TDR Fault Location using a 20Mhz Analogue Scope?

[kolbep]

So, Busy running around at 00:15am trying to locate a fault in a 70M long 35 x 4 core SWA Cable (at a local tourist spot).
It was actually 2 of the 35 x 4 Cables Paralelled, split them so that I could find the fault was in the Red Phase Conductor, at some point it was down to earth. Isolated the Faulty Cable, and then tried to look for any visible damages.
To make it worse, most of the cable is in the ground, under concrete, so that ain't gonna happen.

Eventually I decided (Remember it was just past midnight), that I would just cut the cable at a convenient point, and then megger each side to locate the fault.

Luckily the fault was in the short piece (15m), so I managed to replace that piece, and get them going again.

But anyway, a TDR would have probably saved a lot of time and frustration (and if the cable has to be divided in several places to try localize the fault, it will then have a lot of joints).
And I have had to do this several times over the years.

But TDR's are pricey buggers, especially for us South Africans.....

So I saw this project  :
http://www.epanorama.net/circuits/tdr.html
That can do pulses between 10nS and 5uS.

Now this type of circuit relies on an oscilloscope. I have a 20mHz Analogue Scope (Beckmann Indutries Scope)
Would this oscilloscope be fast enough to work with this type of circuit, to locate faults in cables from 10m up to 100m long,
or is the scope too slow.

Thanks

[singapol]

From the epanorama article :

As you can see I have built my circuit to a piece of verboard (I did not had interrest todesign a special circuit board for this). The components on the circuit board are wired directly from one location to another using as short possible leads as possible. The timing capacitors are all soldered on the back of the range selection switch. The length of the connection wires is quite much as short as possible, and the signal wires that come with ground are wrapped together with their ground wires. The design does not look nice, but it works well. This TDR circuit uses very fast signals (over 50 MHz fundametal frequency and harmonics to well over 100 MHz), so you need to be careful how you place the component so that the the wires on sensitive circuit locations are not too long. Veroboard is not designed for high frequency circuits, but with careful circuit layout you can use Veroboard successfully even on high frequency circuits.

Looking at the TDR signals

You need a fast oscilloscope to get nice measurement results from the TDR measurements. Generally the faster the better. I have made some basic measurements with 20 MHz analogue oscilloscope and 20 megasamples per second digital samplign oscilloscope (osziFox). With such equipment, the measurement accuracy is limited to few meters at the best. You can get information if something is terribly wrong and where is the problem. And you can measure length of long cables (tens of meters with accuracy of few meters). But using such oscilloscpe will not show you all the fine details of the wiring.

If you have access to oscilloscope with 100 MHz or 1 GHz sample rate, you can get much more accurate measurements.

[tggzzz]

Start by thinking about what's happening physically: a voltage step is injected, it propagates and is attenuated, a proportion is reflected, and the reflection is measured.

The maximum distance is determined by the input voltage, the reflection coefficient, the receiver sensitivity, and also the cable attenuation (and the propagation velocity, of course). The longer the range, the longer time it will take for the reflection to be received. Hence the 10ns-10us (not nS to uS, since that's Siemens, not seconds) determines only the maximum distance that might be observed.

The distance resolution is limited by the edge rate and/or the receiver bandwidth. In this case the 20MHz will be the limiting factor. 20MHz => 17ns => 3.5m (propagation velocity = 0.66c).

[David Hess]

...

The distance resolution is limited by the edge rate and/or the receiver bandwidth. In this case the 20MHz will be the limiting factor. 20MHz => 17ns => 3.5m (propagation velocity = 0.66c).

I have done this very thing using an oscilloscope and pulse generator.

In practice an oscilloscope can accurately see a delay to at least 1/10th of its transition time specification and some can do better than this depending on CRT sharpness and other factors.  That gets you to 0.35 meters with 20 MHz of bandwidth assuming that the horizontal timebase is accurate.

[rstofer]

This can be built for a few dollars and it is worth playing with even if your scope doesn't have particularly high bandwidth.

http://www.epanorama.net/circuits/tdr.html

[tggzzz]

...

The distance resolution is limited by the edge rate and/or the receiver bandwidth. In this case the 20MHz will be the limiting factor. 20MHz => 17ns => 3.5m (propagation velocity = 0.66c).

I have done this very thing using an oscilloscope and pulse generator.

