Close and Return
C.3.4 8th International Conference on Insulated Power Cables
A NEW APPROACH TO AVOID NEGATIVE SUPERPOSITION ERRORS FOR PD
MEASUREMENTS ON LENGTHS OF EXTRUDED POWER CABLES
Kay RETHMEIER, Alexander KRAETGE, Michael KRUEGER Omicron, (Austria), kay.rethmeier@omicron.at
ABSTRACT
PD in cables cause double pulses due to the signal
reflection from the far cable end. As a result, the PD
bandpass filter cannot determine the correct value of the
Apparent Charge. As positive superposition errors can be
tolerated, negative signal superposition causes dangerous
under-estimation of the involved PD charge. Considering
the regulations of the standards IEC 60270 and IEC
60885-3 with respect to the PD bandpass characteristics,
the double pulse problem can be solved by using a
synchronous 3-band filter, not affected by resonance
faults, showing no significant under-estimation of the
Apparent Charge value.
KEYWORDS
Partial Discharge, PD, cable, cable length, IEC 60270,
IEC 60885, resonance, negative superposition, apparent
charge, multi-band PD measurement.
INTRODUCTION
Partial discharge measurements are used as one of the
quality control tests for extruded power cables. One
important requirement is to allow an accurate
measurement of single PD faults, independent from their
position in the cable length. Due to their traveling wave
characteristic, PD pulses do propagate in both directions,
causing reflections at the cable's ends. A PD measuring
system located at one end of the cable will therefore react
on these groups of double pulses by superposed
oscillations of the PD bandpass filter system. Depending
on the double pulse distance these oscillations may lead
to an over-estimation (positive superposition error) or to a
dangerous under-estimation (negative superposition error)
of the "real" Apparent Charge q of the PD fault. In the
worst case, a PD fault may not be detectable at all during
the PD measurement. These so-called "forbidden lengths"
of a cable with negative superposition errors, showing an
under-estimation of the Apparent Charge, depend on the
PD fault position in the cable length, but also on the PD
filter settings selected at the PD instrument. For a
systematic analysis of these "forbidden lengths", double
pulse diagrams have to be plotted once a year for every
PD test facility, as demanded by the IEC standard 60885-
3 [1].
OVERVIEW: STANDARDS
PD measurements generally have to be performed
compliant to the IEC standard 60270 [2]. Besides defining
the expression "Apparent Charge" as the most important
value for the PD measurement itself, this standard
describes the measuring setup with HV source, coupling
capacitor, measuring impedance and PD measuring
system. The test object (DUT) is reduced to a discrete
capacitor, characterized by its capacitance (Farad) only.
For cables, the simplification of modeling the DUT as a
single capacitor cannot be tolerated, as PD pulses
propagate along the cable. Therefore, not the complete
Jicable’11 – 19 – 23 June 2011, Versailles - France
C.3.4
cable capacitance is involved in the resulting charge
transfer process. Consequently, the cable has to be
modeled with distributed capacitances or as a
transmission line. This fact was also recognized by the
authors of the IEC 60270 and considered in the
informative annex C ("Measurements on cables, gas
insulated switchgear, power capacitors and on test
objects with windings") of the standard. It is stated, that
the Apparent Charge measured at the terminals of the
cable is not equal to the charge at the PD fault position
due to traveling waves, reflections and resonances.
Finally, the IEC 60270 as a horizontal standard is not
specific enough for HV cables. As a consequence the
"cable" standards IEC 60840 [3] (<150 kV) and IEC 62067
[4] (>150 kV) do not refer to the IEC 60270 at all, but only
to the IEC 60885-3 ("Electrical test methods for electric
cables – Part 3: Test methods for partial discharge
measurements on lengths of extruded power cables"),
where the double pulse problem is discussed in detail.
The topics of double pulse discrimination and double
pulse superposition are also addressed in the IEC
standard 60034-27 [5] (rotating machines) as well as in
the CIGRE guide 366 [6], showing the oscillating PD
signals of double pulses right after passing the PD
bandpass filter system in a very vivid manner.
SUMMARY OF IEC 60885-3
The standard gives valuable background information on
the nature of PD on cables, behaving as traveling waves,
and about attenuation effects reducing the Apparent
Charge level more and more with increasing PD
propagation path. The main focus lies on the double pulse
nature of PD on cables, as PD pulses (nearly) always
appear as pairs due to the signal reflection from the far
cable end. These double pulses can cause resonance
problems of the PD bandpass filter, leading to an
increased Apparent Charge value of up to 200% or to an
reduced value close to 0% of the originally involved
charge. With respect to uncertainties, a charge under-
estimation of -10% is tolerated by the standard. To be
aware of this problem, double pulse diagrams have to be
plotted once a year for every PD test facility or after major
repairs or changes of the PD measuring setup.
SPECTRUM OF DOUBLE PULSES
As the frequency spectrum of a single PD pulse is similar
to a Dirac pulse, broadbanded and more or less of
horizontal development with increasing frequency (see
fig. 1), the frequency spectrum of a double pulse pair
shows regular repeating minima and maxima (see fig. 2).
This frequency behavior is caused by the λ/4 resonance
(and multiples of λ/4) of the cable, where the original
signal at the beginning of the cable line (measuring
position, near end) is eliminated by the delayed and
therefore phase-shifted reflection from the far cable end.
As a result, not all spectral components of the signal can
exist due to this elimination process.
 Close and Return

