Cable Fault Locating: Theory and Practice

Introduction
Electric utilities often face the problem of finding the exact location of a failure in a
distribution cable. These failures often occur at the worst possible time and cause the
maximum amount of inconvenience to utility customers. The utility must quickly find and
isolate the failure to restore electric service. Available fault locating methods use fault
indicators, "thumpers," radars, acoustic detectors or combinations of these. Continuing
research has indicated that subjecting aged cable systems to excessive surges reduces their
remaining life. In fact, anecdotal reports are often heard about aged cables that are thumped
excessively being strong candidates for a failure within the next year. To answer this need for
gentler fault locating methods, the industry has developed more sophisticated methods that
reduce the stress on aged insulation systems. The general approach is to reduce the amount of
thumping necessary to locate a fault while simultaneously reducing the voltages required to
perform the task.
The radar methods used in new instruments are related to TDR and to transient methods
developed in the 1970s. To make the instruments more cost-effective, multiple techniques are
being incorporated in a single instrument. In the following discussion many of the variations
on radar will be discussed. Radar methods dramatically reduce the stress on the cable by
rapidly localizing the fault position. The added benefit is quicker location of the fault.
Once the approximate location of the fault is found using a radar technique, the operator must
pinpoint the exact location of the fault. The most widely practiced technique is to listen for
the small explosion at the fault site when a high energy is discharged through the fault; i.e.,
"thumping" Methods range from listening with a traffic cone to using sophisticated
instruments that incorporate geophones. Not only have radar techniques have been advancing
rapidly but so have the pinpointing methods. Some new technology used for pinpointing
faults will also be discussed.
Fault Process
To understand how faults are located, it is necessary to discuss the properties of the
breakdown process. All faults can be represented electrically by a gap shunted by a resistance
[2]. Although the electrical circuit is simple, variations in the conditions of the two paths can
cover an extremely wide range, the resistance ranging from a dead short to many megohms,
and the gap breakdown voltage varying from zero to many thousands of volts.
While modeling the fault itself is important, so is modeling the faulted cable. When a cable
fails, portions of the cable are destroyed. Depending on the cable type and the available fault
current in the circuit, the damage at the fault site can vary anywhere from a small pinhole to
damage that destroys both the center conductor and neutrals. Traditional classification of
faults as high or low resistance only describes the fault partially. Successful application of
fault locating techniques requires a complete definition that describes the state of the
conductors of the cable as well as the impedance of the fault. A classification scheme that
extends the high- and low-resistance definition to include the state of the cable is:
High-Resistance Fault - High resistance describes the case where the impedance of the fault
(conductor to ground or to another conductor) is greater than 200 O . When a fault is high
resistance, the cable will support some high-voltage application before it breaks down. The
cable may only hold voltage for a fraction of a microsecond but this will be long enough to
generate a surge transient of the type that can be useful in fault locating.
Low-Resistance Fault - Low resistance describes the case where the impedance of the fault
(conductor to ground or to another conductor) is less than 200 O . When a fault is low
resistance the cable will usually not support any high voltage. Any applied voltage is
immediately returned to ground and surge transients are not generated.
Shunt Fault - Shunt fault describes the case where the fault path is from the high-voltage
conductor to ground, hence the designation as a shunt fault. Often both the center conductor
and neutrals are still intact.
Series Fault - Series fault describes the case where the center conductor, neutrals or both are
burned open.
These designations are used in combination to describe the faulted cable. For example, one of
the most common faults is the high-resistance shunt fault. This designation describes a
faulted cable in which the conductors are still intact and the impedance of the fault is greater
than 200 O .
When voltage is applied to a faulted cable, the damaged portion of the cable begins to
conduct. The conduction process is random and depends on the geometry of the faulted
portion of the cable as well as the properties of the materials present at the fault site. In either
the shunt fault normal or series fault case the conduction process does not begin immediately;
there is a statistical time lag. This time lag can extend from tens of nanoseconds to many
microseconds and depends heavily on the over voltage applied to the equivalent gap
modeling the fault. If the instrument uses the time between voltage application and the time at
which the breakdown transient is first observed to determine the distance to the fault,
enormous errors can result. The relationship between the statistical time lag, the fault
transient and methods used to locate faults are often misunderstood [3, 4]. As pointed out in
[2] the statistical time lag must not be included in the time delays used to locate the fault.
A less understood phenomenon also occurs. If the rise time of the high voltage is too slow,
the fault gradually progresses into conduction. In this case, a fault transient is still generated,
but a sharp breakdown is not observed. The implication of this slow conduction process is
that the fault transient does not contain much of the high frequency energy necessary for
effective fault location. From empirical evidence, it is clear that as long as the rise time of
excitation voltage exceeds the statistical time lag of the breakdown, there is little effect on the
frequency content of the fault transient. In other words, the power source must be able to
bring the cable to full voltage before the conduction process begins. This places practical
constraints on the power source and the energy it must be able to supply.
Another important factor that influences the fault location is the impedance of the fault.
When the breakdown occurs, the arc at the fault site creates a low impedance at that point.
Several amperes must pass through the fault to make its impedance low enough to be an
effective reflector. Any additional current above this level lowers the impedance because the
impedance of the arc is an exponentially decreasing function of the current through it. The
practical implication is that the power source must be able to supply, at minimum, several
amperes for the required time.
The analysis of the fault transients for fault location depends on the fact that the arc has an
impedance much less than the characteristic impedance of the cable under test and that this
low impedance is maintained for an adequate time (less than 1 ms). In the surge methods it is
assumed that the fault impedance is temporarily near zero and that the transient repeatedly
reflects back and forth between the fault and the power supply used to induce the arc at the
fault. Sustaining a low impedance at the fault is also fundamental to the arc reflection
method. In this case it is assumed that the fault impedance is temporarily near zero and that
the TDR pulse reflects back from the fault while the power supply used to induce the arc at
the fault sustains the low impedance for tens of milliseconds.
Decay Method
In the decay method, the voltage source has a high impedance in series with it and the voltage
transient in the cable is measured. A high voltage is applied to the cable inducing a
breakdown at the fault. A transient is generated which travels back and forth between the
fault site and the voltage source. The voltage transient is measured using a voltage-coupling
device with a frequency response adequate to resolve both the edges and step portions of the
voltage transient. The propagation of the transient along the conductors can be described in
precise terms. An expression describing the Fourier transform of the voltage at the power
source, including the first N reflected terms, is

