Measuring non 50 ohm transmission lines

[fonograph]

Every scope,every spectrum analyser,every VNA are all 50 ohm.But what if I need to measure some kind of microwave electronic circuit or transmission line that use different impedance? In my VNA thread I kind of touched upon this subject and the answer I was given was that VNA work best at 50 ohm,they can work if the impedance is different but it cant be too far away becose the accuracy decreases fast the more we go away from 50 ohm.What about oscilloscopes and spectrum analysers?

I read that cheap scopes that have only 1M ohm inputs can put on some adapter or something to make it into 50 ohm,but I have no idea how ideal it is,I am no Ku band clairvoyant but I smell signal degradation.I read on some website I cant remember that the impedance of transmission lines ranges from 10 - 380 ohm.

I have no idea how truthful or actual that information is but I know for example that Display Port spec is 100 ohm +-20% so it can be 120 ohm and be within spec.Or usb is 90 ohm +-15% too,thats not the 380 ohm I read about,but is still very big difference between these and 50 ohm.How the funk am I supposed to measure these impedances with any kind of accuracy if the mismatch is going to be so large?

[PhilipPeake]

One of the common ones that you find is 75 vs 50 ohms. A dipole is more like 75 ohms, so a lot of consumer co-ax and devices use 75 ohms.

"Professionals" tend to use 50 ohms. Honestly, I think that choice originated in the US, like their choice of 60Hz for electrical power. As close as I can tell in both cases, because it makes the basic math just a bit easier.

Anyway, the simple answer is that you need to match impedances.
There are many ways to do this, depending upon the type of transmission line. If the line under test is pretty well matched, a simple transformer will do. Unfortunately, with RF matching is often far from perfect, so the matching of 50 to 75 becomes more problematic.

If you want to do it the hard way, you recalibrate your instruments to the impedance of the line under test each time.

[ogden]

Appnote regarding subject:

https://cdn.rohde-schwarz.com/pws/dl_downloads/dl_application/application_notes/1ez53/1EZ53_0E.pdf

[swingbyte1]

Hi
These impedances were chosen by the physics of xmission line theory. 50 gives the best power transfer between the two systems, while 75 transfers higher signal amplitude. While these impedances may not always be the best,a standard few allow you to build more economically

Sent from my Nexus 5 using Tapatalk

[ConKbot]

For spec-ans, there can be an option somewhere for impedance correction. You tell it you fed it with a 75 ohm source, and it applies a correction for what the reflection coefficient would be.  It still doesn't present the load with a proper termination, so a termination sensitive load (I.e. a filter) won't be correct.

For termination sensitive loads, use a minimum loss pad, which presents a 50 ohm impedance on one port, 75 ohm (or whatever your incoming impedance is) on the other, and allow for the external attenuation.

This requires the signals all to be single ended, so a 90/100/120/300 ohm differential pair would still need a bal-un, so you might as well do your impedance matching in the bal-un too.

[G0HZU]

Every scope,every spectrum analyser,every VNA are all 50 ohm.But what if I need to measure some kind of microwave electronic circuit or transmission line that use different impedance?

I'd still use a decent lab 50 ohm VNA for this? The VNA should be OK across a few ohms to several thousand ohms so this covers most RF circuits. The whole point of a full n port VNA is that it can produce an n port model of an unknown network. So you can measure non 50R circuits and produce an s parameter model to use on a computer.

With a modern n port VNA (where n can be 2, 3, 4 or expanded to maybe 12 or more ports) you can measure/model transmission lines up to several hundred ohms and filters (eg crystal filters) up to maybe a few thousand ohms with very good results with no need to mess about with impedance correction factors or baluns or minimum loss pads.