C.3.4 8th International Conference on Insulated Power Cables C.3.4

As a consequence, the Apparent Charge, which is

typically determined by frequency-domain integration
(quasi integration) of a PD bandpass filter, shows a value
which is too high (local maximum of the spectrum, positive
superposition, over-estimation of Apparent Charge) or a
value which is too low (local minimum of the signal
spectrum, negative superposition, under-estimation of the
Apparent Charge), depending on the PD filter
characteristics or on the PD fault position on the cable.

Fig. 4: Signal response of PD bandpass filter, fcenter =

Fig. 1: FFT of a single PD pulse
2 p.u., bandwidth = 1 p.u.
By additionally reducing the PD filter bandwidth to 50%,
the duration of the oscillation is doubled, see figure 5. As
a consequence the probability of influencing the oscillation
of a second pulse by the tail of the first oscillation is higher
here.

Fig. 2: FFT of a PD pulse pair

A systematic way to analyze this behavior is the
evaluation of a double pulse diagram, where a double
pulse calibrator simulates different PD fault positions on a Fig. 5: Signal response of PD bandpass filter, fcenter =
cable by altering the double pulse time distance delta-t. 2 p.u., bandwidth = 0.5 p.u.
This double pulse diagram then shows the areas within a
These curves demonstrate why the double pulse problem
cable, where PD will cause a dangerous low Apparent
affects small banded filters more than broad banded
Charge value (the so called "forbidden length").
bandpasses. To reduce the superposition problems, the
PD filter should be chosen as broad as possible.
THE INFLUENCE OF THE PD BANDPASS
As a second PD pulse causes a time-shifted but almost
FILTER identical oscillation, both oscillations superimpose.
A first PD pulse (as well as a single PD pulse) causes a
modulated sinusoidal oscillation with approximately the
center frequency of the PD bandpass filter (see fig. 3).
The duration of this oscillation depends on the bandwidth
of the PD bandpass filter.

Fig. 6: Signal response of PD bandpass filter, positive

Fig. 3: Signal response of PD bandpass filter, fcenter =
1 p.u., bandwidth = 1 p.u.
By doubling the center frequency of the PD bandpass
filter, the oscillation is faster, but approximately of the
same duration as before (see fig. 4).
superposition of 2 signals
Depending on the time delay, both oscillations can
increase the resulting signal up to 200% (positive
superposition) or decrease the resulting Apparent Charge
value down to 0% (negative superposition, signal
elimination). Figure 6 shows an example of positive signal
superposition of almost 200%, figure 7 shows a nearly
eliminated signal due to negative superposition.

Jicable’11 – 19 – 23 June 2011, Versailles - France

 Close and Return

C.3.4 8th International Conference on Insulated Power Cables C.3.4

nearly be eliminated by choosing broadbanded PD

bandpass filters for the PD measurement (see fig. 8). If
the PD bandpass filter is capable to cover some of the
local positive and negative maxima of the signal's
spectrum, all disturbing effects are more or less averaged,
which leads to an almost correct Apparent Charge value
and A100 value, respectively.