where the term (e ) is the complex propagation constant of the cable. Some simplifying
assumptions can be used to understand this equation. The first is that the reflections
coefficients are
v
(0)= 1 and
v
(x)= -1. This assumes that the source has a high output
impedance in series with it, such as a resistance and therefore the voltage reflection
coefficient is plus one. The fault impedance is considered to be a short circuit with a voltage
reflection coefficient of minus one.
The next assumptions are that t
s
is zero and that (e ) is the propagation constant for a
nondispersive cable. Using these assumptions, the equation simplifies to
2
where -je t
f
, the delay term from the fault to the power source, was used in place of (e ) x.
The terms of the summation inside the brackets time shift the step voltage generated by the
compensating source described in appendix A. The inverse Fourier transform of the first few
terms in the brackets are

As can be seen, this is an alternating series of positive and negative impulses displaced from
each other by twice the time delay to the fault. When these delay terms are convolved with
the step transient, the ideal voltage appears as a square wave having a period that is four
times the delay time to the fault. To determine the delay to the fault, one measures the time
difference between the first rising edge (falling edge) and the next falling edge (rising edge)
of the recorded transient. It should be noted that in this case there is a high impedance in
series with the voltage source so the rise time of the excitation voltage is very slow and is
usually not observed in the transient recording. A dc power source is often used to energize
the faulted cable to produce these waveforms since this type of source can readily produce
very high voltages. An example of a transient generated in this manner is illustrated in Fig. 1.
Current Impulse Method
In the current impulse method, the impulse generator has no intentional impedance placed in
series with it and the current transient in the cable is measured. A surge generator applies a
high-voltage step to the cable under test that induces a breakdown in the cable. The transient
travels back and forth between the surge generator and the fault. The current transient is
measured using a current transformer with a frequency response adequate to resolve only the
edges of the current transient. This method uses a theory similar to that discussed in the
preceding description of the decay method. The basic difference is that this method uses the
current transient induced in the cable rather than the voltage transient. The expression for the
current transient is the same previously described for the decay method if the voltages are
replaced with currents and the voltage reflection coefficients are replaced with current
reflection coefficients. Some simplifying assumptions were used. The first is that the
reflections coefficients are
I
(0)= 1 and
I
(x)= 1. This assumes that the impulse generator
has a low output impedance. Ideally, the impulse capacitor is a short circuit to high
frequencies and therefore the current reflection coefficient is plus one (current and voltage
reflection coefficients are opposite). The fault impedance is considered to be a short circuit
with a current reflection coefficient of plus one.

This equation is almost identical to the simplified equation except that the signs of the terms
in the summation do not alternate between positive and negative values. The Inverse Fourier
transform of the first few terms in the brackets is

This series of positive impulses is similar to the voltage transient case because each impulse
is displaced from the others by twice the time delay to the fault. The difference is when these
delay terms are convolved with the step transient, the ideal current waveform appears as a
staircase wave with each step having a length that is twice the delay time to the fault.
In practice, one does not observe the step current waveform but rather an approximation to
the mathematical derivative of it. To detect the current transient, a high frequency current
transformer is used. This transformer only responds to the rising edges of the step current.
Therefore, the observed transient resembles a series of impulses with a period equal to twice
the delay to the fault. It should be noted that in this case the impulse generator has a fast rise
time and the excitation current is also observed in the transient recording. The first impulse is
the excitation pulse from the impulse generator that travels out into the cable to break down
the fault. The time between this first pulse and the next pulse is the statistical time lag to
breakdown, t
s
, and provides no useful information for locating the fault; it should be ignored.
The procedure used to locate the fault is similar to that used with the voltage transient
measure the time between adjacent impulses, recalling that the first pulses should not be used.
An example of a transient measured using this method is illustrated in Fig. 2.
Time Domain Reflectometry
The theory surrounding the propagation of electromagnetic signals in cables is very complex
and a complete explanation is beyond the scope of this paper. The two most important aspects
of the theory are the velocity of propagation and the characteristic impedance. When a pulse
is applied to a cable, the pulse will travel along it with a speed called the velocity of
propagation. Knowing the velocity of propagation in the cable helps the operator interpret the
measurement results in terms of the distance to the fault. The velocity of propagation is
normally expressed either as a velocity factor or as one-half the actual velocity. Each power
cable has its own velocity and any error made when assuming a velocity for a cable will
translate, in direct proportion, to errors in locating the fault. The velocity in power cables is
discussed in more detail in appendix B.
The other important property of a cable is its characteristic impedance. When a voltage pulse
is applied to a cable, it will travel along it. Along with the voltage pulse there will be a
corresponding current pulse that also propagates along the cable at the same velocity. The
ratio of the voltage pulse, v, to the current pulse, i, is the characteristic impedance, Z
0
6
where x is a position along the cable and t is time. Usually this quantity has the same
dimensions as resistance. When an impedance discontinuity is encountered along the length
of the cable, some energy is reflected back according to the following relationships:
7

where
V
and
I
are the voltage and current reflection coefficients, respectively. The
reflection coefficients are given by
8

where Z
dscnt
is the impedance discontinuity at location x along the cable. It is simplest to
consider as a real valued quantity for discussion purposes; however, in general it is not. As
can be seen, whenever there is a change in impedance along the cable, energy is reflected
back toward the TDR. Using the appearance of a reflected pulse along the trace, in
conjunction with the velocity of propagation in the cable, lets the operator easily find the
fault. Considering these equations, it can be seen that if the cable is blown open (high-
resistance series fault), then
v
is equal to one and thus an upward pulse will appear at the
fault site. If the cable has a bolted fault (low-resistance shunt fault) then
v
is equal to minus
one and thus a downward pulse will appear at the fault site. These are the two most readily
identifiable faults that can be found using TDR. The skilled operator can learn to identify
many other features present in the TDR trace and use these features to perform a more
thorough diagnosis of the cable. An example of a TDR trace is shown in Fig. 3.
The one flaw that TDR has is that it cannot find high-resistance shunt faults. Examining the
impedance discontinuity, Z
dscnt
, in more detail can see this. When the impedance
discontinuity is located along the cable, it is defined as the parallel combination of the fault
impedance and the characteristic impedance of the cable
9

