Professional Documents
Culture Documents
net/publication/257621757
CITATIONS READS
8 1,090
1 author:
SEE PROFILE
All content following this page was uploaded by Dennis van der Born on 05 June 2014.
Polymeric insulation materials have not been used in HVDC cable systems until recently. The
main cause of this being the tendency of polymers to deplete accumulated charge very slowly.
Research into the dynamics of space charge trapping, injection and conduction in polymers
can reveal important information from which new materials can be designed. Solutions can
be found in polymers other than the well-known XLPE or introducing additives into XLPE
based materials.
This thesis investigates the space charge accumulation dynamics of mini cable models consist-
ing of several different XLPE based insulation materials and PE based semiconductive layers.
The main difference between the materials being the type and concentration of additives.
Extra measurements on thin plaque samples of both the semiconductive layers and the insu-
lation materials, such as conduction current measurements and frequency domain dielectric
spectroscopy, are used to obtain more information on the space charge accumulation dynamics
of the mini cable samples.
i
Contents
Abstract i
1 Introduction 1
1.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Space Charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Aim of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Experimental methods 5
2.1 Test specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Mini cables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Thin plaques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Pulsed Electroacoustic method . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 General PEA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.2 Cable geometry PEA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Test setup, equipment and procedure . . . . . . . . . . . . . . . . . . . 17
2.3 Conduction current measurements . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.1 Thin plaque conduction current setup . . . . . . . . . . . . . . . . . . 22
2.3.2 Mini cable conduction current setup . . . . . . . . . . . . . . . . . . . 24
2.4 Frequency domain dielectric spectroscopy . . . . . . . . . . . . . . . . . . . . 25
2.4.1 Measurement equipment and procedure . . . . . . . . . . . . . . . . . 26
3 Theoretical Background 29
3.1 Space Charge accumulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.1 Macroscopic view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.2 Microscopic view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Space Charge accumulation threshold . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Apparent trap controlled mobility and trap depth . . . . . . . . . . . . . . . . 36
3.3.1 Apparent mobility estimation . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.2 Trap depth estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Charge packets at high electric fields . . . . . . . . . . . . . . . . . . . . . . . 40
3.5 Conduction current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.5.1 Thin plaque conductivity . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.5.2 J-E Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.6 Dielectric polarization and relaxation . . . . . . . . . . . . . . . . . . . . . . . 45
ii
Contents iii
Acknowledgements 90
Chapter 1
Introduction
1.1 General
High Voltage DC (HVDC) cable systems have been in use in the power grid in Europe since
the nineteen fifties. One of the first commercial DC cable system to be installed in Europe
was the HVDC link between Gotland and the mainland of Sweden which was built in 1954
and had a maximum power capacity of 10-20MW.
In the past, DC systems were only applied when AC transmission was not possible. DC cable
systems need two DC-AC converter stations at both cable ends. These converter stations
were very expensive and could increase the costs of the entire project too much. In the past
few decades the converter technology matured and the costs of building converter stations
dropped accordingly.
The use of HVDC poses some important technical advantages over HVAC. Undersea cables
are used to connect power grids over long distances. In the Netherlands two cable systems
have been installed which connect the grid Dutch grid either to Norway(NorNed, 600km) or
Great-Britain(BritNed, 250km). When building such a system undersea cables have to be
used because the use of overhead wires is not possible. The use of AC power cables is also not
an option because the capacitive losses are too high at AC for distances longer than 30-50km.
Furthermore, dielectric losses also introduce an extra loss component. When HVDC is used
these capacitive losses are not present leaving only the resistive losses. Although a DC leakage
current flows between the inner conductor and outer sheath the losses related to this current
are very small.
With the use of HVDC remote power plants can be connected to the power grid via long
distance DC cables. Furthermore, renewable energy sources, which do not deliver a constant
power output, can affect the power quality of the grid. Connecting these source via a HVDC
link to the grid can remove this problem.
HVDC links can connect power grids which are not synchronised. Furthermore, power grids
which have different voltages and/or frequencies can be connected with a DC connection.
The stability of the grid can also be increased because the HVDC system is not susceptible
to load changes. In the event of severe load changes, which would cause parts of the AC to
become unsynchronised and seperated, an HVDC link is not affected and will remain able to
1
Chapter 1. Introduction 2
Nowadays the construction of new overhead lines is not accepted by the public because of
the ongoing discussion on the hazard of magnetic fields to the human health and the visual
pollution caused by overheadlines. Therefore underground cables are installed which are
much more expensive than overhead lines. Using DC could partly cut the costs because fewer
conductors are needed(no three phase system) which can also have a smaller diameter than
for AC due to the absence of the skin effect.
The use of polymeric cables shows more advantages over paper-oil and mass impregnated
cables: no oil is used, making repairs and fitting new joints is much easier, higher operating
temperatures are possible and polymeric cables show a larger mechanical rigidity resulting in
the use of a thinner cable armor(resulting in lower costs and lower weight).
Because of these advantages polymeric cables have been used in HVAC system for quite some
time. However, in HVDC systems polymeric cables have not been used until recently because
paper-oil cables showed much higher service reliability. The reason for this is explained in
this section.