In practice an oscilloscope can accurately see a delay to at least 1/10th of its transition time specification and some can do better than this depending on CRT sharpness and other factors.  That gets you to 0.35 meters with 20 MHz of bandwidth assuming that the horizontal timebase is accurate.

That's valid with "ordinary" 2-channel analogue scopes, but only partially true with TDRs.

You are often interested in distinguishing between and measuring multiple discontinuities that are close together. Examples: a cable with two kinks, or where a short section is waterlogged, or where a fault has been removed and spliced, or even just a connector. The transition time (not 10% of it) is the limitation in this case.

Even if there is only one discontinuity, its range is dependent on the propagation velocity, and is often not known accurately. Examples: the coax cable installed two decades ago, and then spliced with a different cable. In this case your best hope is to measure the end reflection and guess. You imply that with your comment "assuming that the horizontal timebase is accurate".

[kolbep]

Thank for the discussion on this.
I will (when i get time), knock one together, and try out out....
Cannot hurt anything, and will gain more Xperience Points anyway....
Could be fun (and a learning excercise)

P

[David Hess]

...

The distance resolution is limited by the edge rate and/or the receiver bandwidth. In this case the 20MHz will be the limiting factor. 20MHz => 17ns => 3.5m (propagation velocity = 0.66c).

I have done this very thing using an oscilloscope and pulse generator.

In practice an oscilloscope can accurately see a delay to at least 1/10th of its transition time specification and some can do better than this depending on CRT sharpness and other factors.  That gets you to 0.35 meters with 20 MHz of bandwidth assuming that the horizontal timebase is accurate.

That's valid with "ordinary" 2-channel analogue scopes, but only partially true with TDRs.

You are often interested in distinguishing between and measuring multiple discontinuities that are close together. Examples: a cable with two kinks, or where a short section is waterlogged, or where a fault has been removed and spliced, or even just a connector. The transition time (not 10% of it) is the limitation in this case.

I do not completely disagree but finding the approximate location of the faults even if one cannot distinguish between them is still feasible with lower bandwidth.  My 14 GHz sampling oscilloscope can see the various imperfections in a BNC connection but my fastest oscilloscope which still has a high impedance input can still tell its approximate location with its 300 MHz bandwidth.

Even if there is only one discontinuity, its range is dependent on the propagation velocity, and is often not known accurately. Examples: the coax cable installed two decades ago, and then spliced with a different cable. In this case your best hope is to measure the end reflection and guess. You imply that with your comment "assuming that the horizontal timebase is accurate".

This applies in the case of a TDR or other high bandwidth instrument as well though.  Horizontal timebase accuracy is typically about 2% but can be calibrated on the spot against a reference like the transmission line being used if necessary.  The old Tektronix TDRs that I am mildly familiar with have an adjustment just to do this.

[C]

A TDR using a fast pulse assumes that the pulse will travel to the other end of cable and back. For a lot of cables this is not valid. A telephone company cable is an example where this will fail after a short distance.

For low frequency cables, I think there is a method that uses directional couplers.
From memory the transmitter uses a directional coupler to put a signal on the wire and s matching directional coupler picks up the returning echo.
 

[tggzzz]

...

The distance resolution is limited by the edge rate and/or the receiver bandwidth. In this case the 20MHz will be the limiting factor. 20MHz => 17ns => 3.5m (propagation velocity = 0.66c).

I have done this very thing using an oscilloscope and pulse generator.

In practice an oscilloscope can accurately see a delay to at least 1/10th of its transition time specification and some can do better than this depending on CRT sharpness and other factors.  That gets you to 0.35 meters with 20 MHz of bandwidth assuming that the horizontal timebase is accurate.

That's valid with "ordinary" 2-channel analogue scopes, but only partially true with TDRs.

You are often interested in distinguishing between and measuring multiple discontinuities that are close together. Examples: a cable with two kinks, or where a short section is waterlogged, or where a fault has been removed and spliced, or even just a connector. The transition time (not 10% of it) is the limitation in this case.

I do not completely disagree but finding the approximate location of the faults even if one cannot distinguish between them is still feasible with lower bandwidth.  My 14 GHz sampling oscilloscope can see the various imperfections in a BNC connection but my fastest oscilloscope which still has a high impedance input can still tell its approximate location with its 300 MHz bandwidth.

