3
$\begingroup$

Thought I'd try doing some actual experiments with twisted pair feedlines and also just some general "non-coax" measurements and testing, but not until after I understood a bit more about using my VNA. Turns out that was probably a good idea as even the basic initial measurements are not what I had expected!

I haven't tracked down any official standards (maybe this is partly where I went wrong?) but "the internet" says that the twisted pairs of network cable have a characteristic impedance of 100Ω. But when I look down a pair of wires with my NanoVNA in TDR mode, the |Z| it says it sees through the length of wire is pretty much right around 50Ω instead:

NanoVNA connected via an SMA jumper, BNC adapter, and banana plug adapter to a coiled-up run of Ethernet cable, the loose end of the cable spread apart

The setup here is obviously not super precise, I'm just using a BNC→banana plug adapter to clamp onto both wires of a twisted pair. The NanoVNA is set to sweep from the bottom of it's range up to 750MHz iirc seemed to match the wire, and I've got it showing |Z| with the "transform" option on.

I don't thing the results are completely bogus, as when I change the end of the pair from an open to a short the initial disturbance of the banana plug remains the same while the disturbance corresponding to the end of the line "flips" as it were:

VNA showing roughly the same graph except for the loose end of the cable is clamped together and the righthand glitch is now upside down

But the reading along the whole line shows ~50 ohms instead of the 100 ohms I was expecting to see. Am I doing something wrong, or is the expected result? Is there a way to see the actual characteristic impedance along a transmission line with a VNA, or does it basically just "see" its own (i.e. the VNA's) characteristic impedance in the absense of a disturbance?

$\endgroup$

2 Answers 2

5
$\begingroup$

Your second guess is correct - it's displaying its own system impedance, not the line impedance.

A VNA's fundamental units of reflection are: 1) Magnitude of the reflection, and 2) Phase of the reflection. It measures these at many frequencies. It doesn't measure Z or L or R.

Everything else you see on the screen is transformed from these numbers, using the system impedance, 50 ohms. The impedance (seen at the point to which you've calibrated) is $Z=Z_0{{1+\rho}\over{1-\rho}}$. Then you can ask for the magnitude of the impedance $|Z|$ in ohms.

By applying an inverse fourier transform (if you've set up the frequency steps correctly), the complex reflection @ frequency can be transformed into complex reflection @ time. This can also be plotted as |Z| over time, or R+L over time, or dB over time, etc.

Now consider the long transmission line which is as you say about 100-110 ohms. There will be a reflection from the initial impedance step, and you see this on the screen. But the line itself doesn't reflect anything, until you reach the open or short at the end. So the reflection coefficient $\rho$ at that time will be zero, and $50{{1+0}\over{1-0}} = 50 \Omega$. You can see that any transmission line will give you that, and indeed even after the last reflection you still see "50 ohms" again.


To accurately measure the $Z_0$ of an unknown line, I recommend a modified $\lambda/8$ procedure.

Note that for a lossless short-circuit line,
$ X_{SC}=Z_{0}\tan({{2\pi\ell }\over{\lambda}})$
And if $\ell=\lambda/8$ then $tan({\pi\over4})=1$, so
$ X_{\mathrm {in} }=Z_{0} $ i.e. the imaginary impedance is equal to the characteristic impedance of the line, not the VNA!

But how do you know where the line is exactly $\lambda/8$ long?

The open-circuit impedance is almost the same:
$ X_{OC}=-Z_{0}\cot({{2\pi\ell }\over{\lambda}})$
where $cot({\pi\over4})=1$ too.

My trick, I'm sure not my own invention, is to do this:

  1. Calibrate the VNA over a wide frequency range, perhaps from 0 to quarter-wave.
  2. Display the Smith Chart, R+jX
  3. Short circuit the far end of the unknown line
  4. Place a marker at the top of the smith chart, where X = your best guess for Z0.
    For example, marker at 1+j100 ohms.
  5. Open circuit the line, and look at the marker value for -X, now at the bottom.
    It won't be quite the same value, eg. 2-j120 ohms.
  6. Adjust the position of the marker to be in the middle, say 2-j110 ohms
  7. Short circuit the line again
  8. and repeat until the value of X and -X doesn't change.
    Now you've found the $\lambda/8$ point, and also the $Z_0$.

Finally, watch out when measuring a balanced line with an unbalanced feed - the whole line becomes an antenna too. The effect will be more pronounced at $\lambda/4$ when it's a resonant monopole, less at 1/8. And with a battery-powered VNA maybe even less of an issue. Some ferrites on the coax wouldn't hurt... as usual just try laying your hands on the circuit to see what changes.

$\endgroup$
1
$\begingroup$

You've got a VNA - so use it as such!

You're measuring the impedance of an open end as seen through x meters of cable, back and forth (plus the SMA-to-BNC-to-Banana-to-twisted pair adapter), which in S-parameters would be called an S11 measurement. That tells you how much gets reflected - but never how much of what gets not reflected gets lost where!

Do a S12 measurement instead. Consciously limit your upper frequency end: your adapter is only electronically negligible for wavelength such that the electrical length of it is much less than a wavelength; if go with a rule of thumb it maybe $\lambda/15$, and say that whole thing is 20 cm long at a guessed velocity factor of 2/3, so, it stops being "no problem" at ca 67 MHz. If you don't have two adapters, or want to bring your frequency range up to maybe 400 MHz, try soldering on SMA connectors, or, and don't tell anyone, carefully remove the insulation put one wire in the center of your SMA Plug, and wrap the other around the thread. Fix with tape. (This should be pretty ok in terms of working, but it's going to be too fiddly to yield the same results twice.)

Of course, it would be better to design a board with two rj45 plugs and two SMA connectors; that would actually allow for calibration with "as close as it gets" cat5 shorts and opens, moving the measurement plane to the front of the rj45, instead of incorporating a fickle mismatch depending on the geometry of how exactly the adapter cable or the coax lies. Such boards aren't hard; dc4hp has one, for example, that can convert an Ethernet patch cable to a loop antenna, here.

Prepare multiple lengths of twisted pair, and measure them.

$\endgroup$
4
  • $\begingroup$ Hi; does this really answer the question? $\endgroup$ Jan 30 at 14:19
  • $\begingroup$ I think it almost answers the question. Explains how to get a S12 measurement. Now if it explained what to expect and how to use the S12 measurement... $\endgroup$
    – user10489
    Feb 3 at 7:25
  • $\begingroup$ as the OP this answer does seem useful and does somewhat answer my initial question ("an S11 measurement [only] tells you how much gets reflected […]") but then afaict seems to maybe conflate line loss ("[…] but never how much of what gets not reflected gets lost where") with the characteristic impedance I was more curious about in this Q&A. Thanks for the ideas and tips though, as I do hope to continue other experiments with twisted pair and even just plain random runs of wire this way to push my own understanding further. $\endgroup$ Feb 3 at 20:26
  • $\begingroup$ I concur, I didn't quite hit the topic! Thank you for the kind words $\endgroup$
    – sina bala
    Feb 4 at 22:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .