See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/255701107

Influence of lightning impulses to insulating

systems - PhD thesis - TECHNISCHE
UNIVERSITT GRAZ
THESIS MAY 2000

READS

187

1 AUTHOR:
Sonja Monica Berlijn
Statnett
36 PUBLICATIONS 111 CITATIONS
SEE PROFILE

Available from: Sonja Monica Berlijn

Retrieved on: 10 February 2016

TECHNISCHE UNIVERSITT GRAZ

DISSERTATION

INSTITUT FR HOCHSPANNUNGSTECHNIK
MIT VERSUCHSANSTALT

Influence of lightning impulses

to insulating systems
Sonja Berlijn

Dissertation
For the academic degree
Doctor of Technical science

Permitted by the
Faculty of Electrical Engineering
of the
Technical University Graz

Submitted by:
1. Expert & Supervisor: O.Univ.-Prof.Dipl.-Ing.Dr. Michael Muhr
2. Expert: Univ.-Prof.Dipl-Ing.Dr. Ernst Gockenbach
Arnhem, The Netherlands May 2000

Preface

Preface and acknowledgements

This doctor thesis describes the theoretical and experimental
investigations performed at KEMA High-Voltage Laboratory and at
the Technische Universitt Graz. The PhD project was started in
November 1997. At this time I was involved in an European project
and in two international working groups which dealt amongst others
with the measurement and evaluation problems of lightning impulse
voltages.
The work performed in the framework of the European project was a
good start for a PhD project and therefore, together with my professor
M. Muhr and my colleague G. Roelofs, I decided to take my PhD in
the area of tests with lightning impulse voltages.
I would like to take this opportunity to express my gratitude to all
who were involved in my projects and my education, to all who
supported me in writing this doctor thesis and to all of my friends and
family who were there to talk or to e-mail to. Of course there are
some persons to whom I would like to express my gratitude in
particular they are listed here below in random order.
All my colleagues at KEMA High-Voltage Laboratory I would like to
thank for the creation of a fine working environment, helping me with
the experimental investigations and for the interest they showed.
Also other colleagues from KEMA TNK and KEMA HPL, I would
like to thank for their help. KEMA, in particular, G. Roelofs, H.
Kempen, T. Klomp and B. Verhoeven, I would like to thank for
making this PhD project possible. G. Roelofs I would like to thank
also for his help, inspiration, support and for the comments he gave to
my work.
During the experimental investigations performed at KEMA I was
helped by M. Kvarngren from KTH. I have learned quit a lot from
him, technically and scientifically, and he became a very good friend.
I would like to thank him for his criticism, for his help and support.
My friend H. Scheefhals, I would like to thank for his patience, his
support, for listening to me all the time and of course for his advice.
The members of CIGRE 33.03, the joint task force SC12/WG33.03
and all other persons involved in the European project (especially
F. Garnacho, P. Simon, E. Gockenbach, P. Werle, K. Hackemack,
J. Blackett, M. Watts and P. Wong). I would like to thank for their
input, their ideas and their co-operation. Without their help this
project would not have been so successful.
I would like to thank my special friend J. Benda, for his help, his
support and his friendship during all the years.
J. Wetzer, F. Pesgens, M. Bongaerts, G. Buhre, R. Verhoeven and
H. Obers I would like to thank for their contributions to this PhD.
I would like to thank Dr. Kachler from Siemens, for providing
samples. W. Lick and G. Puckel from TU Graz, I would like to thank
for performing the investigations on these samples.

I dedicate this thesis to my parents, my

sister and her husband, my uncle Eddie,
my grandparents and my uncle Douwe.

Giuseppe Ryszard GehrardRen

Bruin Miriam Ronald Diane
Hans BerryOma Ernst Toshiji
Kurt Jean JackTonnie Jeroen Bill
Tonnie
TUE Ger Peter Fernando
HannekeJean Patrick Toine Godert
Marcel
Fred Evert Patricia Eddie Malcolm
Hauke MatthiasHannie
Klaus Albert
Trudy Sara
Silvia
Werner
George
Hiske Romy Edwin
EC
Siemens
Jos Frank Mats Cor Anna
Dick
Wouter Kimlie Diny
Mark Herman KEMA Wim
Harrie OrsoHenkMichael Wijnand
Eberhart Pieter
Wolfgang
Cees Rob Ton MarttiPascual
Jet
Theodor Susan RobertHughSven-Erik
Hans-Erik Herbert Martijn Jari
Sanne Marja
John Leo
Anders Paolo Loes
Fried
Sadique
Lars-Olof Anouscka Terry
Andr Martien Bob All I forgot

Thanks
The EC is thanked for the financial
support
of
the
project
Digital
measurement of parameters used for
lightning impulse tests for high voltage
equipment, contract no. PL95120-SMT4CT96-2132.

Last but not least I want to thank my two dogs Romy and Orso for
always being there, keeping me company and making me laugh.

Kurzfassung

Sonja Berlijn, Arnhem, May 2000

vi

Abstract

Abstract
Lightning strokes have caused a considerable number of failures of
high voltage equipment installed in medium and high voltage grids.
In order to prevent these failures several studies have been performed
and several measures have been taken. One of these measures is to
test high voltage equipment with lightning impulse voltages.
The measurement of the applied voltages and the evaluation of the
parameters characterising these lightning impulse voltages have to be
performed accurate and uniform all over the world, since high-voltage
equipment require large capital assets and they are exported all over
the world.

During tests of high voltage equipment with lightning impulse

voltages oscillations or/and overshoot may occur at the front or near
the peak of a lightning impulse voltage. The measurement of
lightning impulse voltages without oscillations or overshoot can
nowadays be performed with sufficient accuracy. However, problems
arise with chopped lightning impulse voltages, in case the lightning
impulse voltage has some small overshoot or oscillation and even
bigger problems arise when the lightning impulse voltage is a nonstandard lightning impulse voltage. The cause for these measurement
problems is that when small oscillations or/and overshoot occur, the
rules for the evaluation of the parameters (as described in the present
standard) are ambiguous. In case of non-standard lightning impulses
no evaluation rules are given in the present standard. Although the
problems existed when the evaluation was performed manually, they
were not recognised. The introduction of digital equipment and
evaluation software showed very clearly the problems and made the
problems needed to be solved.
Several solutions for the problems with the evaluation have been
proposed. Most of the solutions given up to now, based on the
present definition in IEC 60060-1, do not taken into account the
possibilities of modern (measuring) equipment and do not consider
the introduction of new insulating materials as for instance SF 6.
Besides this, the problems with the evaluation of parameters of
standard and non-standard lightning impulse voltages are dealt with
separately.
To find an ideal solution for the evaluation problems that is based on
the breakdown physics and that is applicable to both standard and
non-standard impulses, theoretical and experimental investigations
have been performed by the author at KEMA and by others at other
institutes. Tests were performed on different insulating materials.
The results of these investigations showed that a new approach was
necessary and a new evaluation method, using the k-factor function is
introduced.
When using this method, lightning impulse
measurements can be performed with a better accuracy besides this, it
is possible to use the method for both standard and non-standard
impulses.

vii

Kurzfassung

Keywords
lightning impulse voltage, evaluation methods, standard and nonstandard lightning impulses, k-factor function, test with lightning
impulses, IEC 60060-1

viii

Abstract

Kurzfassung
Blitzeinschlge in Mittel- und Hochspannungsnetze haben in der
Vergangenheit zu einer groen Anzahl von Fehlern an
Hochspannungsgerten gefhrt. Um diesen Fehlern vorzubeugen,
wurden einige Studien und verschiedene vorbeugende Manahmen
durchgefhrt.
Eine dieser Manahmen ist die Prfung der
Hochspannungsanlagen oder -gerte mit einer Blitzstospannung.
Die Messung der angelegten Spannung und die Auswertung der
Parameter, die die Blitzstospannung charakterisieren, mssen sehr
genau und in der ganzen Welt gleich ausgefhrt werden, damit die
Ergebnisse vergleichbar sind.
Bei Blitzstospannungsprfungen an Hochspannungsgerten knnen
Oszillationen und/oder berschwingungen im Anstiegsbereich oder
nahe des Scheitelwertes des Blitzstoimpulses entstehen. Die
Messung
der
Blitzstoimpulse
ohne
Oszillationen
und
berschwingungen kann heute mit zufriedenstellender Genauigkeit
durchgefhrt werden. Probleme bei der Auswertung entstehen bei
abgeschnittenen Blitzstoimpulsen oder wenn der Blitzstoimpuls
kleine berschwingungen oder Oszillationen aufweist.
Noch
schwieriger wird die Auswertung, wenn es sich um einen nicht
standardisierten Blitzstoimpuls handelt. Die Ursache fr die
Probleme beim Auswerten ist darin begrndet, da fr
Blitzstoimpulse mit berlagerten Schwingungen oder Oszillationen
keine eindeutigen Meverfahren in den Standards definiert sind. Im
Fall von nicht standardisierten Blitzstoimpulsen gibt es gar keine
Lsungsanstze in den aktuellen Standards.

In der Vergangenheit sind verschiedene Lsungsanstze fr diese

Probleme vorgeschlagen worden. Die meisten Lsungen basieren auf
den derzeitigen Definitionen in IEC 60060-1. Sie bercksichtigen
aber dabei nicht die Mglichkeiten der modernen Metechnik und die
Einfhrung neuer Isoliermedien, wie zum Beispiel SF6. Auerdem
werden die Probleme bei der Bestimmung der Parameter von
standardisierten und nicht standardisierten Blitzstoimpulsen bisher
separat behandelt.
Um einen optimierten Lsungsansatz fr die Bestimmung der
Parameter
zu
finden,
der
auf
den
physikalischen
Durchschlagsmechanismen beruht und der fr beide Impulsformen
Gltigkeit hat, wurden theoretische und experimentelle Untersuchen
bei der KEMA und anderen Instituten durchgefhrt. Dabei wurden
die Einfle unterschiedlicher Isoliermedien bercksichtigt. Die
Ergebnisse der Untersuchen haben gezeigt, da ein vllig neuer
Lsungsansatz und eine neue Berechnungsmethode, die eine k-Faktor
Funktion benutzt, bentigt wird. Mit dieser Berechnungsmethode ist
es mglich, hhere Genauigkeiten bei der Auswertung von
Blitzstospannungsprfungen zu erzielen. Gleichzeitig kann diese
Methode sowohl fr standardisierte als auch fr nicht standardisierte
Blitzstoimpulse angewendet werden.

Schlgwrter
Blitzstospannung, Auswertung, standardisierter und nicht
standardardisierter Blitzstoimpuls, k-Faktor Funktion, NennstehBlitzstospannungsprfung, IEC 60060-1

ix

Contents

Contents
1

Introduction _______________________________________________________________________ 1

Lightning impulses __________________________________________________________________ 3

2.1 Introduction ______________________________________________________________________ 3
2.2 Lightning ________________________________________________________________________ 3
2.3 Lightning impulse tests _____________________________________________________________ 4
2.4 Implementation of the standards _____________________________________________________ 12
2.5 International work ________________________________________________________________ 18
2.6 Current evaluation methods _________________________________________________________ 20

Disruptive discharge phenomena _____________________________________________________ 25

3.1 Introduction _____________________________________________________________________ 25
3.2 Insulation systems ________________________________________________________________ 25
3.3 Breakdown processes in materials ____________________________________________________ 27
3.4 Influence of lightning impulse voltage to the breakdown __________________________________ 34
Breakdown statistics and statistical techniques ______________________________________________ 39

Goal of investigations _______________________________________________________________ 43

4.1 Introduction _____________________________________________________________________ 43
4.2 Goals and expectations of investigations _______________________________________________ 45
4.3 Insulating systems to investigate _____________________________________________________ 46
4.4 Tests to perform on insulating materials _______________________________________________ 46
4.5 Translation of results to insulating systems _____________________________________________ 47
4.6 Questions/hypothesis ______________________________________________________________ 48

Investigations; the test set-up ________________________________________________________ 49

5.1 Introduction _____________________________________________________________________ 49
5.2 Generating circuit, combined generating circuit versus modified conventional circuit ____________ 49
5.3 Test set-up and test method used for the tests at KEMA ___________________________________ 53
5.4 Test set-up used of the tests at TU Graz _______________________________________________ 78

Investigations; the results ___________________________________________________________ 81

6.1 Introduction _____________________________________________________________________ 81
6.2 Test to be performed ______________________________________________________________ 81
6.3 Comparison measurements in air _____________________________________________________ 84
6.4 Investigations performed in oil at KEMA ______________________________________________ 89
Measurements performed in SF6 at NGC ___________________________________________________ 92
6.6 Measurements performed in PE at the Schering Institute __________________________________ 96
6.7 Measurements performed in air at LCOE ______________________________________________ 99
6.8 Measurements in oil-paper at TU Graz _______________________________________________ 102
6.9 Analysis of results, general discussion________________________________________________ 103
6.10
K factor hypothesis and linearity statement; influence of oscillation frequency and amplitude __ 104

Proposal ________________________________________________________________________ 113

7.1 Introduction ____________________________________________________________________ 113
7.2 Proposal for evaluation method _____________________________________________________ 113
7.3 Proposal for algorithms ___________________________________________________________ 119
7.4 Simulation with wave shapes that occur in testing _______________________________________ 119
7.5 Check of the proposal with the average k-factor function _________________________________ 121
7.6 K-factor in hardware versus k-factor in software ________________________________________ 125
7.7 Final proposal and discussion ______________________________________________________ 126
7.8 Open questions and experiments needed ______________________________________________ 127

Summary and conclusions __________________________________________________________ 129

xi

Contents

xii

List of terms, symbols and abbreviations

Lists of terms, symbols and abbreviations

Symbols
A
B
B
C
Ce
d
E
f
F
Fa
F(x)
H
k
K
k1
k2
Kt
L
O1
p
P
P
P(U)
R
R(t)
sr
ssa
ssg
td
tp
T
T
Tc
T1
T2
T30%
T90%
U
Um
Umc
Up
Urmc
Us
Ur
Ut

Semi-range value
Pressure [mbar]
Confidence limit [%]
Capacitance
Capacitance to earth
Overshoot duration
Electric field strength [V/m]
Frequency [Hz]
Force [N]
Assigned scale factor [1]
Distribution Function
Humidity [g/m3]
Factor by which the amplitude of the applied impulse should be
attenuated
Coverage factor
Air density correction factor
Humidity correction factor
Atmospheric correction factor
Inductance [H]
Virtual origin of a standard lightning impulse voltage
Pressure [bar or Pa]
Power [W]
Probability
Distribution function
Resistance [Ohm]
Residual Curve
Experimental standard deviation
Standard deviation of systematic contributions that have a
rectangular distribution
Standard deviation of systematic contributions that have Gaussian
distribution
Time delay between the origin of the impulse and the origin of the
superimposed oscillations [s]
Time to peak [s]
Temperature [C]
Students t factor
Time to chopping [s]
Front time of a lightning impulse voltage [s]
Time to half value of a lightning impulse voltage [s]
Time at which the lightning impulse is 30% of it peak value
Time at which the lightning impulse is 90% of it peak value
Overall uncertainty
Measured curve [V]
Mean curve [V]
Peak value of a lightning impulse voltage [V]
Reference mean curve [V]
Systematic contributions to the overall uncertainty
Random contributions to the overall uncertainty
Value of the test voltage [V]

xiii

List of terms, symbols and abbreviations

U10
U20
U50

U50(peak)
U50(RMC)
dU/dt
W
Z

(x;,2)
(x;,2)

10% disruptive voltage [V]

20% disruptive voltage [V]
50% disruptive discharge voltage. This is the prospective voltage
value which has a 50% probability of producing a disruptive
discharge on the test object [V]
50% disruptive voltage, taken from the peak voltage of the total
curve
50% disruptive voltage, taken from the peak voltage of the mean or
base curve
Impulse Steepness at each point
Kinetic Energy
Complex Impedance [Ohm]
Overshoot amplitude
Wave integral above a reference voltage
Density function
Distribution function

Terms and abbreviations

BL
Chopped lightning impulse voltage

CIGRE
Disruptive discharge

Flashover
IEC
Impulse

MLM
Non-standard lightning impulse
voltage
OPV
PE
PSM
Puncture
RMC
SD
SE
Sparkover
Standard lightning impulse voltage
Surge
TDG
VPV
XLPE

xiv

Base Line
a lightning impulse voltage during which a disruptive discharge
causes a rapid collapse of the voltage, which then falls to zero or
nearly zero, with or without oscillations
International Conference on Large High Voltage Electric Systems
phenoma associated with the failure of insulation under electrical
stress, in which the discharge completely bridges the insulation
under test, reducing the voltage between the electrodes practically
to zero. It applies to electrical breakdowns in solid, liquid and
gaseous dielectrics and combinations of these.
a disruptive discharge over the surface of a dielectric in a gaseous
or liquid
International Electrotechnical Commission
intentionally applied aperiodic transient voltage or current which
usually rises rapidly to a peak value and then falls more slowly to
zero.
Multiple Level Method
lightning impulses with large overshoots or oscillations
overshoot peak value [V]
Polyethylene
Progressive Stress Method
Disruptive discharge in a solid medium
reference mean curve
Standard Deviation
Single Exponential of Tail
disruptive discharge in a gaseous or a liquid medium
the lightning impulse voltage as defined in IEC 60060-1
transient occurring in electrical equipment or networks in service
Test Data Generator
virtual peak value
Cross-linked polyethylene

Introduction

1 Introduction
Lightning strokes have caused a considerable number of failures of
high voltage equipment installed in medium and high voltage
networks. In order to prevent these failures much research has been
performed and several measures have been taken. One of these
measures is to test high voltage equipment with lightning impulse
voltages to verify the withstand capability.
Lightning impulse voltages represent surges i.e. transients occurring
naturally in high voltage systems under operation. A full lightning
impulse voltage (without oscillations or overshoot) is, currently in
IEC-60060-1 [1], defined by its peak value Up, its virtual origin O1,
and its time parameters, the front time T1 and the time to half value
T2. Tests with lightning impulse voltages are made using a standard
lightning impulse having a front time of 1,2 s and a time to halfvalue of 50 s, described as a 1,2/50 s lightning impulse voltage.

During tests of high voltage equipment with lightning impulse

voltages, oscillations and/or overshoot may occur at the front or near
the peak of a lightning impulse voltage. It is being assumed that these
oscillations and/or overshoot can be of significance and right now a
maximum amplitude of 5% of the peak value is allowed [2].
The measurement of the applied voltages and the evaluation of the
parameters characterising these lightning impulse voltages have to be
performed accurate and uniform all over the world, since high-voltage
equipment require large capital assets and they are exported all over
the world. That is why a lot of effort is put into generating the
lightning impulse voltages as good as possible and into calibration
and maintenance of measuring systems used.
In the past the measurements of lightning impulse voltages were
performed with the aid of analogue oscilloscopes and the evaluation
of lightning impulse voltages was performed manually. The
achievable accuracy was influenced by the thickness of the trace of
the oscilloscope and the accuracy of the test engineer [3].
Nowadays the measurement of lightning impulse voltages is mostly
performed with digital oscilloscopes.
The evaluation of the
parameters is mostly done using computer programs.
The
measurement of lightning impulse voltages without oscillations or
overshoot can nowadays be performed with sufficient accuracy.
However, problems arise with chopped lightning impulse voltages, in
case the lightning impulse voltage has some small overshoot or
oscillation [4]. Even bigger problems arise when the lightning
impulse voltage is a non-standard lightning impulse voltage. The
cause for these measurement problems is that, when small oscillations
and/or overshoot occur, the rules for the evaluation of the parameters
(as described in the present standard) are ambiguous. In case of nonstandard lightning impulses no evaluation rules are given in the
present standard.
Several solutions for the problems with the evaluation have been
proposed. (see paragraph 2.6). Most of the solutions given up to now,
which are based on the present definition in IEC 60060-1, do not
taken into account the possibilities of modern (measuring) equipment
1

Introduction
and do not consider the introduction of new insulating materials as for
instance SF6. Besides this, the problems with the evaluation of
parameters of standard and non-standard lightning impulse voltages
are dealt with separately.
The most ideal solution for the evaluation problems will be a solution,
which is applicable to both standard and non-standard lightning
impulse voltages, which uses the possibilities of modern processing
techniques and which is founded on the breakdown behaviour of
insulating materials.
To find this most ideal solution or a solution based on the breakdown
behaviour of insulating materials or systems, theoretical and
experimental investigations were necessary.
Therefore this research was started. The main topics of the PhD
study of which the results are presented in this thesis are:
1. study of basic literature;
study towards tests with lightning impulses
study and evaluation of already existing evaluation methods
2. theoretical investigation towards the breakdown behaviour of
insulating materials and systems
3. experimental investigation towards the breakdown behaviour of
different insulating materials for different lightning impulse wave
shapes
4. proposal for a new evaluation method
Ad 1:
Besides a summary of the information found in the basic literature
also international developments will be explained. Because some of
the newest information is not available yet in official publications,
results will be presented of activities within CIGRE WG33.03 and
TC12/WG 33.03
Ad 2:
Besides the literature investigation performed at KEMA, also the
results are presented of a literature investigation performed in the
framework of the European project on Lightning Impulses.
Ad 3:
The experimental investigations on different materials were
performed at different locations and within different projects:

In the framework of the European project the insulating materials,

XLPE, SF6, air and oil were investigated.

At KEMA oil was investigated

At LCOE air was investigated

At NGC SF6 was invesigated

At the Schering Insitute XLPE was investigated

As part of this PhD project at the Technische Universitt Graz

configurations representative for transformer windings with
paper-oil insulation were studied.

Lightning Impulses

2 Lightning impulses
2.1

Introduction

In this chapter it is described why and how lightning impulse test are
performed. Attention is paid to the evaluation of parameters
characterising lightning impulse voltage, to the problems that arise
during evaluation and to the current state of the art to solve these
problems.

2.2

Lightning

It has always been the case that people are at the same time astonished
and afraid for lightning and thunder [5]. About 200 years ago (in
1754) it has been proved that lightning that occur during
thunderstorms is an electrical discharge. At this moment still not all
the aspects of thunderstorms are revealed although it is known that
they are caused by the complicated process of air flows. 60%-70% of
the lightning discharges occurs between or in clouds of different
charge. These cloud-cloud discharges can be dangerous to airplanes.
The most dangerous discharges for human beings and equipment are
the discharges between clouds and earth and earth and clouds (also
called lightning strokes) [6]. In Figure 1 a thundercloud is shown in
the stage before lightning occurs. In 90-95% of the cases, especially
in the Netherlands, the cloud is negatively charged at the bottom.
A lightning stroke is an impulse current. The voltage value and the
voltage shape of the stroke are determined by the impedance the
current meets on its way to ground and are therefore situation
dependant. Important factors of lightning strokes are the maximum
current value, the charge, the energy contents and the current
steepness (some typical values for the Netherlands are given in
Table 1) [6].
Table 1 Typical stroke parameters in the Netherlands
Parameter

Normal

High

Extreme

Max current value in kA

100

200

400

Charge in C

100

300

1000

Energy in MJ/Ohm

10

100

Current steepness in kA/s

50

200

300

2
Figure 1 An example of a thundercloud
(rain is not necessarily present)

The number of lightning strokes, the number of damage cases and the
losses due to lightning strokes vary from country to country. In the
Netherlands thunderstorms occur on 107 days per year and on average
about 2 to 3 lightning strokes occur per km2 per year.
Some extreme cases of damage can be mentioned, for instance strokes
that occurred at 13 July 1997 in New York caused a damage of
1 billion US Dollar and it was the reason that 10 million people didnt
have electricity for 20 hours.
Detailed information about the process of lightning, atmospheric
overvoltages and the influence to power grids is given in [5].

Lighting Impulses

2.3
2.3.1

Lightning impulse tests

Background of lightning impulse tests

Serious studies in high voltage engineering began in 1836. In 1886

the first transformer was patented. The invention of this transformer
opened the door to efficient and effective power systems working at
high voltages [2]. After 1895, when it was used in combination with
power prime movers, the use of electricity became more widespread.
Within a few years the practical use of electricity increased and after
the First World War the demand for electricity rose spectacularly [7].
The development of the equipment needed for the transportation of
electricity was rapid up to UHV (Ultra High Voltage) levels (132 kV
in the 1920s, 400 kV in the 1950s, 1100 kV in the 1970s).
Unfortunately, the equipment at these higher voltages was introduced
before the principles were fully understood. The first equipment
installed at 132 kV was prone to fail due to lightning strokes. As
mentioned in the previous paragraph, lightning strokes can cause
black outs and considerable damage. The studies towards the
lightning stroke failures in the 1920s led to a better understanding of
transient voltage distributions within the windings of transformers and
to the introduction of lightning impulse tests [2].

2.3.2

Applicable standards

Tests with impulses are designed to demonstrate the response of the

equipment to transients over a wide frequency range. Lightning
impulse voltages represent transients occurring naturally in high
voltage systems under operation.
To make sure that lightning impulse tests on high voltage equipment
are made in the most correct way and as uniform as possible, experts
have written several standards dealing with the details on lightning
impulse tests.
At least the following four general standards are applicable:
IEC 60060-1, High-voltage test techniques, part 1: General
definitions and test requirements, from 1989
IEC 60060-2, High-voltage test techniques, part 2: Measuring
Systems, from 1994
IEC 61083-1, Digital recorders for measurements in high-voltage
impulse tests, part 1: requirements for digital recorders, from 1991
IEC 61083-2, Digital recorders for measurement in high voltage
impulse tests, part 2: evaluation used for the determination of the
parameters of impulse waveforms, from 1996
IEC 60790, Oscilloscopes and peak voltmeters
In the next paragraphs the relevant parts of the standards applicable to
tests with lightning impulses will be explained in detail. A thorough
understanding of these subjects is necessary for understanding the
remainder of this thesis.

Lightning Impulses

2.3.3

Tests with lightning impulse voltages

When high voltage equipment is tested with lightning impulse

voltages, a number of lightning impulse voltages of certain amplitudes
and shape are applied. A lightning impulse voltage is, currently in
IEC 60060-1 [1], defined by its peak value and its time parameters.
Figure 2 shows the standardised lightning impulse voltage with the
peak value Up (U=1.0), the virtual origin O1, the front time T1 and the
time to half value T2.
Figure 2 Standard lightning impulse
voltage

The impulse is usually produced by an impulse generator consisting

essentially of a number of capacitors that are charged in parallel from
a direct voltage source and discharged in series into a circuit that
includes the test object.
Most lightning impulse tests are made using a standard lightning
impulse, having a front time of 1,2 s and a time to half-value of
50 s, described as a 1,2/50 s lightning impulse voltage or a standard
lightning impulse, that may be chopped to simulate the effect of a
protective gap flashover.
When test are done with lightning impulse voltages, transient
recordings of the applied voltage are taken and in case of testing
transformers also transient recordings of the current at the neutral end
of the winding under test are taken. In general, the object passes the
test if there is no evidence of complete or incipient failure from
audible indications or changes in the voltage or current records [2].
Of course, when applying standard lightning impulse voltages, some
tolerances on the parameters are allowed. If not otherwise specified
by the relevant Technical Committee, the following differences are
accepted between specified values for the standard lightning impulse
and those actually recorded:
Peak value
3%
Front time
30%
Time to half-value 20%

Lighting Impulses

2.3.4

Tests procedures

The lightning impulse voltage tests performed are withstand voltage

tests. In IEC 60060-1, four tests procedures (A, B, C and D) are
described for withstand voltage tests. The recommended test
procedure depends on the nature of the test-object.

Procedure A
Three impulses of the specified shape and polarity at the rated
withstand voltage level are applied to the test object. The
requirements of the tests are satisfied if no indication of failure is
obtained, using methods of detection specified by the relevant
Technical Committee (This procedure is only recommended for tests
on degradable or non-self-restoring insulation.).

Procedure B
Fifteen impulses of the specified shape and polarity at the withstand
voltage level are applied to the test object. The requirements of the
tests are satisfied if not more than two disruptive discharges occur in
the self-restoring part of the insulation and no indication of failure in
the non-self-restoring insulation is obtained by the detection methods
specified by the relevant Technical Committee.

Procedure C
Three impulses of the specified shape and polarity at the withstand
voltage level are applied to the test object. If no disruptive discharge
occurs the tests object has passed the test. If more than one disruptive
discharge occurs the tests object has failed to pass the test. If one
disruptive discharge occurs in the self-restoring part of the insulation,
then nine additional impulses are applied and if no disruptive
discharge occurs the tests object has passed the test. If any detection
of failure in a non-self-restoring part of the insulation is observed
with the detection methods specified by the relevant Technical
Committee during any part of the test, the test object has failed to pass
the test.

Procedure D

U10 U 50 (1 1,3z )
Equation I Relation between U10 and U50

For self-restoring insulation the 10% impulse disruptive discharge

voltage, U10, may be evaluated by using statistical tests procedures
described in appendix A of IEC 60060-1. These test methods permit
either direct evaluation of U10 and U50 or indirect evaluation of U10. In
the latter case U10 is derived from the U50 value using the relationship
given in Equation I. The relevant Technical Committee shall specify
the value to be assumed for the conventional deviation z of the
disruptive discharge voltage. For dry tests on air insulation, without
any other insulation involved, the per-unit value z = 0,03 can be used.
The test object has passed the test if U10 is not less than the specified
impulse withstand voltage.

Lightning Impulses
The following test methods can be used to evaluate U50:
the multiple-level method with n 4 voltage levels and m 10
impulses per level
the up-and-down method with m = 1 impulse per group and n 20
useful applications
To evaluate U10, the up-and-down method, with m = 7 impulses per
group and at least eight useful groups can be used.
In all cases the voltage interval U between levels should be
approximately 1,5 to 3% of the estimated value of U50.

2.3.5

Measurement of lightning impulse voltages

The measurement of the applied lightning impulse voltage is usually

done with an Approved Measuring System. A Measuring System is a
complete set of devices suitable for performing high-voltage or
impulse-current measurements. An approved Measuring System is a
Measuring System that is shown to comply with one or more sets of
requirements set out in IEC 60060-2 by an initial Performance Test
and Acceptance Test, successive Performance Tests and Checks and
inclusion of these results in a Record of Performance.

General requirements
The general requirements for an Approved Measuring System are
given on the next page, more detailed requirements and the methods
for Performance Checks and Tests can be found in IEC 60060-2:
to measure the peak value of full impulses with an overall
uncertainty within 3%
to measure the peak value of chopped impulses with on overall
uncertainty which is depending on the time to chopping as follows:
for front chopped impulses (0,5 s Tc 2 s) within 5%
for tail-chopped impulses (Tc 2s) within 3%
to measure the time parameters which define the waveform with an
overall uncertainty within 10%
to measure oscillations which may be superimposed on an impulse
to ensure that they do not exceed the permitted levels given in
IEC 60060-1
the scale factor shall remain constant over long periods. More
specific, the scale factor of the converting device and the
transmission system shall not vary by more than 1% for the ranges
of the ambient temperature and clearances given in the Record of
Performance
the measuring instruments shall for oscilloscopes and peak
voltmeters comply with IEC 60790 and for digital recorders with
IEC 61083-1.
Evidence that the Measuring System meets the requirements is given
in its Record of Performance.

Lighting Impulses

Measuring System for lightning impulse voltages

A Measuring System for the measurement of lightning impulse
voltages comprises the following components:

Figure 3 Evaluation rules for a Full

lightning
impulse
according
to
IEC 60060-1

converting device with the leads required for connecting this

device to the test object and the connections to earth
transmission system connecting the output terminals of the device
to the indicating or recording instrument with its attenuating,
terminating and adapting impedances or networks
indicating or recording instrument together with any connections
to the power supply
software or the procedure for the evaluation of the parameters

Indicating or recording instruments

In high-voltage laboratories world-wide different indicating or
recording instruments for the measurement of lightning impulse
voltage are being used, distinction can be made between the following
three types of instruments
peak voltmeters
analogue oscilloscopes
digital oscilloscopes or digital recorders
The evaluation of the parameters is done either manually or
automatic, depending on the type of indicating or recording
instrument and the availability of software. How the evaluation of the
three parameters (Up, T1, T2) is done precisely is explained in the next
paragraph.

2.3.6 Evaluation of the parameters characterising lightning

impulse voltages
When the applied lightning impulse voltage has been measured, the
parameters characterising this lightning impulse voltage have to be
evaluated according to the evaluation guidelines given in
IEC 60060-1 [1]. The evaluation of parameters is different for full
standard lightning impulse voltage, for lightning impulse voltages
chopped on the tail or on the front for non-standard lightning impulse
voltages. In the next three paragraphs the definitions in IEC 60060-1
will be explained in detail.

2.3.6.1 Evaluation of the parameters of standard lightning

impulse voltages
For full standard lightning impulse voltages the following parameters
have to be evaluated:
value of the test voltage, Ut (in some cases Ut is equal to Up, in
other cases not, see the explanation below)
front time, T1
time to half value, T2

Value of the test voltage

Figure 4 Evaluation rules for lightning
impulses with overshoot or oscillation

For a lighting impulse without oscillations, the value of the test

voltage is its peak value, see Figure 3 or Figure 2.

Lightning Impulses
With some test circuits, oscillations or overshoot may occur at the
peak of the impulse, see Figure 4 a) to d). If the frequency of such
oscillations is not less than 0,5 MHz or the duration of overshoot not
more than 1 s, a mean curve should be drawn as in Figure 4 a) and
b). For the purpose of measurement, the maximum amplitude of this
mean curve is chosen as the peak value defining the value of the test
voltage.
Overshoot or oscillations in the neighbourhood of the peak, measured
by a system according to IEC Publication 60060-2, are tolerated
provided their single peak amplitude is not larger than 5% of the peak
value.

Figure 6 Evaluation rules for a lightning

impulse chopped on the front

Front Time T1
The front time T1 of a lighting impulse is a virtual parameter defined
as 1,67 times the interval T between the instants when the impulse is
30% and 90% of the peak value Up, (points A and B, Figure 3).

Time to half-value T2
The time to half-value T2 of a lighting impulse is a virtual parameter
defined as the time interval between the virtual origin O1 and the
instant when the voltage has decreased to half the peak value.

Figure 5 Evaluation rules for a lightning

impulse chopped on the tail

2.3.6.2 Evaluation of the parameters of chopped lightning

impulse voltages
For chopped lightning impulse voltages the following parameters
have to be evaluated:
value of the test voltage Ut
front time T1
time to chopping Tc
The value of the test voltage and the front time have to be determined
using the definitions given in the previous paragraph.

Time to chopping
The time to chopping Tc is a virtual parameter defined as the time
interval between the virtual origin O1 and the instant of chopping.
The instant of chopping is that at which the rapid collapse of voltage
which characterises the chopping first occurs.
The instant of chopping is different if the impulse is chopped in the
front or in the tail, see Figure 6 and Figure 5.