Careful examination of this equation and defining the reflection coefficient reveals that as the
fault impedance becomes larger than about four times the characteristic impedance of the
cable, the reflected voltage from the fault diminishes to the point where it is difficult to
distinguish as a return due to a fault.
Arc Reflectometry
The arc reflection test is similar to TDR in the respect that it uses TDR, but it is different
because it augments the TDR test to find high-resistance faults. As mentioned previously,
reflections from high-resistance shunt faults (the most common type) are rarely found using
TDR alone. However, auxiliary equipment can be used to convert the fault impedance into a
short circuit. Once the fault impedance is converted into a short circuit, it is easily identified
using standard TDR methods.
During the arc reflection test, the cable has a high-voltage surge applied to it. This high-
voltage surge has a special wave shape tailored to this application. The wave shape is
oscillatory and has the shape of an exponentially damped sinusoid. The purpose of this shape
is to extend the time that the arc is ignited so that the reflection can be observed on the TDR
screen. Specialized TDRs have been developed that are capable of capturing TDR pulses that
reflect from the momentary arc created during the surge. The benefit of this technique is its
simplicity in interpretation. The pulses displayed on the TDR screen are the standard types
normally seen on a TDR. The fault appears as a significant downward deflection and is
readily identified. Furthermore, the TDR trace prior to voltage application is also displayed
on the screen making it even simpler to distinguish the change that occurs in the TDR trace
when the fault impedance is converted into a short circuit. An example of an arc reflection
trace is shown in Fig. 4.
An added advantage of using arc reflection is that it minimizes the test voltage needed to find
the fault. In essence, the cable under test behaves like a surge protection device; it clamps the
voltage applied to the cable to the fault breakdown voltage. The rise time of the surge
generator, producing the special arc reflection surge, is much slower than the statistical time
lag associated with the breakdown. The result is that the voltage applied to the cable barely
exceeds the breakdown of the cable since the fault breaks down before the voltage rises much
beyond the breakdown level. Because of this clamping, increasing the voltage of the surge
generator does not increase the applied voltage much beyond the breakdown voltage of the
cable; it only increases the current through the fault. This is only true for the arc reflection
wave shape. The standard surge wave shape used for thumping applies voltages proportional
to the output voltage of the impulse generator and can theoretically generate transients up to
four times the applied voltage, peak to peak.
Acoustic Methods
After the approximate position of the fault has been found using a radar method, some means
of pinpointing the fault must be used. Common practice has been to apply repeated high-
voltage surges to the cable and listen for the "thump." To date, this is by far the most
successful method used to pinpoint the precise location of a fault. However, indiscriminate
use of a surge generator can place sound insulation at risk. It is prudent to minimize the stress
to the cable because there is suspicion that excessive impulsing of aged cable shortens its life
[5]. The safest way to pinpoint a fault is to keep the surge voltage as low as possible and to
apply it only as long as is needed. The problem associated with this approach is that the
sound decreases when the voltage is lowered. Increasing the capacitance in the surge
generator can compensate this but the point of diminishing returns is quickly approached
when the increase in size and cost does not justify the added benefit.
To simplify the task of fault pinpointing, the industry has developed acoustic listening
devices that amplify the sound of the thump. Using devices like these can simplify fault
locating greatly, especially in difficult cases where the cable is buried deep. The simplest
acoustic detectors are geophones followed by an amplifier. These devices work well if the
operator is experienced in their operation. The general approach is to listen carefully in the
vicinity of the fault, move the detector, listen again and make a judgment, based on the
sound, whether the geophone was placed closer to or farther from the fault. This requires a
trial-and-error approach that translates into an increased number of thumps being applied to
the cable. Furthermore, when an inexperienced operator uses these devices, the time to locate
the fault increases even more.
To answer these needs, even more sophisticated devices have been developed. One of the
more advanced methods senses the surge in the cable and measures the time between the
surge and the reception of the acoustic pulse produced by the fault breaking down. Measuring
this time delay allows the instrument to indicate a general distance to the fault. Precise
measurements of the distance to the fault are possible in principle, but this would require
knowing the propagation velocity of sound in the soil. The variation in soil content varies so
much that it would be nearly impossible to know the velocity accurately enough to be of any
use. Rough approximations can be used for the velocity to get a measurement of the distance
to the fault, but it is best to use the distance reading as an indication of "closer to" or "farther
from" the fault. Having this additional information reduces the trial-and-error search for the
fault but does not eliminate it.
Yet another technology has been developed that reduces further the amount of trial and error
in pinpointing the fault. New research has developed a directional acoustic detector that
enhances the technology used in units that simultaneously measure the acoustic pulse and the
surge pulse. This new development is capable of determining the direction to the fault using
the times between the surge pulse and the arrivals of the acoustic pulses at two geophones.
Using this method eliminates the trial and error since the instrument always indicates the
direction to the fault; there is no guessing involved.
Choosing the Best Method
The methods previously described all have their advantages. Choosing the best method to use
depends on the type of fault and the equipment available.
For a low-resistance fault, TDR is the best method to use. It is unlikely that the high-voltage
radar methods will be successful in this case. However, skilled operators can interpret current
impulse waveforms to locate low-impedance faults. One advantage in using TDR is that it is
not necessary to use high voltage.
For a series fault, TDR is the best method to use because it is easy to interpret and high
voltage is not required. High-voltage radar will also be successful on series faults because, in
most cases, the series fault is usually a combination series-shunt fault and the fault will break
down to the neutral.
For a high-resistance fault, the best methods are the high-voltage radar methods. TDR will
usually not work because the impedance discontinuity is too small to reliably indicate the
fault position. The choice of the high-voltage radar method is sometimes a matter of
preference but is often dictated by the situation. A more detailed explanation of the
advantages and disadvantages follows.
Transient methods have found success in practice but have deficiencies. The first deficiency
is that, in general, the waveforms are complicated and require a highly skilled operator to
interpret them. The second deficiency is more serious. The velocity dispersion in the cable
can cause the interpretation to depend on the position of the fault within the cable. The source
of this error is directly attributed to the frequency content of the transient. This effect is small
for short cables but is exaggerated for longer cables.
It is useful to compare TDR (Arc Reflection) to the transient methods. In general, both
methods are time domain-based methods for locating faults. The main difference is in the
time domain signal that is used to locate the fault. TDR and Arc Reflection use what is called
an active radar method; the instrument sends out a pulse. In contrast, transient methods use
what is called passive radar; they rely on the fault breaking down to generate a time domain
signal. When the instrument sends out a pulse, it uses an optimized signal that has completely
known characteristics. In contrast, transients generated by a breakdown, while often well
behaved, can often vary wildly, making it difficult to interpret. The most troublesome
variation is the change in frequency content, which makes the waveform susceptible to
dispersion creating location errors. TDR and Arc Reflection pulses are also susceptible to
dispersion errors but to a much lesser extent because the pulses are optimized to eliminate
this effect and do not vary in frequency content. For this reason TDR and arc reflection are
the preferred methods.
There are reasons to use the transient methods. In some situations it is not possible to
establish a reliable arc to use Arc Reflection. In these cases, transient methods are an
excellent backup method when Arc Reflection is not successful. In some cases it is not
possible to cause the fault to break down at a voltage low enough to use arc reflection. When
high voltages are needed that are in excess of what a thumper can generate, a dc power
supply can be used in combination with the decay method. This approach is often used on
high-voltage cables with rating in excess of 35 kV.
In complex situations, where T's, Y's, taps or branches are present, it is extremely difficult to
interpret transients to locate a fault. Even with only a single tap, the multiplicity of reflections
makes it extremely difficult to identify the appropriate positions to place the cursors on the
transient waveform.
In contrast, using TDR or Arc Reflection it is a simple matter to locate a fault in complex
situations because TDR pulses are so repeatable. When using Arc Reflection, the TDR trace
can be saved prior to applying high voltage. When high voltage is applied, the TDR trace is
identical up to the position of the fault. This makes it simple to interpret the trace even when
T's, Y's, taps or branches are present. An example showing the use of differential arc
reflection on a complex circuit is illustrated in Fig. 5.
References
[1] Steiner, J.P., Weeks, W.L., "Advanced Cable Fault Locator," Final Report, EPRI EL-
6765, March 1990, Electric Power Research Institute, Palo Alto, Calf.
[2] Gail, P.F., "Cable Fault Location by Impulse Current Method," Proc. IEE, Vol. 122, No.
4, April 1975, pp. 403-408.
[3] Komoda, M., Kawashima, T., Arakane, M., Aihara, M., Fujiwara, Y., "Development of a
Current Detection Type Cable Fault Locator," Conference Record 90 WM 248-5 PWRD,
IEEE Winter Power Meeting, Atlanta Georgia, Feb., 1990.
[4] Gnerlich, H.R., "Fault Locating: What's the Effect on the Cable?" Electrical World, Vol.
204, No. 6, June, 1990, Supplemental Section, pp. S6.
[5] Hartlein, R.A., Harper, V.S., Ng, H.W., "Effects of Voltage Impulses on Extruded
Dielectric Cable Life," IEEE Trans. on Power Delivery, Volume 4, N0. 2, April 1989, pp.
829-841.
[6] Weeks, W.L., Diao, Yi Min, "Wave Propagation Characteristics in Underground Power
Cable," IEEE Transactions on Power Apparatus and Systems, PAS-103 pp. 2816-2825, Oct
1984.
[7] Diao, Yi Min, "Propagation of Wideband Signals in Power Cables," Ph.D. Thesis, Purdue
University, School of Elec. Eng., West Lafayette, Ind., Dec., 1984, chapters 2-3.
[8] Steiner, J.P., "Digital Measurement of Partial Discharge," Ph.D. Thesis, Purdue
University, School of Elec. Eng., West Lafayette, Ind., May, 1988, chapter 3.