Just connect it directly to the VNA and as long as the network is linear and stable you can measure it and create a decent model to use on a computer to predict the termination impedance, insertion loss, etc etc

[coppercone2]

my VNA allows you to select impedances actually, you can adjust the number with the rotary encoder to whatever, and it has a option for a 1Mohm hardware circuit, but it seems that you can adjust the impedance to whatever you like, but I guess this would require standards. I am not sure what it does internally to the adjustment though. E5100A

I can see the appeal of designing a system to whatever impedance you like, but accurate measurements...

Are there any other popular impedances besides 75 (i associate this with television) and 50 ohms (laboratory)? I know audio likes 600 ohms for some reason.

[David Hess]

I read that cheap scopes that have only 1M ohm inputs can put on some adapter or something to make it into 50 ohm,but I have no idea how ideal it is,I am no Ku band clairvoyant but I smell signal degradation.I read on some website I cant remember that the impedance of transmission lines ranges from 10 - 380 ohm.

For oscilloscopes this is called a "feedthrough termination".  The parallel capacitance of the 1 megohm input, typically between 10 and 20 picofarads, appears in parallel with the feedthrough termination but 1 megohm inputs only support lower frequency operation where this is not a significant problem.  Lower bandwidth oscilloscopes which have a switchable 50 ohm termination use an internal feedthrough termination in parallel with the 1 megohm input.  Higher bandwidth oscilloscope switch between a dedicated 50 ohm and 1 megohm input or only have a 50 ohm input.

[coppercone2]

For spec-ans, there can be an option somewhere for impedance correction. You tell it you fed it with a 75 ohm source, and it applies a correction for what the reflection coefficient would be.  It still doesn't present the load with a proper termination, so a termination sensitive load (I.e. a filter) won't be correct.

For termination sensitive loads, use a minimum loss pad, which presents a 50 ohm impedance on one port, 75 ohm (or whatever your incoming impedance is) on the other, and allow for the external attenuation.

This requires the signals all to be single ended, so a 90/100/120/300 ohm differential pair would still need a bal-un, so you might as well do your impedance matching in the bal-un too.

Is this what my E5100A does when you can put whatever impedance you want? Adjust the reflection coefficents? It is called "set Zo" under the Cal button.

So then I can use pad's at the expense of signal power to measure whatever impedance I want?

[Wolfgang]

Well, 50Ohms are the most common standard, but (for the cable TV guys) a number of T&M equipment for 75Ohms exists.

This includes spectrum analyzers, VNAs and scopes (Keysight, RIGOL, ...). The lower frequency ones have switchable inputs, the higher (GHz) ones need to be ordered with the correct impedance, or you could use a "minimum loss adapter" and calibrate it out.

Scopes with BNC can be feedthru input terminated with 75Ohm loads.

75Ohm cables are also very common, like 50Ohms.

I personally have never seen any RF impedance used in T&M other than 50 or 75Ohms. If you insist on working with a different impedance (say, 100Ohms) make a minimum loss pad at the input and output of your DUT and calibrate it out.

[coppercone2]

does anyone have a full procedure for using pads to measure something, like a youtube video, so I can be sure of the method ?

anyone wanna make one?

[G0HZU]

If you have a full 2 port VNA the best (as in most informative) method would be to just connect the (non 50R) unbalanced transmission line directly to the 50R VNA and export a 2 port s parameter model to a computer and then post process the data to get characteristic Zo and pF/m and L/m and insertion loss and group delay etc etc.

With a 4 port 50R VNA you can do this with balanced (non 50R) transmission lines and export a 4 port model and so you don't need any baluns or matching pads.

[coppercone2]

Still would be nice if someone made a video where all the software was explained, examples were given, to make sure we are doing it right

if someone is bored

[G0HZU]

I don't have a microphone here and I only have an ancient version of Camstudio (plus I'm a hoplessley poor presenter) but I have uploaded an old 'post process' analysis of a 59cm section of 300R twinline if that is of interest? This is the classic 300R twinline ribbon cable used for old TVs and VHF radios. It was measured on my 4 port 50R VNA and the model from the VNA was exported to an RF simulator for post processing in the video. There's no sound because I have no microphone so just follow the mouse pointer.