And then you have to consider where the fault is near a connector. That's not uncommon, if you think in terms of which bits of the cable are likely to be most handled and have most stress applied.

Even if there is only one discontinuity, its range is dependent on the propagation velocity, and is often not known accurately. Examples: the coax cable installed two decades ago, and then spliced with a different cable. In this case your best hope is to measure the end reflection and guess. You imply that with your comment "assuming that the horizontal timebase is accurate".

This applies in the case of a TDR or other high bandwidth instrument as well though.  Horizontal timebase accuracy is typically about 2% but can be calibrated on the spot against a reference like the transmission line being used if necessary.  The old Tektronix TDRs that I am mildly familiar with have an adjustment just to do this.

Curiously I just picked up two TDS 1520 cheap, sight unseen, in the hope I might be able to make one frankenmachine. One turns out to have a broken case at the rear and leaking batteries. I'm charging the other's batteries externally, in the vague hope that I might be able to avoid faking a battery. I'll be very pleasantly surprised if I don't have to do any recapping!

[tggzzz]

A TDR using a fast pulse assumes that the pulse will travel to the other end of cable and back. For a lot of cables this is not valid. A telephone company cable is an example where this will fail after a short distance.

I don't understand that. Can you elaborate?

For low frequency cables, I think there is a method that uses directional couplers.
From memory the transmitter uses a directional coupler to put a signal on the wire and s matching directional coupler picks up the returning echo.

If you are thinking of the "CW burst" type testers, then even in the 70s they were working at hundreds of MHz.

[C]

A TDR using a fast pulse assumes that the pulse will travel to the other end of cable and back. For a lot of cables this is not valid. A telephone company cable is an example where this will fail after a short distance.

I don't understand that. Can you elaborate?

Think of what a fast pulse is. It is a very high frequency with many even higher frequency harmonics.

What is the frequency response of a very long twisted pair like the phone system uses( very low).
When you remove the harmonics of the pulse you have a sign wave. When the sign wave that remains is high for the cable it gets attenuated also. In simple terms the cable eats the pulse and you see no reflection. No reflection = no TDR like response.

For low frequency cables, I think there is a method that uses directional couplers.
From memory the transmitter uses a directional coupler to put a signal on the wire and s matching directional coupler picks up the returning echo.

If you are thinking of the "CW burst" type testers, then even in the 70s they were working at hundreds of MHz.
[/quote]
Sorry, can not remember much about the system other then it used low frequencies and matched directional couplers to put the test signal on the test line and receive the echo. Keep thinking of an SWR meter but connected to a scope, but it was not this simple.
This existed in the 70's

[tggzzz]

A TDR using a fast pulse assumes that the pulse will travel to the other end of cable and back. For a lot of cables this is not valid. A telephone company cable is an example where this will fail after a short distance.

I don't understand that. Can you elaborate?

Think of what a fast pulse is. It is a very high frequency with many even higher frequency harmonics.

What is the frequency response of a very long twisted pair like the phone system uses( very low).
When you remove the harmonics of the pulse you have a sign wave. When the sign wave that remains is high for the cable it gets attenuated also. In simple terms the cable eats the pulse and you see no reflection. No reflection = no TDR like response.

No wonder I didn't understand it; it is wrong on several levels.

Start by realising that (a) phone companies use cables other than twisted pairs, (b) the twisted pairs don't have a very low frequency response, (c) they do use TDRs on cables. And finally replace adjectives such as "high" with numbers (Does "high" = 3.4kHz, 2MHz, 150MHz, all of which have been used by the phone companies)

[vk6zgo]

We had a TU-5 pulser as an accessory to our Tektronix 545B at one job.
The PMG Dept had a writeup on using this for Time Domain Reflectometry.

Playing around with this using various lengths of cable around the TV site,it proved to have quite good accuracy,using the 545B  which was rated as a 35MHz bandwidth Oscilloscope.
We never tried it with anything lower in bandwidth,but I believe it would still be useful with a 20MHz CRO.

PMG/Telecom Aust used "Pulse Echo Testers" (PETs) on landlines from early in the 1960s.
These were a standalone unit including both pulse generator & display.

The "sinewave" testing method referred to by C may be similar to one fairly widely used by Radio Amateurs to determine the electrical length,& by using a correction factor,the physical length of a cable.

Hams normally don't separate the outgoing & reflected waves ,& look at the resultant interference pattern on an Oscilloscope.