∇ · εEρ = ρ (1.1)
Where ε is the permittivity of the corresponding material and ρ is the space charge density
inside the material. If it is assumed that the space charge is not influenced by the applied
electric field E0 the total electric field in the material can be represented by equation 1.2.
The total electric field is thus determined by both the applied electric field and the electric
field induced by the space charges present in the material.
Etot = E0 + Eρ (1.2)
In AC transmission the direction of the electric field inverts periodically causing the flow of
charge also to invert its direction. This inversion happens too quickly for charges to build
up in the materials normally used in AC cables. The efect of space charge can thus be
neglected for AC systems. In DC systems however, the direction of the electric field always
remains the same causing the direction of the charge flow also to remain the same. Due to
these circumstances space charge can build up under DC field. The combination of both the
Chapter 1. Introduction 3
applied field and the space charge induced field can locally cause large enhancements in the
total electric field.
Charges present in the insulation can be depleted by removing the voltage from the conductor
and short-circuiting the conductor to earth. This depletion is not instantaneous, it takes time
for the charges to deplete from the material. The time period for charge depletion can be
as long as several weeks. The main difference between polymeric insulation and paper-oil
insulation can be found in the charge depletion time. Polymeric cables tend to hold the
accumulated charge much longer than paper-oil cables.
The disadvantage of this becomes clear when looking at the situation in which the voltage
polarity on the conductor is inverted, which is a common practice in DC systems. When the
total amount of accumulated charge is still present in the cable the total field in the insulation
will be the sum of the space charge field and the applied field, which has changed direction.
The maximum electric field is in this case usually present at the inner conductor and can be
as high as twice the maximum applied field.
It has become clear that polymeric cables still perform worse than paper-oil cables in the
case of space charge properties, which is the main reason that polymeric cables are still not
in widespread use for HVDC systems. However, Siemens and ABB are using voltage source
converters(VSC) for DC to AC conversion, which makes full control of the output voltages
and currents possible. In this case voltage inversions are not necessary which makes the use of
current polymeric cable technology possible. These systems are called HVDC PLUS(Siemens)
and HVDC Light(ABB).
Ongoing research in the field of space charge dynamics continues to improve the space charge
characteristics of polymeric materials. At this moment polymeric HVDC cables for rated
voltages up to 150 kV are already in use.
The test techniques used in this work include: space charge measurements with the PEA
method, conduction current measurements and frequency domain dielectric spectroscopy.
of the space charge accumulation threshold, the theory of the estimation of trap-controlled
mobility and trap depths, the theory on charge packet dynamics, the significance of performing
conduction current measurements and the basic theory of dielectric relaxation.
Chapter 4 contains the results of the measurements including the acquired space charge
thresholds, the conductivity of the semiconductive layers of the cables, the determined ap-
parent trap-controlled mobility and trap depths and the measured complex permittivity and
loss factor tan(δ) of the insulation materials. The discussion on the measurement results is
also contained in this chapter. Chapter 5 contains some conclusions and observations on the
experimental results. Some recommendations on the future research including space charge
measurements at negative polarity, investigation into charge packet dynamics and time do-
main dielectric spectroscopy are also contained in this chapter.
Chapter 2
Experimental methods
This chapter gives a description of the used test specimens, test techniques and measurement
set-ups. Section 2.1 contains a description of the test specimens. This description includes
the dimensions, used materials and the pre-treatment of the specimens.
Section 2.2 describes the Pulsed Electro-Acoustic method of space charge measurements.
Section 2.3 contains a description of the conduction current measurements used to determine
electric field-conduction current characteristics. Section 2.4 gives a description of the dielectric
spectroscopy set-up used to determine the complex permittivity of a material.
The cables all consist of a solid copper inner conductor, an inner and outer semiconductive
layer and an XLPE based insulation layer. The mini cables all lack the presence of an outer
shield. The outer semicon layer is thus also the outer layer of the cable. The dimensions of
the cables are presented in figure 2.1.
Four cables with different types of semiconductive layers and insulation materials were used
in this work. To create four different cables two types of insulation material and two different
semiconductive materials were used. The two types of insulation material are designated
insulation I and insulation II. Insulation I and II are both XLPE based but have different
5
Chapter 2. Experimental methods 6
Figure 2.1: Cable dimensions. The conductive screen is not included in the cables.
additives thus creating two different XLPE compositions. Insulation I consists of crosslinked
polyethylene with less than 0.5% charge trapping agent added. Insulation II also consists of
crosslinked polyethylene but has less than 1% carbon black added.
The two different types of semiconductive material are designated as semicon C and semicon
D. Semicon C is an ethylene copolymer with polar comonomer which has a “low” concentration
of added carbon black. Semicon D is also an ethylene copolymer with polar comonomer but
has a “high” concentration of carbon black. Note that low and high are between quotation
marks because the terms high and low are in this case relative to each other. In table 2.1 the
four cables are listed with their designations and insulation and semicon type.
