How to check twisted pairs?
Is there any easy way to tell if the signals are correctly paired all the way through? I would think they would have similar problems in large Ethernet installations but I don't see any discussion of how to root out problem cables there.
I can systematically excite each pair and look for coupling to all the other pairs but that seems like a major pain. Hopefully someone knows a shortcut.
I'm not an expert in what you are doing, but could a simpler transmission method reduce the major inconvenience of a miswire? Example, methods which do not need a return wire, radio, fiber optic.
What if you were to apply a waveform to one wire, and look for sympathetic response in another wire? It might be capacitive caused, or induction caused. If the two wires don't belong together then you know you've got a miswire. On the other hand maybe twisted pairs are supposed to minimize interference between them. I may be barking up the wrong tree.
Come to mention it, miswires are easy when attaching ethernet plugs. Seems there ought to be a procedure for technicians to test that each twisted pair is valid, besides an ohmmeter.
The effectiveness of applying waveforms, pulses, etc., depends on what device is at the far end of a pair. Example, whether it is low or high impedance. Or whether crosstalk can occur between pairs, via low impedance path through a device. Etc.
I believe, this can only happen if the technician is color-blind or distracted. Anyway, ethernet or other TP cables can be perfectly checked for correct pair connection with a TDR measurement. But you don't necessarily need a TDR instrument. A lower pair capacitance should already indicate a wrong wire assignment, higher loop inductance would show it clearly.
Let us not forget about the basic philosophy: twisted pairs are expected to pick up noise, etc but identically on both the wires and they are expected to cancel out because of common mode rejection.
It is possible to get shielded twisted pair cables; I have not used them but they are available.
Only four wires are used in normal cases (100MHz) but all the eight wires are used for high speed (1GHz). Normally only two pairs are used.
The way I test it is by shorting one pair at one end and seeing the other end with a multimeter. Mostly the technician makes suitable adapters himself.
The wiring error SherpaDoug is talking about can't be identified with an ohmmeter. But it won't usually happen with a CAT-x cable where all wires have unique color markings. Some twisted pair cables have however one wire in a pair unmarked, wires can be mixed up if you work carelessly.
Best way is to follow the common convention: http://www.incentre.net/tech-support...oding-diagram/
It helps in find faults more easily. If there are joints, there will more problems; best way is to use joint boxes.
There is a limit of 100m on the length of any given segment.
I would think it pretty unlikely that someone terminating a plug at one end, and a socket at the other end would make the identical but opposite identical mistake at both ends that would split up a twisted pair and not be detected during final testing.
What are the odds of terminating say fifty wires in sequence, and transposing wire seventeen with wire thirty five at BOTH ends ?
I do not understand the query clearly; how does the 60Hz and 5kHz come in the picture? Is he user is using the same cable but not for ethernet data?
This is what is meant by split pairs:
Three twisted cable pairs A+B, C+D, E+F
All tests perfectly well with a multimeter between plug and socket.
A goes to A, B to B and so on.......
But pairs two and three will be very seriously compromised by crosstalk, because the guy that fitted the plug and socket to the cable screwed up mightily, and transposed wires D and F at both ends.
Its pretty difficult to do in practice, but its certainly possible, and the results pretty catastrophic for a long data cable.
Under normal conditions, this will not matter simply because the twisted pair D+E does not carry any signal. But the results will be unpredictable if you transpose C and D OR E and F.
Normal ethernet (100 Mbps) uses only four wires and the remaining four wires are not used (standby??)- engineers love to keep spares in hand!!
Oh, I see.
Messing up D and F will be bad - it may not even work for 100m cable.
*update*
How to test it ?
My first approach would be to use an improvised TDR.
Even with a short cable length, a split pair should show up as a major change in cable impedance.
Feed some fast very short square wave pulses down the cable, and look for a reflection at the source. Terminate the far end with a potentiometer and tweak the pot for zero refection.
The reflected pulses will be the same polarity as the transmitted pulse if the termination is too high a resistance. The reflection will be opposite in sign if the terminating resistance is too low a resistance.
At one specific value of terminating resistance the reflection will disappear altogether. It will be a very sharp null. It might be say 115 ohms or something ??
You then build a dummy terminating plug for the far end with as many 115 ohm resistors as there are pairs. Then simply use your TDR at the other end to find any pairs that produce any really high reflections.
It sounds more difficult than it is. Its a very effective way to locate cable opens and shorts as well as split pairs.
Is it possible to set up an improvised TDR using a decent scope and a handful of passive components? If you see the price list of commercial equipments...
Proper TDRs are frighteningly expensive, and just not needed for this.
You can experiment with a 555 timer (or function generator) and a mosfet gate driver chip to drive your cable.
The gate driver will deliver some nice fast edged pulses with nanosecond rise and fall times.
Terminate the cable at both ends, and hook up your oscilloscope directly to the cable at the transmission end.
A dead shorted cable you will obviously see nothing.
An open cable you will see double the amplitude at the transmit end.
No visible reflection at half amplitude means all is well.
The transmitted pulse need not be very short, but it needs to have fast edges.
The reflection may be quite early and overlap the transmitted pulse, and you will still see the reflection on top of the transmitted pulse.
Try it and see, its great fun and requires very few parts.
Most cables are around 100 to 150 ohms, but a 500 ohm potentiometer at the far end will quickly tell you.
Thanks for the help guys (& gals?). I think in my case a loop inductance comparison to other pairs will tell me if there are a couple of suspect pairs. Then I can rig an improvised TDR which should tell me where along the cable the problem is.
FYI this is a tether for a Lamp Ray hull inspection robot. It crawls on the bottom of a ship and uses ultrasound and magnetic pulses to measure the thickness of the steel and paint respectively, to tell if the ship is damaged or if it needs a general dry docking. There is 1000' of main tether and multiple 100' extensions. At that length there is no actual Ethernet but we use an Ethernet extender modem pair for data and other twisted pairs for the 300VDC power bus, power telemetry, a couple of video channels, and some other instrumentation.
We used to use 300W of 220VAC halogen lights fed through a twisted pair with little 60Hz noise in the other signals. Now we have rebuilt the system with LED lights running off the DC power bus and we are seeing MORE 60Hz noise on the other channels. We found the source is 60Hz content in the Ethernet extender modem signal.
These cables are frequently damaged and are then spliced by divers on a pier, not electricians in a shop. If the same diver mistakes the blue wire for the grey wire at each end you can easily get split pair problems.
Again, Thanks for the ideas.
Colour codes are not infallible.
Is that orange, or dirty yellow, or maybe its red ?
Or perhaps its supposed to be brown ?
Another way split pairs can happen if "Fred" terminates the cable and discovers while continuity testing, that two pins are crossed, then proceeds to correct the problem at the wrong end.