency domain data.

[EE-digger]

Can't contribute much but here is an interesting comparison between an original Tek cable and H155 cable.  The writer claims that the H155 is within tolerance limits (for the SG503 output ?).  (translations with Google)

https://amplifier.cd/Test_Equipment/Tektronix/Tektronix_500/SG503.htm

Also, in either groups.io or the Tek wiki, there is mention of Tek's quest for a true 50 ohm BNC connector and the fact that they had to fabricate their own to get it (at the time).

[Pinörkel]

Can't contribute much but here is an interesting comparison between an original Tek cable and H155 cable.  The writer claims that the H155 is within tolerance limits (for the SG503 output ?).  (translations with Google)

https://amplifier.cd/Test_Equipment/Tektronix/Tektronix_500/SG503.htm

Also, in either groups.io or the Tek wiki, there is mention of Tek's quest for a true 50 ohm BNC connector and the fact that they had to fabricate their own to get it (at the time).

Thank you for the reply. I was already aware of that and even tried to order two of these cables from the ham radio shop mentioned in the associated discussion because they were very cheap and worth a try. However, the cables were delivered with 75Ω BNC connectors instead of 50Ω ones. So, I send them back for replacement, but it turned out that the shop insisted on them being 50Ω connectors and they did not even know the difference. That was when my trust in them dropped to zero. ;-)
There is an interesting analysis on mixing 50Ω and 75Ω plugs and sockets on cables and devices (link). Although it is only available in German it should be translatable with one of the free online pdf translators.