2.3.6.3 Evaluation of the parameters of non-standard

lightning impulse voltages
For other impulse shapes then the ones explained in the previous two
paragraphs (see for example Figure 7 e)-g)) the relevant Technical
Committee shall define the value of the test voltage taking into
account the type of the test and test object.
Figure 7 Examples of
lighting impulse voltages

non-standard

Lighting Impulses

2.3.7

Validation of software, IEC 61083-2

When the parameters characterising lightning impulse voltages are

evaluated automatically with the aid of software, this software has to
be validated according to IEC 61083-2. In IEC 61083-2 the test
procedures to be applied to assess the accuracy of software used to
process and read the records of impulses and calibration signals are
specified. More specific, it:
defines the terms specifically related to digital processing
establishes the tests which are necessary to show that the software
is compatible with the requirements of IEC 61083-1 and
IEC 60060-1
specifies limits on estimated values of parameters of the reference
waveforms
gives the requirements for the record of performance
The compatibility with the standard can be demonstrated with
software validation. The software is validated using the TDG (test
data generator). The TDG is a computer program that generates
reference waveforms with specified parameters.
Software may be validated for the evaluation of one or more sets of
impulse parameters (e.g. T1, Up). Any parameter for which the
software is validated shall be evaluated for all reference waveforms in
each selected waveform group. The values of the parameters
determined by the software under test are compared with the limits
given in IEC 81083-2, table 2. The software under test passes the test
for each set, if all values of that parameter set are within the specified
limits. For a full lightning impulse a deviation of around 1% is
allowed.
In the Record of Performance information about the software should
be given as well as the list of groups of impulses for which the tests
were performed and the list of parameters for which the software was
tested and passed.

2.3.8

Measurement uncertainty

The measurement of the parameters characterising a lightning impulse

voltage using approved Measuring Systems, have to be performed
within the uncertainty limits mentioned in IEC 60060-1 (3% or 5%
for the peak value and 10% for the time parameters).
In this paragraph it is explained how the measurement uncertainty of a
measurement made with a Measuring System should be calculated.
An example of a measurement with its
uncertainty is:
1080 kV 20 kV
Estimated confidence level not less
than 95%

10

2.3.8.1 General
Uncertainty is a statement of [9]:
the limits (U) of the range of values within which the true value
of a measurement is expected to lie in relation to the recorded
result and
the probability of the true value lying within these limits; this
probability is expressed as the confidence level (B)

Lightning Impulses
In most measurement, the overall measurement uncertainty will result
from a combination of several contributions that are classified into
two categories according to the method used to evaluate their
numerical values [8]:
systematic contributions
random contributions
The overall uncertainty is calculated from these contributions.
As a high-voltage laboratory, one has to comply with several
standards dealing with measurement uncertainty. These standards are:
IEC 60060-2 annex H [9]
STL-guide [10]
EAL-2
These standards are in general identical, however they differ on a few
points, mainly the calculation of the random contributions. At KEMA
High-Voltage laboratory the measurement uncertainty is calculated
with the method described in IEC 60060-2, since this is the most
precise. Therefore only this standard is explained here.

2.3.8.2 IEC 60060-2 annex H

In this standard the two categories of contributions to the uncertainty
are also:
systematic contributions to the overall uncertainty, Us
random contributions to the overall uncertainty, Ur
The systematic and random contributions to the uncertainty are
calculated separately. The overall uncertainty is calculated from
these two contributions by the formula given in Equation IV.

U U s2 U r2
Equation IV Calculation of the overall
uncertainty out of the two contributions,
when Us and Ur are calculated at the same
confidence level

s sa

a
3

Equation II standard deviation of a

rectangular distribution

2
ss s sa2 ssg

Equation V standard deviation of the

systematic contribution

2.3.8.3 Systematic contributions

According to IEC 60060-2 annex H, systematic contributions are
those that are not evaluated statistically but are estimated by other
means. Two examples are the uncertainty of a calibration and the
resolution of an instrument.

2
2
U s k ss k ssa
ssg

There are two types of systematic contributions:

rectangular (ssa); these systematic contributions have a rectangular
distribution, that is, any measured value between the estimated
limits (a, were a is the semi-range value) is assumed to be equally
probable.
Gaussian (ssg); it is assumed that these systematic contributions
have a Gaussian distribution
The standard deviation for all systematic contributions (s s) is
calculated from these two systematic contributions according to
Equation V.

Equation III systematic contribution to the

uncertainty

The systematic uncertainty is calculated from this standard deviation

by Equation III. In which k is the coverage factor. The coverage
factor k is 2 for a confidence level B of 95%.

Equation VI experimental standard

deviation, where:
n is the number of measurements
xi are the measured values for i = 1 to n
xm is the mean of the measured values

sr

1 n
( xi xm ) 2

n 1 i 1

11

Lighting Impulses

Ur

t sr
n

2.3.8.4 Random contributions

Random contributions are those that are derived statistically from a
repetitive measurement and, being random, will usually be found by
measurement to tend a Gaussian distribution. Each random
contribution is characterised by the experimental standard deviation
(sr) of the sample of measured values, see Equation VI. The
contribution to the uncertainty is then given by Equation VIII. The
students t factor t is 2,26 for 10 measurements.

Equation VIII random contribution to the

uncertainty, where:
t the students t factor

Ur

k sr
n2

Equation VII calculation of the uncertainty

for n220 with the uncertainty obtained
for n20

U k

U
1
ai2 i

3
ki

t sr

k n

If a value of sr has been established from a large number of

measurements (e.g. n 20) for a measuring system to which no
significant change is made, then the uncertainty in a subsequent single
or repeated measurement (n2) is given in Equation VII.

2.3.8.5 Overall uncertainty

The overall uncertainty is calculated from the random and systematic
contributions to the uncertainty using Equation IV. A more general
form for the overall uncertainty U is given in Equation IX.
This implies that the overall measurement for one measurement can
be calculated with Equation X.

Equation IX general formula for the

calculation of the overall uncertainty

2.4
2.4.1

U k

U
1
ai2 i

3
ki

t sr

Equation X measurement uncertainty for

one measurement

Implementation of the standards

Generation of lightning impulse voltages

The generation of a lightning impulse voltage with high amplitude is

generally done with a lightning impulse generator, also known as a
Marx generator. The principle of such a generator is parallel charging
and series discharging of its capacitors, see Figure 8. The peak value,
the front time and the time to half value of the generated lightning
impulse voltages are a function of the impedance of the load (object
under test), the inductance of the whole test circuit and the resistances
and capacitances of the generator [2].
For some arrangements, e.g. large generators, high inductive load or
low resistance of the test circuit, oscillations or/and overshoot occur
at the front or near the peak of an lightning impulse voltage. An
example of such a case is given in Figure 10. It is being assumed that
these oscillations and/or overshoot can be of significance and a
maximum amplitude of 5% of the peak value is allowed. Therefore
effort is put into generating lightning impulse voltages without
overshoot and/or oscillations [11].
For some test arrangements the oscillations and/or overshoot can be
reduced by adapting the test circuit. When adapting the test circuit, a
compromise have to be made between the amplitude of the
oscillations and/or overshoot and the permissible front time (tolerance
30%) and time to half value (tolerance 20%). For some test
arrangements a standard lightning impulse can not be generated [2].

Figure 8 Schematic diagram of an impulse

generator

12

Lightning Impulses

2.4.2

Analogue practise of evaluating parameters

In the past, but also in some cases nowadays, the measurement of

lightning impulse voltages is performed using analogue oscilloscopes.
The evaluation of lightning impulse voltages was performed manually
using pencil and ruler. The achievable accuracy was influenced by
the thickness of the trace of the oscilloscope and the accuracy of the
test engineer and is limited [3], see also Figure 9. Since the
evaluation had to be performed manually, also the number of
parameters that can be evaluated is limited. The parameters described
in IEC 60060-1, are the most practical ones for manual evaluation.
According to IEC60060-1, the definitions of the parameter apply
either to impulses without oscillations or overshoot or to the mean
curve drawn trough the oscillations and overshoot. Because of the
natural filtering of the measurement system, only oscillation
and/overshoot of higher amplitude and low frequency could be seen.
Besides this, the thickness of the oscilloscope trace in conjunction
with the thickness of the line drawn as the mean curve makes it
almost impossible to discriminate between different possible mean
curves [12] in case of oscillations or overshoot within the tolerance
limits. In case of large oscillations or overshoot or non-standard wave
shapes (e.g. transformer tests) discussing took place between the test
engineers to draw the best mean curve. Over a period of sixty years
this manual procedure was the best way to evaluate the parameters of
impulses measured with an analogue oscilloscope.

2.4.3

Figure 10 Typical lightning impulses in

transformer testing

Present-day methods of evaluating parameters

Nowadays the measurement of lightning impulse voltages is mostly

performed using digital oscilloscopes and measuring systems with a
higher bandwidth. Disadvantage of this is that small overshoot and
oscillations can easily be made visible. Advantage of this is that the
evaluation of the parameters can be done using computer programs.
The advantage of using computer programs for the evaluation of
parameters is that the evaluation is performed more uniformly and
faster. However, software packages have difficulties with the
implementation of evaluation rules made for manual evaluation, since
they need more precise instructions. Because of that in some cases,
for instance non-standard lightning impulse voltage, still a manual
evaluation is used.

2.4.4

Results of recent intercomparisons

To maintain international traceability, from time to time international

intercomparisons are organised.
A recent international
intercomparison (1995-1996) [4] showed that the requirements of IEC
60060-2 can be met for smooth standard lightning impulse voltages.
The values of all participating laboratories but one agreed to each
other within 1% for the peak value and within 5% for the time
parameters.
Slightly higher deviations of comparison results,
especially for T1 were obtained when short front (0,84 0,95 s)
impulses were compared. Although an attempt was made to define
the chopped impulses clearly, these comparison results were heavily
influenced by different impulse shapes used in different laboratories.

Figure 9 Example of an analogue

oscillogram and manual evaluation

13

Lighting Impulses
For front chopped impulses, measurement uncertainties in the peak
value and the time to chopping of larger than 6% were seen.
A comparison of software showed that, by using different methods a
large spread of some parameters determined from different mean
curves has been obtained [13]. In paper [17] different methods, all
acceptable to the standard, have been compared and differences from
each other up to 1,7% for the peak value, 18% for the front time and
3% for the time to half value have been obtained when case no 11 of
the TDG of IEC 61083-2 (see Figure 11) was analysed.
Figure 11 Case no. 11 of the TDG of
IEC 61083-2

Both comparisons showed that there are problems with the evaluation
methods given in IEC 60060-1. It seems right now that the hardware
is good enough for the accuracy required nowadays, but that the
software or evaluation methods are the major contribution to the
measurement uncertainty due to the ambiguous definitions

2.4.5

Present-day evaluation problems

The procedure for the evaluation of parameters as described in IEC

60060-1 is given in Figure 12. As can been seen from Figure 12 it is a
very complex and incomplete procedure. The evaluation guidelines,
are ambiguous for standard lightning impulse voltage or not given at
all for non-standard lightning impulse voltages, for both manual and
automatic evaluation.

Figure 12 Flow diagram of evaluation

procedure

Full Lightning

Impulse

Determination

of Mean Curve
No

Determine

Oscillations or
Overshoot ?

Non Standard
Lightning
Impulse

Yes

Special Case

T 1 and T 2

T 1 =1,2 s 30%
T 2 =50 s 20%?
No
Yes

Yes

Standard Lightning

Impulse

No
TC determines
Amplitude
> 5%?

Test voltage

Yes
Determine

T 1 and T 2

No

Freq > 0,5 MHz

Dur. 1 s?

Yes

Test voltage = Peak voltage

of mean curve

No
Test voltage = Peak voltage

14

Evaluate

T 1 and T 2

Lightning Impulses
One of the main problems with the evaluation of parameters
nowadays is the determination of the mean curve. The standard
allows different mean curves, resulting in different values for the
parameters characterising the same lightning impulse voltage [12].
The reason for the ambiguousness in the standard is that the
evaluation guidelines given in IEC 60060-1 were introduced in the
time there were mainly analogue oscilloscopes were available and the
evaluation of the parameters had to be done manually using pencil,
ruler and engineering judgement.
By eye it was decided if overshoot or oscillation appeared and in case
of overshoot and oscillation appeared outside the allowed limits, a
mean curve was drawn. The thickness of the oscilloscope trace in
conjunction with the thickness of the line drawn as the mean curve
made it impossible to discriminate between different mean curves
and therefore the problems with the evaluation guidelines did not
arise. Since then a lot has changed.
in 1994 a new standard, IEC 60060-2 ed. 1994, was introduced in
which the requirements towards measurement uncertainty for the
measurement of lightning impulse voltages are more strict than
they were before. The cause for this is that the interest of
manufacturers and purchasers of high voltage equipment are
higher and because the measuring technique has been improved,
the need for the reproducibility and traceability of the test results
has increased, because of the interest of manufacturers and
purchasers of high voltage equipment and because of the
introduction of EN 45001. The results of the evaluation of the
parameters should be person independent, it should not be possible
that the results are influenced by clients and one should be able to
calibrate the evaluation methods used,
digital recording instruments (digitizers and digital oscilloscopes)
have been introduced and nowadays the evaluation of the
parameters is mainly done by software. The software that
evaluates the parameters is used to save on the one hand time and
on the other hand to be sure that the evaluation of the parameters is
reproducible and traceable. A negative, but also positive side of
using software is that this software needs clear algorithms for the
evaluation of parameters. An advantage of using software for the
evaluation of parameters is that the engineering judgement and the
experience of engineers can be put into the computer (expert
systems) and that it permits a wider range of parameters than was
previously practical.
The three changes mentioned above, are the reason that the
requirements to the evaluation guidelines for standard lightning
impulse voltages have increased and that the need for evaluation
guidelines for non-standard lightning impulse voltages arose. They
are the reason that the problems with the interpretation of IEC 600601 have increased.

15

Lighting Impulses

2.4.6

Problems with the interpretation of IEC 60060-1

Independent whether manual or digital evaluation methods are used,

there are some problems with the interpretation of IEC 60060-1. The
following list presents some of the problems associated with the
definitions and interpretation of the standard. This list certainly not
comprises all of the questions to be resolved, but most of them are
probably included [14].
What is the definition of the mean curve, especially for impulses
with both overshoot and oscillations?
How to determine the T30% and T90% when there are oscillations
close to these times (see case no. 11 in IEC 61083-2)?
For the calculation of the front time, T1, the peak value should be
used as reference for the 100% value of the impulse. When you
have overshoots or oscillations, which peak value shall be used,
the peak value of the original impulse or the peak value of the
mean curve?
How should the amplitude and duration or frequency of
overshoots and oscillations be calculated?
How should a non-standard wave shape, for instance those that
occur during real testing, Figure 7 or Figure 10, be evaluated?
How to know if they are inside the tolerances of IEC 60060-1?
What is the physical background that supports the magical
0,5 MHz value for the frequency of oscillations and the 1 s
value for the duration of the overshoot?
Are the parameters described in IEC 60060-1 really the
parameters that characterise the breakdown process? Could other
parameters describe this process better?
Should the same parameters be used for all materials (XLPE, SF6,
oil, vacuum, etc.)?

2.4.7

Current practise in laboratories

In order to know how the high-voltage laboratories around the world

are facing the evaluation problems and to note which evaluation
methods are used, two different questionnaires (one for manufacturers
and testing laboratories and another one for calibration laboratories)
were prepared in the framework of the European project.
In the questionnaire questions are present about details of the test
circuit that are relevant when oscillations and overshoot are present.
Besides this, questions were asked how these laboratories evaluate
voltage and time parameters in the cases of full and chopped impulses
with oscillations superimposed (f 0,5 MHz) and/or overshoot
(d 1 s).
Only manufacturers and testing laboratories were asked about the
possible influence on the breakdown behaviour of dielectric materials
of these overshoots and/or high frequency oscillations with small
amplitude superimposed to the lighting impulse. In the following
paragraph the results of the questionnaire concerning the evaluation
are presented. In [14] more detailed information can be found.

16

Lightning Impulses

2.4.7.1 Results of the questionnaires

23 manufacturer laboratories, 8 testing laboratories and 12 calibration
laboratories all over the world answered the questionnaires. The main
conclusions are summarised in the next three paragraphs, detailed
information can be found in [15]. In general it can be stated that the
problems experienced in testing laboratories, laboratories of
manufacturers and calibration laboratories are greater than expected.

2.4.7.1.1

Manufacturers laboratories

For full impulses:

the majority of manufacturers of power transformers and cables

use the maximum value as the test voltage (not in accordance with
IEC 60060-1), while the majority of manufacturers of other high
voltage equipment considers the peak value of the mean curve
the majority of the laboratories determine T1 from the original
curve, using the peak value of the mean curve as the 100% value
only some manufacturers of transformers determine T 1 from the
original curve, using the peak value of the original curve as the
100% value
For chopped impulses:
for the two methods to determine the test voltage (maximum
value, or the peak value of the mean curve) the majority of
laboratories uses the maximum value
For the questions to the possible influence of oscillations and
overshoot:
about 30% of the manufacturer laboratories supplied information
of the possible influence on the breakdown behaviour of different
dielectric materials, when overshoots and/or high frequency
oscillations are superimposed on the lightning impulse. There was
no agreement between the received answers.

2.4.7.1.2

Testing laboratories

For full impulses:

all laboratories consider the peak value of the mean curve as the
test voltage, which is in accordance to the standard IEC 60-1
the majority of the laboratories use a mean curve that removes the
oscillations to determine the time parameters
For chopped impulses:

almost all laboratories consider the maximum value as the test

voltage
the majority of the laboratories consider the time to the point
when the voltage decreases suddenly as the time to chopping
The answers from testing laboratories are in a better agreement than
the ones of manufacturer laboratories, and are also in better
agreement with IEC 60060-1.

17

Lighting Impulses

2.4.7.1.3

Calibration laboratories

For full impulses:

half of the laboratories consider the peak value of the mean curve
as the test voltage and the other half considers the maximum
value as the test voltage
the majority of the laboratories use a mean curve that removes
oscillations to determine the time parameters
Calibration laboratories propose also some following alternative
methods for evaluating lightning impulses:
Method 1:
Comparison of the wave shapes of the reference
measuring system and the approved measuring system being
calibrated.
Method 2:
Use of a tolerance band. The tolerance band is
determined taking into account the uncertainty of the reference
measuring system.
For chopped impulses:

2.5

the majority of the laboratories considers the maximum value as

the test voltage
there is no general agreement on how to determine the time to
chopping, although in the majority of the answers the time to
chopping is considered to be the time when the voltage decreases
suddenly

International work

In this paragraph the state of the art will be given of the international
work that is being carried out at this moment on the field of the
evaluation of parameters characterising lightning impulse voltages.
Basically there were two approaches, one approach tried to solve the
problems in the evaluation of the standard lightning impulse voltages
and the other one tried to solve the problems with the evaluation of
the non-standard lightning impulse voltages. The author introduced a
third one.

2.5.1 Current status of evaluation guidelines for standard

lightning impulse voltages
The problems with the evaluation of parameters have been recognised
by CIGRE WG 33.03 and within CIGRE WG 33.03 the evaluation
guidelines for parameters characterising standardised lightning
impulse voltages as given in IEC 60060-1 are being discussed for
several years now. Within the CIGRE WG 33.03, even a special task
force has been given the assignment to investigate the evaluation of
parameters.
The main problem is nowadays considered to be the definition of the
mean curve. When using manual evaluation the mean curve is
drawn by hand, when using software for the evaluation the mean
curve is some mathematical function.
One thought that when one was able to find an unambiguous
mathematical function for this mean curve the evaluation problems
of the parameters characterising standardised lightning impulse

18

Lightning Impulses
voltages would be solved. The definition of the mathematical function
of the mean curve is however not so simple and people have
proposed also different solutions (see Chapter 2.6). Besides new
methods for calculating the mean curve also other methods like the
tolerance band have been introduced. The disadvantage of most of
the proposals explained in Chapter 2.6 is that these do not solve the
problem for non-standard lightning impulse voltages.

2.5.2 Current status of evaluation guidelines for nonstandard lightning impulse voltages
In September 1996 CIGRE WG33.03 and IEC SC 12 proposed to
formate a joint task force with SC 12. The objectives of this joint task
force are:
to set the parameters to be measured and the limits of uncertainty
which should be achieved
to develop and evaluate a Test Data Generator, similar to the one
used in IEC 1083-2, but for lightning impulse voltages that
represent those which can occur in transformer tests
preparation of a report from CIGRE 33.03 to IEC TC42, providing
technical basis for a new IEC publication
At the first meeting of the joint task force a start was made to list the
parameters which are considered to be either essential, or useful, to
determine the result of a test with lightning impulse voltages on a
transformer (see Chapter 2.6). The consideration of these parameters
was based on the breakdown processes due to the effect of the applied
lightning impulse voltage. It was taken into account that the use of
digital recorders and the processing facilities now available permits
the measurement of a wider range of parameters than has been
practical before [16].
This proposal has been discussed in the relevant committees and a
Test Data Generator is being made. The join task force has been set
on hold, because it was found best to wait for the results from the
European research project, that is discussed in the next paragraph.

2.5.3

PhD and European research project

According to the author the two approaches were too much based on
mathematical models and assumptions. When maintaining these two
approaches a solution or two solutions that would be accepted by the
many parties would never be found. To solve the problems with the
evaluation for both standard and non-standard lightning impulse
voltages as discussed in CIGRE WG 33.03 and the joint taskforce
TC12/WG33.03, the relevancy of the parameters had to be proven.
This required research. Therefore the author, as an employee of
KEMA Nederland BV, initiated to propose, together with FFII/LCOE,
the Schering Institute and NGC, to the European community DG XII
in the SMT fourth framework, a research project. The title of this
project is Digital measurement of parameters used for lightning
impulse test for high voltage equipment. The project has started at
January 1 1997 and the author was the project leader and thought
about what to do in the project. The duration of the project was two
and a half years.

19

Lighting Impulses
The aim of this project was to:
define one or more sets of parameters to characterise lightning
impulse voltages
write a report in which the relevancy of the proposed parameters
are explained
establish unambiguous algorithms to calculate these parameters
write a proposal for the relevant parts of IEC 60060-1 and
IEC 61083-2
At KEMA the author was responsible for carrying out the work, she
was the supervisor of the students involved in the project, built the
test set-up together with M. Kvarngren and carried out most of the
investigations and tests herself.
Besides the work performed in the framework of this European
project, more work has been performed. The author made an
additional literature investigation and was responsible for and
evaluated the measurement results of the investigations on paper-oil
samples provided by Siemens made at TU-Graz.
All the work explained before in this paragraph is presented in this
thesis.

2.6

Current evaluation methods

In this paragraph an overview is given from the current

automatic/mathematical evaluation methods.
The current
mathematical evaluation methods and the developments in these can
be divided into two categories:
evaluation methods using IEC 60060-1 parameters and its
definitions
evaluation methods using other parameters and definitions
In [17] an overview is given from the methods available, which will
be summarised here and is supplemented there where necessary. At
the end of this paragraph some conclusions are drawn.

2.6.1

Methods based on IEC 60060-1

The definitions in IEC 60060-1 apply to impulses without oscillations

or overshoot and to the mean curve drawn through the oscillations or
overshoot. To decide whether or not the parameters have to be
evaluated from the original curve or the mean curve, a mean curve has
to be drawn and the frequency or the duration of the overshoot have to
be determined. In other words, the frequency or the duration of the
residual curve, R(t), has to be analysed. R(t) is the difference between
the original curve W(t) and the mean curve M(t). Another possibility
is to determine the main frequency directly by means of FFT.
The first step in the analysis is to detect and to remove oscillations,
second step in the evaluation is the detection and removal of the
overshoot. There are different methods available for determining
oscillations and overshoot. They will be explained separately.

20

Lightning Impulses

2.6.1.1 Methods for determining oscillations

There are two types of mean curves that can be used for determining
whether or not the parameters shall be calculated from the mean
curve:
Global mean curves
Local mean curves
Regardless of the criterion applied it is necessary to use a numerical
method to calculate the mean curve. Many options have been
developed, the majority on the basis of fitting to mathematical models
and others by means of digital filtering.

2.6.1.1.1

U mc p1 t p2 t p3 t p4
3

Fitting to mathematical models

Equation XI third degree polynomial

Mathematical models most frequently used for determining mean

curves and some newly introduced ones are:
straight line
circumference arcs [18]
third degree polynomials, see Equation XI [19], [20], [21]
double exponential functions, see Equation XII [22], [23], [24],
[25]
quadruple exponential functions, [26], [27], [28]
double exponential functions with sin and/or cos terms, see
Equation XIV [29]
tension splines [30]
wavelet packett transform [31]

2.6.1.1.2

U mc A e t e t

Equation XII double exponential function

Digital filtering

There are two methods of digital filtering which have been applied for
the evaluation of parameters of impulses:
moving window filtering [32], [33]
low pass FIR filtering [34]

2.6.1.1.3

F U m i U c i
i 1

Choosing the reference mean curve

Before analysing the R(t) it is clear that the best reference mean curve
should be chosen, preferably by a mathematical procedure.

Equation XIII Least mean square error of

the measured impulse Um and the mean
curve Umc

There are several methods for determining the best RMC (reference
mean curve), this is to minimise the:
absolute mean value of the residual curve within a specific time
interval (MP)
root mean square of the residual curve within a specific time
interval (RMSP)
least mean square error between the measured impulse and the
RMC (LMS), see Equation XIII [29]
Also generetic algorithm models have been used to select the
reference mean curve [35].

U mc A e t e t B 1 cost e t C sint e t

Equation XIV double exponential function

with sin and cos terms

21

Lighting Impulses

2.6.1.2 Methods for determining overshoot

Before overshoot can be determined, it is useful or even necessary to
remove in advance any oscillation or superimposed noise. An
overshoot can be detected analysing the peak zone of the lightning
impulse voltage. There are two methods for detecting overshoot:
fit the tail with a single exponential function. If the impulse is
above the single exponential function that fits the tail of the wave
there is an overshoot, other wise there is no overshoot
compare the lightning impulse voltage with a specific mean curve.
If the lightning impulse voltage is above the reference mean curve,
there is an overshoot, otherwise there is no overshoot
Both method have some disadvantages, it is said that with the first
method an overshoot can exist below the single exponential function,
with the second method, the result depends on the specific mean
curve used.
The two methods result into three possibilities [36], as shown in
Figure 13. In [17] some alternative methods for determining the
duration of an amplitude of the overshoot are given, using the virtual

Figure 13 Three different possibilities to

determine the overshoot duration

peak value.

2.6.2

Other methods

Since the methods investigated do not show acceptable agreements,

two different methods have been proposed:
the method using a tolerance band
introducing new parameters

2.6.2.1 Tolerance band

The major problem with the method described in the previous
paragraph is, that in any case a mean curve has to be drawn and that
different mean curves can result in different values for the parameters.
A method that avoids the use of a mean curve is the tolerance band
method. Inside this tolerance band any real standard lightning
impulse voltage should be found. This band of a 1,2/50 impulse must
be obtained from the values for the tolerances given in IEC 60060-1.
In Figure 14 an example is given of a tolerance band for a 1,2/50
lightning impulse voltage. This tolerance band method can easily be

22

Lightning Impulses

Figure 14 Tolerance band for a 1,2/50 lightning impulse voltage (on the left the zoom of the full tolerance band on the right)

calculated with a computer and requires less computing time than the
calculation of reference mean curves.

2.6.2.2 New parameters

During lightning impulse tests with transformer non-standard wave
shapes are measured, see Figure 10. For these wave shapes no
evaluation methods are given. Therefore within the joint taskforce
SC12/WG33.03, a possible set of parameters to be measured during
impulse test on transformers were specified and defined. This set of
parameters has to be proven relevant by the transformer industry and
by experimental investigations.
The set of parameters consists of parameters that are needed for tests
with other types of apparatus and parameters that are required for
transformer tests only [37].
parameters needed for tests on other types of apparatus
base level
origin
virtual origin
peak value
time to chopping
average rate of rise (front time)
parameters required for test with transformers
extreme value of a full impulse
extreme value of the undershoot of a chopped impulse
time to extreme value
time interval Tk
wave integral
average rate of voltage collapse during chopping

2.6.3

Conclusions

The present rules indicated in IEC 60060-1 are insufficient and many
different interpretations are possible. In addition, very high accuracy
is required for the present definition and for the set of the limit of the
overshoot or oscillation. Therefore there are two possibilities, either
to give better and more useful definitions of the parameters of interest
or to completely change the approach [17]. In case of the first option
better and clearer requirements should be included in the standard to
determine the parameters of an impulse. But independent of the
method chosen, more extensive discussion and experimental evidence
is needed to establish the requirements for the evaluation of
parameters characterising lightning impulse voltages to be included in
the revision of IEC 60060-1.

23

Lighting Impulses

24

Breakdown

3 Disruptive discharge phenomena

3.1

Introduction

Before the goal of the investigations will be explained and the tests
performed and their results will be described, a theoretical
investigation towards the disruptive discharge behaviour of insulating
systems, or better materials, will be presented.
Disruptive discharges (also called breakdowns) are phenomena
associated with the failure of insulation under electrical stress, in
which the discharge completely bridges the insulation under test,
reducing the voltage between the electrodes practically to zero. It
applies to electrical breakdowns in solid, liquid and gaseous
dielectrics and combinations of these. The theoretical investigation
was also focussed mainly on the influence of the wave shape of a
lightning impulse to the discharge. At the end of this chapter
attention is paid to statistical techniques used when performing
breakdown tests.

3.2
3.2.1

Insulation systems
General

Dimensioning an insulation system requires exact knowledge of the

type, magnitude and the duration of the electric stress while
simultaneously considering the ambient conditions. The design and
construction of high-voltage equipment requires also experience in
the laws of electric fields.
Insulation arrangements for high voltage usually contain liquid, solid
or gaseous insulating materials which strength is many times that of
atmospheric air. For the practical application of these materials not
only their physical properties, but also their technological and
construction features must be taken into account [38]. The properties
of the insulating materials must be know, so that the optimum design
of the insulating system can be chosen. A particular problem in this
respect is that the determination of the properties of the insulating
materials is done with model samples under standardised conditions.
Extrapolation of these models to real insulating systems is not
unconditionally possible. For instance in case of transformers (which
will be explained in more detail in the next paragraph), the stress is
not evenly distributed and the samples used for testing are relatively
simple models. In any case, before the breakdown behaviour of
insulating systems or samples can be fully understood, the breakdown
behaviour of insulating materials should be known.

3.2.2

Transformers

3.2.2.1 Lightning impulse tests on transformers

One apparatus that is difficult to test with lightning impulse voltages
because of the superimposed oscillations and overshoot that occur is a
transformer. For the dimensioning of a transformer, the magnetic,
electric and thermal fields are important.
25

Breakdown
Particularly in the region of the windings, strong electric fields occur
in transformers, were high temperatures and short-circuit forces can
also act simultaneously.

Figure 15 Example for winding insulation

of an oil-transformer
1 high-voltage winding
2 low-voltage winding
3 pressboard tubes
4 shielding rings
5 angle rings
6 spacer blocks

One of the important parts of the dimensioning and design of a

transformer are its windings.
In the insulation of windings,
essentially three characteristic insulation problems occur:
High-voltage winding against low-voltage winding
High-voltage winding against the yoke
High-voltage winding against high-voltage winding
An example for winding insulation is presented in Figure 15.
High voltage windings are intricate structures, being weakly damped
networks comprising capacitances, self and mutual inductances. If an
impulse voltage appears across the terminals of the transformer,
characteristic oscillations are excited which can result in
impermissable high local stress of the insulation inside the windings.
For this reason, extensive investigations have been performed to
determine the voltage distribution occurring in different types of
windings. [41]
A greatly simplified but still useful assumption is that in the first
instant of stress with a steep lightning impulse voltage only the
capacitances of the network are effective. A simple equivalent circuit
for the calculation of the initial distribution is shown in Figure 16. It
is a homogeneous iterative network containing only the main
capacitances C as well as the earth capacitances Ce of a high-voltage
winding.

Figure 16 Capacitance equivalent circuit to

calculate the initial distribution

From this model it can be concluded that in case of steep front

impulses, the first winding section is overstressed compared with the
normal linear voltage distribution which reappears as the normal
operating condition after a transient mechanism has faded out.
The deviation of the initial distribution from the final distribution is
of great significance to impulse voltage stressing of a winding during
the subsequent transient mechanism. Since damping of the oscillation
is not possible because of low losses required during normal
operation, one must aim not to excite dangerous oscillations at all.
The less the initial and the final distribution differ from one another,
the smaller the possibility of oscillations occurring.

3.2.2.2 Effect of the wave shape of lightning impulses

Because of the costs involved in experimenting with testing of real
transformers, element models and statistical techniques have been
developed to transfer the results of the tests on models to a real
transformer.
Many investigations performed towards the influence of the
frequency to the breakdown voltage are performed using oscillatory
waves (or damped sine waves) in the framework of this investigation
they are not so useful. However, one of the experiments performed
that is useful, are experiments performed on a turn to turn model with
high frequency oscillations superimposed on an impressed wave [39].

26

Breakdown
The goal of this investigation was to
find a method to transfer the
breakdown behaviour under standard
lightning impulse conditions to that
under restriking surge (Figure 17) and
intruding lightning surge voltages (Figure 18) as are occurring in
modern substations.

Figure 17 Restriking surge voltage at

reactor current interruption

The result of this investigation was that the following method could
be used. Determine the time to peak (of the first peak) and the peak
voltage of the wave shape that occurs. In case of Figure 17 this would
be 0,42 s and 1129 kV. From the results presented in Figure 19 it
follows that the transfer factor is 1,27. This implies that the stress
caused by the surge presented in Figure 17 is equal to the stress
caused by a standard impulse with an amplitude of
1129/1,27 = 889 kV.

3.3

Figure 18 Intruding lightning impulse

surge

Breakdown processes in materials

Discharge phenomena are complicated and not fully understood. The

discharge depends on many things like atmospheric conditions,
voltage type, polarity, gap distance and electric field distribution in
the gap [40].
The breakdown mechanisms are different for the different types of
insulating materials. Four types of insulating materials can be
distinguished: gases, liquids, solids and vacuum. The discharge
mechanisms will be discussed in general for the four types of
insulating materials in the next paragraphs. In paragraph 3.4 the
attention will be given to the influence of the wave shape of a
lightning impulse to a breakdown.
Figure 19 V-t characteristics

3.3.1

Gases

The physical mechanisms associated with the development of a

breakdown in gases are of principle significance to all types of
breakdowns. Therefore, they will be explained first, before the
breakdowns in other insulating materials will be discussed.
Gas discharges can be sustained by means of external influences as
irradiation or heating, or even by the discharge itself [41]. In the first
case one speaks of a non-self-sustaining discharge in the second case
of a self-sustaining discharge.
The resistance of a gas-filled gap collapses to low values once the
voltage for complete breakdown is reached. The type of discharge
that then occurs and its duration depend upon the yield of energising
current source [38].

3.3.1.1 Non-self-sustaining discharges

3.3.1.1.1

Avalanche

When a free electron, with charge q, is available in an electric field

(E), the Coulomb force on this electron is given by: F = qE. If this
electron traverses without collision, a distance x corresponding to a
partial voltage U, the increase in its kinetic energy is: W = qU.

Figure 20 Electron
homogeneous field

avalanche

in

27

Breakdown
For a sufficient long mean free path and appropriate field strength,
electrons can, in collision with other neutral molecules, have such a
large kinetic energy W, that the molecule is ionised with the release
of further electron. If the ionisation condition (E Wi/e) is fulfilled,
an independent multiplication process of the electrons by collision
ionisation sets in. This exponential increase of the electron number is
called an electron avalanche [41]. At the head of the avalanche the
electrons may be very densely packed. Behind the head of the
avalanche the positive ions remain, see Figure 20.

3.3.1.2 Self-sustaining discharges

Figure 21 Electron generation according to

the Townsend mechanism

The voltage on an electrode arrangement in gases can only be

increased to the point where a breakdown occurs, i.e. a changeover to
a self-sustaining discharge. This differs from the non-self-sustaining
discharge in that the charge carriers required along the path are
created by the mechanism itself rather than liberated by external
ionisation.