The far end of the cable is either left open or is short circuited.
A signal generator is fed to a coaxial tee,one leg of which is connected to the 'scope input,& the other to the cable to be tested.
The generator frequency is adjusted until the display reaches either a minimum  or maximum amplitude.

If the far end is shorted & the 'scope shows a minimum,the cable is  an electrical half wavelength or any multiple of  a half wavelength.

If,with the same conditions at the far end,a maximum is seen,the cable is a quarter wavelength,or an odd multiple of that length.
The opposite situation applies for an open circuit at the far end.

By adjusting the generator,the lowest frequency which shows the "quarter wavelength" result may be found-- twice that frequency will give the lowest "half wavelength" result,

Techs testing a line will usually already know if the fault is an open or a short,& only need to see a peak in the reflected signal,so directional couplers may be used,making the test setup more convenient.

[tggzzz]

I had wondered if finding the resonant frequencies was the technique C was thinking of. I decided against that guess since, as I'm sure you are aware, TDRs are used for far more than simply the electrical length. In particular, they are used to see imperfections in the cable, i.e. small changes in the reflection coefficient.

That's the kind of technique behind the "CW burst" techniques I mentioned. The "burst" part is absolutely necessary to allow the imperfections in the connections to the cable to be ignored. I know it is necessary since an engineer on the next bench to me made such a product in the late 70s
http://www.sciencemuseum.org.uk/online_science/explore_our_collections/objects/index/smxg-8057026

[C]

So, Busy running around at 00:15am trying to locate a fault in a 70M long 35 x 4 core SWA Cable (at a local tourist spot).

Some pictures


specifications
https://www.armouredcable.net/35mm-4-core-armoured-cable.html
Current Rating    154A

If above is correct,
Simple to say NOT a RF cable.
It may or may not match the specifications of some local loop telephone cable when using a TDR.
http://www.generalcable.com/getmedia/b9ccd73f-391d-4ef5-b6da-e2f0f29df0e5/G4283

If you look at the extremes and what effect they could have you have a better chance of getting something to work.

It is a good idea to know how a circuit work.
Think of why some TDR's use a fast level change.
As stated in previous post a fast pulse will be nothing at a distance. Think of a 1 Hz square wave with very fast rise and fall times. That 1 Hz component forces a change down a very long cable. It shows on oscilloscope. In both cases what is lost is the fast rise and fall times.
With out the fast change you have more length of the cable creating a reflection. 
Think of that C-Rual cable at very long lengths, say 10 miles (15.7Km).
You have two conductors separated by an insulator, a Capacitor. But not a simple capacitor, an infinite number of capacitors separated by a small resistance.
Some test equipment used this fact with a very slow charge rate and a meter.
 It takes time for that last cap start to change. A short or open will effect the source end after some time. A change in the cable will also effect the meter.

Now think of that signal that is causing all the reflections on the oscilloscope. Due to the cable it becomes an Unknown signal. You can use the results but have no idea of what created the results. The signal changes with distance and reflections created.

Now look at the test setup.
A signal generator connected by cable A to a Tee which is connected to a oscilloscope.
Cable B from the Tee to the cable under test.
The oscilloscope is showing the sum of voltage at the Tee vs time.
The change seen on the oscilloscope is limited by the complete system, the slowest response wins.

To test the above cable you need more then short or open. This is a huge range. If you are looking for future problems you need even more detail that is hidden in the sum voltage at the Tee connector.

So with the above in mind, how do you get better results?
The easy way would be not to test A cable, but test TWO cables. Using the difference between the two cables to highlight the problems in one.
A small change would be using one conductor to test a second conductor.
If you think it through, the easy way would be a positive transition that matched a negative transition with the oscilloscope in A+B. For two cables you could use two positive outputs. With two conductors you would get the most detail with one being inverted with respect of the other.

Cable B has many functions. It serves as a time delay such that an amplifier has time to come out of overload for large signal from signal generator . If you attach a measured length of the cable under test to Cable B before the actual cable under test you get a calibrated time reference if you can see the changes at the connections.

When the fast change wave to create reflections to read becomes a problem, a known wave to create reflections could still function.   

A TDR is trying to maximize reflections over a short length of the cable while keeping other reflections to a minimum.
Said a different way, a TDR is trying to identify reflections from a short length of cable while removing all other reflections.