Fig. 1: Voltage decay transient generated by a fault in a cable. Note the placement of the
cursors on consecutive edges of the transient waveform.


Fig. 2: Current impulse transient generated by a fault in a cable. Note the placement of the
cursors on consecutive peaks of the transient waveform.


Fig. 3: TDR trace of a low impedance discontinuity in a faulted cable. Note the placement of
the cursors on the beginning of the downward deflection.


Fig. 4: Arc reflection trace of a high impedance shunt fault. The illustration shows a TDR
trace prior to voltage application and a TDR trace of the negative reflection from the arc.
Note the placement of the cursors on the beginning of the downward deflection.


Fig. 5: Differential arc reflection trace of a high impedance shunt fault in a complex circuit.
The upper graph shows a TDR trace prior to voltage application and a TDR trace of the
negative reflection from the arc. The lower graph shows the difference that clarifies the
analysis significantly. Note the placement of the cursors on the beginning of the downward
deflection in the lower trace.
Appendix A: Fault Transients
The most important factors influencing the analysis of the fault transient are the generation
and propagation of the transient in the cable under test. This analysis includes modeling the
fault process using transmission lines before and after breakdown as well as the fine details of
the propagation of the transient along the cable. The fundamental principles in the following
discussion do not depend on the type of high-voltage source used, but the discussion is
clearer if one is specified.
When a source energizes a cable under test, it applies a voltage, V
0
, to the cable. For the
purposes of this discussion it is assumed that the source has an output impedance Z
S
and that
the cable has a characteristic impedance Z
0
. It is also assumed that the cable has a high
resistance shunt fault. The source energizes the cable and after a random time delay, the fault
breaks down and suddenly there is a short circuit (ideally) at the fault site. This short circuit
divides the cable into two separate and essentially independent lengths. If the fault is not too
close to the ends of the cable, the initial current through the fault is

The factor of two is due to the inrush of current from the two separate lengths of cable on
either side of the fault. This is only the initial current; the final current depends on the current
available from the voltage source energizing the cable.
From fundamental considerations, any current entering the fault must also exit it. The current
transient entering the fault at breakdown is a step current. This same step current exits the
fault (Kirchhoff's current law) and travels along the cable in a direction opposite to the
current entering the fault. The voltage transient associated with the current exiting the fault is