3.3.1.2.1

Townsend Mechanism

The breakdown of gases at low pressures and small spacings can be

described by the Townsend mechanism. Thereby electrons of
external origin accelerated by the field can form new charge carriers
by collision ionisation, provided their kinetic energy exceeds the
ionisation potential of the gas molecules concerned. An electron
avalanche is built up which travels from the cathode to the anode. If,
as a consequence of the avalanche, a sufficient number of new ions
are formed near the cathode, complete breakdown finally takes place
[38].
Figure 22 Formation of a streamer in a
parallel plate configuration

This mechanism is also explained in Figure 21. When the number of

secondary initial electrons (no(es-1)) exceeds the number of original
initial electrons (no) the current in the configuration rapidly increases
and the gap breaks down. The condition for ignition is given by
(es-1)1. In electronegative gases has to be replaced by where
= - e, where e is a factor representing the loss of electrons by
attachment.

3.3.1.2.2

Streamer Mechanism

Investigations have shown that the exponential growth of an

avalanche cannot be increased at will since the avalanche becomes
unstable at a critical length xk (measurements in a homogeneous field
have shown that for air the avalanche becomes unstable when
xk20). The original field becomes more and more distorted due to
space charges, particular in the front of the avalanche, causing further
ionisation.
The mechanism occurring is called the streamer
mechanism and will characterise the breakdown. Breakdowns in
gases at higher pressures and with larger spacings take place by the
streamer mechanism according to Raether, Loeb and Meek [38].
The streamer mechanism is characterised by the fact that photon
emission at the tip of an electrode avalanche induces new avalanches
and initiates the growth of a streamer at a very fast, abrupt rate,
compared to the primary avalanche.

28

Breakdown
The transition of a discharge from Townsend growth to streamer
growth can, for a given spacing, be promoted by several parameters
such as the product of pressure and gap distance, high field strengths
caused by steep fronted impulses, density of the gas, inhomogenity of
the field, ionising radiation [38].
A schematic description of the formation of a streamer is shown in
Figure 22 and Figure 23. In Figure 22 the following is illustrated: A:
and external photon triggers an avalanche, B: a positive ion strikes the
cathode and starts an avalanche C. D: the avalanche tip reaches the
anode. E: photons originating from the avalanche produce free
electrons both from the cathode and in the gas. F: the positive space
charge close to the anode increases the geometrical electrical field
and a streamer is bound to be formed. The sequence is followed in
Figure 23. G: plasma of positive ions and electrons forms the streamer
channel. H: streamer tip. I: production of free electron by photons, J:
streamer close to the cathode [40].

3.3.1.2.3

Leader

Figure 24 Development of a breakdown

according
to
the
streamer-leader
mechanism

For longer gaps (>2 m) the sparkover is caused by a leader bridging

the gap. For short gaps (<2 m) the sparkover is usually caused by
streamers. The leader is usually one narrow irregular channel
propagating towards the cathode, but often at an angle to the applied
electric field [40]. Very large spacings, as they are unavoidable in
outdoor high-voltage installations, usually occur in combinations
having strong inhomogeneous electrical field. Particularly when the
electrode with the stronger curvation is stressed by a positive
switching impulse, a discharge process occurs which is denoted as the
streamer-leader mechanism. This mechanism leads to bridging of
large gaps at comparatively low mean field strengths [41].
The propagation of the leader for an impulse with critical time to peak
is continuous. At the leader tip, the leader corona is preceding the
leader. The leader channel is usually not luminous, except when the
voltage increase is lower than with critical time to peak. Then the
leader can stop and suddenly re-strike and light up. After the first
corona a dark period follows because of the divergent electric field
that stops the propagation of the streamers. When the second corona
starts, a leader can be formed from the streamer stem [40].

Figure 25 Current density in technical

insulating oils in a dc field as function of
the duration

This difficult process is visualised in Figure 24 [42]. Part a) of the

figures shows the voltage as function of the time, part b) the current
as function of the time and part c) shows the visualisation of he leader
as function of the time.

3.3.2

Liquids

In most applications, besides insulating voltage carrying parts,

insulating liquids have to satisfy further requirements. For example
they also serve to cool the windings and cores in transformers or to
extinguish the arc in circuit breakers. The behaviour of technical
insulating liquids in the electric field, differs fundamentally from that
of gases and solids. It is critically
Figure 23 Continued sequence of streamer governed by impurities, by the aging
condition as well as by space charges.
formation from Figure 22

29

Breakdown
As a consequence of this there is no unified breakdown theory,
although certain mechanisms are beyond doubt. [41] Two of the most
important types, the intrinsic breakdown and the breakdown by fibrebridging, will be described.

3.3.2.1 Intrinsic breakdown

As in gases and solid insulating materials, an avalanche breakdown is
also possible in very pure liquids. Only the reduction of the mean
free path, should be compensated by a higher electric field [41]. The
breakdown in clean and dry oil can be described with the streamer
theory [43,44].
In an electrode configuration in a d.c. field a current density S appears
which only reaches a nearly constant value within a few minutes (see
Figure 25). The current density S is not only depending on the
duration of the applied voltage it is also a function of the electric field
strength (see Figure 26). Since the mobility of the charge carriers
doesnt change, the only reason for the increase in the current density
with the increase of the field strength can be the carrier density. The
presumption that, similar to a breakdown in liquefied gases, the

charge carrier multiplication may be attributed to collision ionisation

in the gaseous or vapour part of the liquid, leads to the description of
this breakdown as a masked gaseous breakdown.

3.3.2.2 Thermal breakdown/fibre-brigde breakdown

Every technical liquid insulating material contains macroscopic
contaminants in the form of fibrous elements of cellulose, cotton, etc.
Particularly when these elements have absorbed moisture from the
liquid insulating material, forces act upon them, moving them to the
region of higher field strength as well as aligning them in the
direction of E. In this way fibre bridges come into existence. A
conducting channel is created which can be heated due to the
resistance loss to such an extent that the moisture contained in the

elements evaporates. In Figure 27 a

fibre-bridge formation is presented. Figure 26 Current density in technical
The breakdown that then sets in at insulating oils in a dc field as function of
the field strength

30

Breakdown
comparatively lower voltages, can be described as local thermal
breakdown at a defect. This mechanism is of such great technical
significance that in electrode arrangements for high voltages pure oil
sections have to be avoided. This is achieved by introducing
insulating screens perpendicular to the direction of the field strength.
In the extreme case consistent application of this principle leads to
oil-impregnated paper insulation, which is the most important at very
highly stressable dielectric for cables, capacitor and transformers.
[38]

3.3.3

Solids

Solid insulating materials are an inevitable component of every

technical insulating system since they are necessary to act as the
mechanical support of the electrodes against each other. Their
properties, as may be expected from their number, are very different.
The experimental determined values of the breakdown strength of a
solid insulating material are, owing to the many possible breakdown
mechanisms, strongly dependent upon the electrode configuration in
which they have been measured [38]. An important parameter for the
electric strength of solid insulating materials is the temperature. In
the following a few breakdown mechanisms will be explained shortly.

Figure 28 Illustration of the thermal

breakdown of solid insulating material

3.3.3.1 Intrinsic breakdown

The mechanism which leads to sudden loss of insulating capability
even after a short period of stressing, without appreciable pre-heating,
and without partial discharge is called intrinsic breakdown. Except in
the case of strongly inhomogeneous field configurations, a complete
breakdown can occur in a few nanoseconds. Intrinsic breakdowns
must be anticipated in homogeneous, predominantly crystalline
materials and especially during short-period stressing [41]. At high
field strength a large conductivity is observed, which must lead to the
conclusion that there is an increase in free electrons in the conduction
band. If the field strength continues to increase, the electron current
density reaches such high local values that it leads to the heating of
the insulating material due to the Joule effect.
If the electrons are accelerated by the electric field as in gas, collision
ionisation can result. Under favourable conditions, an avalanche can
be created.

3.3.3.2 Thermal breakdown

In solid insulating materials thermal
breakdown can be either total or local.
Figure 27 Model experiment showing
It can be explained by the temperature
fibre-bridge formation in insulating oils
dependence of the dielectric losses;
their increase can exceed the rise in the
heat being conducted away, Pab, and so can initiate a thermal
destruction of the dielectric. Figure 28 shows the curves of the power
Pdiel fed in at different voltages and the power Pab which can be drawn
from the test object, as function of the temperature which is
assumed constant throughout the entire dielectric. Thermal
breakdown then occurs when no stable point of intersection for the

31

Breakdown
curves of the input and output power exists. Point A represent a
stable working condition and point B, on the other hand, is unstable.
If the voltage is increased at a constant ambient temperature u, both
point of intersection move closer until, at U=Uk, they coincide in C.
This voltage is referred to as the critical voltage; or above Uk a stable
condition is impossible. [38].

3.3.3.3 Partial discharge breakdown

Partial discharges can occur in gas-filled cavities of a dielectric, but
also at electrodes with small radius of curvation when these are not
completely imbedded in the solid insulating material. Partial
discharges can, through long period mechanisms, lead to a complete
breakdown of the insulating, particularly during stresses with
alternating voltages. In Figure 29 a sketch of a canal which arise due
to stressing with a chopped lightning impulse voltage is shown.

3.3.4

Vacuum

All breakdown theories for gaseous, liquid and solid insulating

materials assume that the insulating material is made conducting by
the ionising process. In high vacuum (p10-5 mbar), the mean free
path is so large that collision processes in the rest of the gas become
meaningless for the breakdown mechanism. Rather the mechanisms at
the electrodes are critical to the breakdown behaviour. Many
breakdown hypotheses have been developed to explain the
mechanism in vacuum gaps (e.g. the cathodic breakdown, the anodic
breakdown and the clump hypothesis).

3.3.4.1 The cathodic breakdown hypothesis

The cathodic breakdown mechanism [45] assumes that the field
emission current, at a micropeak on the cathode, above a critical

current density leads to so much Figure 29 Partial discharge channel in

heating that the micropeak evaporates unfilled epoxy resin moulding
explosively. In the metal vapour then
formed ionising collisions processes
take place. If sufficient charge multiplication is achieved breakdown
of the vacuum gap occurs across the ionised metal vapour cloud.

32

Breakdown

3.3.4.2 The anodic breakdown hypothesis

In the anodic breakdown hypothesis it is assumed that the electrons
released from the cathode by field emission and accelerated in the
electric field, heat up the anode to such an extent that the anode
material vaporises. The metal vapour is ionised by collision
processes and reinforces the electron emission back at the cathode.
For a sufficiently high vaporisation rate at the anode, a gas breakdown
occurs within the metal cloud. [46]

3.3.4.3 The clump hypothesis

According to the clump hypothesis, vacuum breakdown is initiated by
free metal particles residing on the electrodes that are torn and
accelerated by field forces and vaporise on collision with the opposite
electrode. [47]

3.3.4.4 Effect of electrode material and temperature on the

breakdown voltage
As mentioned, the properties of the electrode material have a great
influence on the breakdown voltage in vacuum. In Figure 30 and
Figure 31 two examples are presented.

Figure 30 Effect of the electrode material

on the breakdown voltage with positive
and negative dc voltage in weakly
inhomogeneous field.

Figure 31 Effect of electrode material and

electrode temperature on the breakdown
voltage with alternating voltage

33

Breakdown

3.4 Influence
breakdown

of

lightning

impulse

voltage

to

the

In the previous paragraph general breakdown processes have been

explained. In this paragraph, attention is paid to breakdown processes
when lightning impulse voltages are applied and attention is paid to
the influence of the polarity and the wave shape of the lightning
impulse voltages to the breakdown voltage. Since they might be
different for different type of materials, they are explained separately.

3.4.1

Gases

The literature has been investigated with the goal finding as much
information as possible about breakdown behaviour under lightning
impulses in general for all types as gases. Especially for air and SF6 a
literature study has been performed towards the effect of polarity
and wave shape.

3.4.1.1 Breakdown under lightning impulse voltages

Experience has shown that the development of a breakdown requires
a finite time. This fact is important for
short-duration stresses, such as
Figure 32 Formation and shape of voltagelightning impulses. It is known that a
time curves
complete breakdown occurs after a
combined interval (td) of the statistical
time lag (ts) and the formative time lag (ta) [41]. (td = ts+ta)
The statistical time lag is the time between the reaching of the
breakdown field strength and the moment an initial electron appears
(this time varies from experiment to experiment). The formative time
lag is the time interval from the beginning of the first electron
avalanche to the formation of the highly conductive canal that leads to
a collapse of the voltage. The formative time lag is depending on the
applied voltage magnitude and the homogenity of the electric field.
If an electrode arrangement is stressed with a large number of
identical pulses, one obtains the values for the breakdown voltage
(Ud) and the time lag (td). If the measurement is repeated with
impulses of different steepness, one obtains the results that are
presented in Figure 32. The lower limiting curve (curve 1)
corresponds to a 0% breakdown probability, the upper curve (curve 2)
corresponds to the 100% breakdown probability.

Many thanks to LCOE, Fernando Garnacho and Pascual Simon, who have
contributed to the results presented here

34

Breakdown
For the calculation of the voltage-time curves, the equal area criterion
[48] has proven to be a useful assumption in many cases.
For weakly inhomogeneous fields, the reference voltage (Ud in
Figure 32, Ub in Figure 33) becomes equal to the inception voltage Ue.
If the reference voltage is known, the equal area criterion, therefore
permits an approximate calculation of the impulse voltage-time curve
of an electrode configuration for a few measured values.

Effect of non standard overvoltages

Various approaches have been proposed to evaluate the strength
under non-standard overvoltages in air. The approaches may be
grouped into three main categories namely [49]:
Physical approaches, which, in general form, evaluate the time to
breakdown, Tb for a given applied voltage, as the sum of times
necessary for the development of the various phenomena involved
in the discharge (time to corona inception Ti, time necessary for
streamers propagation Ts, time necessary for leader propagation Tl)
Integration methods (e.g. constant area criteria) which considers
the time integral of the difference between the applied voltage and
a fixed voltage (in some cases this difference is raised to a power,
n, different from unity). Breakdown is assumed to take place when
the integral reaches a fixed value, depending on the configuration
and voltage polarity.
Simple formulae interpolating the various sets of data available.
It has to be seen if these approaches can be used to evaluate the
parameters.

Effect of polarity
For SF6; plate-plate, rod-plate and for needle plate configurations, the
breakdown voltage for negative lightning impulse voltage is higher
than the breakdown voltage for positive lightning impulse voltage (in
case of the rod-plate configuration and the needle-plate configuration,
the high voltage is applied to the rod) [50], [51].
Also for switching impulses and a non-uniform field, the breakdown
voltage is for negative impulses higher than for positive [51].

Effect of front time

For SF6, the breakdown voltage in case of a steep fronted impulse (T 1
around 50 ns) is lower than for standard 1,2/50 lightning impulse
voltage [52]. In this coaxial electrode configuration this was valid for
both positive and negative impulses.

Figure 33 Equal area criterion

Effect of oscillatory waves

Tests performed in a strongly non-homogenous field in SF6, with a
gap distance of 60 mm under a pressure of 0,12 and 0,3 MPa, showed
that the breakdown voltage for oscillating impulses with a frequency
lower than 1 MHz is higher than that for a standard 1,2/50 lightning
impulse voltage [53]. Oscillations with a frequency of 1 to 8 MHz
have a similar breakdown voltage as standard 1,2/50 impulses.
Oscillations with a frequency above 8 MHz result in a lower

35

Breakdown
breakdown voltage as for a standard 1,2/50 impulse. In all cases the
damping of the oscillating wave was around 40%.

Effect of oscillations superimposed

When there are oscillations of relevant amplitude (40%)
superimposed to lightning impulse voltages, the discharge mechanism
depends, at least for SF6, on the frequency of the oscillation and the
pressure of the gas. For instance for 1 bar and oscillation frequencies
in the kHz range up to 1 MHz all investigations have shown higher
breakdown levels compared to lightning impulse voltages. The
reason is an interrupted leader propagation during the negative cycle
of the oscillation. With rising frequency there is a reduction of the
breakdown voltage, for instance the breakdown voltage of the 10
MHz oscillation is lower than the lightning impulse oscillation. In
[54] the breakdown voltage as function of oscillation frequency and
gas pressure is presented.

3.4.2

Liquids and combinations with paper

The application of oil as insulating material occurs almost exclusively

in combination with cellulose in the form of paper of pressboard.
Oil-paper insulation was for a long time the most important highvoltage technical insulating material, without which the often used
concept of transformers, capacitors and cable would be inconceivable.
Therefore a literature investigation was made to the effect of the wave
shape to oil-paper insulation. But to be able to understand the
behaviour of the combined insulating materials, an investigation

Figure 34 Impulse voltage-time band of

the rod-rod configuration in oil for
negative impulses voltages

toward the effect of the wave shape in oil was made.

Besides depending on the impurities, the electric strength of liquids
depends on several other parameters, particularly on pressure and
stress duration. General properties of oil and oil-paper are described
in [41].
During impulse stressing the breakdown field strength is many times
the value for alternating voltages [55]. The impulse voltage-time
curve of an electrode configuration in transformer oil presented in
Figure 34 gives an idea of the effect of the stress duration.

36

Breakdown

Effect of polarity
For oil, the breakdown voltage in a strong non-uniform field is for
positive oscillating waves lower than for negative oscillating waves
[56]

Effect of wave shape

In a highly non uniform field, the breakdown voltage of oil for
oscillating waves (frequencies of several Hertz to 100 kHz) is higher
than for standard 1,2/50 lightning impulse voltages. Furthermore it
was observed that the negative oscillatory wave has a higher
breakdown value than a positive one. [56].
When for oscillating waves of a frequency higher than 200 kHz, the
damping is decreased, the breakdown voltage will decrease.
The breakdown voltage (for different impregnants, heating of paper,
thickness of the paper) of oil-paper insulation is lower for steep
fronted lightning impulse voltage than for lightning impulse voltage.
Many investigations with oil-paper insulation have been performed
using damped sine waves, but for this investigation that is of no
relevance.

3.4.3

Solids

A literature investigation about polyethylene had shown that the

breakdown behaviour of polyethylene (PE) or cross-linked
polyethylene (XLPE) depends on several parameters. Besides the
quality, the age, the shape and the dimensions of the specimen also
the test arrangement concerning the field configuration and the shape
of the used test voltage as well as the pre-stress, the pre-conditioning
and the frequency and duration of the stress have an influence [57,58,
59 and 60]. Additional parameters like the temperature or the test
environment can also influence the breakdown strength. However,
less information could be found about the dependence of the
breakdown behaviour of PE or XLPE on standard lightning impulses
superimposed by oscillations or overshoots with various amplitudes
and frequencies or respectively overshoot duration.
Investigations using standard lightning impulses (1,2/50 s) have
shown that there is no significant difference in the impulse strength
between PE, XLPE and PE containing voltage stabilisers [61]. The
nominal value of the mean breakdown strength for these materials is
at least 100 kV/mm at room temperature [62], but the strength
decreases clearly for rising temperatures.
Investigations on the treeing inception of XLPE, which is a prestadium of the breakdown process, in a needle-plane arrangement
under lightning impulse voltages (1,2/50s) had the findings that
there is a strong dependence of the tree inception voltage on the
polarity of the needle-electrode, whereas the time between the
impulses as well as the shape of the lightning impulses show almost
no influence [63 and 64]. Varying the front time T1 from 0,6 s to

Many thanks to the Schering Institute Hannover, Peter Werle, Klaus Hackemack
and Ernst Gockenbach who performed this study.

37

Breakdown
2,3 s and the time to half value T2 from 6 s to 95 s does not
influence the tree inception voltage, which is similar for other solid
insulation materials like e. g. ceramic [65, 66 and 67].
The conclusion of the literature investigations about research
concerning the breakdown strength of polyethylene under lightning
impulse stresses allow the statements given in the next three
paragraphs concerning the main parameters of a breakdown test.

Field configuration / polarity

The field configuration has an influence on both treeing inception
voltage and the breakdown strength of polyethylene, whereas the
polarity only has an influence, if a non-homogenous field
configuration is used. The voltage strength using negative impulse is
higher than at for positive voltages.

Shape of the voltage stress

The shape of the impulse voltage concerning the rise and fall time has
almost no influence on the impulse voltage strength as well as the
time between the impulses.

Maximum voltage
The maximum voltage of the lightning impulses is a decisive
parameter. For rising temperatures or different environmental test
conditions the maximum voltage of a lightning impulse that a
polyethylene specimen is able to withstand can be reduced.
About the influence of superimposed oscillations or overshoots at the
lightning impulse on the breakdown behaviour of polyethylene no
information is available. However, the investigations indicate that the
breakdown process depends like for other solid materials on the lifetime-law, which describes an exponential relation between the
breakdown strength and the time of the applied voltage stress.

3.4.4

Vacuum

Effect of polarity
For vacuum the negative impulse has a lower breakdown voltage than
the positive impulse. The difference between the breakdown voltage
for negative and positive impulses is larger for a lightning impulse
voltage than for a switching impulse. The difference between the
breakdown voltage for positive and negative impulses is also a
function of the gap distance. For a smaller gap distance the difference
is smaller. [68]

Effect of waveshape
In [69] breakdown voltage values are reported for different wave
shapes for a stainless steel hemispherical electrode configuration. A
summary of the results is presented in Table 2.

38

Breakdown

Figure 35 Distribution function P(U) as

function of the breakdown voltage

F ( x) P( x x)

f (t )dt

Table 2 Effect of wave shape

Equation XV Distribution function,
probability and density function

T1/T2

1,2/50 s

250/450 s

0,75/4,5 ms

1/4,5 ms

2,6/15 ms

3,6/15 ms

2,0 mm gap

80 kV

70 kV

80 kV

95 kV

150 kV

150 kV

1,5 mm gap

75 kV

65 kV

70 kV

80 kV

140 kV

120 kV

1,0 mm gap

45 kV

45 kV

45 kV

45 kV

80 kV

105 kV

0,5 mm gap

20 kV

20 kV

20 kV

20 kV

3.5
3.5.1

Breakdown statistics and statistical techniques

Introduction

Physical phenomena can be either exactly reproducible or be subject

to statistical scatter. By applying statistical methods it is possible for
relative few measurements to make a statement about the behaviour of
an ensemble with a certain uncertainty. Comprehensive treatment of
this subject with emphasis to high-voltage tests is made in [78]. Here
only the headlines are presented.
Electric breakdowns belong to the second group (the physical
phenomena that are subject to statistical scatter) and therefore during
repeated measurements of a breakdown voltage one should expect a
certain scatter in the values. The values for the breakdown voltages
can be approximately determined by experiments.

39

Breakdown
If n identical samples will be subjected to a certain voltage stress, the
number of breakdowns nd will depend upon the voltage magnitude U.
The observed breakdown probability (P) nd/n is a function of U as
shown in Figure 35. Ud-0 or U0 is the withstand voltage. Ud-50 or U50 is
the 50% breakdown voltage, Ud-100 or U100 is the assured breakdown
voltage.
In technical set-ups in general the distribution function P(U) is not
known, but experience showed that the actual distributions, in case of
medium probabilities, are very well approximated by a normal
distribution function. For very small probabilities, the Weibull
distribution function has proven to be good.
The distribution function P(U) is often referred to as F(x). The
distribution function at a point x, indicates the probability with which
the variate X will assume a value below the boundary x. The
distribution function of continuous variates can be represented in the
form as given in Equation XV.
The normal distribution function and density function are given by the
formulas presented in Equation XVI and Equation XVII. In Figure 36
the normal distribution is presented in a graphical way.

3.5.2
f ( x)

Application

When an (almost) infinite number of measurements could be

performed, one should be able to deduce the type and the function of
density and distribution function almost exactly. However in
practical applications, one has a limited number of measurements
available from which the distribution function has to be determined.

d
F ( x)
dx

This implies of course then directly that there will be a certain

uncertainty in the determined type of function and the values of the
function.

( x; ; 2 )

e ( x )

/ 2 2

Equation XVI Density function of the

normal distribution

When the 50% breakdown voltage has to be determined, a normal

distribution function will be suitable in most cases. For a normal
distribution function the (two-sided) confidence limits for the
parameters and 2 are given by two complex formulae presented in
[38]. This approximation for the confidence limits can also be
presented in a graphical format, this is presented in Figure 37.
Figure 36 Density ((x)) and distribution
function ( (x)) of standard normal
distribution, =0, 2=1

( x; ; )
2

1
2

(t )
e

2 / 2 2

dt

Equation XVII Distribution function of the

normal distribution

40

Breakdown

3.5.3

Independence

The statistical evaluation of the measurement results is of course only

valid under the condition that the measurements represent a random
selection, i.e. they are independent from each other. When evaluating
the measurement results, one of the checks that has to be performed is
therefore the test of independence. There are several independence
tests, which are comprehensively treated in [78].

Figure 37 Approximation for the

confidence limits of a normal distribution
function

41

Goal of investigations

4 Goal of investigations
In Chapter 2 of this thesis, lightning impulse tests, applicable
standards, their interpretation and implementation have been
described in detail. In Chapter 3 of this thesis, the breakdown process
in materials and statistical techniques to be used when performing
breakdown tests have been explained in detail. In this Chapter the
goal of the investigations, the insulating systems and the materials to
be investigated will be described. At the end of this Chapter, some
research questions and hypothesis are formulated.

4.1

Introduction

In this paragraph a summary of the Chapters 2 and 3 and some

considerations about the development of the measuring techniques
will be given. These three paragraphs are an introduction to the need
for, the goal of and some expectations of the results of the
investigations carried out in this study.

4.1.1 International standards, their implementation and

interpretation
Tests with impulses are designed to demonstrate the response of the
equipment to transients over a wide frequency range. Lightning
impulse voltages represent transients occurring naturally in high
voltage systems under operation. There are at least four standards
that are applicable when tests with lightning impulse voltages are
performed, namely IEC 60060-1, IEC 60060-2, IEC 61083-1 and
IEC 61083-2. They deal with the wave shape of the lightning impulse
voltages to be applied, the parameters by which it is characterised, the
measurement of the lightning impulse voltages and last but not least
with how to evaluate the parameters characterising the lightning
impulse voltage.
As may have become clear from the detailed explanation in Chapter 2,
great difficulties (with measurement uncertainty and uniformity) are
experienced when either lightning impulse voltage with overshoot
and/or oscillations or non-standard lightning impulse voltage have to
be measured and evaluated. It appears that the rules for the evaluation
of the parameters are ambiguous, that even for some cases no
evaluation rules are given and that some laboratories use methods quit
far from the present standards.
When performing tests, the aim is always to generate standardised
lightning impulse voltages (without oscillations or overshoot).
However, during tests of high voltage equipment with lightning
impulse voltages, oscillations or/and overshoot may occur at the front
or near the peak of a lightning impulse voltage. This is due to the
interaction between the test object and the generator. Especially in
case of testing power transformers, so-called non-standard lightning
impulse voltages are generated. It is being assumed that oscillations
and/or overshoot can be of significance and right now a maximum
amplitude of 5% of the peak value is allowed [70].
Since power transformers are large capital assets, it is important for
this equipment to test and to measure the applied lightning impulse
43

Goal of investigations
voltage with a good accuracy. Precisely for these cases where nonstandard lightning impulse voltages occur, no evaluation rules are
given. Most of the time these measurements are carried out by
experienced engineers, who use engineering judgement to evaluate
the lightning impulse voltages. However two engineers may obtain
different measurement results. This assumption is being supported by
a questionnaire made amongst different laboratories. The result of
this questionnaire was quit surprising, many laboratories do not only
use very different evaluation methods, but also methods quit far from
the present standards (see Chapter 2.4.7).
Not only in the case of lightning impulse voltages with large
superimposed oscillations and/or overshoot differences between the
measurement results may occur even when using evaluation methods
according to present standards. Results of round-robin tests showed
that even large differences (up to 2% for full and 7% for chopped
impulses) occur when lightning impulse voltages with small
overshoot and/or oscillations are evaluated.
Up to now, several evaluation methods have been proposed, but as
may have become clear, no satisfactory solution has been found.

4.1.2

Breakdown phenomena and statistical techniques

A criterion from which it can be concluded whether an object passes

or fails a test with lightning impulse voltage, is the occurrence of a
complete or partial breakdown. New parameters or new evaluation
methods shall not be introduced before the relation to the breakdown
behaviour is known.
From the questionnaire made amongst
manufacturers it appeared that there is not a well-established physical
background about the relevancy of different parameters presently
used. The information supplied by the laboratories on this subject is
contradictory [71]. Therefore a literature investigation towards the
breakdown process of different materials was necessary.
In Chapter 3 the breakdown processes of materials and insulating
systems have been described in general. It was obvious that wave
shape and polarity are important parameters. However not so much
information was found about the breakdown behaviour for lightning
impulses with overshoot and/or oscillations. With the information
found it not possible to deduct the breakdown behaviour for lightning
impulses with overshoot and/or oscillations and from that determine a
suitable evaluation method.
For performing breakdown tests it is well know that one has to apply
certain statistical methods to evaluate the results.

4.1.3

Developments in measuring technique

In the past the measurements of lightning impulse voltages were

performed using analogue oscilloscopes. The evaluation of lightning
impulse voltages was performed manually. The achievable accuracy
was influenced by the thickness of the trace of the oscilloscope and
the accuracy of the test engineer. Because of the natural filtering of
the measurement system, only oscillations and/overshoot of high
amplitude and low frequency could be seen. When oscillations or

44

Goal of investigations
overshoot could be seen, discussing took place between the test
engineers to draw the best mean curve.
Nowadays the measurement of lightning impulse voltages is mostly
performed using digital oscilloscopes and measuring systems with a
higher bandwidth. Because of this, small overshoot and oscillations
can easily be made visible. The evaluation of the parameters is
mostly done using computer programs, that have difficulties with the
implementation of evaluation rules made for manual evaluation, since
they need more precise instructions. Because of that in some cases
still a manual evaluation is used. The advantage of using computer
programs for the evaluation of parameters is that on the one hand
more parameters can be evaluated than were previously practical or
possible and on the other hand that the evaluation is performed more
uniformly.
The developments in the measuring technique have made it possible
to measure with more accuracy and so the requirements towards the
accuracy have increased.
Intercomparisons and questionnaires
however showed that the present evaluation rules are ambiguous,
causing large differences in the obtained measurement results. It
seems right now that the hardware is good enough for the nowadays
required accuracy, but that the software or evaluation methods are the
major contribution to the measurement uncertainty.
Some blame the computers and digital recorders for the problems
occurring nowadays, but the problems are not new. The introduction
of digital recorders and computers made it only possible to see and to
recognise the problems.

4.2

Goals and expectations of investigations

From the previous paragraphs it is obvious that it is appropriate to

investigate the introduction of new parameters and evaluation rules of
these parameters. However, the introduction of new parameters and
evaluation rules or algorithms of new parameters cannot be done
without physical evidence that they are indeed the most relevant ones.
The deviations from the standard lightning impulse voltage are caused
by oscillations and overshoot. Therefore it should be investigated
what the influences are of oscillations and/or overshoot on insulating
materials and systems and what the most relevant parameters are.
One of the problems that arise today, is that one has to determine
whether it is a standardised or not standardised wave shape. Because
of this, algorithms that are applicable to all types of lightning impulse
voltage (standardised and non-standardised) are preferable to
algorithms specific for one specific type of impulse.
The goal of the investigations is therefore to find parameters and clear
evaluation rules, that are applicable to both standardised and nonstandardised lightning impulse voltages, taking into account the
possibilities and limitations of modern computers and that are proven
to be of physical relevancy for all types of practical insulating systems
in case lightning impulse voltages are applied.

45

Goal of investigations

4.3

Insulating systems to investigate

Problems with the evaluation of parameters characterising lightning

impulse voltages occur in many different types of laboratories;
calibration laboratories, manufacturing laboratories and testing
laboratories. The problems can be categorised differently. For
calibration laboratories it is often possible to generate very smooth
and standardised lightning impulses. For these cases more simple
algorithms can be sufficient. It can be assumed that the type of
parameters used are of less importance as long as the evaluation
method is clear for people all over the world. For the two latter types
of laboratories, it is often not possible to generate smooth
standardised lightning impulse voltages. Besides this the parameters
must have a relation with the breakdown behaviour.
The majority of problems occur during the testing of high voltage
equipment (i.e. in manufacturers and testing laboratories).
Transformer manufacturers and laboratories who tests transformers,
complain the most about the present situation in the standards.
Apparently for them the need for new parameters, which they assume
to be relevant for transformer testing as well as clear evaluation rules
is high. Since transformers are also high capital assets, it seems to be
appropriate to investigate these insulating systems in this study.
Cables, switchgear and air insulation should not be forgotten, since
they are of importance too. It is important to introduce parameters that
are applicable to all types of insulating systems. However they will
not be investigated in the framework of this PhD study, but they will
be investigated in the framework of the European project. However
the results of the investigation performed in the European project will
be used for the introduction of new parameters.
One of the ways to prove the relevancy of parameters is to perform
breakdown tests. When breakdown tests are performed, it is not usual
perform them on insulating systems. One aspect is the costs of such
investigations and the other aspect is that the results of these
breakdown tests may depend on too many factors to draw conclusions
or to make relations towards other insulating systems. Therefore the
way to start this investigation is to study the insulating materials and
from the results make a translation to practical insulating systems. In
this projects the insulating materials XLPE, SF6, air, vacuum, oil and
oil-paper are chosen.

4.4

Tests to perform on insulating materials

Before performing breakdown tests on insulating materials to find out

which parameters are relevant and which are not, it is important to
select carefully the measuring methods, the experimental parameters,
the test equipment and of course make a test planning [72].
The goal of the investigations is therefore to find parameters and clear
evaluation rules, that are applicable to both standardised and nonstandardised lightning impulse voltages, taking into account the
possibilities and limitations of modern computers and that are proven
to be of physical relevancy for all types of practical insulating systems
in case lightning impulse voltage are applied.

46

Goal of investigations
Right now, there are two problems with the evaluation of lighting
impulse parameters:
In the standard for some cases a mean curve has to be drawn for
the evaluation of the peak value.
Cable and transformer manufacturers do not agree that the peak
value of this mean curve is a relevant parameter. They claim that
the absolute peak value is a more relevant parameter.
Therefore the major problem is nowadays the definition and the
validity of the mean curve in case oscillations and overshoot occur.
For the international acceptance, one has to prove first whether a
mean curve is really relevant. Perhaps there are also other parameters
that are more relevant. When the present peak value of the mean
curve is proven to be a relevant parameter, then it is appropriate to
look into ways to define this parameter without having to draw a
mean curve. This is because it seems not to be possible to define a
mean curve which is valid for all lightning impulse voltage that occur
during testing of high voltage equipment.
As a way to prove which parameters are relevant and which are not,
breakdown tests and the determination of the U50% were chosen. A
list of parameters that were thought or assumed to be relevant was
made, see Chapter 5.3.5. In order to be able to prove whether the
mean curve is relevant, lightning impulse voltages with oscillations
and overshoot of different frequencies and amplitudes superimposed
have to be applied. Above all the wave shapes generated should
cover the wave shapes that occur during actual testing of
transformers.
There are several ways to design a generating circuit in which these
wave shapes can be generated. One of the important criteria is, that in
case different amplitudes or frequencies are applied the mean curve is
not varied. The only circuit for which this was possible is the circuit
described in Chapter 5.2.2, it is the generating circuit that consists out
of two generating circuits. One circuit generates the standard
lightning impulse voltage and the other generates the oscillations or
overshoot. This circuit has many advantages, one disadvantage might
be that the circuit has some influence to the result, so this should be
checked first
Since oil and oil paper are the insulating materials used, attention
should be paid to the statistical spread in the measuring results, due to
the properties of oil and oil-paper. This means that sufficient checks
and comparison measurements have to be made.
Detailed attention to the generating circuit, the test cell, the measuring
circuit and to the experimental parameters will be paid in Chapter 5 of
this thesis.