Fig. 7: Signal response of PD bandpass filter,

negative superposition of 2 signals
As the PD filter parameters can be varied in some limits
given by the IEC 60270, the PD filter can be chosen to be
minimum impacted by this effect. Here, broadbanded PD
filters are strongly recommended. Fig. 8: Wide band bandpass (blue area) covering
multiple resonances in the FFT
As the PD fault position of the cable, and therefore the
pulse pair time distance (respectively the signal phase
shift), is not known before the PD test, there is no CREATING DOUBLE PULSE DIAGRAMS
systematical way to select the PD filter settings to The IEC 60885-3 clearly defines the procedure to create
guarantee immunity against negative superposition errors. double pulse diagrams. With no DUT (HV cable)
connected, double pulses have to be injected by a double
CLASSICAL METHODS TO SUPPRESS THE pulse calibrator directly between the HV connection of the
REFLECTION coupling capacitor and ground. The cable is simulated by
a parallel resistor (typically: 50 Ohm).
As the reflected PD pulse causes the problematic reaction
of the PD filter, a suppression of this pulse solves the As a first measurement, the time distance of the double
problem of negative superposition errors. Pulse pulse calibrator has to be set to t100 = 100 µs. For such a
suppression can be theoretically realized by an electronic long time delay between 2 pulses, no influence is
switch, closing the PD signal input path shortly after the expected on the PD bandpass filter. The measured
detection of the first pulse. Another method is to terminate Apparent Charge A100 therefore can be used as a
the far cable end with the characteristic impedance of the reference value. All other values recorded in the following
cable. Both methods are very difficult to realize in practice, test can be related to this A100 value. The time delay then
as the signal termination by the characteristic impedance, has to be reduced in steps. All the value pairs (delta-t and
for example, does not consider the fact, that there already relative charge At /A100) can be plotted into a diagram. As
is a significant change of the traveling wave impedance of time difference and cable length are connected by the PD
the cable due to the large test termination. A broadbanded pulse propagation speed, the whole double pulse diagram
signal termination therefore is not possible at all. can also be drawn with the relative charge value against
the cable length in meters. Different types of diagrams are
thinkable, as given by the IEC 60885-3 (see figures 9-11),
ELIMINATION OF NEGATIVE showing only positive errors, both, positive and negative
SUPERPOSITION ERRORS BY USE OF A 3- errors or even an oscillating curve.
BAND PD FILTER
A more sophisticated method to minimize the negative
superposition error is the use of 3 PD bandpass filters
within the filter range of the IEC 60270 synchronously for
the PD measurement, as possible with digital state-of-the-
art PD instruments [7]. The PD signal is recorded and
digitized in the time domain. Then the digital signal is
applied to 3 digital bandpass filters. As every single PD
bandpass filter is affected by the double pulse problem, a
smart combination of these 3 filters (weighted sum) shows Fig. 9: Double pulse diagram type 1 [1]
only a minimal influence and gives a reliable result and
therefore can be used for the PD measurement on cable
lengths.

ELIMINATION OF NEGATIVE
SUPERPOSITION ERRORS BY USE OF
BROADBAND PD FILTERS
By leaving the narrow regimentations of the IEC 60270
behind, the problem of negative pulse superposition can Fig. 10: Double pulse diagram type 2 [1]

Jicable’11 – 19 – 23 June 2011, Versailles - France

 Close and Return

C.3.4 8th International Conference on Insulated Power Cables C.3.4

Fig. 11: Double pulse diagram type 3 [1]

In order to reduce the "forbidden length" of a given cable,
the PD testing time can be split into 2 parts. After the first
part, where only the "allowed" length of the cable can be
accessed, the cable has to be turned around and
energized from the second termination. As a Fig. 13: Double pulse diagram of digital PD system
consequence, the double pulse diagram is mirrored in the
middle of the cable length. A "forbidden length" of the test
part 1 may become an "allowed" length in test part 2 (see
fig. 12).

Fig. 14: Double pulse diagram, 160 kHz bandwidth

Fig. 12: Double pulse diagram for measurements from

both sides [1]

EXAMPLES OF DOUBLE PULSE DIAGRAMS

Double pulse diagrams from test setups as well as from
different HV test facilities are shown to demonstrate the
superposition problem.