The polarity of the voltage transient is opposite the voltage at the fault just prior to
breakdown. The voltage reversal occurs because the current transient exiting the fault has the
same polarity as the current entering it, but travels in the opposite direction thus generating a
negative voltage. The negative voltage step traveling away from the fault ensures that the
other fundamental boundary condition is met: the voltage at the fault (short circuit) must be
zero (Kirchhoff's voltage law).
These transients can be thought of as having been produced by a low-impedance voltage
source suddenly being switched in (step voltage generator) at the fault site at the instant the
fault breaks down. This hypothesized stepped voltage source is such that it superposes with
the voltage present at the fault just prior to breakdown making the voltage at the fault zero
after breakdown. Hence this voltage source is equal in magnitude but opposite in polarity to
the voltage prior to breakdown. The added voltage source shown in Fig. A1 is sometimes
known as a compensating voltage source; it generates the breakdown transient. It should be
noted that this breakdown transient is only related to the voltage source through its
magnitude. This transient if often incorrectly described as a transient generated by the voltage
source waveform reflecting from the fault; it is not. The transient is generated by the
breakdown of the cable.
With an adequate circuit model for fault generation in hand it is now possible to describe the
remaining aspects of the fault transient. Consider the specific example of a cable having total
length L with a fault at position z=x. An impulse generator is used to energize the cable and
is located at position z=0. The breakdown voltage of the fault is V
0
and the statistical time lag
to breakdown is t
s
. As shown in the circuit in Fig. A1, step-like transients propagate from the
fault site at the instant of breakdown.
Consider the length of cable between the fault and the power source. When the fault transient
arrives at the power source (z=0) it is reflected with an amplitude determined by the
reflection coefficients

where the argument of the reflection coefficient is indicative of the position in the cable.
After reflecting from the power source the transient travels back toward the fault. Upon
reaching the fault (z=x) the transient is again reflected but this time with reflection
coefficients given by

The transient travels back and forth, reflecting each time it reaches either the power source or
fault, until its energy is dissipated.


Appendix B: Non-ideal Propagation
Coaxial power cables are optimized for power delivery, not signal propagation; consequently
any measured transient signals have poor fidelity. This poor signal fidelity is due to the
propagation characteristics of the power cable and complicates the analysis of the transients.
The propagation of the transient through the cable is determined by the propagation
parameter, (e ), which is, in general, a function of frequency

The frequency dependent attenuation parameter, o (e ), determines the loss characteristics of
the cable and is responsible for the high frequency losses in the cable. The primary causes of
the high frequency loss are the semiconducting layers used for the strand and insulation
screens. The loss in the cable, k (e ), as a function of frequency, is given by

where z is the length of cable through which the signal travels. At high frequencies this loss
becomes severe and as can be seen from Fig. B1, for longer cables, most of the signal energy
above a few megahertz is lost. This attenuation of the higher frequencies limits the precision
of the time delay measurements used to locate the fault.
In addition to the high attenuation suffered by the transient as it propagates along the cable,
its phase also becomes distorted. The phase change, u (e ), caused by propagation through
the cable is given by

The frequency dependent phase parameter, | (e ), determines the propagation velocity of the
cable. If the phase parameter is strictly a linear function of frequency, then the cable has a
linear phase characteristic and there is a single value for the velocity