4.5

Translation of results to insulating systems

After the behaviour of insulating materials towards lightning impulse

voltage of different shape is known, these results should be translated
to insulating systems. This might not be easy since the behaviour of
the materials can differ for homogenous and inhomogeneous fields.
The design of an insulating system is much more complex than simply
47

Goal of investigations
either homogeneous or inhomogeneous. One has to consider the
voltage distribution in a complex insulating system such as a
transformer in case a lightning impulse voltage is applied.
What at least will be done is to check the hypothesis on samples that
are representing some typical parts of transformers.

4.6

Questions/hypothesis

Following the discussion in this chapters the following questions and

the hypothesis can be formulated:
are the present parameters the most significant ones?
is it indeed true that oscillations and/or overshoot with an
amplitude of more than 5% are of significance?
is it indeed the absolute peak value that is one of the most
important parameters when testing transformers and
cables?
is it true that the peak value of the reference mean curve
is one of the most important parameters when testing air?
how can the results on insulating materials be translated to
insulating systems?
is it possible to define a set of parameters applicable to both
standard and non-standard lightning impulse voltages?
is it possible to define a set of parameters applicable to all
insulating materials/systems?
is it possible to translate algorithms used for manual evaluation
into digital algorithms?

48

Investigations; the test set-up

5 Investigations; the test set-up

5.1

Introduction

The goal of this work is to find parameters and clear evaluation rules
that are applicable to both standard and non-standard lightning
impulse voltages, taking into account the possibilities and limitations
of modern computers.
The goal of the theoretical and experimental investigations is to prove
that the parameters chosen are of physical relevance for different
types of practical insulating systems in case lightning impul7se
voltages are applied. In previous Chapters the results of theoretical
investigations have been presented. The result of this theoretical
investigation was that experimental investigations were necessary.

The experimental investigations were carried out simultaneously in

several laboratories (KEMA, NGC, LCOE, Schering Institute and TU
Graz) on different materials and samples (oil, SF6, air, XLPE and
transformer winding models). The investigations performed in the
framework of this PhD-project are focussed on oil and oil-paper
insulation and are performed at KEMA and TU Graz. At KEMA, the
author was responsible for the test set-up and the measurements and
performed most work herself.
The relevancy of the parameters will be proven by breakdown tests.
For this purpose a test set-up comprising of a generating circuit, a test
cell, measuring- and evaluation systems had to be designed and built.
In this chapter the test set-ups used for the investigations are
described in detail. But before they are explained, first two possible
generating circuits, the modified conventional generated circuit and
the combined generating circuit will be explained.

5.2 Generating circuit, combined generating circuit versus

modified conventional circuit
To generate the different lightning impulse voltages a generating
circuit was needed. Initially, it was the idea to build only a slightly
modified conventional lightning impulse generator. By varying the
different components of this generator, lightning impulse voltages of
different wave shapes, i.e. different front times, tail times,
superimposed overshoots and oscillations could be generated.
For reasons explained in paragraph 5.2.1, this modified conventional
generator had too many disadvantages to perform the first basic
investigations on insulating materials.
Because of the disadvantages it was decided to build a generating
circuit (further called combined generating circuit) comprising of two
generators. Generator 1 generates the standard lightning impulse and
the generator 2 the oscillations or overshoots. The two voltages are
superimposed in the test-cell. This circuit was much more suitable for
the first basic investigations, although some problems arose. The
combined generating circuit is explained in paragraph 5.2.2.

49

Investigations; the test set-up

1
2 LC

Equation XVIII Frequency of oscillation

5.2.1

Modified conventional circuit

5.2.1.1 Principle
The principle of the generating circuit based on a conventional
generator, called the modified conventional circuit, is shown in Figure
38. For this conventional circuit, simulations have been made to
dimension the components for the different lightning impulse voltage
to be generated. The circuit has been built and impulses have been
generated. The results are explained in detail in [73 and 74] and
summarised in this paragraph.

Rs

Figure 38 Modified conventional circuit

Cs1 and Cs2 charging capacitors
Cp
parallel capacitor
L
coil
Rs
series resistor
Rp
parallel resistor
Rf
front resistor

Cs2

Rf

Object Divider

Rp

Cp

Cs1

Trigger circuit

Computer, digitizer
and attenuator

By varying the front resistor Rf, the front time will mainly be varied.
A smaller front resistance will result in a smaller front time. By
varying the parallel resistor, Rp, the time to half value will mainly be
varied. By increasing the parallel resistance the time to half value
will increase.
The frequency of the oscillations or the duration of an overshoot can
be varied by varying the inductance and the capacitance, because in
general this frequency in given by Equation XVIII.

Figure 39 Simulation of an impulse with

overshoot; L = 520 H, Cs = 150 nF,
Rp = 520 , Rf = 800

Overshoot can be considered as a well damped oscillation. The

damping is mainly determined by the resistance of the circuit. To
keep the time to half value and the front time on a certain value, an
additional resistance Rs is introduced. The series resistance has
practically no influence on the front time and the time to half value,
but has influence on the damping of the oscillations.

5.2.1.2 Simulations and test results

Since the generating circuit had to be built with the components
available at the High-Voltage Laboratory of KEMA, the dimensioning
of the components in the circuit is somehow limited. Several
simulations have been made using the circuit shown in Figure 38.
The capacitance of the object will be in the range of 10 to 20 pF,
since this is so small, a capacitor, Cp, is placed in parallel. For Cp, the
values 500 pF, 1 nF and 2 nF can be chosen.
The value of the inductance L can be varied between 10 and 3000 H,
but not in indefinite steps. For Rp, Rs and Rf more or less any value
can be chosen between 0 and 2000 .

Figure 40 Simulation of an impulse with

oscillation; L = 270 H, Cs = 150 nF,
Rp = 520 , Rf = 250

50

Some of the simulations mare are shown in Figure 39 to Figure 44.

With the values of the components found with these simulations,
some real measurements were done. There was a small discrepancy

Investigations; the test set-up

between the simulations and the measurements, caused by the nonideal behaviour of the components.
After some simulations and measurements it became clear that a test
circuit based on a conventional circuit had too many disadvantages
for the first investigation (see the list presented in the next paragraph).

5.2.1.3 Advantages and disadvantages of the modified circuit

Disadvantages of this modified conventional circuit are that:
the position of the oscillations superimposed cannot be varied, for
instance one cannot generate oscillations on the peak without
generating oscillations of the front too
it is difficult to generate overshoot and oscillation at the same time
it is impossible to generate oscillations or overshoot without
influencing the shape of the mean curve
it is practically impossible to generate oscillations of a certain
frequency with different amplitudes
Advantages of this conventional circuit are that:
the circuit is similar to the one used during normal testing
only one measuring circuit is needed
the construction is simple

5.2.2

Combined generating circuits

Figure 42 Simulation of a standard

lightning impulse; L = 10 H, Cs = 300 nF,
Rp = 230 , Rf = 430

5.2.2.1 Principle
One of the most important requirements for the generating circuit is
that the mean curve does not change when oscillations and
overshoot are generated and that oscillations of different frequencies
and amplitudes can easily be generated. Since the influences of the
oscillations and overshoot are expected to be small, one has to be
absolutely certain that the differences in measuring results are not
caused by the differences in mean curves. Therefore the generating
circuit shown in Figure 43 was designed (combined generating
circuits). Generator 1 generates the full lightning impulse voltage and
generator 2 generates the overshoots or oscillations. At the object the
two voltages are superimposed to each other.
By varying the resistors in generator 1, Rp and Rf, the front time and
the time to half value of the lightning impulse voltage can be varied.
By varying the inductance Ls and the resistor Rd of generator 2,
Generator 1, lightning impulse

Figure 41 Simulation of an impulse with

overshoot; L = 750 H, Cs = 300 nF,
Rp = 350 , Rf = 1400

Figure 44 Simulation of an impulse with

oscillation; L = 750 H, Cs = 300 nF,
Rp = 250 , Rf = 200

Figure 43 Combined generating circuit

Divider 1

Rf
Cs

Rp

Cp
Test cell

Ampl. 1
Atten. 1
Trigger box

Digital
Oscilloscope

Computer

Atten. 2
Ampl. 2

Ls
Rd

Generator 2, oscillation or overshoot

Divider 2

51

Investigations; the test set-up

respectively the frequency of the oscillations and the damping of
these oscillations are varied.

5.2.2.2 First simulations and measurements

The use of a combined generating circuit was proposed by Fernando
Garnacho in August 1997, when all partners within the EC project
came independently from each other to the conclusion that a
conventional circuit could not be used for the investigations to be
performed. The first simulations were made by LCOE. These
simulations showed that it should be possible to build such a circuit
and that the circuit could meet the requirements for the investigations
to be performed. Disadvantage was that the test set up itself was
made much more complex.

5.2.2.3 Advantages and disadvantages

The advantages are:
mean curve can be maintained constant while varying overshoots
and oscillations
the mean curve can be obtained exactly
frequency, amplitude and damping of oscillations can be varied
independently from each other
amplitude of the oscillations can be kept constant while varying
the amplitude of the lightning impulse voltage
The disadvantages are:
the two circuits have to be synchronised
two measuring systems are needed
influence of the two circuits on each other during withstand and
breakdown

5.2.3 Influence of using the combined generating circuit

instead of the modified conventional circuit on the test results
One of the concerns towards using the combined generating circuit
instead of a modified conventional circuit, was the possible influence
on the breakdown process. Performing either calculations or
measurements should make it clear if there was an influence. It was
found the best to check it by measurements.
Four different types of measurements have been performed using a
15 mm sphere-plate gaps in air. It was expected that there should be
no difference in the results obtained. The situations were the
following:
1. Two measurements have been performed applying a standard
impulse by using only one generating circuit (generator 1)
connected to the upper electrode. The lower electrode was
connected to ground. This situation is identical to a situation
where the modified conventional circuit would be used.
Measurement no. 981002-1, 980929-3
2. One measurement has been performed applying a standard
impulse by using one generating circuit (generator 1) connected to
the upper electrode. The lower electrode was connected via a 30
Ohm resistor to ground. Measurement no. 980929-4.

52

Investigations; the test set-up

3. One measurement has been performed applying a standard
impulse by using one generating circuit (generator 1) connected to
the upper electrode. The lower electrode was connected to the
not energised generator 2. Measurement no. 980929-5.
4. One measurement has been performed applying a standard
impulse by using one generating circuit (generator 1) connected to
the upper electrode. The lower electrode was connected to the
generator 2. With that generator an oscillation of 2,5 MHz and
10% amplitude was applied. The oscillation was superimposed at
the start to the impulse. Measurement no. 980930-1.

Zobject

Zg1

Zg2

Figure 45 Equivalent circuit of the

generating circuits and the test cell

The result of these five measurements performed was that the average
U50 was -42,95 kV. All results were within 0,2% of that average. In
addition to this no influence on the time to breakdown could be
detected.
Therefore, the conclusion is that using the combined generating
circuits instead of the modified conventional circuit has no influence
on the results of the breakdown tests.

5.3 Test set-up and test method used for the tests at
KEMA
5.3.1

Generating circuit

Z object Z g1
Z onject Z g 2
Equation XIX requirements to be met for
minimum influence between the two
generating circuits

The generating circuit used for the breakdown testing of oil and air at
KEMA is the combined measuring circuit. In this paragraph the
circuit used, its design criteria, the test and simulations made and
some of the problems encountered are described in detail. Most of
the work was performed by the author, but she was assisted by others.

5.3.1.1 General requirements

The generating circuit needs to be stable
The two generators should work independent from each other.
To prevent that generator 1 (Zg1 in Figure 45) influences
generator 2 (Zg2) and vice versa, the requirements of Equation
XIX, where Zobject, is the impedance of the object under test (the
test cell) have to be met.
The circuit has to be built in such a way that in case of a
breakdown no components of the generating and measuring circuit
are damaged. Most critical part of the circuit was the 10 kV fast
divider and the resistors parallel to this divider in generator 2.
Therefore a spark-gap was placed in parallel with these, see Figure
46.
It should be possible to generate all required wave shapes.
It should be safe for people to work with.

Figure 46 Protection sphere gap for the

divider and the resistors in generator 2

5.3.1.2 Design and dimensioning of basic components

One thing that became once more very clear from all measurements
and tests performed is that the components do not seem to be what
one think they are. When working with these high-frequencies,
resistors are not pure resistors, capacitors are not pure capacitors and
wires not simply wires. The equivalent circuit of the generating
circuit is presented in Figure 47.
53

Investigations; the test set-up

54

Investigations; the test set-up

Figure 47 Equivalent circuit of the

generating circuit used at KEMA

Figure 48 Oscillation generated with a

bifilar high-voltage resistor

5.3.1.3 Problems associated with interference

Because of the breakdown test that had to be performed, problems
with interference were expected and experienced. The problems with
the interference can be divided into two categories, interference to the
measuring system, interference to the generating circuit. To limit the
interference to a tolerable level, several measures were taken.

55

Investigations; the test set-up

For the digitizer, the computer, the power supply and the low voltage
parts of the dividers, a special designed (by TU Eindhoven) open
EMC cabinet was built, see Figure 75. The measuring cables were
shielded partly by hollow litze-cable and partly by solid cable tubes.
Because of the interference caused by the spark gaps, the cables from
the trigger delay box to the trigger boxes were put in the same solid
cable tubes.
The initial set-up was optimised by performing some breakdown in air
and making some small modifcations to the screening and
connections. The final set-up had no problems with interference;
either they were not present anymore or within the required limits.

5.3.1.4 Problems with the generation of high-frequency

oscillations
From the first measurements it appeared that the beginning of the
oscillations was very steep (the wave shape had more the shape of a
cosine than sine). Since it was not known if the steep front had
influence on the breakdown behaviour (it could be that a breakdown
was provoked by the steep dV/dt above a certain voltage level and not
by the voltage level itself), the problem needed to be investigated.
First idea was that the problem was caused by inductances in the
circuit of generator 2. After some investigations (at low voltage) it
appeared that it was caused by the internal inductance of the resistor
in generator 2 placed between high voltage and ground (in parallel
with the divider). Some simulations were made in addition to these
measurements and besides that they showed good agreement with the
measurements performed, they showed that the inductance of the
high-voltage resistor was 1,44 H. In Figure 49 the wave shape of the
oscillations generated with a low voltage resistor is shown, in Figure
48 the wave shape of the oscillations generated with a high-voltage
resistor is shown. To avoid these high frequency oscillations, special
non-inductive high-voltage resistors were purchased and used.

5.3.1.5 Other aspects of the circuit

Many aspects played an important role in the proper functioning of
the test circuit. Some of them are described in the previous
paragraphs, other ones will be shortly mentioned here.
As is well known, the distance of the spark gaps influences the shape
of start of the impulse and oscillation. In the set-up, the gap distance
had to be more or less perfect. (If the gap distance to large at
oscillation part, the first part of the oscillation was disturbed. If the
gap distance was to small, a spontaneous trigger could occur). This
meant that when the voltage to be generated had to be changed, the
gap distances needed to be changed too. This seems an unimportant
matter, but in this case it is important since it was not known if it
could have an influence on the test results.

Figure 49 Oscillation generated with a low

inductive low voltage resistor

56

Investigations; the test set-up

In the trigger-box overvoltage protection were placed, to prevent
spontaneous triggers and to protect the devices inside the trigger-box.
The overvoltage protections placed close to the digitizer, intentionally
placed to protect the digitizer, appeared to have a bandwidth limiting
Figure 52 Drawing of the components of
the delay trigger box

Figure 50 Photograph of the delay trigger

box

character and had to be removed, risking damage to the digitizer.

5.3.2

Trigger delay

One of the advantages of the combined generating circuit is that the

two generators can be triggered independently of each other. This
implies that oscillations or overshoot can be superimposed on any part
of the impulse, on the front, on the peak on the tail etc.
To be able to trigger the two generating circuits independent from
each other a delay trigger box had to be designed and built. Because
of interference on the cables from the delay trigger box to the trigger
unit near the trigger gaps (caused by breakdowns and the triggering of
the trigger gaps) and because the output signal of the delay trigger
box had to be in the 100 Volt range, this was not an easy task.
A drawing of the components of the delay trigger box is shown in
Figure 52, a photograph of the delay trigger box is shown in Figure
50, the connections to the 100 Volts supply and to the trigger units of
generator 1 and 2 is shown in Figure 51. The delay can be varied
from 5 s till +5 s.
Because of the influence of the position of the superimposed
oscillations or overshoot on the U50, it was very important that the
delay between the two circuits is very stable (see Figure 88). The
stability of the delay trigger box itself had been checked and the result
was that the output signals were stable within a few ns.
During investigations on the stability of the whole set up it appeared
that the major contribution to the variation of the delay was caused by
instability of the trigger gaps, but still the trigger was stable within
50 ns, which was very good.

5.3.3

Test cell

The test cell was designed for performing breakdown tests on the
insulating materials, oil and air (at normal pressure or below).

Figure 51 Connections of the delay trigger

box to the 100 V DC supply and the
trigger units of generator 1 and 2

57

Investigations; the test set-up

Since it is expected that the differences in U50 of different wave
shapes are very small it is very important that the test conditions and
the quality of the insulating medium is very stable. Therefore one of
the requirements of the test cell is that the test conditions are known
and to some extend adjustable. The most important requirements are
listed below, other relevant requirements for the electrodes and the
test conditions of the dielectric media are listed in the following
paragraphs.
General requirements:
the test cell should preferably be at eye height and good working
height
the test cell should be either made of transparent material for
instance glass or plastics or another insulating material with a part
of it of transparent material. This for being able to check the oil
between the electrodes, the electrodes and to check if breakdown
occurs
the test cell should have two electrodes, one to apply the standard
lightning impulse voltage and the other one to apply the
oscillations or overshoot
the high voltage connections to these electrodes should be in such
a way that no flash-overs occur on the outside of the test cell
the divider of the superimposed oscillations and overshoots should
be placed in a very well defined environment and as close as
possible to the electrode that is connected to generator 2
the connections to the high voltage and to ground should be proper
and robust
it should be possible to evacuate the air above the oil
it should be easy to open and to close the test cell

Figure 53 Drawing of the test cell

58

Investigations; the test set-up

5.3.3.1 Electrode configuration

One of the requirements for the test cell is that the two electrodes
should be replaceable and that different electrode configurations can
be used. For instance sphere-sphere or plane-plane or point-plane.
Another requirement is that it should be possible to adjust the
distance between the electrodes from 1 mm to 10 mm and measure
this very accurately (in m range). The electrode material should be
such that several breakdown tests can be made with one electrode
without having to replace or polish it. The distance between the
electrodes and the wall of the test cell should be at least 6 cm.
For the homogenous field configuration two different types of
electrodes were made, spheres and plates. The sides of the plates
were smoothed according to the Rogowski profile (y = 1/2 * + e(x1)
). A photograph of the two types of electrodes is shown in Figure
54.

Figure 54 Photograph of the sphere and

the plane electrode within the test cell

In this thesis project mainly the sphere-sphere electrode configuration

was used. A field calculation for sphere-sphere electrodes and an
electrode distance of 15 mm and air at atmospheric pressure as
insulating material is made, see Figure 55 and Figure 56.
Figure 55 Field calculation with sphere
electrodes at a distance of 15 mm

59

Investigations; the test set-up

Figure 57 Sphere electrode with tungsten

inlay after a number of breakdowns

To be able to replace the electrodes and fix them in the test cell, the
body of the electrodes is made of brass. The effective surface of the
electrode (about 1 cm in diameter) had been replaced by a tungsten
inlay because of severe surface deformation of the brass caused by
breakdowns. To check how large the surface deformation of the
tungsten inlay would be a number of breakdowns (around 100) have
been applied to the electrodes. From both the visual check with the
aid of a microscope, see also Figure 57, and the results of the
measurements it can be concluded that the deformation of the
tungsten can be neglected.

Figure 56 Same as Figure 55, but a

magnification of the area in between the
electrodes

60

Investigations; the test set-up

5.3.3.2 Test conditions

The test conditions that have influence on the breakdown voltage are
different for the different insulating materials. In the following
paragraphs most of the test conditions that have to be kept constant,
should be influenced or adjustable or should be fulfilled are
explained.

5.3.3.2.1

Air

During the test with air, the pressure, the humidity, the temperature

and the availability of free electrons have influence on the test results.
Therefore these factors should be maintained constant or known.
Also the test cell should fulfil the criteria of IEC 60052 [75]. It
should be possible to place a UV lamp so the gap can be radiated.

5.3.3.2.2

Oil

A number of factors are influencing the breakdown voltage of oil. For

instance the gas content, relative humidity, particle content and, of
course the type of oil.

61

Investigations; the test set-up

Even though oil is a self-restoring medium, it is known that when
breakdown tests are performed after every breakdown the oil between
the electrodes should be replaced since discharges break the oil into
low-molecular gaseous hydrocarbons and elementary carbon [78].
One way to do this is to replace the oil in the whole test vessel after a
number of breakdowns, another way is to circulate the oil.
All methods have some disadvantages. When circulating the oil,
space-charges can be left or gas bubbles can be created, when
changing the oil after each breakdown a larger scatter in results is
expected, when changing the oil after a number of breakdowns, space
charges as well as dirt between the electrodes can be left.
Investigation should make clear what is the best procedure. Since it
was not known on forehand it should therefore be easy to circulate
and change the oil.

5.3.3.3 Design of the cell

From all requirements and with the aid of the first hand-written
design, KEMA-TNK made a professional drawing and suggested
some improvements. This professional drawing of the test cell is
shown in Figure 53. A photograph of the test cell (without oil and
vacuum pump) is shown in Figure 58.

5.3.4

Measuring circuit

For measuring the applied voltage to the test cell that is generated by
the combined generating circuit, two Measuring Systems are
necessary. The Measuring System that measures the lightning
impulse is henceforth named MS1 and the Measuring System that
measures the superimposed oscillations and overshoot is henceforth
named MS2. The applied voltage to the test cell is the difference of
the two measured voltages. The voltage measured with MS1
represents the base curve, the voltage measured with MS2
represents the residual curve R(t) in case the base curve is taken as
the mean curve.

5.3.4.1 Requirements
In principle both measuring should fulfil the requirement of
IEC 60060-2. This implies that the overall measurement uncertainty
should be within 3%. Besides the following requirements:
the cable length of the two measuring systems should be equal and
the time delay between the two measuring systems should be
checked.
the divider of MS2 should have protection in case of a breakdown
both measuring systems should not influence the generating circuit
MS1 should be an approved measuring system and should be able to
measure standard full lightning impulse voltages and chopped
lightning impulse voltages with a maximum of 100 kV.
Figure 58 Photograph of the test cell

62

Investigations; the test set-up

MS2 should measure voltages with a maximum amplitude up to 20
kV. The voltage can be either an overshoot with duration up to 5 s
or oscillations with a frequency of 5 MHz. MS2 should be able to
detect parasitic high frequency oscillations, much higher than 5 MHz,
caused by the dimensions of the generating circuit, since they might
have an influence. The bandwidth of MS2, assuming an RC circuit
and requiring that the scale factor does not deviate more than 1% at
5MHz from the scale factor at for instance 100 kHz, should be 35
MHz.
It was known that the IEC 60060-2 requirements might be difficult to
meet for MS2, but it should be tried.

5.3.4.2 Design and construction of the measuring systems

As discussed in the previous paragraph, for the investigations in oil,
two measuring systems are needed. One to measure the standard
lightning impulses (MS1) and another one to measure superimposed
oscillations and overshoot (MS2).

Figure 59 RCR divider of MS1

63

Investigations; the test set-up

5.3.4.2.1

MS1

For MS1, one part of a mixed (RCR) divider for LI, SI, AC and DC
was taken. The original divider comprises of three HV-parts, when
used together (in series), the rated voltages of the complete divider are
1550 kV LI, 1175 kV SI, 343 kV DC and 220 kV AC.
Photo of the divider (one HV-part of the original) and its connection
to the test cell is shown in Figure 59.
Most difficult in the construction of this measuring system was to get
rid of the interference that was picked up at the bottom of the divider
and the tuning of the low voltage part. It appeared from measurement
that the scale factor of the measuring system was lower for higher
frequencies and therefor amplified the disturbances. By tuning the
low voltage part and to use two zener diodes (that were intentionally
meant to protect the inputs of the digitizer, but appeared to have a
bandwidth limit effect), the interference could be reduced with a
factor hundred to an acceptable level without disturbing its
capabilities of being able to measure standard LI, see Figure 60.

Figure 60 Effect of input protection

The dark line is the impulse measured using a filter, the gray line is
the impulse measured not using the filter. The topline is the difference
between them.

5.3.4.2.2

At first it was thought that for this measuring system a commercial

available high voltage probe could be used. It appeared to be that the
dynamic behaviour of this probe was not fulfilling the requirements
and that they became even worse when the cable length was extended.

Figure 61 Drawing of the D/I divider

High voltage part

MS2

Since it is very difficult to build a high frequency high voltage

divider, two different types of measuring systems were designed and
built. One design was based on the conventional resistive divider and
the other design was based on a D/I (differentiating/integrating)
system. Because of several reasons explained below, the resistive
divider was chosen as divider for MS2.

Ri

Vh

Vout
R
Cp

Ci

Low voltage part

Figure 62 Equivalent circuit of D/I system

Vout

RC
Vh
Ri Ci

Equation XX Relation between in and

output voltage of the D/I divider

64

D/I measuring system

Together with TUE, a D/I system was built. A drawing of the divider
built is shown in Figure 61. The equivalent circuit of the system is
shown in Figure 62. The relation between the output voltage of the
divider (Vout) and the input voltage on high voltage side Vh is given in
Equation XX. Detailed information about the design of this D/I
system can be found in [76].
The design of the D/I system had the disadvantage that the scale
factor of the divider was very high (when it was dimensioned for the
high frequency, large bandwidth requirements). This made it difficult
to measure the overshoots and oscillations, since the measured
amplitude of the signal was only a few times larger than the
magnitude of the internal noise of the oscilloscope.

Investigations; the test set-up

Resistive divider
Because of the requirements to the bandwidth (at least 35 MHz) and
the value of the estimated parasitic capacitance, the resistance of the
divider should be in the range of 5 k. The resistances should be low
inductive. After some investigations it turned out that the optimal
design and construction for the divider is the one that is shown in
Figure 63. Before the construction and the design of this divider was
as it is shown in this figure, a lot of investigations and measurements
were carried out. It appeared that very small things had major
influence on the frequency characteristic and the measurement itself,
e.g. type of coaxial cable, aluminium tape placed on the corners of the
test cell. Also there was a difference between the values obtained by
using a spectrum analyser to establish the frequency characteristic and
the values obtained by using a sinusoidal wave shape generator. But
in both cases the divider appeared to meet the requirements.
A photograph of the divider is shown in Figure 64.

Figure 63 Design and construction of the

divider of MS2
168 matching resistors
aluminium hat
perspex tube, 5 k high voltage resistors
50 , low voltage part
20 meter coaxial cable, 50 RG 214
50 , 15 times, attenuator

5.3.4.2.3

Digitizer

The digitizer used is a Nicolet Power Pro 610, see Figure 75. MS1 is
connected to Channel 1 of the digitizer. MS2 is connected to Channel
2 of the digitizer. The digitizer has a bandwidth of 35 MHz and a
minimum sampling time of 13,3 ns per sample. To reduce
interference to the digitizer and the low voltage part an EMC-screen
was built, see Figure 75.

Figure 64 High frequency resistive divider

of MS2

65

Investigations; the test set-up

Technical Data of LIR1/46.05

On the two measuring systems several measurements have been

performed. One of these measurements is a comparison measurement
with a Reference Measuring System, named LIR1/46.05. Some data
of this Reference Measuring System is given in the column besides
the text. The results of some of the measurements performed are
explained in the next paragraphs.

Nominal Epoch
Up: 100 kV 500 kV
T1: 0,8 s 2,5 s
Scale factor
F = 15580
Overall Measuring Uncertainty
LI 0,84/50 Up:0,93%, T1:6,7%, T2:2,3%
LI 5/50
Up:0,93%, T1:1,6%, T2:2,0%
Last Performance Test
PTB 07-02-1996
Details of divider
Manufacturer
Type
Resistance HV arm 5 k
Resistance LV arm
Damping resistor

5.3.4.3 Calibration of the measuring systems

5.3.4.3.1

Calibration of MS1

For the calibration of MS1, the following properties were checked:

scale factor, linearity, dynamic behaviour and interference.

Scale factor
Haefely
R500 REF
19,1
400

Details of coax cable

Hber-Suhner, G07273D, 75 , 20 meter
Details of attenuator
HV-arm
71,58
LV-arm
1,966
Details of digital recorder
Manufacturer
Nicolet
Type
Power Pro 610
Bandwith
25 MHz
Sample frequency 76,9 MHz

First, the scale factor has been measured using DC on 7 July 1998.
The scale factor was according to that measurement 19526. For
linearity tests and the comparison measurement, the scale factor was
set to 19530. A comparison measurement against LIR 1 was carried
out on 18 January 1999 at 96 kV. The result was that the scale factor
was 19950 for 100 kV positive polarity. The measurement uncertainty
of this calibration was 0,94%. From the linearity measurements it
appeared that the best scale factor would be 20060. This value was
used for all breakdown tests performed later.

Linearity
A linearity test of MS1 in comparison with LIR 1 was performed for a
voltage level of 2,5 kV up to 96 kV. The result of this linearity test is
presented in Figure 65. The contribution to the measurement
uncertainty caused by the non-linearity is 0,921%.

Dynamic behaviour
The dynamic behaviour of the measuring system was checked by
measuring the rise-time by applying a high voltage step (rise time of
2,5 ns) and by measuring the scale factor in a frequency range 50 kHz
to 20 MHz when a sinusoidal wave of 10 Volt amplitude.
A photograph of the latter measurement in presented in Figure 69.
The results of the latter test are presented in Figure 66. The error
caused by the dynamic behaviour of the measuring system in the
range of 50 kHz to 0,7 MHz, i.e. for time parameters from 5 s to 0,5
s is 1,5%.
1,1

1,05

Figure 65 Results of linearity test of MS1

relative scale factor [1]

0,95

0,9

0,85

0,8

0,75

0,7

10,0E+3

66

100,0E+3

1,0E+6
Frequency [Hz]

10,0E+6

100,0E+6

Scalefactor [1]

Investigations; the test set-up

1,4
1,2
1
0,8
0,6
0,4
0,2
0
10,0E+3

Figure 66 Scale factor of MS1 in the range

of 50 kHz to 20 MHz

100,0E+3

1,0E+6

10,0E+6

100,0E+6

Frequency [Hz]

Interference
Investigations performed demonstrated that the most important
contribution to the interference to MS1 is the triggering of the spark
gaps and is picked up by the low voltage part of the divider. The
interference caused by the measuring cable itself is negligible.
In case a 5,8 kV oscillation of frequency 2 MHz is generated, the
MS1 measures a voltage of 56,9 mV. When multiplied with a scale
factor of 20060 this equals to 1,14 kV. In relation to the expected 70
kV LI, this means that the interference is 1,6 %. (see Figure 67)
In addition to this measurement, also a simultaneous measurement
(MS1 and MS2) was performed while applying a lightning impulse
voltage. The amplitude of the interference is in the same range as the
previous measurement (see Figure 68). The applied voltage is 12,5
kV.

Figure 69 Test set-up used for the

measurements of the dynamic behaviour
using the sine wave generator

Figure 68 Interference measured. Grey is

MS1 and black is MS2.

Figure 67 Interference measurement of

MS1. Black is generated oscillations, grey
measured voltage by MS1.

67

Investigations; the test set-up

5.3.4.3.2

Calibration of MS2

For the calibration of MS2 (the home-built resistive divider), the

following properties were checked: scale factor, linearity, dynamic
behaviour, short term stability and interference.

Scale factor
First, the scale factor has been measured using DC on 7 July 1998.
The scale factor was according to that measurement 3077. For the
linearity test and the comparison measurement, the scale factor was
set to 3040. A comparison measurement against LIR 1 was carried out
on 18 January 1999 at 19 kV. The result was that the scale factor was
2996 for 19 kV negative polarity. The measurement uncertainty of
this calibration was 0,94%. From the linearity test it appeared that the
scale factor of 3040 was a good scale factor.
2,0%

Figure 70 Results of linearity test of MS2

1,5%

1,0%

Error

0,5%

0,0%

-0,5%

-1,0%
Positive polarity
Negative polarity

-1,5%
0

10

12

14

16

18

20

Voltage [kV]

Linearity
A linearity test of MS2 in comparison with LIR 1 was performed for a
voltage level of 2,5 kV up to 19 kV. The result of this linearity test is
presented in Figure 70. The contribution to the measurement
uncertainty caused by the non-linearity is 1,62%.

Dynamic behaviour

Figure
71
Dynamic
behaviour
measurements using a spectrum analyser

68

To determine the dynamic behaviour three types of measurements

were performed; a step response measurement, an automatic
frequency response measurement using a spectrum analyser (see
Figure 71), a manual frequency response measurement using a
sinusoidal wave shape generator. The result of these measurements
was that according to the automatic frequency response measurement,
the response of the divider was within 0,6 dB from 100 kHz to 50
MHz. The ratio at 100 kHz and 50 MHz are about the same level,
maximum ratio difference is at 30 MHz. According to the manual
frequency response measurements, the 3 dB point is at 25 MHz (see
Figure 72). For the step response measurement a pulse with a front
time of 2,5 ns is applied, the front time of the pulse measured by the
divider is 4 ns. Taking the result of the manual frequency response
measurement, which results in the smallest bandwidth, the error

Investigations; the test set-up

caused by the dynamic behaviour is 0,7% in the range of 100 kHz to 2
MHz.
1,1

Figure 72 Scale factor as function of the

frequency for MS2

1,05

relative scale factor [1]

0,95

0,9

0,85

0,8

0,75

0,7

10,0E+3

100,0E+3

1,0E+6

10,0E+6

100,0E+6

Frequency [Hz]

Interference
For MS2, also three interference measurements have been performed
(interference on the measuring cable, interference caused by generator
1 and interference caused by generator 2). The interference caused by
the trigger of the gap when applying a standard impulse was the
largest. The measuring system measured 12 mV (see Figure 68). The
oscillations to be measured will be between 4 and 14 kV, therefore
the interference is maximum 0,9%.

Short-term stability
To establish the short-term stability of MS2, 60 1,2/50 impulses of
19 kV with a time interval of approximately half a minute were
applied to MS2. Simultaneous readings were take of LIR1 and MS2.
The results are presented in Figure 73. The average scale factor of
the 60 applied impulses is 0,9872. The maximum and minimum scale
of the graph is this value with respectively +1% and 1%. From the
figure it can be seen that the short term stability of MS2 is well within
the required limits. The average scale factor of the first 10 impulses
is 0,9866 the average scale factor of the last 10 impulses is 0,9883
this means that the uncertainty contribution of the short term-stability
is 0,09%. At the same time the uncertainty of the short-term stability
for the time parameters was determined, for T1 this was 0,27% and for
T2 this was 2,15%.
0,9972

Figure 73 Stability of the scale factor of

MS2 compared to LIR1 at 19 kV for 60
impulses

0,9947

scale factor

0,9922
0,9897
0,9872
0,9847
0,9822
0,9797
0,9772
1

11

16

21

26

31

36

number of impulses

41

46

51

56

69

Investigations; the test set-up

5.3.4.4 Validation of the software

The software used has been validated. It was found that differences up
to 0,4% can occur in the peak value due to the nature of noise. [77]
Differences up to 0,8% in the peak value can occur when high
frequency oscillations occur on the peak. This will not be the case.