When measured with a decent 4 port VNA you can get a good idea of the various parameters of the transmission line. In the video the Zo is predicted from the (low frequency) open circuit capacitance of 7.51pF and the short circuit inductance of 724nH.
Zo = sqrt(L/C)   = sqrt(724e-9 / 7.51e-12) = approx. 310 ohm but I used 309R in the video.

The Zo of 309 R is then used to set the Zo of the test setup and the insertion loss is then predicted when in  a 309 ohm environment on the simulator. You can see that 309 ohm is the best Zo for lowest loss when I wiggle the Zo either side of 309R. So I think this is a fairly good way of testing a non 50R tline with a 50R VNA.

You can also do some of this with E and H field probes and a spectrum analyser and tracking gen and this was how I used to measure the Zo of (balanced) twisted pair (enamelled copper wire) transmission lines for use in HF/VHF transformer design. The idea would be to optimise the twists/cm to get 50R or 75R as required. It's an old E/H field trick I developed before I had a VNA here at home and the tline under test remains 'balanced' because it is tested in mid air by the E and H field probes. So there is no ground connection to cause unbalance. But the E and H field probe method can only indicate pF/m and L/m and Zo and velocity factor. But it usually works very well.

The video below is a bit basic and nerdy but it does show how powerful a decent VNA can be. However, I don't think you would get data this good from a budget VNA. I also think I'm pushing the limits of my old E5071 VNA here. Note that I use various low loss 'non 50R' transmission lines to test and optimise my VNA cal kits. By designing lines with differing Zo it's possible to really test the accuracy of the correction factors in a typical VNA cal kit. But I used a commercial (Agilent) Ecal module and a custom fixture calibration data file for the initial VNA calibration in the video. This gives really good results as you can see in the video when compared against the basic tline model that sits below the VNA data file of the real 300R twinline. This is based on a library model of a transmission line with Zo of 309 ohm.

https://youtu.be/aAtVTIVmrBc

[coppercone2]

lol 'old' vna 

Good video but now if you consider that vna 'old' I want to see is done on 'prehistoric' 

It's good to see the graphs and stuff to get an idea if your measurement is completely insane due to some kind of mistake, I have had people looking at me like my VNA was broken because I was not sure what I was doing when I sent them pictures.

I noticed sudden increases in correct measurement difficulty with first the electrometer, microwave antenna and then the VNA, which I still am working on understanding. Those lab devices seem to blow everything out of the water when it comes to making hard measurements. I think the first hurdle people tend to run into is using the high Z mode on multimeters, then oscilloscope loading/capacitance/iamgoingcrazyneedexpensiveprobes, then cleanness issues from Electrometer, calibration/setup issues from VNA and general antenna black magic problems, I mean as far as making commonly useful lab measurements go, I don't want to downplay metrology, as any measurement when precise is difficult, just I mean in the general order a person in electronics tends to go through the arsenal of measurement that is useful for common circuits that approach high performance but are not necessarily cutting edge. When I went from doing some measurements on power supplies, looking at some digital signals with a oscilloscope to trying to measure photodetector circuits and hybrid amplifiers it was just like WTF IS GOING ON HERE??, even if those circuits were basic by their cost and design specifications.

[G0HZU]

I think my E5071 VNA is about 15 years old now so it's no youngster

I've got various low loss (non 50R) unbalanced tlines here made with polished silver that I use to check my various mechanical VNA cal kits and this can show up even tiny irregularities in the corrections in the cal kit file. When I first got my E5071 VNA I spent the first few weeks optimising my various test fixtures and mechanical cal kits using low loss non 50R lines.