Calibrator directly connected to PD Input of

the PD Measuring System
By connecting the double pulse calibrator directly to the
PD input port of the measuring system, the principal
behavior of different PD bandpass filters can be shown by
various double pulse curves (see fig. 13).
By using a fully digital PD system, an automated software
algorithm can change the PD filters comfortably to save
time. The PD bandpass filter was set to two different
bandwidths (300 kHz and 160 kHz) and to center
frequencies of 100 kHz up to 300 kHz. Figure 13 shows
the group of curves for the 160 kHz bandwidth on the right,
the group of curves for the 300 kHz bandwidth with a high
visual density on the left part of the figure. Figures 14 and
15 show the two bandwidths individually.
Fig. 15: Double pulse diagram, 300 kHz bandwidth
As expected the broader PD bandpass filter with 300 kHz
bandwidth shows the better double pulse behavior. For
pulse distances of 2 µs and above the Apparent Charge is
determined correctly. For the 160 kHz bandwidth, the PD
pulse delay must not be below 4 µs to ensure a correct
charge evaluation. Additionally, the derivation from the
100% level is higher for the 160 kHz bandwidth. Figure 16
shows the evaluation of additional bandwidth, like the
9 kHz bandwidth, which is in accordance to the IEC 60270
(narrow band PD measurement) and which is required for
RIV measurements [8].

Jicable’11 – 19 – 23 June 2011, Versailles - France


C.3.4 8th International Conference on Insulated Power Cables
Fig. 16: Double pulse diagram, different bandwidths
Figure 17 shows an enlarged view for time distances up to
5 µs.
Fig. 17: Double pulse diagram, different bandwidths,
zoom
Again, it can be seen, that the Apparent Charge value is
heavily affected for narrow band PD filters. The curve for
the 9 kHz bandwidth shows quantization errors for time
differences above 20 µs and therefore only visualizes the
envelope curve. In figure 17 it can be seen that the
bandwidth of 3 MHz does not show any notable evidence
of superposition errors.
Calibrator Connected to Coupling Capacitor
The frequency properties of the whole PD measuring
setup can be significantly changed by the coupling
capacitor and by the measuring impedance. Therefore the
double pulse diagrams have to be taken with the complete
measuring circuit. Figure 18 shows curves for different
filter settings.
Jicable’11 – 19 – 23 June 2011, Versailles - France
Close and Return
C.3.4
Fig. 18: Double pulse diagram with 3-band filter
The black curve was generated by a 3-band PD bandpass
filter with 160 kHz bandwidth. The comparison to the
regular bandpass filter, the 3-band filter does not show a
significant under-estimation of the Apparent Charge with
only ca. 90% as absolute minimum. Therefore, this filter is
ideal for PD measurements on cable length in accordance
to the IEC 60885-3 and IEC 60270.
PD Test Facility of Cable Manufacturer
The double pulse diagram shown in figure 19 was
recorded at the HV lab of a cable manufacturer. It shows
the result of a 3-band PD bandpass filter of 650 kHz
bandwidth. The injected charge of the double pulse
calibrator was 50pC. The measured Apparent Charge was
not below 50pC for any double pulse time difference. The
limit for the negative superposition error defined by the
standard IEC 60885-3 is equal to 45pC (horizontal red line)
in this case.
Fig. 19: 3-band double pulse diagram of HV test lab
Double Pulse Generation by RG58 BNC
Cables
Reproducible double pulses can be generated either by
double pulse generators or by coaxial measuring cables
of variable length, causing a signal reflection at the open
end. Besides the fact, that the cheap solution of using
measuring cables is not covered by any IEC standard, it
describes the behavior of PD pulses on HV cables even
more realistic compared to the double pulse calibrators. A
first pulse is generated by a standard calibrator, traveling
in both directions, to the PD detector and to the far cable
end. Reaching the far cable end, the pulse component
changes its direction and travels back through the cable,
finally reaching the PD detector with a time delay related
to the cable length. During this pulse propagation, the
reflected pulse component is attenuated to a lower level,
as it also can be expected for real PD. Because of this
similarity it is worse to have a closer look to double pulse
diagrams created with BNC refection lines. Figure 20
shows such a double pulse diagram.
The high readings at the left side of the diagram at a
delta-t of 0 µs can be neglected, as no reflection
processes took place and therefore the result is not
comparable to all the other time differences. Also, the
curve of the 9 kHz bandwidth as well as the curve of the
30 kHz bandwidth are most properly subject to bad
quantization.