5.3.4.5 Check of the time difference between the two

measuring systems
The time delay between the two measuring systems is checked by
applying a chopped impulse. In Figure 74 it can be seen that the time
difference is within a few nanoseconds. The black signal is the signal
measured with MS2, the grey signal is the signal measured with MS1.

Figure 74 Time delay between MS1 and

MS2

70

Investigations; the test set-up

5.3.4.6 Properties and measurement uncertainty

The measurement uncertainty is calculated using the method of
IEC 60060-2, as explained in paragraph 2.3.8. This implies that the
general formula for the calculation is the one that is given in Equation
XXI. With the results of the calibration and the information found in
the description of the equipment used, the measurement uncertainty
for one measurement is calculated with Equation XXII.

5.3.4.6.1

MS 1

With:
a1 =
a2 =
a3 =
a4 =
a5 =
a6 =
a7 =
a8 =
a9 =
sr =
Uk =

0,55%
0,92%
estimated: 0,1%
estimated: 0,2%
estimated: 0,1%
no effect: 0%
1,5%
1,6%
0,4%
0,15%
0,94% for k=2

U k
sr

1 2
U ts
Rn R x2 k r
3
k k n

Fm

n 1

Equation XXI General formula for the

calculation of measurement uncertainty
n = number of measurements
Fi = scale factor of measurement i
Fm = average scale factor of n
measurements
sr = experimental standard
deviation
t = student t factor
k = coverage factor
Uk = measurement uncertainty of the
reference measuring system or the
calibration
aI = systematic contribution to the
measurement uncertainty

Therefore, the measurement uncertainty for k=2, B=95% for MS1 for
one measurement is 3,1%.
U k

5.3.4.6.2

MS2

U
1
ai2 i

3
ki

t sr

With:
a1 =
a2 =
a3 =
a4 =
a5 =
a6 =
a7 =
a8 =
a9 =
sr =
Uk =

1,47%
1,62%
0,09%
estimated: 0,2%
estimated: 0,1%
no effect: 0%
0,7%
0,9%
0,4%
0,066%
0,94% for k=2

The measurement uncertainty for k=2, B=95% for MS2 for one
measurement is 3,1%

Equation XXII Measurement uncertainty

for one measurement
a1 =
a2 =
a3 =
a4 =
a5 =
a6 =
a7 =
a8 =
a9 =
sr =

correction for the scale factor

non-linearity
short term stability
long term stability
temperature effect
proximity effect
dynamic behaviour
interference
software
experimental standard
deviation of the calibration

71

Investigations; the test set-up

5.3.5

Control software and evaluation methods

For the evaluation of the parameters of the impulses applied during

the experiments and the control of the digitizer a special software
package, with the name PARAMETER, was developed by FFII/LCOE
(one of the partners in the European project). Besides the standard
IEC 60060 parameters some research parameters are evaluated. The
parameters to be evaluated have been proposed and discussed within
the European project team. How the software controls the digitizer
and which parameters are used and how they are evaluated is
explained in the next three paragraphs.
Figure 75 Digitzer,
evaluation software

computer

and

5.3.5.1 Control of the digitizer and acquisition of the signals

The software can control a Nicolet Power Pro 610 or a Tektronix
RTD 710A. Signals either from only Channel 1, from only Channel 2
or from Channel 1 and 2 can be acquired. Because it might be
necessary to evaluate the measured curves once more after all
measurements were carried out to test the validity of a new parameter,
it was decided in the project that all measured curves should be
stored. The software acquires the signals from the chosen Channel(s)
and saves the data to disk. Channel 1 is saved as a *.c1 file, Channel
2 as a *.c2 file and in case of a two channel measurement the applied
voltage over the test cell is the difference between the two channels is
saved as a *.c12 file, taking into account the two different scale
factors. In all case of the Nicolet Power Pro, the used record length of
the signals is 6300 samples per channel, the sampling rate is 13,3 ns
for each channel.
Once the signals have been acquired, displayed on the screen and
stored to disk the parameters can be evaluated and displayed on the
screen as well, see Figure 76. The complete set-up of computer,

Figure 76 Measured signal and evaluated

parameters displayed on the screen

digitizer and software is shown in Figure 75.

The parameters evaluated are stored in an ASCII-file (*.pc1, *.pc2,
*p12 format), so that after measurements some automatic evaluation
of this output files can be made, for instance to calculate U50 or a
mean area above a certain level. Which parameters are evaluated is

72

Investigations; the test set-up

explained in the paragraph 5.3.5.2. How the U50 is calculated is
explained in the paragraph 5.3.5.4.

5.3.5.2 Evaluated parameters

The evaluated parameters can be divided into two categories:
1.

2.

IEC 60060 parameters

BL, base line
T1, front time
T2, time to half-value
Tc, time to chopping
O1, origin
Ut,, test voltage
Uc, voltage at Tc
research parameters
Uprmc
Up
f, frequency of the oscillation
AI, amplitudes of the superimposed oscillation
, wave integral above a reference voltage
OPV, overshoot peak value
VPV, virtual peak value
Wave integral above VPV
, overshoot amplitude
d, overshoot duration
In addition to these parameters, some curves can be drawn.
BL, Base Line
RMC, Reference Mean Curve
R(t), Residual Curve
SE, Single Exponential of Tail
dU/dt, Impulse Steepness at each point

5.3.5.3 Evaluation procedure

The flowchart presenting the steps made to evaluate the parameters
(IEC 60060 and research) and a detailed description how each
parameter is evaluated are presented in Appendix 1.
In case of a two channel measurement and in case generator 1 has no
influence on generator 2 and vice versa, the residual curve is the
signal recorded at Channel 2. The mean curve is the signal recorded
at Channel 1.

5.3.5.4 Software for the determination of the U50

For the determination of U50 a statistical analysis software
programme, with the name MLM 53, version 5.2/5.3, is used. MLM
53 is programme that analyses statistically the measurement results of
multiple level tests and has been developed for impulse testing. This
programme has been developed by the Technische Universitt
Dresden, Insitut fr Hochspannungs- und Hochstromtechnik, in
cooperation with Dr. W. Hauschild. Detailed information about how
the data is statistically analysed can be found in [78].

73

Investigations; the test set-up

Once the measured data have been entered, a graph showing the
distribution function (breakdown probability as function of the
voltage) with the maximum likelihood, the confidence limit of this
function, the U50 and the sigma can be displayed on the screen or
printed, see Figure 77.

Figure 77 Example of a distribution

function of the measurement series:
44,4 kV, 10 impulses, 0 breakdowns;
44,8 kV, 10 impulses, 0 breakdowns;
45,1 kV, 10 impulses, 1 breakdowns;
45,25 kV, 10 impulses, 6 breakdowns;
45,5 kV, 10 impulses, 8 breakdowns;
45,8 kV, 10 impulses, 10 breakdowns.

With some of the measurement series (in oil and in air) it has been
investigated which distribution showed the best results. In most cases
in air the Weibull and the normal distribution function showed similar
and good result. In most cases for oil, the normal distribution showed
good results. To make sure that the difference between the test results
in the work at KEMA and within the European project was not
influenced by the difference in distribution function, one distribution
function had to be chosen. Therefore, in all cases in this thesis and in
the European project a normal distribution function is chosen for the
approximation function.

5.3.6

Stability of the whole set-up

To check the stability of the whole set-up, several measurements in

air have been performed, only some of the results are presented here.
To check how the mounting and de-mounting of several components
and setting the gap distance, changing it, and setting it back again
influenced the results two measurements in air were performed. The
result of Measurement 980929-3 was: U50 = -42,866 0,132 kV, the
result of Measurement 981002-1 was:U50 = -42,988 0,169 kV. The
conclusion was that the influence of changing some of the
components and the gap distance was within the standard deviation
(0,4%) of the determination of the U50 and even within 0,3%.

45,6122

U50 in [kV]

45,5122

If the voltage is set to a certain value 10 times, eg. 41,7 kV, the
average of these 10 voltages was 41,6758, the maximum voltage
differed 0,12%, the minimum voltage 0,16% from this value.

45,4122
45,3122
45,2122
45,1122
45,0122
990119 990120-2 990121-2 990205-1 990205-3 990208-2

Measurement number

Figure 78 Measurements in air with a

sphere-sphere gap of 15 mm

74

To check the reliability of the system several comparison

measurements with IEC 60052 as reference were made. The results
are presented in Figure 78. The result was that the difference of the
average of the measured values to IEC 60052 less than 1,3%. The
values do not differ more than 0,36% of each other.

Investigations; the test set-up

5.3.7

Test method and procedure

After the generating circuit, the test cell, the measuring systems and
the evaluation methods have been explained, the test method and the
test procedure of the breakdown tests will be explained.
It appeared that after all problems associated with the generating
circuit, the measuring systems and the software had been solved,
finding a suitable and reliable test procedure was more difficult than
expected. The results of the several investigations made are not
presented in this thesis, only the procedure finally chosen is explained
here.

5.3.7.1 Test method and procedure for air

For the tests in air the multiple level test for establishing the U50 is
chosen. This because of the limitations of the generating circuit.
The electrodes are placed in an area with enough fresh air, since it
appeared from measurements that inaccurate and unreliable results
were obtained when the electrodes were used inside the plastic tube.
Therefore, for these measurements, the plastic tube is removed.
Before starting a test series, the environmental conditions are
measured and noted. In some cases UV-lamps were used, in some
cases not. This is indicated when the measurement results are
presented.
The tests in air are made using the sphere-sphere electrodes, without
tungsten inlay. It was checked my measurement and by visual
inspections, that the electrode was not damaged noticeable.

5.3.7.2 Test method and procedure for oil

5.3.7.2.1

Test method

From the many investigations performed towards the best test method
for oil, it could be concluded that the best suitable and most reliable
and independent method was the multiple level test for establishing
the U50. Because of the limitations of the test set-up it was not always
possible to have 6 levels, but only 5 levels (one with all breakdown,
one with all withstands and three with withstands and breakdowns)
with equal intervals can be obtained. Each level consists of at least
10 shots.

5.3.7.2.2

Test procedure

One of the conclusions drawn from the investigations performed

towards the best test procedure is that every test series (this is a
multiple level test, comprising of at least 50 shots) should be done
using fresh and conditioned oil.
Before the test series is started at least two breakdowns should be
made. After these breakdowns, three withstands should be made and
after two minutes rest the test series can be started. After a long break
at least one breakdown; followed by three non-breakdown is made.
Since the tests are independent from each other, it is not so important
at which voltage level the test series is started. But, because of the

75

Investigations; the test set-up

degradation of the oil after too many breakdowns (more than
approximately 100), it is preferred to start in case of an unknown U 50
at a low voltage level and not at a too high voltage level.
Because of the memory effect of the oil, it was found that one cannot
simply use the multiple level method. After thorough investigations a
special procedure for the multiple level test was found to be the best.
This procedure is used for further tests and is explained below.
After each withstand one minute should be waited before the next
impulse is applied. After five withstand in a row, a breakdown should
be made (this impulse is not used as a shot for the test series). After
every breakdown three withstands, far below the U10, should be made
(these three withstands are not used as shots for the test series). After
a breakdown followed by three withstands two minutes should be
waited before the next impulse is applied.

5.3.7.2.3

Treatment of the oil

As mentioned before, for each test series fresh and conditioned oil
will be used.
Because at KEMA no treatment unit for oil was available the fresh oil
is take from a barrel of new oil and it is not regenerated. After the oil
was used for one test series, it was removed from the test cell and
discarded.
The conditioning of the oil comprises of the following steps:
1. filling the test cell with oil
2. conditioning of the oil before starting the test series
3. conditioning of the oil during the test series

Filling the test cell with oil

The air and oil in the test cell are at normal surrounding pressure. In
case there is still oil in the test cell, the oil is taken out completely of
the cell by opening the bottom valves. The oil is flowing out by
gravity forces.
Before filling with fresh oil, the test vessel and the electrodes are not
being cleaned. It was checked that the electrodes, which are from
tungsten, are not deformed. Besides that, the electrodes do not get
dirty from the relatively few breakdowns. Cleaning it may introduce
more dirt.
After that the used oil is removed, all valves are closed and the test
vessel is pumped vacuum. When it is vacuum, the valves to the barrel
of oil are opened and the oil is being sucked in the test cell, see
Figure 79. The fresh oil is pumped into the test vessel from a barrel
of oil. This barrel is normally closed and only opened for a short time
during the filling of the test cell.

Figure 79 Filling the test cell with oil

76

When there is enough oil in the vessel, the valves to the barrel will be
closed and the test vessel will be pumped vacuum, so that the air
inside the oil gets out. After this, the oil is being circulated, for some
time by the pumping system, and it is again pumped vacuum. This is
repeated for two times.

Investigations; the test set-up

Conditioning of the oil before starting the tests series

When the fresh oil in the barrel (which should be done no longer than
16 hours before the test series is started) and has been brought to
vacuum for three times, the test cell is brought back to surrounding
pressure by opening the valves. Before the start of the test series, the
electrode distance are checked. Half an hour before the first test the
oil pump of the circulating system is started and the DC supply of the
three ring electrodes is switched on (see explanation later).
Before each test series, the temperature, the air pressure and the
humidity are measured and the values are written down.
When the barrel was opened the first time and when the last oil was
taken out, an oil sample was taken and an analysis was made.
A photo of the vacuum and oil pump, the DC supply and the tube
connections is shown in Figure 80.

Conditioning of the oil during the test series

During the test, an oil pump continuously circulates the oil between
the electrodes. This is done for three reasons:
1. to get rid of the dirt in the gap, but for that you do in principle not
need continuously pumping

Figure 80 Vacuum and oil pump

2. every time the pump starts, you might get some small air bubbles,
to prevent these air bubbles, the pumping is continuously
3. to get rid of the charges in the oil. For this the oil is pumped
through a hose containing three ring electrodes. The outer ones
are on earth potential, the middle on 30 V DC, see Figure 81
(Investigations pointed out that 1 kV DC was decreasing the
properties of the oil, AC voltage was equally worse. If 30 V DC is the
optimum, we do not know since we are not making investigations to
this and the 30 V DC gave very stable values)

Influence of the oil pollution caused by breakdowns to the

breakdown voltage
Two oil samples were taken. One sample, sample no. 61534 was of
the fresh and new oil. The other sample, sample no. 61535 was taken
after a significant number of breakdowns (at least 50). The results of
the analysis performed on these two samples at KEMA are presented
in detail Appendix 5. The breakdown voltage for 50 Hz, changes
from 75 19 kV per 2,5 mm for fresh oil to 69 17 kV per 2,5 mm
for the used oil. The change is well within the measurement
uncertainty of the measurement, so no significant influence can be
detected.

Figure 81 The three ring electrodes

connected to the 30 V DC supply

Standard deviation
Because of several physical reasons, the standard deviation in the
obtained results is around 4 kV for a breakdown voltage of 70 kV
(around 5%). For the purpose of this project it is rather large, but
comparing the results to values heard from other institutions
performing breakdown tests in oil (around 20%), it is rather low.

77

Investigations; the test set-up

Figure 82 The two generating circuits, the
test cell and the two dividers

5.3.8

Conclusions

The preparation of a reliable generating circuit, measuring circuit, test

cell and test methods at KEMA took at least 1 year. A lot of
investigations towards the best generating circuits, measuring circuits,
test cell, test methods and their properties were performed. For sure
the tests performed with the designed and built test set up and test
method give reliable and reproducible results.
In Figure 82 a photo of the set up built at KEMA (without the
pumping system of the oil) is shown.

5.4
5.4.1

Test set-up used of the tests at TU Graz

Generating circuit used for the tests at TU Graz

The circuit used at TU Graz was the modified conventional circuit.

This circuit was sufficient for the type of investigations to be
performed.
For the generated lightning impulse voltages of different types, two
different circuits were used. Schematic descriptions of these two
circuits are presented in Figure 83 and Figure 84.

Figure 83 circuit used for the standard

impulses and for the standard impulse with
superimposed 200 kHz oscillations and 1,3
s overshoot

Figure 84 circuit used for the standard

impulse with superimposed 1 MHz
oscillations

78

Investigations; the test set-up

A photograph of the circuit used to generate the damped sine wave of
which the results are not presented in this report is presented in Figure
85.
The values of the different components used in the different circuits
are presented in Table 3.
Table 3
Waveshape

Cs in nF

Re in

Rd in

Ld in H

Cb in nF

a) Standard
Impulse

35

400

375

b) Overshoot
1,4 s
35

400

300

165

c) 200 kHz
Oscillations

35

400

41

13,6

d) 1000 kHz
Oscillations

35

400

50

78

0,24

e) Damped
sine wave 200
kHz
35

400

2,5

5,6

5.4.2

Test vessel

A photograph of the test vessel in which the samples are placed is

given in Figure 86.
Figure 85 Photograph of the circuit used
for the one of the experiments (200 kHz
damped sine wave)

Figure 86 Test vessel with sample

79

Investigations; the test set-up

5.4.3

Samples

Two different types of samples were used for the investigations:

Sample type A: The two copper strips, wrapped with paper, are only
on two places wrapped together
Sample type B: The two copper strips, wrapped with paper, are over
its whole length wrapped together
The samples were placed at the disposal by Siemens. For which they
are thanked.

5.4.4

Test method and procedure

Since the samples consist of, amongst others, paper material it is very
important that the paper contains as less as possible moisture, since
this has a effect on the breakdown behaviour. It is also very important
that each sample is of the same quality. If this is not the case, it will
be almost impossible to evaluate the results because of the wide
spread in the results.
Figure 87 The samples

To reduce the moisture, the samples were dried under vacuum, 1 mbar
at 105C (2C) for 24 hours. After the samples are dried, they were
places in oil for 24 hours. After that this process was completed the
tests were started.
To find the breakdown voltage the progressive stress method was
used with 3 shots at the same level. For each test series at least 10
samples were used.
The tests were started at 85% of the U20 (20% breakdown voltage)
and the voltage was raised in steps of 5% until breakdown occurred.
After each shot 2-3 minutes was waited before the next impulse was
applied.

5.4.5

Digitizer and evaluation software

For the tests performed at TU-Graz, the same digitizer and evaluation
software was used as that used at KEMA. Also for the evaluation of
the U20, the same software was used as that one used at KEMA.

80

Investigations; the results

6 Investigations; the results

6.1

Introduction

In the previous chapter the test set-up, the test methods and test
procedure used at KEMA and TU Graz are explained in detail. In this
chapter the results of the investigations performed at KEMA and at
TUG are explained. The results of the investigations performed by
the other partners in the European project, Schering Institute, LCOE
and NGC, are explained as well. The author was the project-leader
and initiator of the European project, performed the measurements at
KEMA and evaluated the results performed at TU-Graz.

Before the measurement results will be presented, the terminology

used will be explained. When the statement, there is no influence is
made, it means that two or more measurements performed do not
differ from each other. When the statement, there is no significant
influence is made, it means that there is a difference between two or
more measurement results, only that the difference between them is
within the standard deviation of the U50. In case the statement, there
is an influence is made, it means that the difference between two or
more measurement results is larger than the standard deviation of the
U50.
For the measurements from which the results are presented in this
chapter, with the exception of the measurements performed at TU
Graz, the combined generating circuit as described in the previous
chapter is used. The peak value of the curve generated by generator
1, the part that generates the standard impulse is referred to as the
peak value of the RMC, Up(RMC). The peak value of the difference
between the two applied waves (the full impulse and the oscillations)
is called the peak value of the total curve Up. For the measurements
performed at TU Graz, the curves measured when applying only the
standard impulse could be used for fitting the curve, or in case of
overshoot the curve was fitted with a exponential function that fitted
the tail. The peak value of the measured curve is referred to as Up.
The peak value of the fitted curve is called Up(RMC).

6.2

Test to be performed

Before the breakdown tests were started at the different location, took
place on which test should be performed. Redundant tests should be
prevented, but on the other hand there were a lot of research questions
that had to be solved. In this paragraph the test that were intended to
be performed are explained. Of course, like in every other research
project, on the road, some more or some fewer tests are performed,
since the information available from the tests already performed
changes the view and the number of open questions.

Comparison test between the laboratories

To check the influence of systematic and random errors, the
laboratories participating in the European project performed a
comparison measurement using a sphere-sphere gap in air. The test

81

Investigations; the results

should be performed with a 1,2/50 with 2 MHz oscillations
superimposed so that the peak values coincide.

Check of the validity of the parameters used in IEC 60060-1

for standard lightning impulse voltages
The standard full lightning impulse voltage is a 1,2/50 s impulse. A
tolerance of 30% is allowed for the front time, 20% for the time to
half-value and 3% for the peak value. One of the goals of the
experimental investigations will therefore be to prove that, indeed, the
U50 is within certain tolerance limits, when the front time and the time
to half-value are varied between or even outside these limits.

Peak value of the curve versus peak value of the mean

curve as test voltage
When evaluating the parameters of a measured lightning impulse
voltage according to the present IEC 60060-1, in some cases the test
voltage is the peak value of the mean curve. It seems not possible to
define a mathematically mean curve that is suitable/acceptable for
all lightning impulse voltage that are generated during normal testing.
Besides this not all persons agree with the relevancy of the test
voltage obtained by taking the peak value of this mean curve.
Therefore another goal of the experimental investigations will be to
investigate whether the peak value of the curve or the peak value of a
mean curve is the most important parameter when evaluating the test
voltage.

Oscillations versus overshoot

In the present IEC 60060-1 no difference is made in the evaluation of
lightning impulse voltage with overshoot and the evaluation of
lightning impulse voltages with oscillations. It is necessary to
investigate whether this assumption is correct. Therefore tests have to
be performed with both oscillations and overshoot.

The magic 500 kHz and 1 s limit

At the moment in the IEC 60060-1 the test voltage is equal to the peak
value in case the frequency of the oscillation is below 500 kHz (or the
overshoot longer than 1s) and it is equal to the peak value of a
mean curve in case the oscillation frequency is above 500 kHz (or
the overshoot shorter than 1 s). Why is this sharp limit and is it on
the right frequency? To answer this question investigations with
oscillations and overshoot of different frequencies and duration have
to be performed.

Standard and non standard wave shapes

Right now, a lightning impulse with oscillations or overshoot of
amplitude larger than 5% is called a non-standard impulse. Is it really
so that the evaluation is different? Is it so that oscillations of 5%
amplitude do not have a significant effect on the 50% breakdown
voltage and oscillations of more than 5% do? To answer this question
investigations with oscillations and overshoots of different amplitudes
have to be performed.
82

Investigations; the results

New parameters
Another assignment in this project is to investigate totally new
parameters. In case that one finds out that the present parameters are
either not relevant or too difficult to implement one does not have to
perform all test over again, since one has the data already available.
Therefore it was also agreed upon that all test data should be stored
digitally so that post processing was possible.

Summary of the tests to be performed

To find all the answers to the questions mentioned before and in
Chapter 4.6, standard lightning impulse voltages with different front
times and time to half value, standard lightning impulse voltages with
superimposed oscillations and overshoot of different amplitudes and
frequencies have to be generated.
An overview of the tests that initially have been thought of
performing is given in Table 4.
Table 4 Overview of test to be performed on oil
Parameters to vary
Wave shape Oscillation
front/tail
frequency

Overshoot
duration

Amplitude of oscillation or Position of the

overshoot
oscillation or overshoot

1,2/50

200 kHz

0,625 s

20%

on the peak

1,56/50

500 kHz

1 s

10%

on the front

1,84/50

800 kHz

2,5s

5%

1,56/40

2 MHz

5 s

0,84/40

5 MHz

The oscillation has at least three cycles of amplitude >5% of the

amplitude of the first peak.

83

Investigations; the results

6.3

Comparison measurements in air

Before the investigations in solid, liquid and gaseous insulation media

could be performed, the test set ups of the different partners had to be
tested on reproducibility and correctness by a comparison
measurement. All partners performed a breakdown test in air using
sphere-sphere gaps and applied a 1,2/50 s impulse with a
superimposed 2 MHz oscillation at the peak of the impulse.
The disruptive discharge of external
insulation depends on the atmospheric
conditions [1]. By applying correction
factors, a disruptive discharge voltage
measured in given test conditions may be
converted to the value, which would have
been obtained under the standard reference
conditions. The disruptive discharge is
proportional to Kt that results from:

Kt k1 k2
Equation XXIII atmospheric correction
factor
The measured disruptive discharge
voltages U are corrected to U0
corresponding to the standard reference
atmosphere by dividing by Kt:

It was agreed upon that the results should be converted to standard

reference conditions (see explanation to the left). The initial
measurements performed showed some inconsistency. To find the
reason for this inconsistency several additional measurements were
performed by KEMA and LCOE.

6.3.1

Results of test performed at KEMA

6.3.1.1 Initial measurement

Two measurements (measurement no. 990208-1 and 990208-2) were
performed for the inter-comparison measurement (in a 15 mm spheresphere gap, sphere diameter 6 cm, minimum radiation). T 1 and T2 of
the full impulse were respectively 1,18 s and 49,2 s. The waves of
measurement no. 990208-1 had oscillations superimposed and the
waves of measurement no. 990208-2 were smooth impulses. The
superimposed oscillation had an amplitude of 4,4 kV, its frequency
was 1,85 MHz and it was placed around the peak (1,6 s delay).

U0 U / Kt

Up(RMC),50 of measurement no. 990208-1 was 43,4 kV with a standard

deviation of 0,6 kV and UP,50 was 47,8 kV with a standard deviation
of 0,5 kV. This means that the amplitude of the oscillations was 10%.

Equation XXIV correction to reference

atmosphere

Up(RMC),50 of measurement no. 990208-2 was 45,2 kV with a standard

deviation of 0,12 kV.

air density correction factor k1

The superimposed oscillations caused that the Up(RMC),50 decreased

from 45,2 kV to 43,4 kV.

k1 m
b 273 t0

b0 273 t

The result of the inter-comparison test performed at KEMA was

therefore that the influence of superimposed oscillations of frequency
1,85 MHz, amplitude 10%, placed around the peak (delay 1,6 s) to
the Up(RMC),50 is 4%.

Equation XXV
humidity correction factor k2

k2 k w
Equation XXVI

84

6.3.1.2 Additional measurements

Influence of trigger delay
To study the influence of the trigger delay, in case oscillations are
superimposed, to the Up(RMC),50, a measurement series was performed.
The electrode configuration used was a sphere-plate gap at a distance
of 15 mm in air. The air is irradiated with UV light. The standard
impulse had a 1,6/50 shape. The superimposed oscillations had a
frequency of 2,5 MHz and amplitude of 10%. The standard deviation
in the determination of the Up(RMC),50 was around 0,4%. In Figure 88
the results are presented and a trend-line is added.

Investigations; the results

-41

Figure 88 Influence of the trigger delay

time between the start of the full impulse
and the start of the oscillation of frequency
2,5 MHz and amplitude 10% for sphere
plate (15 mm) configuration in air with
UV light to the Up(RMC),50

Up(RMC),50 in [kV]

-41,5
-42
-42,5
-43
-43,5
-44
-44,5
0

delay time in [s]

From the results presented in Figure 88 it can be seen that

superimposed oscillations have no influence on the Up(RMC),50 when
they are placed near or before the origin of the impulse. They
however have an influence on the Up(RMC),50 when they are positioned
close to the peak. Therefore the conclusion of this measurement
series is that the trigger delay has an influence on the Up(RMC),50.
The influence is the largest when the delay is around 3 s for a
1,6/50 s impulse and 2,5 MHz oscillation frequency. At that point
the influence to the Up(RMC),50 is expected to be around 4,8% (a
Up(RMC),50 of around -42 kV in case oscillations superimposed on the
peak versus a Up(RMC),50 of around -44 kV in case of a standard
impulse).

Influence of irradiation towards the 50% breakdown voltage

of standard impulses
To investigate the influence of irradiation to the 50% breakdown
voltage a measurement series in air was performed. During this
measurement series a sphere-sphere gap with a gap distance of 15 mm
and impulses of positive polarity were used. The results are presented
in Figure 89. It can be seen that the first three measurements have a
higher U50 value than the last three measurements.
The average value of the first three measurements differs 0,3% from
the average value of the last three measurements. From this it might
be concluded that UV irradiation has the effect that the U50 increases
a little. The difference is however within the standard deviation of
these measurements (0,4%) and therefore the conclusion is that no
significant influence of the irradiation on the U50 can be seen.
Just for information of some readers, the difference between the
IEC 60052 values and the average values of the measurements is

45,61

Figure 89 Measurements performed using

a sphere-sphere gap of 15 mm in air to
investigate the influence of radiation

U50 in [kV]

45,51
45,41
45,31
45,21
45,11
45,01
990119

990120-2 990121-2 990205-1 990205-3 990208-2

Measurement number

85

Investigations; the results

1,1%.

Influence of the irradiation on the Up(RMC),50 of impulses with

oscillations superimposed
To investigate the influence of the irradiation towards the influence
on the Up(RMC),50 impulses with superimposed oscillations a
measurement series was performed. The oscillation superimposed
had an amplitude of 4,4 kV (around 10%), a frequency of 1,85 MHz,
and it was placed 1,6 s after the start of the impulse. The results are
presented in Figure 90. (The results of these measurements are
normalised to 10% oscillations amplitude).
During measurement 990120-1, the gap was irradiated with UV light,
during measurement 990205-2 no special precautions were taken and
during measurement 990208-1 the gap was covered so it could not be
irradiated by other sparks gaps close to it.
49,0

Figure 90 Influence of irradiation on the

Up(RMC),50 of in case oscillations are
superimposed

48,0

U50 in [kV]

47,0
46,0
Up(RMC),50
45,0

Up,50

44,0
43,0
42,0
990120-1

990205-2

990208-1

Measurement number

From the measurements presented in Figure 90, it can be concluded

that the influence of superimposed oscillations of frequency 1,85
MHz, amplitude 10%, placed around the peak is influenced by
irradiation. The influence on the Up(RMC),50 becomes significant
smaller, when the irradiation is less, a change from 6,2% to 4,5 %.
The general additional conclusion of the measurement series
performed is that under the used test conditions the influence of
superimposed oscillations of frequency 1,85 MHz, amplitude 10%,
placed around the peak on the Up(RMC),50 is around 5%.

6.3.2

Results of test performed at NGC

Two measurements carried out in March 1999, showed the following

results. The plate electrodes used had a diameter of 100 mm, a flat
area of 55 mm and the distance between the electrodes was 15 mm.
The T1 and T2 of the impulse were 1,2 s and 50 s. The frequency
of the oscillation superimposed was 1,84 MHz, its amplitude 5,25 kV
and the delay was 2,5 s.
The Up(RMC),50 of the impulse with oscillations was 49,8 kV with a
standard deviation of 1,07 kV. The Up,50 of the impulse with

86

Investigations; the results

oscillation was 54,8 kV with a standard deviation of 1,07 kV.
Therefore the amplitude of the oscillations was 10,5%.
The Up(RMC),50 of the full impulse was 51,31 kV with a standard
deviation of 1,04 kV.
The conclusion of this measurement was that oscillations of
frequency 1,84 MHz, 10,5% amplitude and with a delay of 2,5 s
have an influence of 2,9% to the Up(RMC),50.

6.3.3

Results of test performed at LCOE

Initial measurement
Two measurements carried out in March 1999, showed the following
results. The spheres used had a diameter of 150 mm and the distance
between the spheres was 35 mm. The T1 and T2 of the impulse were
1,19 s and 50,17 s. The frequency of the oscillation superimposed
was 2,14 MHz, its amplitude 9,76 kV and the delay was 2,46 s.
The Up(RMC),50 of the impulse with oscillations was 98,64 kV with a
standard deviation of 0,86 kV. The Up,50 of the impulse with
oscillation was 109,25 kV with a standard deviation of 1,27 kV.
Therefore the amplitude of the oscillations is 10%.
The Up(RMC),50 of the full impulse was 98,0 kV with a standard
deviation of 0,62 kV.
The conclusion of this measurement was that oscillations of
frequency 2,14 MHz, 10% amplitude and with a delay of 2,46 s have
no significant influence on the Up(RMC),50 (the influence is in the range
of the standard deviation).

Additional measurements
In order to study the influence of the trigger delay on the Up(RMC),50,
the trigger delay has been changed between 1,2 s and 3s for an
impulse with superimposed oscillations of a fixed frequency (2 MHz)
and fixed amplitude (A1 = 10%). This study has been carried out for
the homogeneous field.
When the trigger delay, to, is inside the interval 1,2 s and 3,0 s the
breakdown voltage of the mean curve Up(RMC),50 decreases. The
highest decrease of around 3% was obtained for to = 2,3 s, see
Figure 91.
Figure 91 Influence of the trigger delay
towards the influence of superimposed
oscillations

87

Up(RMC),50/U1,2/50 [1]

Investigations; the results

1,01
1
0,99
0,98
0,97
1

1,5

2,5

trigger delay time [us]

6.3.4

Results of test performed at the Schering Institute

Two measurements carried out in March 1999, showed the following

results. The spheres used had a diameter of 20 mm and the distance
between the spheres was 15 mm. The T1 and T2 of the impulse were
1,2 s and 50 s. The frequency of the oscillation superimposed was
1,88 MHz, its amplitude 3,06 kV and the delay was 2,58 s.
The Up(RMC),50 of the impulse with oscillations was 33,91 kV with a
standard deviation of 0,89 kV. The Up,50 of the impulse with
oscillation was 36,94 kV with a standard deviation of 0,95 kV.
Therefore the amplitude of the superimposed oscillations is 9%.
The Up(RMC),50 of the full impulse was 33,62 kV with a standard
deviation of 0,93 kV.
The conclusion of this measurement was that oscillations of
frequency 1,88 MHz, 9% amplitude and with a delay of 2,58 s have
no significant influence (smaller than the standard deviation) on the
Up(RMC),50.

6.3.5 Analysis of the results of the comparison measurement

in air
In four different laboratories test on sphere-sphere gaps been carried
out under different test conditions. The goal of the test was twofold.
The first goal was to check the influence of oscillations of frequency
2 MHz and 10% amplitude. The second goal was to check the
stability of the whole test set-up and the consistency of the results
obtained by different partners.
The initial results showed some inconsistency, the influence of 10%
oscillations varied between 0% and 4%. Investigations performed
showed that the influence was depending on the trigger delay used
and the irradiation. Because of this, discussions took place about
what the best suitable trigger time delay should be for the other tests
to be carried out. This resulted in the fact that the trigger delay had to
be such that:

88

Investigations; the results

The front time, T1, obtained for the curve with the superimposed
oscillations is as close as possible to the T 1 obtained for the curve
without oscillations
Up Up(RMC) + 0,7*Oscillation amplitude
Trigger delay time, td, is shorter than the time to peak tp(RMC)
If more than one td fulfils the requirements, the most natural curve
should be chosen
The same td should be obtained for all amplitudes
It also resulted in the decision that when results of tests carried out
were discussed the overshoot value in percent is not the amplitude of
the superimposed oscillation itself, but ((Up-Up(RMC))/ Up(RMC))*100%.
The general conclusion of the test carried out in air was that
oscillations of frequency 2MHz and of 10% amplitude superimposed
on the peak of a 1,2/50 smooth impulse in a sphere-sphere gap cause
that the 50% breakdown voltage (Up(RMC),50)is decreased by around
3%.

6.4

Investigations performed in oil at KEMA

The measurements in oil were carried out using the test set-up and
76

U50 in [kV]

74
72
70
68
66
64
62
0

0,5

1,5

2,5

Front time in [us]

procedures as explained in Chapter 5. The measurements are carried

out using transformer oil. The oil was purchased at SMIT
transformers Nijmegen and it was processed as if being used for
filling transformers. Details about the oil used can be found in
Appendix 5.
The gap distance of the sphere-sphere electrodes was set to 1 mm.
The test method and procedure used are the ones described in detail in
Chapter 5.

Influence of the front time to the 50% breakdown voltage

Investigations performed towards the influence of the front time
showed the results are presented in Figure 92. In this figure the
measured U50 values, the standard deviations and a trend-line are
presented.
The results presented in Figure 92 shows that the U50 varies between
69,5 kV for a front time of 0,3 s and 69 kV for a front time of 2 s
(69,2 kV for 1,2 s). This means that the influence of variations of

89
Figure 92 Influence of the front time
towards the U50 of transformer oil.

Investigations; the results

the front time between 0,3 and 2 s on the U50 is not more than 0,5%.
The conclusion is that there is no significant influence on the 50%
breakdown voltage of oil in a 1 mm sphere-sphere gap when the front
time is varied from 0,3 s to 2s.

Influence of the time to half value on the 50% breakdown

voltage
The results of the investigations performed towards the influence of
the time to half value are presented in Figure 93. No significant
influence on the 50% breakdown voltage can be detected when the
time to half value is varied from 40 s
Figure 93 Influence of the time to half
value towards the U50 of transformer oil

80

U50 in [kV]

76
72
68
64
60
30

35

40

45

50

55

60

time to half value in [us]

to 50 s.

Measurements performed with oscillations and overshoot

A number of measurements have been performed to investigate the
influence of superimposed oscillations of different frequencies. The
measurements performed are the measurement numbers: O990622a,
O990525a, O990519a, O990518a, O990517b and O990316a as
described in Appendix 4. Because of the standard deviation
performing tests with 5% oscillations was of no use. Therefore only
oscillations of 10% and higher were superimposed.

Overshoot versus oscillations

When an overshoot of amplitude 17,5% and duration 1 s is
superimposed the Up(RMC),50 is 65,24 kV. When an oscillation of
frequency 500 kHz and amplitude 17,5% is superimposed the
Up(RMC),50 is 65,48 kV. The Up(RMC),50 for a standard impulse is 69,2
kV. This means that the influence of an oscillation of amplitude
17,5% on the Up(RMC),50 is 5,48% and the influence of an overshoot of
amplitude 17,5% on the Up(RMC),50 is 5,72%. The difference between
the two results is 0,25% and therefore there is no significant
difference between the influence of oscillations and overshoot.
Therefore only the influence of oscillations is investigated further.

Influence of the amplitude of superimposed oscillations on

the 50% breakdown voltage
Investigations have shown that there are certain oscillations
frequencies have an influence on the Up(RMC),50 and there can be
certain oscillation frequencies that do not have an influence on the
Up(RMC),50. It is clear of course that only in the first case the amplitude
might also have an effect on the Up(RMC),50.

90

Investigations; the results

The measurement results presented in the previous paragraph indicate
that the amplitude of a superimposed oscillations of 500 kHz has an
influence to the Up(RMC),50. It is clear that in case no oscillation is
applied, the influence on the Up(RMC),50 is 0%. If the amplitude is
17,5%, the influence on the Up(RMC),50 is 5,48%.
Question is now, if a oscillation of around 8,5% has an influence on
the Up(RMC),50 of 2,7% or a different influence. And in the latter case
is this influence larger or is it smaller? Do oscillations of amplitude
smaller than 5% affect or not?
With the measurement results obtained a linearity statement could be
validated (see Figure 94). How it works will be explained by
presenting three fictious measurement examples.
In case a standard impulse (without oscillations) is applied, the
obtained Up(RMC),50 is 69,2 kV, indicated in Figure 94 by the point R.
This value is called the test voltage. This point is, of course, equal to a
standard impulse with superimposed oscillations with amplitude 0%.
The points L (Up,50) and M (Up(RMC),50) in Figure 94, correspond to
measurement with a superimposed oscillation of a certain frequency
of amplitude 21,5%. As can be seen from the figure point L is above
the test voltage and point M is below the test voltage. It appeared that
for another measurement, presented by the points A (Up,50) and
B (Up(RMC),50) in Figure 94, that correspondent to an oscillation of
10,4% the test voltage was also in between these two values and
more, that A = x*(L-R)+R and that B = x*(M-R)+R. Where x is a
certain value between 0 and 1.
With these equations, it is easy to interpret and to convert a
measurement result obtained for an amplitude of 20% to that for an
amplitude of 10%.
L
78,2

Up(rmc), 50

76,2

Up,50

Figure 94 Linearity statement

U50

74,2

72,2
70,2
R
68,2
B
66,2
M
64,2
1

Measurement no. with increasing oscillation amplitude

Influence of the frequency of superimposed oscillations on

the 50% breakdown voltage
With the validated linearity statement, the measurement results of the
influence of oscillations superimposed could be interpolated, the
results are only for 20% oscillations presented in Figure 95.
From the results the following conclusions can be drawn:

91

Investigations; the results

For a oscillation frequency of 2 MHz and amplitude up to 20% the
oscillations have no influence on the 50% breakdown voltage
Oscillations of a frequency of 200 kHz have an influence on the
breakdown voltage. In case an oscillation of 20% amplitude is
superimposed on a standard lightning impulse the Up(rmc),50 is 10%
lower than in case a standard lightning impulse with no
oscillations superimposed is applied.
Oscillations of a frequency of 500 kHz have an influence on the
breakdown voltage. In case an oscillation of 20% is superimposed,
the Up(rmc),50 is reduced by 6%.
Oscillations of a frequency of 800 kHz have an influence on the
breakdown voltage. In case an oscillation of 20% is superimposed,
the Up(rmc),50 is reduced by 5,6%.

Peak value versus peak value of a mean curve

For of oscillations between 200 kHz and 2 MHz with amplitudes up to
20%, the 50% breakdown voltage for a standard impulse is in
between the peak value of the total curve and the peak value of the
mean curve. It is therefore not solely the amplitude of the peak value
nor the amplitude of the peak value of a mean curve that is of
physical significance. It seems that: Ut= kUp + (1-k)Up(RMC)
80

Figure 95 Influence of 20% oscillations

towards the 50% breakdown value of
transformer oil (69,2 kV is the breakdown
voltage for standard impulses). The
trendline a logaritmic fitting of the
measurement results

Up,50 [kV]

75

70

65

60
50

250

450

650

850

1050

1250

1450

1650

1850

2050

Oscillation frequency [kHz]

80

Up(RMC)50 [kV]

75
70
65
60
55
50

250

450

650

850

1050

1250

1450

Oscillation frequency [kHz]

6.5

92

Measurements performed in SF6 at NGC

1650

1850

2050

Investigations; the results

At NGC tests were carried out in SF6 gas at a pressure of 4 bar. A
drawing of the vessel used is presented in Figure 96. A drawing of
the electrode configuration used for the uniform field is presented in
Figure 99, the configuration used for the non-uniform field is
presented in Figure 98. The test method used was the up and down
method and for each measurement series, 40 to 50 lightning impulses
were applied.

Influence of the front time

The influence of the front time in the
interval 0,5 s and 2 s on the 50%
breakdown voltage was investigated.
The results are presented in Figure 97.
The standard deviation of the
measurements performed is around 1%.
In the range 0,5 to 2,5 s the influence
of the front time to the 50% breakdown
voltage is within the maximum
variation, 2%, occurring with positive
non-uniform fields.

Influence of the amplitude of superimposed oscillations

The influence of superimposed oscillations on the 50% breakdown
voltage was investigated for uniform
field positive polarity and for nonFigure 96 The SF6 pressurised test vessel
uniform field for both negative and
positive polarity. The measured results
are presented in Figure 101, Figure 100 and Figure 102.

93

Investigations; the results

From the results it can be seen that in general the results for the
different oscillation amplitudes (5%, 10% and 20%) are quite
analogue. There are however small differences as can be seen in the
figures. These differences are caused by the standard variation and
because the amplitude of each measurement point is not exactly 5%,
10% or 20%. The general conclusion is however that also for SF 6,
linearity statement can be validated.

120

Figure 97 Influence of the front time

towards U50 in SF6

110

U50 [kV]

100
90
80
70

Uniform field, pos.

60

Non uniform field, pos

Non uniform field, neg.

50
0

0,5

1,5

2,5

T1 [us]

Influence of the frequency of superimposed oscillations

From the results presented in Figure
101, Figure 100 and Figure 102 it can be
seen that the frequency of the oscillation
has an influence.
More specific the following conclusions
can be drawn:
For uniform and non-uniform field
both polarities, oscillations of 200
kHz have influence on the 50% breakdown voltage. For instance,
in case an oscillation of 20% amplitude is superimposed, the
Up(RMC),50 is decreased by 20%. Or
the 50% breakdown voltage is equal Figure 99 The uniform field electrode
system.
to the Up,50.

Figure 98 The non-uniform field electrode

system.

94

Investigations; the results

For uniform and non-uniform field
2 MHz have some influence. In
amplitude is superimposed, the
approximately 5%.
For uniform and non-uniform field
5 MHz have some influence. In
amplitude is superimposed, the
approximately 5%.

both polarities oscillations of

case an oscillation of 20%
Up(RMC),50 is decreased by
both polarities oscillations of
case an oscillation of 20%
Up(RMC),50 is decreased by

Peak value versus peak value of the mean curve

For oscillations of frequency between 200 kHz and 5 MHz with
amplitudes up to 20%, the 50% breakdown voltage for a standard
impulse is in between the peak value of the total curve and the peak
value of the mean curve. It seems that also for SF6 Ut= kUp + (1k)Up(RMC).

95

Investigations; the results

85

Figure 101 Influence of oscillations,

uniform field, positive polarity, in SF6

U50pk, A=5%
U50pk, A=10%
U50pk, A=20%
U50mc, A=5%
U50mc, A=10%
U50mc, A=20%

80
75
U50 [kV] 70

U50ref
U50ref-5%
U50ref+5%

65
60
55
0,1

1
Oscillation frequency [MHz]

10

110

Figure 100 Influence of oscillations, nonuniform field, positive polarity, in SF6

U50pk, A=5%
U50pk, A=10%

100

U50pk, A=20%
U50mc, A=5%

90

U50mc, A=10%

U50 [kV]

U50mc, A=20%

80

U50ref
U50ref-5%

70

U50ref+5%
60
0,1

10

Oscillation frequency [MHz]

Figure 102 Influence of oscillations, nonuniform field, negative polarity, in SF6

130
U50pk, A=5%

U50 [kV]
120

U50pk, A=10%
U50pk, A=20%
U50mc, A=5%

110

U50mc, A=10%
U50mc, A=20%

100

U50ref
U50ref-5%

90

U50ref+5%
80
0,1

10

Oscillation frequency [MHz]

6.6 Measurements performed in PE at the Schering

Institute
At the Schering Insitute, University Hannover, tests have been
performed on PE. A drawing of the test cell is presented in Figure
103. The test method used was the multiple level tests.

Influence of the front time and time to half value

Figure 103 The test cell used for the

experiments with PE

96

The influence of the front time towards the 50% breakdown voltage
for PE is presented in Figure 104. Conclusions of these investigations
were that:
Front times T1 between 0,5 s and 1,56 s do not have a
significant influence on the breakdown behaviour of the tested PE
specimens.
The time to half value T2 is not a significant parameter for this
kind of test because almost all breakdowns in the tested PE
specimen occur at the front or at the peak.

Investigations; the results

200
kV

Figure 104 Influence of the front time T1

on the U50%-voltage

175
U50%
150

125

100
0

0,5

1,5

T1

Overshoot versus oscillations

There seemed to be no difference between superimposed oscillations
and overshoots because the breakdowns appear at the front or at the
peak of the superimposed oscillation or overshoot. Therefore it is
assumed that the results obtained for oscillations are also valid for
overshoot.

Influence of the amplitude of superimposed oscillations

Also for PE, the amplitude of oscillations have influence on the
Up(RMC)50. The measurement results are presented in Figure 105,
Figure 106 and Figure 107. From these results it can be seen that also
for PE, the linearity statement can be validated for the three
200

U50%

Upeak

200 kHz

kV

Figure 105 U50 for LI superimposed by

200 kHz
oscillations
with
various
amplitudes

190
180
170
160
Upeak (mean curve)
150
0

10

15

Amplitude of oscillation
200

U50%

Upeak

800 kHz

kV
190

Figure 106 U50 for LI superimposed by

800 kHz
oscillations
with
various
amplitudes

180
170
160

Upeak (mean curve)

150
0

10

15

Amplitude of oscillation

frequencies used.
97

Investigations; the results

200

Figure 107 U50 for LI superimposed by

2 MHz
oscillations
with
various
amplitudes

U50%

Upeak

2000 kHz

kV
190
180
170

Upeak (mean curve)

160
150
0

10

15

Amplitude of oscillation

Influence of the frequency of superimposed oscillations

The conclusion drawn from the measurement results is that the
frequency of the superimposed oscillations has an important influence
on the Up,(RMC),50.
The measurement results obtained of the
investigations performed are presented in Figure 108, Figure 109 and
Figure 110.
Superimposed oscillations with a low frequency (200 kHz) are
influencing the breakdown behaviour in a strong way and in this case
besides the frequency the amplitude of the oscillation is another
important parameter. Thus for superimposed impulses with a low
frequency the Up-value has to be the test voltage. Superimposed
oscillations with a high frequency (2 MHz) have no great influence on
the breakdown behaviour and the amplitude of the oscillations can
vary in a wide range without influencing the breakdown voltage. The
frequency at which the 50% breakdown value is exactly in between
the two peak values is around 900 kHz.

Peak value versus peak value of the mean curve

For oscillations of frequencies between 200 kHz and 2 MHz with
amplitudes up to 15%, the 50% breakdown voltage for a standard
impulse is in between the peak value of the total curve and the peak
value of the mean curve. It seems that also for PE Ut= kUp + (1k)Up(RMC).
200

Figure 108 U50 for LI superimposed by

oscillations with amplitude of 5% and
various frequencies

U50%

Upeak

Aosc= 5 %

kV
190

180

JB

170

J
J

U
(mean curve)
peak
160
150
0

98

200

400

600

800

1000 1200 1400 1600 1800

kHz 2000
Frequency of oscillation

Investigations; the results

200

U50%

Upeak

Aosc= 10 %

kV
190

180

Figure 109 U50 for LI superimposed by

oscillations with amplitude of 10% and
various frequencies

B
JB

170

160

Upeak (mean curve)

150
0

200

400

600

800

1000 1200 1400 1600 1800

kHz 2000
Frequency of oscillation

200

U50%

Upeak

Aosc= 15 %

kV

190
180

Figure 110 U50 for LI superimposed by

oscillations with amplitude of 15% and
various frequencies

B
J

JB

170
Upeak (mean curve)

160
150
0

6.7

200

400

600

800

1000 1200 1400 1600 1800

kHz 2000
Frequency of oscillation

Measurements performed in air at LCOE

At LCOE tests have been performed in air. Only part of the

measurements performed are presented in this thesis, a more detailed
presentation of the measurement results will be given in [79].
Drawings of the test cells used are presented in Figure 111 and Figure
113. Many precautions have been taken to have reliable and stable
results, these measures are explained in detail in [15].

240

150

35
900

325

Influence of the front time

To study the influence of the front time, T1 within and outside the
present permitted limits of IEC 60060-1 and to establish a correlation
between the mean slope in the front of the impulse and the Up,50 a
study in air has been carried out. The study was carried out for four
combinations: two polarities (positive and negative) and two kinds of
electrical fields (homogeneous and non-homogeneous fields). The
results are presented in .

325

Figure 111 Test cells for air, uniform field

99

Investigations; the results

Tables and figures corresponding to the influence of the front time on the U p,50%
Homogeneous field.Positive Polarity
T1 ( s )

Homogeneous field.Negative Polarity

T1 ( s )

Non Homogeneous field.Negative Polarity

T1 ( s )

Non Homogeneous field.Positive Polarity

T1 ( s )

Up,50%(kV)

z (kV)

Up,50%(kV)

z (kV)

Up,50%(kV)

z (kV)

Up,50%(kV)

z (kV)

0,5

100,32

0,29

0,5

101,14

0,30

0,5

108,89

1,61

0,5

62,30

2,88

0,8

99,56

0,42

0,8

98,82

0,29

0,8

107,02

0,25

0,8

64,76

2,15

1,2

98,85

0,40

1,2

98,73

0,30

1,2

107,19

0,67

1,2

62,67

2,19

1,56

98,76

0,38

1,56

98,95

0,18

1,56

107,57

1,05

1,56

61,73

3,41

98,94

0,32

98,38

0,27

107,39

0,41

59,96

1,51

105

115

115

112,5

112,5

110

110

70

104
103

67,5

102
Up,50%(IEC52) = 98,5 kV

65

105

Up, 50%

107,5
Up,50

100

Up

107,5

Up,50

Up,50%(IEC52) = 98.0 kV

101

105

62,5

99
102,5

102,5

100

100

98

60

97
98,5

97,598.0

96
95

95

0,5

0,75

1,25 1,5

1,75

0,5

T1 ( s )

57,5

97,5

0,75

1,25

1,5

T1 ( s )

1,75

95
0,5

0,75

1,25
T1 ( s )

1,5

1,75

55
0,5

0,75

1,25

1,5

1,75

T1 ( s )

Figure 112 Influence of the front time to U50 for air

Conclusion of the investigations performed is that the present limits

of the IEC 60060-1 for T1 (0,84 s 1,56 s) can be extended to the
interval 0,8 s 2,0 s without any significant variation of the
breakdown voltage. For front times of 0,5 s there is an increase of
the breakdown voltage but it is within the voltage standard tolerance
3%.
There is a very close agreement between the results for the
homogeneous field and the standard values of IEC 52. The differences
are between 0,5% and 0,7%.
The standard deviation, z, of the Up, 50% is bigger in the nonhomogeneous field than in the homogeneous field. In the nonhomogeneous field, z is bigger with positive than with negative
impulses.

Influence of the frequency of superimposed oscillations

To study the influence of the frequency of superimposed oscillations
several tests have been carried out by LCOE. The tests are carried out
for the two field configurations (homogeneous and nonhomogeneous) and the two polarities (positive and negative).

290

4
150

10
40
50
900

325

260

Figure 113 Test cell for air, non-uniform

field

100

The influence of superimposed oscillations of amplitude around 5%

on the 50% breakdown value are presented in Figure 114 and Figure
115.
The conclusions that can be drawn from the experiments performed
are the following:
Superimposed oscillations with amplitudes of less than 5% and
frequencies of not less than 0,5 MHz do not have a significant
influence on the breakdown voltage.

Investigations; the results

Superimposed oscillations with amplitudes of less than 5% and
frequencies of less than 0,5 MHz have a significant influence on
the breakdown voltage except for the non homogeneous field
negative polarity
The IEC 60060-1 criterion for evaluation of L.I. is satisfied for the
homogeneous field and non-homogeneous field with positive
polarity, but not for the homogeneous field with negative polarity
Taking into account that the breakdown voltage for the nonhomogeneous field is much larger for negative than for positive
polarity the IEC 60060-1 criterion for evaluation of L.I. can be
used without decreasing the insulation level.

Analysis of parameters associated with the area

For air some additional investigations with the results of the test have
been performed. One of the investigations performed was the
verification of the equal area criterion. A detailed explanation of this
study is presented in [15]. The conclusion of this study was that the
equal area criterion was validated.

Peak value versus peak value of the mean curve

Also for air, the conclusion can be drawn that the 50% breakdown
voltage for a standard impulse is in between the peak value of the
total curve and the peak value of the mean curve. It seems that also
U50 vs f. 2,0% <b <5,1%

Figure 114 Influence of 5% oscillations

upon the U50 in a homogeneous field in air

1,1
Up,50%
1,05

U50%

0,95
Up(RMC),50%
0,9
Positive Polarity
Negative Polarity Up

0,85

0,8
0

Positive Polarity
Negative Polarity Up(RMC)
0,5
1
1,5

2,5

3,5

4,5

f(MHz)

for air Ut= kUp + (1-k)Up(RMC).

U50 vs f. 1,9%<b <7,2%

Figure 115 Influence of 5% oscillations

upon the U50, non-homogeneous field
negative polarity in air

1,1
Up,50%

1,05

U50%

0,95
Up(RMC),50%

0,9

0,85
Up
Up(RMC)

0,8
0

0,5

1,5

2,5

3,5

4,5

f(MHz)

101

Investigations; the results

6.8

V( ud ) 1 m 1

Measurements in oil-paper at TU Graz

Two different types of samples have been tested, Sample type A and
Sample type B. As explained in paragraph 5.4.4, the progressive
stress method is used and at each voltage level, three voltage
applications are done. As can be found in [72], if the test data is used
as input in the PSM statistical program it is not the U50 value that is
found, but the U20. That it is like that can be calculated with the
performance function in Equation XXVII. If the equations is known
for m=1, one can calculate by using this formula the equation for
m=3.

S (ud ) S (ud u )
1 S (ud u )

For the determination of the influence of the oscillations frequency on

the breakdown value, it doesnt matter so much if one takes the 50%
or the 20% breakdown value, as long as one takes the same in the
comparison. The only thing that can be different is the standard
deviation.

Equation XXVII performance function

On both samples 4 measurements have been carried out:

standard impulse
superimposed 1 MHz oscillation
superimposed 200 kHz oscillation
superimposed 1,37 s overshoot
The results are summarised in the table below, Table 5.
Table 5
Wave shape

Sample type A

Sample type B

Up,20

SD

Up(RMC),20

SD

Up,20

SD

Up(RMC),20 SD

[kV]

[kV]

[kV]

[kV]

[kV]

[kV]

[kV]

[kV]

Standard impulse

-172,6

8,0

-172,6

8,0

-168,0

7,2

-168,0

7,2

With 1 MHz oscillations

-182,0

8,1

-151,4

6,6

-191,0

5,7

-149,5

4,4

With 1,37 s overshoot

-180,5

4,7

-112,0

3,2

-179,0

8,3

-111,9

5,3

With 200 kHz oscillations

-178,0

7,5

-152,2

6,7

-184,0

4,3

-158,1

3,4

185 kHz damped sine wave

-168,6

7,6

-173,6

10,3

Influence of the frequency of superimposed oscillations on

the 20% breakdown voltage
From the measurement results is became clear that the frequency of
the superimposed oscillation had an influence, the higher the
frequency, the smaller the influence on the Up(RMC),20.

Peak value versus peak value of the mean curve

If the linearity statement is also valid for these types of samples then
also for these type of samples, the test voltage should for certain
oscillating frequencies be in between the peak value and the peak
value of the mean curve. It seems that also for the samples
investigated Ut=kUp + (1-k)Up(RMC).

102

Investigations; the results

6.9
6.9.1

Analysis of results, general discussion

Front time

The general conclusion that can be drawn from the measurements

performed is that the 50% breakdown value does not change
significantly when the front time varies in the interval
0,3 s and 2,0 s for homogeneous gaps in oil
0,5 s and 1,56 s for homogeneous gaps in PE
0,8 s and 2,0 s for homogeneous and non-homogenous gaps in
air
0,5 s and 2,5 s for homogeneous and non-homogenous gaps in
SF6
For some measurements the intervals were limited by the
measurements performed (oil, PE and SF6) for air the limits were set
by the variation of the front time.
Therefore it can be concluded that for the tested configurations the
present front time limits in the IEC 60060-1 are valid. It can be
discussed it they can be extended from 0,84-1,56 s to 0,51,56 s,
since there is no significant influence to 50% breakdown voltage.

6.9.2

Time to half value

Only for oil and PE, a conclusion can be drawn concerning the time to
half value.
For oil the conclusion is that no significant influence on the 50%
breakdown voltage can be detected when the time to half value is
varied from 40 s to 50 s.
For PE the time to half value T2 is not a significant parameter for this
kind of test because almost all breakdowns in the tested PE specimen
occur at the front or at the peak.

6.9.3

Oscillations versus overshoot

For the insulating materials oil, PE, SF6 and air it has been confirmed
that no difference can be detected between the influence of overshoot
and oscillations of corresponding frequency and duration.
Therefore all conclusions drawn concerning oscillations are also valid
for overshoot.

6.9.4

Position of the oscillations or overshoot

The position of the oscillations, i.e. on the peak or on the front has an
influence on the 50% breakdown value for air and for oil. The
influence increases when the position of the oscillations is closer to
the peak of the lightning impulse. Most relevant seems to be the
increase of the peak value of the curve with oscillations compared to
that without oscillations.

103

Investigations; the results

6.9.5 Influence of oscillation frequency and amplitude on

the breakdown voltage
Depending on their frequency, amplitude and position, oscillations or
overshoot have a significant influence on the 20% or 50% breakdown
voltage. In general an oscillation or overshoot of high amplitude and
low frequency or long duration has more influence than an oscillation
or overshoot of small amplitude and high frequency or short duration.
No cases were reported where the breakdown voltage of a standard
lightning impulse was clearly below the Up(RMC) or clearly above the
Up of an impulse with oscillations superimposed. Therefore, the
conclusion can be drawn that the test voltage is, for all materials and
samples investigated, in between the peak value and the peak value of
a mean curve. It should be noted that in the cases investigated the
mean curve is a standard impulse that is generated by a generating
circuit and not a mathematical function.
For all materials and samples investigated it could therefore be
demonstrated that Ut = kUp + (1-k)Up(RMC). It is of course the question
if k is a constant, or that it is a function of some parameters or if it is
the same for all materials investigated etc. The next paragraph will
deal with the k-factor and the linearity statement.

6.10 K factor hypothesis and linearity statement; influence

of oscillation frequency and amplitude
One of the most important parameters when evaluating a lightning
impulse voltage is some kind of parameters that characterises
magnitude of the impulse voltage applied. When this parameter and
the evaluation method for this parameter is well know, the other
parameters characterising a lightning impulse can be defined.
Since the front time (within certain limits) and the time to half-value
do not have a very significant influence parameters like dU/dt and the
area underneath the impulse do not seem to be the best parameter to
choose to define the magnitude of the impulse.
In the previous paragraph it has been demonstrated that the test
voltage, Ut seem to be a parameter is relevant and significant. This
because it is not affected by oscillations and by changes in the front
time and time to half value. As has been demonstrated too the test
voltage seems in case of oscillations and overshoot present, to be a
function of the peak value of the applied curve and the peak value of
the applied base or mean curve.
Utest= kUp + (1-k)Up(RMC)
Equation XXVIII k-factor hypothesus

Therefore in earlier paragraphs it has already been introduced that U t

is a function of Up and Up(RMC), see Equation XXVIII and that for
some materials a linearity statement could be validated. In the next
paragraphs, the k-factor and the linearity statement will be explained.

6.10.1 k-factor; influence of the oscillating frequency

For all insulating materials it has been confirmed that in case of
superimposed oscillations of very low frequencies the test voltage
should be equal or close to the peak value of the total curve. This
means that oscillations of low frequency have almost full influence
and that k = 1.

104

Investigations; the results

In case of very high frequencies the test voltage is equal to the peak
value of the reference mean curve. This means that oscillations of
high frequency have almost no influence and that k = 0.
In all other cases k is in between 0 and 1.
The factor k can in theory be a random variable or a function of:
1. the oscillating frequency
2. the position of the oscillation or overshoot
3. the field configuration
4. the insulating material
5. the amplitude of the oscillation or overshoot
The k-factor can also be different for oscillations and for overshoot.
In the next part it will be explained and discussed that the k-factor
will only a function of the field configuration and the insulating
material. It is namely so that:
earlier evidence is given that that there is no significant difference
between overshoot or oscillation. Therefore k is no function of
overshoot and oscillations.
when the amplitude and the frequency of the oscillation were kept
constant, the position of the oscillations has an influence, at least
for air, but it is plausible that it is the same for other materials.
Therefore the amplitude of the oscillations in combination with the
time delay are important parameter. It is plausible that it is more
the voltage difference between the peak voltage of the complete
curve and the voltage of the mean curve used. Therefore the Up
and the Up(RMC) were introduced.
it is proven in some cases and assumed in some other cases that the
k-factor is not a function of the oscillation amplitude. This will be
treated in detail in the next paragraph
The k-factor as function of the frequency will be presented for
different materials in the one of the next paragraphs. But first a more
fundamental discussion is presented.

6.10.2 Linearity statement; influence of the oscillations

amplitude
The amplitude of the oscillation is an important parameter for
frequencies where oscillations have an influence on the 50%
breakdown voltage. If the amplitude of an oscillation of a certain
frequency increases than the influence to the breakdown voltages
increases as well.
It is to find out what kind of relation the amplitude of the oscillation
has with the influence to the 50% breakdown voltage. It can in theory
be that 20% oscillations of a certain frequency have full influence,
while 10% oscillations only have partly influence. In this case there
is no linear relation with the test voltage for different oscillation
amplitudes and the k factor will be a very complex function.
Therefore it had to be proven first that the influence of oscillations is
a linear one.
A linearity statement has been presented earlier and for oil (up to
40%), for PE (up to 15%), for SF6 (up to 20%) and for transformer
samples (up to 60%) the linearity statement has been validated. It is

105

Investigations; the results

plausible that the linearity statement is valid for other insulating
materials also.
It was expected that things might change drastically when the
amplitude of the oscillations starts to be more dominating than the
basic impulse. But the test with the transformer samples shows that
in case a 185 kHz damped sine wave the theorem might still be valid.
It is now important that the two hypotheses, the linearity statement
and the k-factor function are not two contradicting statements, else it
will not be possible to find an suitable evaluation method. This prove
is given in the next paragraph.

6.10.3 Evidence of the matching of the linearity and the kfactor statement
One of the ways to demonstrate that the two hypotheses are not
contradicting to each other is to use some mathematics.
From the k-factor hypothesis it follows that:
(1)R=Ut=k(A)+(1-k)B
(2)R=Ut=k(L)+(1-k)M
From the linearity hypothesis it follows that:
L
78,2

Up(rmc), 50

76,2

Up,50

U50

74,2

(3)A=x(L-R)+R
(4)B=x(M-R)+R

72,2

filling (3) and (4) in equation (1) gives the following result:

70,2
R
68,2

R=k(x(L-R)+R) +(1-k)(x(M-R)+R)

B
66,2
M
64,2
1

Or xR=x(k(L)+(1-k)M)

Measurement no.

This means that the two hypotheses are not contrary to each other.
Figure 116 Linearity statement and kfactor

Since both the k-factor and the linearity hypotheses are validated with
measurement results, it can be concluded that both hypotheses
correspond and are therefore correct.
This seems a very simple statement, but it is rather fundamental.

6.10.4 K-factor as function of the frequency for different

materials

Figure 117 k-factor for transformer oil,

homogeneous field

Since it is plausible that the k-function is not a function of the

oscillation amplitude for oscillations at least up to 15%, but perhaps
even up to 60%, for each material and field configuration a k-factor
can be determined. The results are now presented.
The measurement results and the most likely k-factor as function of
the frequency for oil are presented in Figure 117.

106

k-factor [1]

Investigations; the results

1,2
1
0,8
0,6
0,4
0,2
0
-0,2
10

100

1000

10000

oscillation frequency [kHz]

107

Investigations; the results

k-factor [1]

The measurement results and the most likely k-factor as function of

the frequency for SF6 are presented in Figure 118 and Figure 119.
1,2
1
0,8
0,6
0,4
0,2
0
-0,2

Figure 118 k-factor for SF6, homogeneous

field

10

100

1000

10000

k-factor [1]

Oscillation frequency [kHz]

1,2
1
0,8
0,6
0,4
0,2
0
-0,2

Figure 119 k factor for SF6, nonhomogeneous field

10

100

1000

10000

Oscillation frequency [kHz]

k-factor [1]

The measurement results and the most likely k-factor as function of

the frequency for PE are presented in Figure 120.
1,2
1
0,8
0,6
0,4
0,2
0
-0,2

Figure 120 k-factor for PE, homogeneous

field

10

100

1000

Oscillation frequency [kHz]

108

10000

Investigations; the results

k-factor [1]

The measurement results and the most likely k-factor as function of

the frequency for air are presented in Figure 121 and in Figure 122.

Figure 121 k-factor for air, homogeneous

field

1,2
1
0,8
0,6
0,4
0,2
0
-0,2
10

100

1000

10000

Figure
122
k-factor
inhomogeneous field

for

k-factor [1]

Oscillation frequency [kHz]

air,

1,2
1
0,8
0,6
0,4
0,2
0
-0,2
10

100

1000

10000

Oscillation frequency [kHz]

The measurement results of the tests performed on to type of winding

models, Sample A and Sample B, and the most likely k-factor
function are presented in Figure 123 and Figure 124.

k-factor [1]

Figure 123 k-factor for Sample A

1,2
1
0,8
0,6
0,4
0,2
0
-0,2
10

100

1000

10000

Oscillation frequency [kHz]

Note that the test with the damped sine is not presented in this figure.
In case a pure 185 kHz oscillations is applied, the Up,20 = 168,6 kV, in
this case the k-factor is 0,956. Because of the standard deviation it is
difficult to make a statement about this result, but it seems to fit very
well.

109

k-factor [1]

Investigations; the results

1,2
1
0,8
0,6
0,4
0,2
0
-0,2

Figure 124 k-factor for Sample B

10

100

1000

10000

Oscillation frequency [kHz]

Note that the test with the damped sine wave is not presented in this
figure. In case a pure 185 kHz oscillations is applied, the Up,20 =
173,6 kV, in this case the k-factor is 0,956. The test voltage should
be 168 kV. With the k-factor one calculates 166 kV, which is a very
good result, considering the standard deviations.

6.10.5 K-factor; summary

The k-factor as function of the frequency for different insulating
materials have been determined using the measurement results and
have been presented in the previous paragraph.
In Figure 125, the most likely k-factors as function of the frequency,
fitted according to a logaritmic function, for all materials investigated
and the present IEC 60060 k-factor are presented in one graph.
It is surprising to see how small the difference is between them, with
the exception of inhomogeneous fields in air, and how the present
IEC 60060-1 crosses them somewhere halfway.
1,20
oil

1,00

air inhom
Figure 125 k-factors
of all materials
hom oscillations or
investigated valid airfor
SF6 hom frequency and
overshoot of corresponding
PE
amplitudes up to 15-60%.

k-factor [1]

0,80
0,60
0,40

IEC 60060

0,20

SF6 inhom
Sample A
Sample B

0,00
-0,20
10

100

1000

10000

Oscillation frequency [kHz]

Since the k-factors presented in Figure 125 are so relative close to

each other it seems possible to determine some kind of average kfactor function. This average k-factor together with all k-factors
determined and the IEC 60060-1 factor is presented in Figure 126. In
this figure it can also be seen that in case a k-factor with a shape of
IEC 60060 is kept it is best to change the frequency where it goes
from 1 to zero to around 700 kHz.

110

Investigations; the results

1,20

Figure 126 best fitting of all k-factors

1,00
together with all
the k-factors
k-factor [1]

0,80
0,60
0,40
0,20
0,00
-0,20
10

100

1000

10000

oil
air inhom
air hom
SF6 hom
SF6 inhom
IEC 60060
PE
proposal(1)
Sample A
Sample B

Oscillation frequency [kHz]

Because all k-factors are best fittings of the measurement point for
each material it is interesting to see how the average k-factor will fit
between all the measurement results. This is presented in Figure 127

1,20

proposal (1)
oil
air hom
SF6 hom
SF6 inhom
PE
sample A
Sample B

k-factor [1]

1,00
0,80
0,60
0,40
0,20
0,00
-0,20
10

100

1000

10000

Oscillation frequency [kHz]

Figure 127 Measurement results and the average k-factor

6.10.6 K-factor implementation

Now the k-factor has been presented it is the question if and how it
can be implemented in a good way. This will be the main subject of
the next Chapter.

111

Investigations; the results

112

Proposal

7 Proposal
7.1

Introduction

The measurement of the applied lightning impulse voltage and the

evaluation of the parameters characterising the applied lightning
impulse voltages should be performed accurate and uniform all over
the world.
The rules for the evaluation of the parameters as described in the
present IEC 60060-1 are ambiguous and in case of non-standard wave
shapes they are not described at all.
The goal of the investigations performed was to find parameters and
clear evaluation rules, that are applicable to both standard and nonstandard lightning impulse voltages, taking into account the
possibilities and limitations of modern computers, that are proven to
be of physical relevance for all type of practical insulating systems in
case lightning impulse voltages are applied.

Because of the tremendous number of people using the present

evaluation method, the evaluation method proposed should be simple,
straight forward, easy to understand and apply.
Besides that, it should, preferably, be possible for laboratories that are
still using analogue measuring equipment to implement the method.
In this chapter a method will be proposed based on the findings
presented in the previous Chapter that, hopefully, will meet all these
requirements.

7.2

Proposal for evaluation method

During tests with lightning impulse voltages the most important

parameters is the test voltage. The test voltage is a parameter that is
described for all components used in the different applicable
standards for long time.
Even in case from a scientific point of view it would be the best to
introduce another parameter that replaces the test voltage, it would be
difficult to implement this proposal. The investigations have however
shown that the test voltage is a relevant parameter.
One of the other reasons to choose the test voltage as an important
parameter is that the equal area criterion is not validated for all
insulating materials investigated, besides this the Ub level is different
for different materials and configurations.
Therefore it would be the best to introduce a new algorithm to
determine the test voltage.
The investigations performed have demonstrated that the front time
and the time to half value can vary within wider limits without
significantly influencing the test results. For the test voltage value
this is different. Therefore of all proposals that can be made, the best
one is that one who is able to evaluate the peak value the most
accurate.

113

Proposal

7.2.1

Test voltage

Results of the investigations performed

By the breakdown tests performed on several insulating materials
(using Measuring Systems with an extremely wide bandwidth), it has
been demonstrated that the test voltage is an important parameter. It
can be used for the characterisation of lightning impulse voltages, if
one can find an appropriate evaluation procedure and algorithm.

U t k U p (1 k ) U p ( RMC )
Equation XXIX

As is demonstrated in the previous chapter, the test voltage is a

function of the peak value of the applied lightning impulse voltage
and the peak value of the mean curve or better base curve of the
impulse applied. In mathematical words, the test voltage can be
defined by Equation XXIX.
In the previous chapter it has been explained that the k-factor in this
equation is only a function of the oscillation frequency or overshoot
duration and the insulating material (see Figure 126). It is plausible
that this statement is valid for all materials and oscillation amplitudes
up to at least 15%, but for some materials even up to 60%.

1,20

IEC 60060
proposal(1)

1,00
0,80
0,60
0,40
0,20
0,00
-0,20
10

100

1000

10000

Figure 128 average k-factor versus IEC

60060-1

In the previous chapter, the best k-factor for the different insulating
materials investigated has been calculated. From the several graphs
presented, it is concluded that the behaviour of the breakdown voltage
as function of the oscillation frequency and the amplitude of the
superimposed oscillations are more or less similar, but not identical
for all materials. In Chapter 6.10.4 the average k-factors, the k-factor
for each material investigated and the IEC 60060-1 factor are
presented. In Figure 128 the average k-factor and the IEC 60060
factor are presented once more. An interesting fact is that the IEC
60060-1 and the average k-factor cross at 0,56 at 500 kHz.
The average k-factor is:
1 for frequencies below 80 kHz
0 for frequencies above 5 MHz
given by the formula k=-0,2479*ln(freq in kHz) + 2,1056 for
frequencies in between these two values.

Present situation of IEC 60060-1

Up to now the evaluation procedure for lightning impulses with
oscillations depends on the frequency concerning the determination of
the test voltage. This fixed limit is generally a problem for digital
recorded data, because only the very small change of one sample will
influence the evaluation results, that means the test voltage. Even if
the amplitude of oscillation is limited up to 5% of the peak of the
mean curve, the maximum deviation can also be 5% only by the
change of one digital sample (i.e. the oscillation frequency changes
from 499,46 to 499, 51 kHz). Because the overall measurement
uncertainty has to be within 3%, this is a problem. Besides this, as
can be seen from Figure 128, even if one is able to determine the
frequency or the duration this accurate, one makes an error. From this
it is even clearer that the present situation has to be changed.

114

Proposal

How to implement the k-factor; possible solutions

In order to prevent the problems with the present IEC, which will
exists if the frequency limit is given with a certain tolerance, it is
better to introduce a smooth transition. The introduction of the kfactor makes such a smooth transition from the low frequencies to the
high frequencies possible. First publication of this k-factor was made
on the ISH and CIGRE WG 33.03 and CIGRE SC 33. [14]
One big advantage of implementing the k-factor it that the values for
k can be defined according to the behaviour of the different materials
or an average can be taken. Besides this the characteristics of the
Measuring Systems used can be taken into account.
It might be clear that in case a Measuring System is used with a
bandwidth much better than the k-factor function. The k-factor needs
to be implemented fully in the evaluation method.
In case a Measuring System is used with a limited bandwidth other
measures are necessary. Lets first assume that Measuring Systems
having sufficient bandwidth will be used. In this case there are
several ways to implement this k-factor, for instance:
Use a modified IEC 60060 evaluation method
Use completely other procedure, for instance filtering
The problem that still exists in the first option is that a mean curve
has to be determined. This is a problem, since one cannot find a
mean curve valid for all possible measured curves.
Therefore the idea was raised to implement the k-factor by using
filtering techniques. One can partly filter the measured curve or filter
the measured curve totally.
Since the front time can vary within a wide range before influencing
the test voltage it is assumed that the k-factor is also valid in case
oscillations or overshoot is superimposed on other wave shapes
besides a standard 1,2/50.
If filtering techniques can be used, this would be a big step forward,
because the determination of a mean curve is avoided, a smooth
transition is guaranteed and it is very easy to program different kfactors or filters if one wishes too.
The idea for using filtering techniques was checked initially with a k
defined as 1 for 0,5 MHz, equivalent with the actual IEC standard,
going from that point linearly down to 0 for the frequency of 2 MHz.
The result of this initial check is presented in paragraph 7.2.3 and
showed good results.

7.2.2 Proposal for evaluation procedure using filtering

techniques
The proposal for the evaluation procedure using filtering techniques
to implement the k-factor is presented in Figure 129.
The amplitude check is necessary, since the linearity statement has for
some materials only been validated for oscillations up to 15%
amplitude.

115

Proposal

Figure 129 proposal for evaluation

procedure using filtering techniques

7.2.3

To check whether the idea to implement the k-factor by using filtering

techniques can be used, the proposal with the k-function filter has
been tried by the Schering Institue on a three cases of the IEC 610832 TDG and two measurements performed in the frame work of the
project.

1,2
1
0,8

H(w)

Initial check of the proposal

0,6
0,4
0,2
0
0

500

1000

1500

2000

2500

Frequency [kHz]

Figure 130 Characteristics of the filter

used for the initial check of the evaluation
proposal

A filter with the characteristics presented in Figure 130 was used.

The results of the initial check are presented in Figure 131, Figure
132, Figure 133, Figure 134 and Figure 135. On the right side of
these figures four numbers are presented; Up,filt (in V and in %) and
Up,orig (in V and in %). Up,filt is the peak value of the filtered curve, Up,
orig is the peak value of the original curve.
The first figure, Figure 131, shows the well-known case 11 of the
IEC-TDG form the IEC 61083-2. The filtered curve shows a slight
overshoot but no more oscillations. The difference in the peak is very
small due to the low amplitude of the superimposed oscillation. The
result of the filtering is within the reference values given in IEC.

Figure 131 Original and filtered signal of a

measured lightning impulse (case 11 of
IEC TDG)

1,2

original

MV

filtered

Up, filt = 0,944 MV

0,8
1

(100 %)

0,6

Up, orig = 0,949 MV

0,4
0,9

(100,5%)

0,2
0

0,8
9

10

11

12

-0,2
0

116

10

15

20

25

13

30

Proposal
Figure 134 and Figure 135 and Figure 133 are respectively the cases
13 and 14 from the IEC-TDG, which represent measured impulses of
lightning impulse tests performed on a transformer. The filtered
curve of the impulse with a long duration overshoot shows no
influence of the filter and the test voltage is the peak value of the
measured curve.
The filtered curve for an impulse with short duration shows a
reduction of the test voltage of about 1,5 % which is in accordance
with the used practice of evaluation of such kind of impulses, but not
in accordance with the present standards.
Concerning the proposed evaluation procedure it is not of importance
that the amplitude of the superimposed oscillations is above 5% and
therefore outside the allowed range.
0,2

original

-0,8

MV

filtered

Figure 132 Original and filtered signal of a

measured lightning impulse with long
duration overshoot (case 13 of IEC-TDG)

-0,9

Up, filt = 1.066 MV

-0,2

(100 %)

-1

-0,4

Up, orig = 1.071 MV

-0,6

-1,1
10

11

12

13

(100,5%)

14

-0,8
-1
-1,2
0

10

15

20

25

30

0,2

original

-0,7

MV

Figure 133 Original and filtered signal of a

measured lightning impulse with short
duration overshoot (case 14 of IEC-TDG)

filtered

-0,8

U
-0,2

-0,9

-0,4

-1

Up, filt = 1,056 MV

(100 %)

-1,1

-0,6

10

11

Up, orig = 1,071 MV

12

(101,4%)

-0,8
-1
-1,2
0

10

15

20

25

30

117

Proposal
Figure 134 and Figure 135 are examples of impulses with
superimposed oscillations, as obtained in the European project. The
oscillation frequency is respectively 2 MHz and 0,8 MHz.
As can be seen from the figures, filtering the curves with the k-factor
works quite good and easy. Of course the proper function of the
implementation of the k-factor by filtering has to be proven in a more
accurate way, but for the cases tested the results look quite promising.
Since the filtered curve is very smooth, the definitions of the
parameters for T1 and T2 can be kept the same.

Figure 134 Original and filtered signal of

an analytic, smooth lightning impulse
superimposed
with
an
oscillation
(f = 2000 kHz, Aosc = 20 %)

0,2

MV

-0,7

-0,8

original
filtered

Up, filt = 1,032 MV

-0,9

-0,2

(100 %)

-1
-0,4

-1,1

Up, orig = 1,168 MV

-1,2

-0,6

10

11

(113,2%)

12

-0,8
-1
-1,2
0

10

15

20

25

30

Figure 135 Original and filtered signal of

an analytic, smooth lightning impulse
superimposed
with
an
oscillation
(f = 800 kHz, Aosc= 10 %)

0,2

original

-0,7

MV

filtered

-0,8

-0,9

-0,2

Up, filt = 1,054 MV

-1

(100 %)

-0,4
-1,1

Up, orig = 1,073 MV

-1,2

-0,6

10

11

12

(101,8%)

-0,8
-1
-1,2
0

10

15

20

25

118

30

Proposal

7.3

Proposal for algorithms

Since using filtering techniques looks promising and if this method is

applicable, then the present definitions for front time and time to half
value can be maintained. The definition of the test voltage and the
peak voltage will change slightly. The test voltage is then the
maximum voltage of the filtered curve. The peak voltage is then the
maximum voltage of the recorded curve.
The proposal is to use the average k-factor function instead of all the
different k-factor functions for the different materials.
In principle there are several ways of implementing filtering
techniques:
implementation of the k-factor in the hardware of the Measuring
System
implementation of the k-factor in the software of the Measuring
System
combination of the implementation in the hardware and the
software of the Measuring System
In case the bandwidth of measuring systems used for performing HV
tests is large enough, the k-factor has to be implemented fully in a
filter.
In case the bandwidth of the measuring system in not sufficient, the kfactor function has to be partly implemented in a filter, since it is
already partly implemented in the limited bandwidth of the measuring
system.
In the following paragraphs it is assumed that the bandwidth of the
measuring systems used is large enough. I.e. frequencies of 4 MHz
can be measured with an overall uncertainty smaller than 3%, or with
a dynamic behaviour error smaller than 1%. Assuming that a
measuring system can be approximated by a first order filter, this
implies that the measuring system has a bandwidth of 28 MHz or
larger.
Because this is certainly not valid for all measuring systems used, the
k-factor function will in the practise be a multiplication of the
bandwidth of the measuring system and the filter programmed into the
software. To demonstrate the validity of the k-factor function, it is
fully implemented into a software filter in the next paragraph.

7.4

Simulation with wave shapes that occur in testing

The proposal will be checked with some wave shapes that occur in
testing. The four examples that are selected are presented in the next
paragraph and the voltage levels are summarised in Table 6.
Table 6 Summary of the examples
Example no.

Up

Uprmc

Ut

1 (LI with 1MHz oscillations superimposed)

168,50

139,36

150,72

2 (LI with 200 kHz oscillations superimposed)

166,86

103,36

153,53

3 (standard LI)

165,25

165,25

165,25

119

Proposal
4 (LI with overshoot)

120

175,15

149,50

Proposal

Example 1
A wave shape typical for transformer testing is presented in Figure
136. The peak voltage of the curve is 168,50 kV, the peak voltage of

Figure 136 a 1,3/47 s impulse with

superimposed oscillations of frequency
1 MHz and 20% amplitude

the RMC is 139,36 kV. The frequency

of the oscillations is 1 MHz. The kfactor for this wave shape is 0,39. This
means that the test voltage is 150,72 kV.

Example 2
Another wave shape typical for transformer testing is presented in
Figure 137. The peak voltage of this wave is 166,86 kV, the peak
voltage of the RMC is 103,36 kV. The frequency of the oscillation is
200 kHz. The k-factor is 0,79. This means that the test voltage will

Figure 137 standard lightning impulse

with 200 kHz oscillations of 60%
amplitude superimposed

be 153,53 kV.

Example 3
The example shown in Figure 139 is a standard lightning impulse.
The peak voltage and the test voltage are equal in this case and are
equal to 168,25 kV.

Example 4
A standard lightning impulse with superimposed overshoot is
presented in Figure 139. The duration of the overshoot is 1,36 s.
This means that the test voltage is equal to 165,97 kV.

7.5 Check of the proposal with the average k-factor

function
To check the proposal a computer program was written in Labview.
A recorded signal can be selected as input, one of the filters available
can be selected and an output file will be created. The output file can

121

Proposal
then be read into the PARAMETER program and the parameters like
the test voltage, front time and time to half-value can be evaluated.
Figure 139 Standard lightning impulse
voltage with peak voltage 165,28 kV and a
front time of 1,28 s and a time to half
value of 47,69 s.

Figure 138 Lightning impulse voltage with

overshoot, peak voltage is 175,15 kV,
peak voltage of the RMC is 149,50 kV.
The duration of the overshoot is 1,3649
s.

The filters that are available within Labview and might be useful are:
FIR windowed filter, with windows:
- none
Hann
Hanning
Triangular
Blackman
Exact Blackman
Blackman-Harris
Kaiser-Bessel
Flat top
Bessel filter
Elliptic filter
Inverse Chebyshev filter
Butterworth filter
Chebyshev filter
A few of these filters have been used to check the proposal, the results
are presented in Table 7.
Table 7 Results of the different filters
Filter type

Waveshape

Chebychev

Chebychev

Chebychev

Chebychev

order 1

order 1

order 1

order 2

60 kHz

100 kHz

90 kHz

200 kHz

Butterworth Bessel
order 2
290 kHz

400 kHz

Test voltage Peak Voltage of the filtered curve [kV]

Front time [s]
Time to half-value [s]

Standard impulse

1MHz oscillations

200 kHz oscillations

1,369 us overshoot

165,25

150,72

153,53

165,97

164,58

165,06

165,03

169,35

165,27

1,77

1,47

1,52

1,55

2,06

1,83

47,14

46,80

46,83

44,46

47,60

46,89

144,53

155,32

152,67

150,79

140,77

143,66

0,85

0,69

0,71

1,03

1,56

1,20

44,61

39,93

41,05

41,17

46,59

45,12

159,20

163,82

163,16

165,79

154,10

158,36

1,92

1,87

1,88

1,88

1,97

1,96

27,95

28,33

28,34

27,18

30,01

28,32

162,30

168,58

155,34

1,27

1,17

1,37

39,66

37,22

42,50

From the results presented in this table it can be seen that the type of
filter can have significant influence on the front time. This is logical
because the standard impulse has the same frequency components as
that of superimposed oscillations that have to be reduced slightly with
the k-factor function. Different types of filters have different effects
on the wave shape. For instance compare the results of filtering of the
standard impulse using the Chebychev order 1, 100 kHz to that using
the Bessel order 2, 400 kHz. Although the peak voltages are within

122

163,76

Proposal
10% when 1 MHz oscillations are applied the front time is twice as
large in the latter case.
From the results presented it can be concluded that it must be possible
to design a filter (combined in the measuring system and the software)
that can be used during impulse tests and that implements the k-factor.
In this thesis this investigation is not further made. Most important
was to demonstrate that it is possible to use a filter. Which filter type
in combination with the measuring system should be used should be
the topic of another study.
From the results presented in Table 7 it can be seen that a first order
Chebychev filter with a cut-off frequency of 90 MHz is not a too bad
choice. This original and the filtered wave shapes using this filter on
the four examples are presented in Figure 140 to Figure 143.

123

Proposal

Figure 140 Standard impulse

Figure 141 1 Mhz oscillations

Figure 142 Overshoot

Figure 143 200 kHz oscillations

124

Proposal

7.6

K-factor in hardware versus k-factor in software

Now the implementation of the k-factor by using filtering techniques

has proven to be possible. It can be discussed how it should be
implemented.
In principle there are several ways of implementing filtering
techniques:
implementation of the k-factor in the hardware of the Measuring
System
implementation of the k-factor in the software of the Measuring
System
combination of the implementation in the hardware and the
software of the Measuring System
In case the bandwidth of measuring systems used for performing HV
tests is large enough (28 MHz or larger.), the k-factor has to be
implemented fully in a filter.
In case the bandwidth of the measuring system in not sufficient, the kfactor function has to be partly implemented in a filter, since it is
already partly implemented in the limited bandwidth of the measuring
system.
With the Measuring Systems available nowadays, the k-factor
function will in the practise be a multiplication of the bandwidth of
the measuring system and the filter programmed into the software.
This is explained in Figure 144 and Figure 145. In Figure 144 the kfactor function and the frequency behaviour of two different
Measuring systems are presented. Measuring System 1, has a
bandwidth of 360 kHz. It is clear that in this case no software filter is
needed. In case of Measuring System 2, which has a bandwidth of 1
MHz (which is a normal bandwidth for Impulse Measuring Systems)
a software filter is needed. This software filter is presented in Figure

1,20
Figure 144 K factor function and the
frequency characteristics of two different
Measuring Systems

1,00

H(freq) [1]

0,80
0,60
0,40
0,20

Meetsysteem (1)
Proposal

0,00

Meetsysteem (2)

-0,20
10

100

1000

10000

Oscillation Frequency [kHz]

145.

125

Proposal

1,20

Figure
145
k-factor,
frequency
characteristic of the Measuring System and
the necessary software filter

1,00

H(freq) [1]

0,80
0,60
0,40
0,20

Proposal
Meetsysteem (2)

0,00

software filter

-0,20
10

100

1000

10000

Oscillation Frequency [kHz]

7.7

Final proposal and discussion

As final proposal the evaluation method presented in Figure 129 is

proposed. It can be discussed whether the whole impulse or only part
of it should be filtered. In case the whole impulse will be filtered, the
present definitions of the time parameters can be kept.
Only the limits of the tolerated time intervals might have to change,
but since the time parameters can be varied in a wider range than
mentioned in the present standard before they start to have a
significant influence on the breakdown voltage, this should not be a
too difficult problem to solve.
With the implementation of the k-factor the measurement uncertainty
for lightning impulses with overshoot and oscillations can be reduced
significantly. It is just an estimation to say that it can be reduced
from 5% to 2%.
It seems possible to use the average k-factor for all insulating
materials. What should be discussed is which tolerance in the
insulation design is tolerated. The maximum difference between the
average k-factor and the individual ones is 0,3. This means that for
instance an oscillation with an amplitude of 20%, the test voltage is
6% to low or to high. Whether this is acceptable should be discussed
in an international forum. If taking the present tolerances into
account then oscillations with an amplitude up to 10% are allowed.
This is a big step forwards.

126

Proposal

7.8

Open questions and experiments needed

After the many investigations performed, many questions have been

answered, but as usual some new questions have also arisen.
In this paragraph the open questions will be listed in random order.

Influence of time to half value in oil when the time to half value is
changed in a wider interval

What is the k function for other insulating materials not

investigated

How does the k-function affect if a certain k function is chosen to

the other materials and the measurement uncertainty

Up to which amplitude are the k-factors valid (up to 15-60% has

been investigated)

What requirements should be set to Measuring Systems

To verify whether the k-factor is also valid for other than the
investigated materials and for insulating systems, more experimental
investigations have to be performed for other complex insulating
system and other materials. But considering the measurement results
obtained so far, no results that are far from the measurement results
obtained within this project are expected.
The results obtained from new investigations performed can be
incorporated in the table for the k factor or new k-factors can be
defined.
One of the most important topics to solve before the k-factor can be
implemented is the bandwidth of measuring systems used. If the
bandwidth of measuring systems used is equal to the k-factor than it is
of course obviously that filtering techniques should not be used and
that always the peak value of the impulse should be taken as the test
voltage.
In case the bandwidth requirements of Measuring Systems are stricter
than the k-factor, than filtering techniques can be used. Which filter
will suit the best has to be investigated.
Many investigations have already been performed towards the use of
filtering techniques in impulse measurements. In [80] some of the
techniques with their advantages and disadvantages are described as
well what to do in case of chopped impulses.
Perhaps it needs to be proven that the method is valid also for
oscillation amplitudes >20%, but it seems not to be necessary.

127

Summary and conclusions

8 Summary and conclusions

Lightning strokes have caused a considerable number of failures of
high voltage equipment installed in medium and high voltage
networks. In order to prevent these failures much research has been
performed and several measures have been taken. One of these
measures is to test high voltage equipment with lightning impulse
voltages to verify the withstand capability.
In Chapter 2 tests with lightning impulses have been explained in
detail. Attention was also paid to the problems with the evaluation of
the parameters characterising lightning impulses during testing and to
the state of the art in solving these problems.

The cause for these measurement problems is that, when small

oscillations and/or overshoot occur, the rules for the evaluation of the
parameters (as described in the present standard) are ambiguous. In
case of non-standard lightning impulses no evaluation rules are given
in the present standard.
Several solutions for the problems with the evaluation have been
proposed, but most of them had some disadvantages that are not
acceptable. The most ideal solution for the evaluation problems will
be a solution, which is applicable to both standard and non-standard
lightning impulse voltages, which uses the possibilities of modern
processing techniques and which is founded on the breakdown
behaviour of insulating materials.
To find this most ideal solution that is based on the breakdown
behaviour of insulating materials or systems, theoretical and
experimental investigations were necessary. Therefore the author
initiated an European project and started this PhD study. The goal of
these projects was to investigate if the new approach was possible and
to find new ideas about how to solve the problem. In the two projects
theoretical and experimental investigations have been made by the
author. The results of these investigations and investigations
performed by others in the European project were presented.
In Chapter 3 the results of the theoretical investigation towards the
disruptive discharge behaviour of insulating systems and materials
were presented. The result of this investigation was that the
information found was not enough so that a good proposal for a new
evaluation method could be given.
In Chapter 4, the investigations needed because of the evaluation
problems and the lack of information about the breakdown physics
when lightning impulses with overshoot and oscillations are applied
are formulated. Open questions and hypotheses are formulated too.
The experimental investigations were performed simultaneously at
different laboratories (KEMA, NGC, LCOE, Schering Institute) and
later at TU Graz. For the measurements that were carried out
simultaneously the generating circuit, the software were chosen the
same to be sure that the measurement results were not influenced by
these factors. At TU Graz a less complicated generating circuit could
be used, but the software used was the same as for the other
measurements. In Chapter 5 the test set up used at KEMA and at

129

Summary and conclusions

TU Graz and the problems associated with these are described in
detail.
The experimental investigations (breakdown tests) were carried out to
find the performance of different insulating materials and systems
against lightning impulse voltages of different shape. It was
especially the task of the author to give ideas and to look at the
performance and to draw the conclusions of the results.
In Chapter 6 the results of the experimental investigations performed
at KEMA, TU Graz, LCOE, NGC and the Schering Institute were
explained in detail. The materials XLPE, oil, oil-paper, SF6 and air
were investigated. From the experimental results the following
conclusions are drawn.
It was concluded that the present limits for the front time in
IEC 60060-1 are valid. The peak value doesnt change significantly
when the front time is varied within certain limits. It can be discussed
that the limits can be extended from 0,84 s-1,56 s to 0,5 s-1,56 s.
The time to half value have only been studied for PE and oil. For PE
it seems not to be a significant parameter and for oil no significant
change in the breakdown voltage can be detected when the time to
half value is varied from 40 s to 50 s.
There was no detectable difference in the influence of oscillations and
the influence of overshoot on the breakdown voltage. Therefore all
conclusions for oscillations are also valid for overshoot.
The position of the oscillations, i.e. on the peak or on the front has an
influence on the 50% breakdown voltage for air and for oil. The
influence increases when the position of the oscillations is closer to
the peak of the lightning impulse. Most relevant seems to be the
increase of the peak value of the curve with oscillations compared to
that without oscillations.
The amplitude and the frequency of the oscillations have an effect on
the breakdown voltage. In general an oscillation with an high
amplitude and a low frequency has more influence than an oscillation
with a low amplitude and a high frequency.
Because in all cases the test voltage was in between the peak valeu of
the applied curve and the peak value of the base curve, the idea with
the k-factor arose. Therefore a linearity statement and a k-factor
hypothesis have been introduced and proven to be valid for most
materials investigated.
It has been proven that the test voltage is, for different materials and
for oscillations with an amplitude up to at least 15%, a function of the
peak voltage and the peak voltage of the mean curve.
But since for paper oil insulation the theorem seems to be valid for
amplitude at least up to 60% but perhaps even for pure oscillations, it
is not expected to be valid for much higher amplitudes.
The best k-factor for the different materials investigated has been
determined. Because the results were so close to each other an
average k-factor could be determined. The author was the first to
calculate and to propose this average k-factor.

130

Summary and conclusions

The question if and how this k-factor could be implemented is
answered in Chapter 7. There are several ways to implement this kfactor function, one of the most easy ways is to use filtering
techniques which the author proposed during one of the meetings held
about the European project. One thing that should absolutely not be
forgotten is the bandwidth of the measuring system used. The kfactor is namely implemented in the measuring system, which
includes the hardware and the software.
If the bandwidth of the measuring systems follows the k-factor
function then in all cases the peak value has to be taken. In case the
bandwidth of the measuring systems used is more then sufficient, the
k-factor can be implemented by for instance using filtering
techniques. In case of manual evaluating the determination of the test
voltage can easily be done by calculation.
Therefore it is certainly possible to evaluate standard and nonstandard impulses using the same evaluation procedure. Before a
final method can be presented a few questions (presented in Chapter
7) should be answered in discussions to be held internationally, e.g. In
WG 33.03.
By implementing the k-factor the problems associated with the
present IEC 60060 can be solved. The evaluation of standard and
non-standard impulses can be done using the same evaluation method.
The evaluation method is the same for all insulating materials and
insulating systems, only different k-factors might be used in the
future. One of the most important advantages of the presented
method is that it is no longer necessary to calculate the mean curve
and that the method can be used independent of the technology
position the laboratory. It is besides this easy to understand and to
apply.

131

Appendices

Appendices
Appendix 1
Appendix 2
Appendix 3

133

Appendices

Appendix 1
Evaluation of parameters by
the software package PARAMETER
The following flowchart (Figure 146) presents the steps to calculate
the standard and complementary parameters of lightning impulses.
The user enters by visual examination of the wave if it is a full,
chopped in the front or on the crest or chopped in the tail impulse.
Figure 146 flowchart of steps to take to
evaluate parameters

Impulse acquisition

Identification of the type of the impulse

Determination of the base line

( BL) and the peak value ( Up)

Step 1

RMC and parameters of the oscillations

Step 2

Determination of the parameters of the overshoot

Step 3

Determination of time parameters and the voltage

at the instant of chopping ( Uc)

Output parameters

Step 4

Step 5

Step 1: Determination of the base line (BL), origin (O), and

the peak value (Up)
The base line (BL) is the average of consecutive samples proceeding
the time when the impulse is applied. In order to reduce the influence
of noise, the number of samples used must be enough.
Note: In order not to include in the average of the base line samples corresponding to
the beginning of the impulse, some samples (~ 10) before the origin (0) are not
included. A number of 100 samples is considered to be enough to determine the base
line (BL).

134

Appendices
The origin (O) is the time when a sample of the recorded impulse first
exceeds the base level by 2% of the peak value.
Note: In order not to consider as the origin (0) a sample with an amplitude > 2% Up
due only to superimposed noise, the amplitude of the next samples must be also
greater than the 2% Up.

The peak value (Up) of an impulse is the absolute maximum voltage

value after the superimposed noise is removed.
For full or chopped in the tail impulses the peak value, Up, is
determined by the difference between the mean value of the samples
in the time interval around the peak zone and the base line adding
0,5 digital levels. The first and the last samples whose amplitudes are
identical define the time interval around the peak zone. These
amplitudes should not be greater than the maximum absolute
amplitude minus one digital level, but as large as possible.

I=I 2

W(I)

U p=

I=I1

( I 2 I 1 + 1)

BL + 0,5

Figure 147 Zoom around the crest of full

or chopped in the tail impulses
Note: The addition of 0,5 digital levels is to compensate the averaging effect between
approximately the same number of samples of levels MAX and MAX-1 for smooth
impulses without noise.

In order to avoid with this method a peak value lower than the real
one when high frequency superimposed oscillations appear, the two
following conditions must be satisfied:
1. |MAX W (I1)| < NOISE BAND, Noise band is defined as the
maximum difference between any two samples of the base line
(BL)
2. The samples I2 and I1 must belong to the same oscillation. If these
two conditions are not satisfied the peak value Up is the maximum
absolute amplitude (MAX).

Figure 148 Conditions

Condition 1 not fulfilled

Condition 2 not fulfilled

135

Appendices
For impulses chopped in the front or on the crest, in general, it is not
possible to remove the noise in the peak zone by means of the above
criterion (see Figure 149). In these cases, the greatest amplitude of the
impulse will be the peak value.

Step 2: RMC and parameters of oscillations:

The RMC is directly acquired by means of a channel of the recorder.
By means of the second channel the oscillations to be superimposed
are acquired. The subtraction of both channels of the recorder is the
impulse with superimposed oscillations.
Figure 149 Zoom around the crest of
impulses chopped in the front or on the
crest. The average of the samples included
in the time interval (I2 I1) can not be
considered correct as peak value. In this
case the greatest amplitude will be the
peak value

Frequency, f, and amplitudes, Ai, of oscillations:

The main frequency, f, and the amplitudes, Ai, of the oscillations can
be evaluated by determining the main frequency and amplitudes of the
wave acquired by means of the channel 2, for instance, by means of
the FFT. The amplitudes Ai are calculated after the superimposed
noise is removed. If during the record length the CH2 curve is not
damped inside the noise band, in order to avoid calculation errors, it
is advised that the FFT is calculated with an integer number of
periods of CH2.

Wave integral above a reference voltage Us, :

In addition, another parameter for oscillations on the crest is defined:

wave integral above a reference voltage, .
This parameter, , is the sum of the areas enclosed by the recorded
impulse above (or below for negative impulses) the horizontal line
that passes through a reference voltage level (Us).

Step 3: Determination of parameters of the overshoot

This step only can be applied to full impulses or chopped in the tail
impulses.
An overshoot is present in an impulse if, after removing the eventual
superimposed noise and oscillations by means of the RMC, the
resulting curve is above (or below for negative impulses) the single
exponential function, SE, that fits the tail of the recorded impulse.
SE(t) = A e-(t-to)/
Where: A, to and are the parameters calculated to achieve the best
fitting (using the same criterion of minimisation as the one
used for the RMC) to the tail of the impulse.
Note: The single exponential fitting is determined for samples of the recorded
impulse after the 0,7 Up voltage in the tail. For impulses chopped in the tail
the fitting is considered to be suitable only if enough samples are recorded
after this level.

Once an overshoot has been identified, the following parameters can

be calculated:

136

Overshoot peak value, OPV:

Virtual peak value, VPV:
Wave integral above the VPV:
Overshoot amplitude:
Overshoot duration:

Appendices
The algorithms to determine these parameters are described in
internal Document of the European project: Doc. Partner 2/10 Rev 1
and are not explained here.

Step 4: Determination of time parameters, and voltage at the

instant of chopping, Uc
The three time parameters that can be determined are: T1, T2, Tc.
In order to calculate T1 and T2 it is necessary to determine the instants
t30, t90, t50, when the impulse reaches the voltage levels: 30% Up, 90%
Up and 50% Up.

Front time, T1 :

If there is noise in the front it is necessary to remove it by using a

moving window filter. Even without noise an appropriate filter
improves the results due to a lower quantisation error.
If there are no oscillations the instants t30 and t90 are determined after
filtering by means of the straight line between the first sample just
below (or above for negative slope) the voltage level and the last
sample just above (or below for negative slope) this level, (see Figure
150).
If there are oscillations, the instants t30 and t90 can be determined
replacing the impulse with the superimposed oscillations by a mean
curve at the front, MC. The MC used, is determined using as the time
interval, from origin to the instant corresponding to the peak value,
and it pass through the origin, O, and the peak value, Up.
The front time is a virtual parameter calculated as:

Virtual origin, O1:

Figure 150 Explanation of 90% Up

The virtual origin, O1, is the instant where the straight line which
passes through the points (t30, 30% Up) and (t90, 90% Up), cuts the
baseline, BL.

Time to half value, T2:

The algorithm to determine T2 is defined in internal Doc. Partner 2/10

Rev1.

Time to chopping, Tc, and voltage at the instant of chopping Uc:

For lightning impulses chopped in the front or on the crest, the instant
to chopping, Tc, is the intersection point of the straight line through
90% and 30% of the peak voltage and the horizontal line that passes
through the impulse peak voltage (see Figure 151).

Figure 151 Lightning impulse chopped on

the front
Note

The times t90%, t30% at the zone of decay voltage are determined by linear
interpolation between the closest samples.

137

Appendices
For impulses chopped in the tail, the instant of chopping, T c, can be
determined also as the intersection point of two straight lines. One
straight line is the regression line of the ten samples before Tc, taking
into account that, when oscillations appear, the ten samples are taken
from the RMC. The second straight line is the one defined by the
points of the 90% and 30% of the voltage at the instant of chopping,
Uc at the zone of decay voltage.
Note:

The oscillations are considered disappeared if the amplitude of the residual

curve
R(t)
is
less
than
2% Up during the time interval used to calculate the regression line.