[TheUnnamedNewbie]

Hi
These impedances were chosen by the physics of xmission line theory. 50 gives the best power transfer between the two systems, while 75 transfers higher signal amplitude. While these impedances may not always be the best,a standard few allow you to build more economically

This is not really true. The main reason I hear cited for 75 ohm in TV is twofold: it matches dipoles, and you can show with some math that a coax cable with dielectric that has a relative permittivity of 1, the lowest loss is around 77 ohm. Since TV applications mainly care about loss (long distance coaxial runs) they supposedly went with 75 ohm for this reason.

The highest power handling capability with an air dielectric is around 30 ohm. 50 Ohm was thus chosen as a kind of compromise, between low loss and power handling. (In addition, the losses get more and more challenging with lower impedance, at least if your main source of loss is metal loss)

[swingbyte1]

Ok, I was being concise because I hate typing on a phone. There is nothing special about 50 ohms or 75 ohms other than it was chosen a long time ago by the then available technologies for the applications of the day. Characteristic impedance is determined by the geometry and dielectric properties of the transmission line - coax, twin ribbon, strip-line etc. In the early days commercial intrests standardised on a few such impedances that were convenient -50 75, 300. Heres an article with a bit of the history https://www.belden.com/blog/broadcast/50-ohms-the-forgotten-impedance and another and a bit more https://www.microwaves101.com/encyclopedias/why-fifty-ohms.
If one is really into this consult Kraus

[Wolfgang]

Hi,

you probably know that they still sell it today. I have an E5071C and I like it a lot.

[G0HZU]

Yes, we have several of the ENA5071B/C VNAs at work in the ATE/Production area and they are pretty decent in terms of performance. I think our C versions are fairly new and they are a nice VNA to use

Mine is the older B version and it doesn't have any internal bias tees fitted and this is a bit of a pain sometimes. I'm not sure if it's possible to retro fit them but I think one of our C versions at work has the bias tee option fitted. So here at home I have various external bias tees that I use.

Obviously, when measuring the s parameters of a transistor the VNA has to be able to measure these non 50R (active) devices and I usually get excellent results when doing this with a VNA either here at home or at work. The accuracy obviously suffers up into the microwave region but I'm not sure there is a viable alternative...

[Wolfgang]

... you could use impedance transformers with known characteristics if your measurment impedances are far off 50Ohms, like power amplifier inputs, crystal filters, or the like.

Keysight also has test fixtures for more exotice devices that could be deembedded.

[bson]

It's pretty easy to pad the source impedance to anything, 50 to 100 by adding 50, or to 1050 by just adding a 1k resistor in a suitable arrangement and calibrating.  At least up to where physical dimensions approach quarter wavelength.  From there it should be easy to recalculate S parameters for an arbitrary nearby source/reference impedance.  For example, S11=Vrefl/Vout=(ZL-Z0)/(ZL+Z0); and with Z0 known... Calculate ZL=Z0*(1-S11)/(1+S11), then recalculate S11 for any other Z0.  Call it Z0alt (with some assistance from Mathematica):

\$  S_{11alt} = \frac{(S_{11}-1) Z_0+(S_{11}+1) Z_{0alt}}{(S_{11}-1) Z_0-(S_{11}+1) Z_{0alt}} \$

I'm pretty sure if the impedances are ballpark this can be repeated for other measurements.

[G0HZU]

Just to repeat again, with a decent 2 port 50R VNA (eg an old used HP/Agilent or R&S VNA) you should be able to just measure the device directly via the 50R ports (after a full 2 port 50R calibration) if it's something fairly mundane like a crystal filter that normally requires a termination of 1k impedance in parallel with a few pF. You can just measure it direct in 50R and then export the data to a computer as a 50R s2p model. Then work out the correct terminating impedance via the RF simulator and design the matching network on the simulator based on the s2p data model from the VNA. The simulator will then be able to tell you the terminating impedance, the passband ripple, insertion loss, return loss and passband group delay with very good results. The bonus is that you have a decent 2 port model you can use for incorporation into a larger system model on the RF simulator.

That's my experience anyway and I've done this many times in the past with very good results. There's usually no need for transformers or minimum loss pads or 1k ohm padding resistors.