where c is the natural velocity of propagation. The constants C and L are the cables per unit
length capacitance and inductance, respectively. For power cables the phase function deviates
from this ideal behavior and the cable is considered to be dispersive. Transients traveling
through a dispersive medium become distorted, even in the absence of any frequency
dependent attenuation losses. The phase distortion occurs because the different frequency
components of the transient travel at different velocities. The effect of this dispersion is to
spread the transient in time, making it difficult to choose a reference point on the waveform
that can be used to measure the arrival time of the transient. This dispersive property also
makes it difficult to assign a simple propagation velocity to power cables. A ratio of the
velocity of propagation, u , in a power cable to the natural velocity of propagation, c, in
polyethylene as a function of frequency is shown in Fig. B2. The change in velocity, over the
frequency range shown, is quite dramatic and is influenced strongly by the properties of the
semiconducting screens. The dispersive character shown in Fig. B2 is for a specific cable,
however, other types of power cable behave in a similar manner as has been theoretically
predicted [6, 7].
Fault location methods are implemented in the time domain and it is difficult to translate the
effects of dispersion into this domain. The fault transient comprises many frequencies, each
traveling at a different velocity. The effects on the transient are to shift its peak, skew its
edges, and basically to tilt the pulse. The effects are most readily quantified as a time delay
bias that describes them as a simple time shift [8]. It is not possible to assign a single correct
velocity to the cable but rather an apparent velocity has to be assigned that is a weighted sum
of velocities at the individual frequencies composing the transient. In general, time delay
measurements in dispersive media are difficult to translate into locations because only an
apparent velocity is available.
Some qualitative comments can be made about the propagation characteristics of power
cables based on the work in [6]. For frequencies less than 15 kHz, the surrounding earth has a
dramatic effect on the propagation characteristics. The propagation velocity is nearly flat, to
within a few percent, for frequencies in the range 0.1 to 50 MHz. At frequencies in excess of
100 MHz, the velocity rapidly approaches the natural velocity, c, of the dielectric. The work
in [6] considered a cable with a solid shield. Many power cables have helically wound
concentric neutrals that only approximate a full shield. In this case, significant energy
propagates in the region external to the concentric neutrals (earth), even at the higher
frequencies, causing further distortion of the signal. These qualitative remarks suggest that
the analysis of fault transients should only use frequencies in excess of 100 kHz to reduce the
errors due to dispersion. The data of Fig. B2 bear this out with only a three-percent change in
velocity over the range 0.1 to 10 MHz.
Many factors influence the propagation velocity of power cables. In [6] it was found that the
composition of the semiconducting material dramatically influences the dispersion. The
relative dimensions of the semiconducting shields and dielectric determine the shape of the
dispersion curve. Furthermore, the composition of the semiconducting material used in the
shields is not uniform along the length of the cable. Since the velocity is heavily influenced
by the semiconducting shields this adds the further dimension that the velocity is not uniform
along the length of the cable, even for a single frequency.
There are other factors that influence the propagation velocity. In cables that have concentric
neutrals, the pitch of the neutrals and the percent shielding contribute to the velocity. The
condition of the soil also influences the velocity because a portion of the electromagnetic
field propagates external to the cable. The amount of neutral corrosion affects the proportion
of signal propagating external to the cable again influencing the velocity. Another effect
occurs when power cables age. Water is often absorbed into the dielectric and this absorption
is not uniform along the length or through the cross section. The dielectric constant of water,
at room temperature, is approximately 80 and can significantly alter the equivalent dielectric
constant of the cable, affecting the velocity. Large errors can also occur if the cable system
has different types of cable spliced in along its length making the velocity change in a
stepwise fashion. Variations among manufacturers, even for new cables with the same rating,
make it virtually impossible to predict the velocity with accuracies better than 1 percent.
These effects can compound to give unacceptable location errors.
Interpretation Errors
The velocity dispersion in the cable can cause the interpretation of transient data to depend on
the fault's position within the cable. A simple example is shown in Fig. B3. The transient in
Fig. B3 is a current impulse waveform recorded from a fault at 786 m in a 1289 m 15 kV,
1/0, power cable. The time delay between the impulses measured from this recording is
10.275 ms. An independent set of measurements using a time domain reflectometer gives the
time delay as 9.452 ms. The 8.7 percent discrepancy is attributed to the velocity dispersion in
the cable and results in a 68.4 m location error.
The source of this error is directly attributed to the frequency content of the transient. The
second impulse in the sequence of impulses shown in Fig. B3 was extracted, and the Fourier
transform of this individual pulse was calculated. The magnitude of this transform is shown
in Fig. B4. The lower 3-dB cutoff of the individual pulse is 25 kHz and the upper 3-dB cutoff
is 180 kHz. The corresponding points on the dispersion curve in Fig. 3 have relative velocity
factors that differ by 11 percent, respectively. This variation in velocity over the frequencies
composing the pulse is responsible for tilting it. The frequency content of the TDR pulse is
much higher on the dispersion curve thus causing the inconsistency between the TDR
measurement and the current impulse measurement.
Often the current impulse waveforms are so complex that the operator has to use the impulses
at the tail end of the transient where they have settled into nearly a single frequency
oscillation. The impulses on the tail have even larger contributions from the lower
frequencies causing a further error. The measurement method is almost equivalent to
interpreting the waveform as an oscillation whose period is equal to the round-trip
propagation delay from the fault to the voltage source. Using this interpretation and referring
to the dispersion data in Fig. 3, it is immediately clear where the source of the error is the
cable itself. Using this interpretation one can directly see that the error depends on the
location of the fault. If the fault is near to the voltage source, then the frequency of oscillation
is high, placing the velocity in a near constant portion of the dispersion curve. However, if
the fault is far from the end of a long cable, this places the velocity much further down on the
curve. In fact, the longer the cable is, the worse the error becomes because the velocity begins
changing rapidly at the lower frequencies.







Vissza a foldalra






Locating That Underground Cable Fault
Feb 1, 1998 12:00 PM, Dennis Koep

ShareThis
What method or combination of methods is best for locating underground cable faults?
Finding the location of an underground cable fault doesn't have to be like finding a needle in a haystack. There are
many locating methods, coupled with new detection technologies, that make this task much easier and less time
consuming. However, you should understand that there is no single method or combination of methods that is "best."
Your selection of the appropriate method for the situation and your skill in employing that method are the keys to safely
and efficiently locating cable faults without damaging the cable. Let's see what's involved.
Basic cable fault locating methods. There are two basic methods of locating an underground cable fault.
Sectionalizing This procedure, as shown in Fig. 1, risks reducing cable reliability, because it depends on physically
cutting and splicing the cable. Dividing the cable into successively smaller sections will enable you to narrow down the
search for a fault.
For example, on a 500-ft length, you would cut the cable into two 250-ft sections and measure both ways with an
ohmmeter or high-voltage insulation resistance (IR) tester. The defective section shows a lower IR than the good
section. You would repeat this "divide and conquer" procedure until reaching a short enough section of cable to allow
repair of the fault. This laborious procedure normally involves repeated cable excavation.
Thumping When you supply a high voltage to a faulted cable, the resulting high-current arc makes a noise loud enough
for you to hear above ground. While this method eliminates the sectionalizing method's cutting and splicing, it has its
own drawback. Thumping requires a current on the order of tens of thousands of amps at voltages as high as 25kV to
make an underground noise loud enough for you to hear above ground.
The heating from this high current often causes some degradation of the cable insulation. If you're proficient in the
thumping method, you can limit damage by reducing the power sent through the cable to the minimum required to
conduct the test. While moderate testing may produce no noticeable effects, sustained or frequent testing can cause the
cable insulation to degrade to an unacceptable condition. Many cable fault locating experts accept some insulation
damage for two reasons: First, when thumping time is minimal, so is the cable insulation damage; secondly, there is no
existing technology (or combination of technologies) that can entirely replace thumping.
Newer fault locating technologies. There are some relatively new methods of locating cable faults that use rather
sophisticated technology.
Time Domain Reflectometry (TDR) The TDR sends a low-energy signal through the cable, causing no insulation
degradation. A theoretically perfect cable returns that signal in a known time and in a known profile. Impedance
variations in a "real-world" cable alter both the time and profile, which the TDR screen or printout graphically represents.
This graph (called a "trace") gives the user approximate distances to "landmarks" such as opens, splices, Y-taps,
transformers, and water ingression.
One weakness of TDR is that it does not pinpoint faults. TDR is accurate to within about 1% of testing range.
Sometimes, this information alone is sufficient. Other times, it only serves to allow more precise thumping.
Nevertheless, this increased precision can produce substantial savings in cost and time. A typical result is "438 ft 5 10
ft." If the fault is located at 440 ft, you only need to thump the 20-ft distance from 428 ft to 448 ft, instead of the entire
440 ft.
Another weakness of TDR is that reflectometers cannot see faults-to-ground with resistances much greater than 200
ohms. So, in the case of a "bleeding fault" rather than a short or near-short, TDR is blind.
High-voltage radar methods There are three basic methods for high-voltage radar, ranked here in order of popularity,
with the most popular described first: arc reflection, surge pulse reflection, and voltage pulse reflection. The arc
reflection method, as shown in Fig. 2 (on page 64N), uses a TDR with a filter and thumper. The filter limits both the
surge current and voltage that can reach the cable under test, thus allowing minimal stress to the cable. Arc reflection
provides an approximate distance to the fault (when there is an ionizing, clean arc produced at the fault and the TDR in
use is powerful enough to sense and display a reflected pulse).
The surge pulse reflection method, as shown in Fig. 3, uses a current coupler and a storage oscilloscope with a
thumper. The advantage of this method is its superior ability to ionize difficult and distant faults. Its disadvantages are
that its high output surge can damage the cable, and interpreting the trace requires more skill than with the other
methods.
The voltage pulse reflection method, as shown in Fig. 4 (on page 64P), uses a voltage coupler and an analyzer with a
dielectric test set or proof tester. This method provides a way to find faults that occur at voltages above the maximum
thumper voltage of 25kV.
The open neutral and cable fault locating Bare neutrals corrode quickly in contaminated soil that holds corrosive
chemicals or excessive moisture. Open neutrals often thwart the effectiveness of high-voltage radar. Beware: In the
existence of an open neutral, nearby telephone or CATV cables will complete the circuit.
One test to detect an open neutral requires shorting a known good conductor to a suspect neutral, as shown in Fig. 5
(on page 64P), then measuring the resistance with an ohmmeter. If the reading is 10 ohms or higher, you can suspect
an open neutral. Remember, other objects can complete the circuit.
Another test uses a TDR. The trace on an open neutral will show a much flatter positive pulse than it will for an open
conductor. On lower-end TDRs, this pulse may not be visible. When the conductor is completely open, the trace will
almost never include a reflected pulse indicating the end of the cable.
If the TDR displays an open neutral, then an AC-voltage gradient test set can locate the break in a direct-buried
unjacked cable. The test set's transmitter forces AC current to flow through the neutral, and the conducting earth
surrounding the damaged section acts as an electrical jumper. An A-frame , as shown in Fig. 6 (on page 64P), then
detects the resulting voltage gradient in the soil.