C.3.4 8th International Conference on Insulated Power Cables
Fig. 20: Double pulse diagram of RG58 reflection line
Comparison of Different Injection Methods
Figure 21 shows the differences that can be observed
when injecting the double calibrator pulses to the PD
measuring system in different ways.
Fig. 21: Comparison of double pulse diagrams for
direct injection, injection via measuring impedance
CPL, injection via coupling capacitor Cc
The Figure shows curves for 2 different PD bandpass filter
bandwidths, 300 kHz (blue curves) and 160 kHz (red
curves). The pulses were injected either directly (broken
lines), via the measuring impedance CPL (solid lines) or
via the coupling capacitor Cc (dash-point-lines). The thick
black line shows the result of the 3-band filter.
It can be seen that the differences of different injection
methods are not significantly high. This result may
indicate that useful double pulse diagrams can be created
directly with the PD instrument, not connected to the final
HV setup.
THE MEANING OF THE IEC 60885-3
The IEC 60885-3 is the attempt to make people aware of
the problem of negative superposition errors when making
PD measurements on HV cables. In the last
consequence, the well familiar value "Apparent Charge"
as an indicator for the quality of the DUT is lead to
absurdum, as this value Qiec can alter between zero and
200% of the "real" Apparent Charge (hypothetically
measured at the PD fault location on the cable length),
only depending on the cable length, the PD fault position
and strongly on the filter parameters of the PD instrument.
Jicable’11 – 19 – 23 June 2011, Versailles - France
Close and Return
C.3.4
THE PRACTICAL MEANING OF THE
IEC 60885-3
As a matter of fact, most of the cable manufactures
(besides some exemplary exceptions) do not apply the
IEC 60885-3. This standard needs a fundamental
maintenance to make it applicable for PD test engineers.
At the moment, the 2nd edition of the IEC 60885-3 is under
preparation by IEC TC20 WG16. Unfortunately, the
changes will only be of minor nature.
SUMMARY
The Apparent Charge level cannot be used to compare
PD measuring results of HV cables, as this value is
strongly influenced by the PD filter settings of the
measuring system as well as by the PD fault position on
the HV cable itself. By simply turning the cable around,
the Apparent Charge of a cable PD fault will be different.
This difference is slightly caused by the changed PD
propagation path and the resulting signal attenuation, but
mostly caused by unpredictable resonance phenomena of
the PD bandpass filter. Double pulse diagrams may help
to visualize the problem and to be aware of critical cable
length or critical PD fault positions on a cable. Finally, the
IEC standard 60885-3 describes this problem without
giving practical solutions to solve it.
The problem of negative superposition errors can be
solved either by the use of broadbanded PD bandpass
filters (e.g.: 3 MHz bandwidth) or by using sophisticated
algorithms combining 3 digital filters synchronously. As
using a broadbanded PD filter with 3 MHz bandwidth is
not covered by the standard IEC 60270, the use of
combined quasi-broadbanded PD filters with bandwidths
between 100 kHz and 400 kHz can be tolerated. With a
synchronous 3-band filter the negative superposition error
can be limited to -10%, as demanded in the IEC 60885-3.
REFERENCES
[1] IEC 60885-3: "Test methods for partial discharge
measurements on lengths of extruded power cable",
st
1 edition, 1988
[2] IEC 60270: "High-voltage test techniques – Partial
discharge measurements", 3rd edition, 2000
[3] IEC 60840: "Power cables with extruded insulation
and their accessories for rated voltages above 30 kV
(Um = 36 kV) up to 150 kV (Um = 170 kV) – Test
methods and requirements ", edition 3.0, 2004
[4] IEC 62067: "Power cables with extruded insulation
and their accessories for rated voltages above
150 kV (Um = 170 kV) up to 500 kV (Um = 550 kV) –
Test methods and requirements", edition1.1, 2006
[5] IEC 60034-27: "Off-line partial discharge
measurements on the stator winding insulation of
rotating electrical machines", 1st edition, 2006
[6] CIGRE WG 1.33: Guide 366: "Guide for Partial
Discharge Measurements in Compliance to
IEC 60270", 2008
[7] Omicron electronics: "MPD 600 – Product brief and
technical specifications", Klaus, Austria 2010
[8] ANSI/NEMA CC1-2009: „Methods of Measurement of
RIV of High Voltage Apparatus”, 2009