The voltage at the instant of chopping, Uc is the voltage value of the

original impulse at the instant Tc. In order to apply this method it is
necessary to have an initial value for Tc. The initial value, Tc, is
defined as the sample when the voltage decreases at least 10% Up in
30 ns and at least 3% Up in 10 ns.
With the initial value Tc and the corresponding Uc (Tc) value it is
possible to determine the two straight lines previously described. The
intersection point between both straight lines corresponds with the
final instant of chopping Tc, and the corresponding Uc (t) can be also
calculated.

138

Appendices

Appendix 2 Table with measurement data

Results of tests performed in air
Date of test
[yy-mm-dd]

Test no.

Temp.
[C]

Air
press.

Humidity

Medium

[gH2O/m3]

[hPa]

gap
distance

wave
shape

Electro
de
conf.

U50(RMC)

sd

[kV]

[kV]

U50(RMC)/k

[mm]

[kV]
98-09-29

98-09-29

98-09-29

98-09-30

98-09-30

98-09-30

98-10-01

98-10-02

980929-3 23

980929-4 23

980929-5 23

980930-1 22

980930-2 23

980930-3 23

981001-1 22

981002-1 22

1003

1003

1003,5

1000

999,5

999,5

1004

1007

14

14

14

13

14

14

13

10,5

air+UV

air+UV

air+UV

air+UV

air+UV

air+UV

air+UV

air+UV

15

15

15

15

15

15

15

15,1

1,6/50

1,6/50

1,6/50

sphereplate

-42,123

sphereplate

-42,172

sphereplate

-42,172

1,6/50+ sphere2,5MH plate

z
on
start
1,6/50+ sphere2,5MH plate
z
on
front
1,6/50+ sphere2,5MH plate
z
on
peak

-42,988

-43,028

-43,007

-42,875

-42,886

-41,022

1,6/50+ sphere2,5MH plate

z
on
tail

-41,439

1,8/50

sphere-

-42,323

plate

-42,866

0,169 Scalefactor
19530
Scalefactor
19530
Scalefactor
19530
Scalefactor
19530

Scalefactor
19530

Scalefactor
19530

Scalefactor
19530

0,132 Scalefactor
19530

139

Appendices

Date of test

Test no.

[yy-mm-dd]

Temp.
[C]

Air
press.

Humidity

Medium

[gH2O/m ]

[hPa]

gap
distance
[mm]

wave
shape

Electro
de
conf.

U50(RMC)

sd

[kV]

[kV]

U50(RMC)/k
[kV]

99-01-19

990119-1

air+UV

15

sphere- 45,197
sphere
45,385

99-01-20

990120-1

air+UV

15

sphere- 42,659
sphere
42,879

99-01-20

990120-2

air+UV

15

sphere- 45,218
sphere
45,452

99-01-21

990121-1

air+UV

15

sphere- 45,195
sphere
45,281

99-01-21

990121-2

air+UV

15

sphere- 45,263
sphere
45,309

99-02-05

990205-1

air

15

sphere- 45,191
sphere
45,364

99-02-05

990205-2

air

15

sphere- 42,911
sphere
43,069

99-02-05

990205-3

air

15

sphere- 44,990
sphere
45,151

99-02-08

990208-1

air+cover 15

sphere- 43,228
sphere
43,422

99-02-08

990208-2

air+cover 15

sphere- 45,010
sphere
45,212

140

Lightning Impulses

Appendices
Results of tests performed in oil
All tests are performed using a gap distance of 1 mm and spheresphere configuration
Date of test

Test no.

[yy-mm-dd]

Temp Air
Humidity
press.
[C]
[gH2O/m3]
[hPa]

wave
shape

oscillatio
ns

oversh
oot

U50(RMC)

U50(peak) sd (rmc)

Sd (peak)

[kV]

[kV]

[kV]

[kV]

99-03-05

O990305a

19

986

1,2/50

no

69,494

5,506

99-03-11

O110399a

18

1015

1,2/50

no

68,815

1,748

99-04-22

O990422a

22

1002

11

1,2/50

no

71,659

3.051

99-04-26

O990426a

23

1015, 10
5

0,85/50

no

69,593

4,113

99-04-28

O990428a

22

1020

0,5/50

no

66,932

4,500

99-05-04

O990504a

22

1014

0,54/49

no

69,106

2,707

99-05-06

O990506a

21,5

1012

11

1,5/51

no

67,987

5,481

99-05-06

O990506b

21,5

1012

11

1,9/51

no

69,000

4,556

99-05-12

O990512a

21

1006

11

0,85/49

no

68,731

3,71

99-05-12

O990512b

21

1006

11

1,2/40

no

66,599

3,621

99-05-17

O990517a

21

1016

0,31/47

no

71,360

2,725

1999-06-22 O990622a

21

1018 10

1999-05-25 O990525a

1,2/50

1 s

65,237

76,638

4,248

4,192

1,2/50

1,2 s

60,612

87,99

5,357

6,104

1999-05-19 O990519a

22

1008, 9
5

1,2/50

2 MHz

70,002

77,632

4,116

3,948

1999-05-18 O990518a

22

1008, 9
5

1,2/50

800 kHz

65,514

77,656

4,136

4,036

1013 8

1,2/50

500 kHz

65,178

77,541

4,494

4,349

1028 9

1,2/50

200 kHz

61,54

75,196

5,261

5,15

1999-05-17 O990517b
1999-03-16 O990316a

20

141

Appendices

Appendix 3 Oil analysis results

142

Lightning Impulses

Appendices

143

Appendices

144

Lightning Impulses

Appendices

145

References

References
The list of references is divided into two parts, public available
references and non-public available references. If the reference
number you are looking for is not in the first list, it will be in the
second. For example the reference numbers 1 to 4 and 7 to 9 are in the
public available list the missing numbers in the non-public available
list.

Public available references

1

IEC: High voltage test techniques, part 1: General definitions

and test requirements IEC 60060-1: 1989.

Ryan, H.M.: High voltage engineering and testing IEE power

series 17, 1996, ISBN 0 86341 293 9

Benda-Berlijn, S.M., Scheltinga, L.: The past, the present and

the future of figure 10 in IEC 60-1 ERA conference, Milan,
1996, paper no. 5.3

Hllstrm, J. Aro, M., Karlsson, H., Gunnarsson, L.-O.,

Wedberg, M., Pesavento, G., Gobbo, R., Benda-Berlijn, S.M.,
Scheltinga, L., Schon, K., Lucas, W., Blanc, I., Gournay, P.,
Sacepe, B., Weck, K.-H., Kster, H.-J., Garnacho, F., Tangen,
K.O., Rizzi, G., Neves Gomes, F., Muhr, M. and Renger, H.:
European intercomparison of h.v. impulse measuring systems
with digital recorder ISH, Montreal, 1997, Volume 4, page 712

Buiter, H.: The history of the Power Cable in the Netherlands

Stichting
Historie
der
Techniek,
ISBN 90-73192-16-1

ISO/TAG/WG3: January 1993, Guide to the expression of

uncertainty in measurement

IEC:Amendment to IEC 60060-2- Annex H Procedure for

estimation of uncertainty in high-voltage measurement
IEC 60060-2, annex H: 1996

13

Schon, K., Lucas, W., Cherbaucich, C., Rizzi, G., Deschamps,

F., Ribot, J., Garnacho, F., Perez, J., Gomes, N., Dias, C., Aro,
M., Valve, P., Claudi, A., Lehmann, K., Strauss, W. and
Notkonen, E.:International comparison of software for
evaluating HV impulses and step respones 8th ISH, Yokoham
1993, paper no 51.01, pp. 289-292

14

Berlijn, S., Kvarngren, M., Garnacho, F., Simon, P.,

Gockenbach, E., Werle, P., Hackemack, K., Watts, M and
Wong, K.C.P.: Todays problems with the evaluation methods
of full lightning impulse parameters as described in IEC
60060-1, ISH 1999.

18

Perez, J., Garnacho, F., Martinez, J.:Circumference Arcs: a

method to calculate a mean curve in lightning impulses
acquired with a digital recorder Paper 51.03 ISH 93 Yokohama

147

References

148

19

Zhang, Y.X., McKnight, R.H and Fenimore, C.: A method for

fitting and smoothing digital data Paper 50.06, 6th ISH, New
Orleans 1989

20

Fenimore, C. and Zhang, Y.X.:Robust estimation in parameter

extraction IEEE panel session on digital techniques in HV
tests, Long Beach, California, 1989

21

Li, Y.M., Kuffel, J. and Janischewskyj, W.:Computer

procedures for the evaluation of parameters of digitally
recorded impulses Paper 62.07, 7th ISH, Dresden 1991

23

Songlin, Q., Xikang, G.: The evaluation of waveform

parameters of digitally recorded lightning impulses by
regression method ISH 1993, Paper no 51.04, Yokohama

24

Valve, P.,Aro, M.:Curve fitting in parameter estimation of

high voltage impulses Helsinki University of Technology, high
voltage institute, Espoo-Finland, report Tkk-SJT-17, november
1995, 26 p.

26

McComb, T.R., Lagnese, J.E.:Calculating the parameters of

full lightning impulses Using Model-based Curve fitting IEEE
Transactions on Power delivery, 6 1991-4), p.1386-1392

27

Prez Sanz, J.:Aplicacin del Filtro de Kalman a la

determinacin de los parmetros de impulsos tipo rayo en
sistematas de medida con registro digital, conforme a la norma
IEC 60-1. Tesiis Doctoral, Madrid, 1994

28

Garnacho, F., Perez, J.:Evaluation of lightning impulse

parameters. Proposal to establish the criteria to determine the
reference mean curve in lightning impulses ISH 95, paper
4547, Graz

29

Brede, P.A., Werle, P., Gockenbach, E., Borsi, H.:A new

method of determining the mean curve of lightning impulses
according to IEC 60060-1 11th ISH, London, 1999, page
1.74.S21-1.77.S21, Volume 1

30

Kato, T.:Mean-curve fitting using tension splines for impulse

parameter evaluation 11th ISH, London 199, page 1.78.S211.81.S21, Volume 1

31

Angrisani, L., Daponte, P.:A wavelet based approach for

measurements on high voltage impulses, 11th ISH, London
1999, page 1.205.S21-1.208.S21, Volume 1

32

Schon, K.:Digital filtering of HV lightning impulses IEEE

Panel Session on digital techniques in HV tests, Long Beach,
IEEE summer meeting California 1989

33

Dias, C.P., Moura, A.M., Valve, A.A.:Digital filetring of HV

impulse test results 7th ISH, Dresden 1991, paper 62.08

35

Wong, K.C.P., Ryan, H.M., Blackett, J., Watts, M.W.:Digital

measurement of lightning impulse parameters using curve
fitting algorithms 11th ISH London 1999, page 1.193.S211.196.S21, Volume 1

References
38

Kind D:An introduction to high-voltage experimental

technique Friedr. Vieweg & Sohn Braunschweig, ISBN 3-52808383-2

39

Kawashima, T., Okabe, S. and Zaima, E. Evaluation of steepfront overvoltage waveforms for oil-filled transformers 11th
ISH 1999, London, Volume 1, p. 1.279.P4-1.282.P4

40

Kvarngren, M.:Screening of sharp points in compact

transmission towers TRITA-EEA-9706 ISSN 1100-1593

41

Kind, D., Krner, H.:High-voltage insulation technology

Friedr. Vieweg & Sohn Braunschweig, ISBN 3-528-08599-1

42.

Lemke, E.:Ein Beitrag zur Abschtzung der rumlichzeitlichen Entwicklung des Durchschlags in langen
Luftfunkstrecken bei Schaltspannungen Wiss. Zeitschrift der
TU Dresden 26 (1977), 1, p.133-136.

43

Torshin, Y.V.:Experimental investigation of common

regularities of breakdown in transformer oil 7th ISH vol.2,
paper no 2102, p.17-20 Dresden, Germany August 26-30, 1991.

44

Breitfelder, D., Buckow, E., Knorr, W. and Peschke,

W.:Dielectric strength under transformer oill under impulse
and hhigh frequency voltage stress 7th ISH, Vol. 2, paper no.
2212, p1-4, Dresden Germany, August 26-30, 1991

45

Alpert, D, Lee, D.A., Lyman, E.M. Tomaschke,

H.W.:Initiation of electrical breadkwon in ultra-high vaccum
J.Vac.Sci.Technol.Vol1 (1964), p. 35-50

46

Maitland, A. Influence of the anode temperature in the

breakdown voltage and condition characteristics of a vacuum
gap Brit.J.A[[;.Phys.Vol13 91962), p.122-125.

47

Cranberg, L.:The initiation of electrical breakdown in

vacuum J.Appl.Phys.,Vol23 (1952), P.518-522

48

Kind, D.:Die Aufbauflche bei Stobeanspruchungen

technischer Elektrode Anordnungen in Luft Dissertation TH
Mnchen 1957

51

Manjunath, B.S., Dwarakanath, K., Naidu, M.S.:Influence of

polarity and field non-uniforminty on the rbeakdwon voltage
characteristics of SF6 under impulse voltages1994 IEEE
International symposium on electrical insulation, p.485-488,
Pittsburg, PA (USA) June 5-8, 1994

52

Qiu, Y., Lu, C.Y. and Zhang, M.:Voltage-time characteristics

of SF6 and SF6-N2 under lightning and steep fronted impulses
Conference record of the 1992 International symposium on
Electrical insulation, p.310-313 Baltimore, MD (USA), June 710, 1992, IEEE, New York, 1992

53

Heers, B., Stepken, A.:Investigation on discharge development

in SF6 for oscillating impulses and fast oscillating
overvoltages7th ISH, Vol. 3, paper no. 3106, p./29-32,
Dresden, Germany August 26-30, 1991

55

Strigel, R.Elektrische Stofestigkeit Springer, Berlin 1955

149

References

150

56

Venkataseshaiah, C, Sarathi, R and Aleyas, M.V.:Studies on

the dielectric strength of transformer oil under oscillatory
impulse voltagesConference recod of the 1996 iEEE
international symposium on electrical insulation. Vol2, p. 554556, Montreal Quebec, Canada June 16-19, 1996

57

Asmuth, P.:Das Durchsclagverhalten

Polyethylen
nach
unterschiedlicher
Dissertation, 1982, Universitt Hannover

58

Gockenbach, E., Schiller, G.: Influence of the material

conditions on the behaviour of XLPE-samples at voltages of
different shapes Proceedings of the 1996 IEEE International
symposium on Electrical Insulation, part2, Montreal, 1996

59

Kahle,
M.:
Der
Einflu
von
Gleichspannungsvorbeanspruchungen
und
Erholezeit
auf
die
Stodurchschlagfestigkeit hochpolymer Folien Elektrie, 1971,
heft 25, S. 315-317

60

Nakagawa, S. et. al.:Breakdown characteristics of XLPE cable

at high temperature 8th ISH Yolohama 1993

61

Peschke, E.F., Kalkner, W., Schrder, G.: Impulse strength of

PE and XLPE insulations 3th ISH, Milan 1979

62

Selle, F.: Durchschlagverhalten von Kunstoffisolierungen bei

berlagerungen von Gleich- und Stospannung Dissertation
Universitt Hannover, 1987

63

Rittner, K., Brner, G.: treeing inception voltage of XLPE at

Lightning Impulse voltage ETEP, Vol.2, No. 4, July/August
1992

64

Zieschang, R.: Zum Kanaleinsatz und Durchschlag in

Polyethylen bei Belastung mit Blitzimpulsspannung
Dissertation TU Dresden, 1993

65

marx, E.: Versuche ber die Spannung von Isolatoren mit

Spannungssten Elektrotechnische Zeitschrift, 1924, Heft 25,
S 652-654

66

Toepler, M.: Stospannung, berschlag und Durchschlag bei

Isolatoren Elektrotechnische Zeitschrift, 1924, Heft 40, S
1045-1050

67

Watson, D.B.: The impulse strength of polyethylene as

function of voltage rise time Journal of Physics, No.4, L19L20, 1971

68

Siodla, K and Zubielik, P.:Electric strength of the vacuum

interrupter under ligthning and switching impulse voltage XVth
International symposium on discharges and electrical insulation
in vacuum. p. 632-635. Darmstadt, Germany, September 6-10,
1992

69

Niessen, F.H.M.:Literature study on vacuum insulated

devices TUE-Netherlands, 8 August 1983, ESA Technical
Management, the Netherlands 1983

70

Ryan, H.M.: High voltage engineering and testing IEE power

series 17, 1996, ISBN 0 86341 293 9

von vernetztem
Vorbelastung

References
72

Hauschild, W and Mosch, W.:Statistical techniques for highvoltage engineeringPeter Peregrinus Ltd., ISBN 0 86341 205
X

73

Pesgens, F.:Towards an improved method for measuring

lightning impulse voltages , Thesis 1999 TUE-Netherlands

75

IEC 60052, sphere gaps

76

Bongaerts, M.:The influence of parameters of a lightning

impulse on the breakdown behaviour of insulating materials
14-8-1998, EH.98.A.152, thesis, Technical University
Eindhoven

Non-public available references

5

Pack, Dipl.-Ing. Dr.techn. Stephan :Transiente Spannungen in

Hochspannungssystemen der Energieversorgung 2.berarbeite
Auflage, Mrz 1998 (page 101-114).

Henneveld, B.: Bliksem en bliksembeveiliging HOMMEMA

Engineering B.V., april 1990.

10

STL, TC Guide to establish traceability of high-voltage

measuring systems used in high-voltage testing to standards of
measurement, Provisional STL Guide, draft September 1996

11

Claudi, A. Neue Entwicklungen in der Impulsprftechnik

Haefely Trench Symposium, 25-26 May 1998, Stuttgart, paper
no. 4.

12

Garnacho, F., et.al.... Evaluation procedures for lightning

impulse parameters in case of waveforms with oscillations
and/or overshoot IEEE 1996

15

Berlijn, S.M., Garnacho, F., Simon, P., Gockenbach, E., Werle,

P., Hackemack, K., Watts, M.:Final report, digital
measurement of parameters used for lightning impulse tests for
high voltage equipment contract no. PL-951210-SMT-CT962132, 17 September 1999

16

Hughes, R.C., Benda-Berlijn, S., Hanique, E., McComb, T.R.:

A survey of parameters for comment and further suggestions,
CIGRE 33-96 (WG 03) IWD 11

17

Garnacho, F., Perez, J. Aro, M, Valve, P, Schon, K. Kato, T.

McComb, T., Pesavento, G., Hauschild, W., Gockenbach, E.,
Ramirez, M. Munoz, P., Larzeler, B. and Xin Zhang,
Y.:Evaluation procedures for lightning impulse paramters in
case of waveforms with oscillations and/or overshootCIGRE
33-96 (WG03) 10 IWD.

22

Denis, E. and Morin, C.:Two methods for curve fitting using

double exponential model CIGRE 33-93 (WG03) 40 IWD

25

Kato, T.Double exponential function fitting for impulse

parameter evaluationCIGRE 33-93(WG03)30IWD

34

Lee, H., Rungis, J.:A non subjective spectrum filtering method

of impulse waves. CIGRE 33-93 (WG03) 7 IWD

151

References

152

36

Garnacho, F. et al.:Mean curve definition and evaluation

procedure of standard impulse parametersCIGRE 3395(WG03) 25 IWD

37

Hughes, R.C. and Hall, A.C.: Characerisation of hv impulses

with non-conventional waveforms, report to Study Committee
12 CIGRE 33-97 (WG03) 19 IWD

49

Report of LCOE, e.g. document 2/12 14/01/98

50

Wetzer, J. van der Laan, P.:Hoogspanningstechniek

1Eindhoven University of Technology, Eindhoven the
Netherlands, Dec 1994

54

Garnacho, F., Simon, P. P2/12

71

Benda-Berlijn, S., Kvarngren, M., Garnacho, F., Gonzales, A.,

Simon, P., Gockenbach, E., Werle, P., Hackemack, K., Watts,
M and Wong, P.:An investigation into the relevancy of the
wave shape parameters used for lightning impulse tests
CIGRE WG 33.03, 1998, CIGRE WG33.03 1998 IWD 15.

74

Gruntjes, R. :Simulation of the generated impulses by a simple

generator project document P1/11

77

F. Garnacho en P. Simon:Software validation Document

P2/16, 12/3/98

78

Hauschild.W., Mosch.W.:Statistical techniques for highvoltage engineering ISBN 0 864341 205 X

79

Simon, P.: Investigacin de los parmetros caractersticos de

comportamiento de medios dielctricos frente a sobretensiones
tipo rayo no normalizados en alta tensin". Dissertation to be
presented at a later time. Professor: F. Garnacho.

80

Bosch, E., Tuin, J.:Self-adjusting digitaal filter voor

stootspanningsproeven aan transformatoren SMIT Nijmegen,
A.P. E.92.2.26

References

153

References

154

IEC: High voltage test techniques, part 1: General definitions

and test requirements IEC 60060-1: 1989.

Ryan, H.M.: High voltage engineering and testing IEE power

series 17, 1996, ISBN 0 86341 293 9

Benda-Berlijn, S.M., Scheltinga, L.: The past, the present and

the future of figure 10 in IEC 60-1 ERA conference, Milan,
1996, paper no. 5.3

Hllstrm, J. Aro, M., Karlsson, H., Gunnarsson, L.-O.,

Wedberg, M., Pesavento, G., Gobbo, R., Benda-Berlijn, S.M.,
Scheltinga, L., Schon, K., Lucas, W., Blanc, I., Gournay, P.,
Sacepe, B., Weck, K.-H., Kster, H.-J., Garnacho, F., Tangen,
K.O., Rizzi, G., Neves Gomes, F., Muhr, M. and Renger, H.:
European intercomparison of h.v. impulse measuring systems
with digital recorder ISH, Montreal, 1997, Volume 4, page 712

Pack, Dipl.-Ing. Dr.techn. Stephan :Transiente Spannungen in

Hochspannungssystemen der Energieversorgung 2.berarbeite
Auflage, Mrz 1998 (page 101-114).

Henneveld, B.: Bliksem en bliksembeveiliging HOMMEMA

Engineering B.V., april 1990.

Buiter, H.: The history of the Power Cable in the Netherlands

Stichting
Historie
der
Techniek,
ISBN 90-73192-16-1

ISO/TAG/WG3: january 1993, Guide to the expression of

uncertainty in measurement

IEC:Amendment to IEC 60060-2- Annex H Procedure for

estimation of uncertainty in high-voltage measurement IEC
60060-2, annex H: 1996

10

STL, TC Guide to establish traceability of high-voltage

measuring systems used in high-voltage testing to standards of
measurement, Provisional STL Guide, draft September 1996

11

Claudi, A. Neue Entwicklungen in der Impulsprftechnik

Haefely Trench Symposium, 25-26 May 1998, Stuttgart, paper
no. 4.

12

Garnacho, F., et.al.... Evaluation procedures for lightning

impulse parameters in case of waveforms with oscillations
and/or overshoot IEEE 1996

13

Schon, K., Lucas, W., Cherbaucich, C., Rizzi, G., Deschamps,

F., Ribot, J., Garnacho, F., Perez, J., Gomes, N., Dias, C., Aro,
M., Valve, P., Claudi, A., Lehmann, K., Strauss, W. and
Notkonen, E.:International comparison of software for
evaluating HV impulses and step respones 8th ISH, Yokoham
1993, paper no 51.01, pp. 289-292

14

Berlijn, S., Kvarngren, M., Garnacho, F., Simon, P.,

Gockenbach, E., Werle, P., Hackemack, K., Watts, M and
Wong, K.C.P.: Todays problems with the evaluation methods

References

of full lightning impulse parameters as described in IEC

60060-1, ISH 1999.
15

Berlijn, S.M., Garnacho, F., Simon, P., Gockenbach, E., Werle,

P., Hackemack, K., Watts, M.:Final report, digital
measurement of parameters used for lightning impulse tests for
high voltage equipment contract no. PL-951210-SMT-CT962132, 17 September 1999

16

Hughes, R.C., Benda-Berlijn, S., Hanique, E., McComb, T.R.:

A survey of parameters for comment and further suggestions,
CIGRE 33-96 (WG 03) IWD 11

17

Garnacho, F., Perez, J. Aro, M, Valve, P, Schon, K. Kato, T.

McComb, T., Pesavento, G., Hauschild, W., Gockenbach, E.,
Ramirez, M. Munoz, P., Larzeler, B. and Xin Zhang,
Y.:Evaluation procedures for lightning impulse paramters in
case of waveforms with oscillations and/or overshootCIGRE
33-96 (WG03) 10 IWD.

18

Perez, J., Garnacho, F., Martinez, J.:Circumference Arcs: a

method to calculate a mean curve in lightning impulses
acquired with a digital recorder Paper 51.03 ISH 93 Yokohama

19

Zhang, Y.X., McKnight, R.H and Fenimore, C.: A method for

fitting and smoothing digital data Paper 50.06, 6th ISH, New
Orleans 1989

20

Fenimore, C. and Zhang, Y.X.:Robust estimation in parameter

extraction IEEE panel session on digital techniques in HV
tests, Long Beach, California, 1989

21

Li, Y.M., Kuffel, J. and Janischewskyj, W.:Computer

procedures for the evaluation of parameters of digitally
recorded impulses Paper 62.07, 7th ISH, Dresden 1991

22

Denis, E. and Morin, C.:Two methods for curve fitting using

double exponential model CIGRE 33-93 (WG03) 40 IWD

23

Songlin, Q., Xikang, G.: The evaluation of waveform

parameters of digitally recorded lightning impulses by
regression method ISH 1993, Paper no 51.04, Yokohama

24

Valve, P.,Aro, M.:Curve fitting in parameter estimation of

high voltage impulses Helsinki University of Technology, high
voltage institute, Espoo-Finland, report Tkk-SJT-17, november
1995, 26 p.

25

Kato, T.Double exponential function fitting for impulse

parameter evaluationCIGRE 33-93(WG03)30IWD

26

McComb, T.R., Lagnese, J.E.:Calculating the parameters of

full lightning impulses Using Model-based Curve fitting IEEE
Transactions on Power delivery, 6 1991-4), p.1386-1392

27

Prez Sanz, J.:Aplicacin del Filtro de Kalman a la

determinacin de los parmetros de impulsos tipo rayo en
sistematas de medida con registro digital, conforme a la norma
IEC 60-1. Tesiis Doctoral, Madrid, 1994

155

References

156

28

Garnacho, F., Perez, J.:Evaluation of lightning impulse

parameters. Proposal to establish the criteria to determine the
reference mean curve in lightning impulses ISH 95, paper
4547, Graz

29

Brede, P.A., Werle, P., Gockenbach, E., Borsi, H.:A new

method of determining the mean curve of lightning impulses
according to IEC 60060-1 11th ISH, London, 1999, page
1.74.S21-1.77.S21, Volume 1

30

Kato, T.:Mean-curve fitting using tension splines for impulse

parameter evaluation 11th ISH, London 199, page 1.78.S211.81.S21, Volume 1

31

Angrisani, L., Daponte, P.:A wavelet based approach for

measurements on high voltage impulses, 11th ISH, London
1999, page 1.205.S21-1.208.S21, Volume 1

32

Schon, K.:Digital filtering of HV lightning impulses IEEE

Panel Session on digital techniques in HV tests, Long Beach,
IEEE summer meeting California 1989

33

Dias, C.P., Moura, A.M., Valve, A.A.:Digital filetring of HV

impulse test results 7th ISH, Dresden 1991, paper 62.08

34

Lee, H., Rungis, J.:A non subjective spectrum filtering method

of impulse waves. CIGRE 33-93 (WG03) 7 IWD

35

Wong, K.C.P., Ryan, H.M., Blackett, J., Watts, M.W.:Digital

measurement of lightning impulse parameters using curve
fitting algorithms 11th ISH London 1999, page 1.193.S211.196.S21, Volume 1

36

Garnacho, F. et al.:Mean curve definition and evaluation

procedure of standard impulse parametersCIGRE 3395(WG03) 25 IWD

37

Hughes, R.C. and Hall, A.C.: Characerisation of hv impulses

with non-conventional waveforms, report to Study Committee
12 CIGRE 33-97 (WG03) 19 IWD

38

Kind D:An introduction to high-voltage experimental

technique Friedr. Vieweg & Sohn Braunschweig, ISBN 3-52808383-2

39

Kawashima, T., Okabe, S. and Zaima, E. Evaluation of steepfront overvoltage waveforms for oil-filled transformers 11th
ISH 1999, London, Volume 1, p. 1.279.P4-1.282.P4

40

Kvarngren, M.:Screening of sharp points in compact

transmission towers TRITA-EEA-9706 ISSN 1100-1593

41

Kind, D., Krner, H.:High-voltage insulation technology

Friedr. Vieweg & Sohn Braunschweig, ISBN 3-528-08599-1

42.

Lemke, E.:Ein Beitrag zur Abschtzung der rumlichzeitlichen Entwicklung des Durchschlags in langen
Luftfunkstrecken bei Schaltspannungen Wiss. Zeitschrift der
TU Dresden 26 (1977), 1, p.133-136.

References

43

Torshin, Y.V.:Experimental investigation of common

regularities of breakdown in transformer oil 7th ISH vol.2,
paper no 2102, p.17-20 Dresden, Germany August 26-30, 1991.

44

Breitfelder, D., Buckow, E., Knorr, W. and Peschke,

W.:Dielectric strength under transformer oill under impulse
and hhigh frequency voltage stress 7th ISH, Vol. 2, paper no.
2212, p1-4, Dresden Germany, August 26-30, 1991

45

Alpert, D, Lee, D.A., Lyman, E.M. Tomaschke,

H.W.:Initiation of electrical breadkwon in ultra-high vaccum
J.Vac.Sci.Technol.Vol1 (1964), p. 35-50

46

Maitland, A. Influence of the anode temperature in the

breakdown voltage and condition characteristics of a vacuum
gap Brit.J.A[[;.Phys.Vol13 91962), p.122-125.

47

Cranberg, L.:The initiation of electrical breakdown in

vacuum J.Appl.Phys.,Vol23 (1952), P.518-522

48

Kind, D.:Die Aufbauflche bei Stobeanspruchungen

technischer Elektrode Anordnungen in Luft Dissertation TH
Mnchen 1957

49

Report of LCOE, e.g. document 2/12 14/01/98

50

Wetzer, J. van der Laan, P.:Hoogspanningstechniek

1Eindhoven University of Technology, Eindhoven the
Netherlands, Dec 1994

51

Manjunath, B.S., Dwarakanath, K., Naidu, M.S.:Influence of

polarity and field non-uniforminty on the rbeakdwon voltage
characteristics of SF6 under impulse voltages1994 IEEE
International symposium on electrical insulation, p.485-488,
Pittsburg, PA (USA) June 5-8, 1994

52

Qiu, Y., Lu, C.Y. and Zhang, M.:Voltage-time characteristics

of SF6 and SF6-N2 under lightning and steep fronted impulses
Conference record of the 1992 International symposium on
Electrical insulation, p.310-313 Baltimore, MD (USA), June 710, 1992, IEEE, New York, 1992

53

Heers, B., Stepken, A.:Investigation on discharge development

in SF6 for oscillating impulses and fast oscillating
overvoltages7th ISH, Vol. 3, paper no. 3106, p./29-32,
Dresden, Germany August 26-30, 1991

54

Garnacho, F., Simon, P. P2/12

55

Strigel, R.Elektrische Stofestigkeit Springer, Berlin 1955

56

Venkataseshaiah, C, Sarathi, R and Aleyas, M.V.:Studies on

the dielectric strength of transformer oil under oscillatory
impulse voltagesConference recod of the 1996 iEEE
international symposium on electrical insulation. Vol2, p. 554556, Montreal Quebec, Canada June 16-19, 1996

57

Asmuth, P.:Das Durchsclagverhalten

Polyethylen
nach
unterschiedlicher
Dissertation, 1982, Universitt Hannover

von vernetztem
Vorbelastung

157

References

158

58

Gockenbach, E., Schiller, G.: Influence of the material

conditions on the behaviour of XLPE-samples at voltages of
different shapes Proceedings of the 1996 IEEE International
symposium on Electrical Insulation, part2, Montreal, 1996

59

Kahle,
M.:
Der
Einflu
von
Gleichspannungsvorbeanspruchungen
und
Erholezeit
auf
die
Stodurchschlagfestigkeit hochpolymer Folien Elektrie, 1971,
heft 25, S. 315-317

60

Nakagawa, S. et. al.:Breakdown characteristics of XLPE cable

at high temperature 8th ISH Yolohama 1993

61

Peschke, E.F., Kalkner, W., Schrder, G.: Impulse strength of

PE and XLPE insulations 3th ISH, Milan 1979

62

Selle, F.: Durchschlagverhalten von Kunstoffisolierungen bei

berlagerungen von Gleich- und Stospannung Dissertation
Universitt Hannover, 1987

63

Rittner, K., Brner, G.: treeing inception voltage of XLPE at

Lightning Impulse voltage ETEP, Vol.2, No. 4, July/August
1992

64

Zieschang, R.: Zum Kanaleinsatz und Durchschlag in

Polyethylen bei Belastung mit Blitzimpulsspannung
Dissertation TU Dresden, 1993

65

marx, E.: Versuche ber die Spannung von Isolatoren mit

Spannungssten Elektrotechnische Zeitschrift, 1924, Heft 25,
S 652-654

66

Toepler, M.: Stospannung, berschlag und Durchschlag bei

Isolatoren Elektrotechnische Zeitschrift, 1924, Heft 40, S
1045-1050

67

Watson, D.B.: The impulse strength of polyethylene as

function of voltage rise time Journal of Physics, No.4, L19L20, 1971

68

Siodla, K and Zubielik, P.:Electric strength of the vacuum

interrupter under ligthning and switching impulse voltage XVth
International symposium on discharges and electrical insulation
in vacuum. p. 632-635. Darmstadt, Germany, September 6-10,
1992

69

Niessen, F.H.M.:Literature study on vacuum insulated

devices TUE-Netherlands, 8 August 1983, ESA Technical
Management, the Netherlands 1983

70

Ryan, H.M.: High voltage engineering and testing IEE power

series 17, 1996, ISBN 0 86341 293 9

71

Benda-Berlijn, S., Kvarngren, M., Garnacho, F., Gonzales, A.,

Simon, P., Gockenbach, E., Werle, P., Hackemack, K., Watts,
M and Wong, P.:An investigation into the relevancy of the
wave shape parameters used for lightning impulse tests
CIGRE WG 33.03, 1998, CIGRE WG33.03 1998 IWD 15.

References

72

Hauschild, W and Mosch, W.:Statistical techniques for highvoltage engineeringPeter Peregrinus Ltd., ISBN 0 86341 205
X

73

Pesgens, F.:Towards an improved method for measuring

lightning impulse voltages , XX9thesis) XX TUE-Netherlands

74

Gruntjes, R. :Simulation of the generated impulses by a simple

generator project document P1/11

75

IEC 60052, sphere gaps

76

Bongaerts, M.:The influence of parameters of a lightning

impulse on the breakdown behaviour of insulating materials
14-8-1998, EH.98.A.152, thesis, Technical University
Eindhoven

77

F. Garnacho en P. Simon:Software validation Document

P2/16, 12/3/98

78

Hauschild.W., Mosch.W.:Statistical techniques for highvoltage engineering ISBN 0 864341 205 X

79

Simon, P.: Investigacin de los parmetros caractersticos de

comportamiento de medios dielctricos frente a sobretensiones
tipo rayo no normalizados en alta tensin". Dissertation to be
presented at a later time. Professor: F. Garnacho.

80

Bosch, E., Tuin, J.:Self-adjusting digitaal filter voor

stootspanningsproeven aan transformatoren SMIT Nijmegen,
A.P. E.92.2.26

159