The mini cables are produced by The Dow Chemical Company. The cables are extruded and
crosslinked. After the crosslinking procedure the cables were preconditioned at 60 ◦ C in a
vacuum oven for 5 days to expel crosslinking by-products which have an adverse effect on
space charge behavior.
The mini cables were already used in a project prior to this thesis for space charge measure-
ments. Because of this, some space charge is still present in the bulk and insulator-semicon
interface of the cables. Therefore, before using a mini cable in this work the cable was first
preconditioned in an oven at 80 ◦ C for 3 days to deplete the remaining charge.
Each semicon sample is a square of 20x20 cm and has a thickness of approximately 0.6 mm.
Note that the thickness is approximate because the thickness of the plaques is not uniform
across the area of the plaques. The variation of thickness is in the order of 0.05 mm.
Because the material used to produce the plaques is from the same lot used to produce
the mini cables the properties(variation in conductivity, dispersion etc.) of the plaques are
assumed to be the same as the properties of the semiconductive layers in the mini cables.
There are two types of semiconductive materials used in the mini cables thus there are also
two types of semiconductive plaques. The designation is the same as used for the minicables
so the plaques are also designated semicon C and semicon D.
Each insulation sample has a circular shape with a diameter of 40 mm and has a thickness of
0.4 mm. Two types of insulation material are used for he mini cables, thus there are also two
types of insulation plaques. The insulation plaques are also designated insulation I and II.
In the PEA setup a material specimen is placed between two electrodes. The specimen
is assumed to be free of space charge. A DC electric field is applied on the specimen by
applying a high DC voltage(U0 ) between the HV and earth electrode. The DC field causes
the formation of surface charges on the electrodes and space charge in the material.
During a measurement a short high voltage pulse (Up (t)) is applied between the electrodes via
decoupling capacitor C. The applied pulse causes an electrostatic force F of short duration
on the charges in the material and on the electrodes as shown in equation 2.1. Because of
this force the charges move slightly which causes acoustic waves with pressure p to form at
the location of the charges(2.2). The acoustic waves will then travel through the material and
earth electrode to a piezoelectric sensor. The sensor placed at the end of the earth electrode
detects the acoustic waves resulting in a small voltage signal as shown in equation 2.4.
p = ρbEp (2.2)
u = kp = kρbEp (2.4)
u = Kρ
The amplitude of the signal is proportional to the density of the space charge. Furthermore,
the sign of the signal is dependent on the polarity of the charges. Positive charges will be
converted to a positive voltage and negative charges will cause a negative voltage at the sensor
output. Figure 2.3 shows such a voltage signal with respect to the charges shown in figure
2.2.
Note that the earth electrode is thicker than the HV electrode. Because the earth electrode
is thick the electrode acts as a delay block for the acoustic waves. Because of the delay the
acoustic waves arrive later at the sensor than the disturbance created by the HV pulse. The
reason for this is to prevent electromagnetic interference from the applied HV pulse occurring
at the sensor location when the acoustic waves arrive.
Reflections of acoustic waves at the sensor can cause interference which distorts the signal.
Therefore a material with the same acoustic impedance as the sensor in combination with an
absorber is placed behind the sensor to absorb the acoustic waves passing the sensor. Note
that the acoustic waves do not only travel in the direction of the earth electrode. The waves
will also travel to the HV electrode where the waves are reflected. However these reflected
waves will arrive later than the waves traveling directly to the earth electrode and will not
distort the desired signal.
Chapter 2. Experimental methods 9
Figure 2.3: PEA voltage signal resulting from the situation shown in figure 2.2
Deconvolution
Because the voltage signal at the sensor is too small to measure with an oscilloscope directly
the signal is fed into an amplifier. The amplifier-sensor system however behaves as a high-
pass filter causing the resulting signal to be distorted[5]. The high-pass filter characteristic
usually causes overshoot in the signal which can be incorrectly interpreted as space charge as
shown on the left side of figure 2.5. To correct the voltage signal a deconvolution technique
is used based on the deconvolution technique of Jeroense[2]. An in depth description of the
deconvolution technique used in this work is contained in appendix A.
Figure 2.5: Left: Voltage signal with distortion from high-pass characteristic, Right: Corrected volt-
age signal
frequency components. Due to this effect the pressure pulse is also wider than the original
pulse(at the location of the space charge).
Figure 2.6: Attenuation and dispersion of acoustic waves originating from different space charge
locations[1]
Due to these effects the acoustic signal detected by the sensor is not directly related to the
space charge distribution. Thus the measured voltage signal at the scope also needs to be
corrected for the attenuation and dispersion in the material. The method used to correct the
signal is described in depth in Appendix A[6].
Chapter 2. Experimental methods 11
Calibration
After the corrections presented in the previous sections the signal is still a voltage signal
vsatt (t)[mV ]. The space charge distribution ρ(x)[C/m3 ] still has to be determined from the
voltage signal. The determination is done by a calibration procedure based on the known
charge distribution at the earth electrode which will be explained in this section.
The relation between the measured voltage signal at the output of the sensor and the charge
distribution in the sample is defined by a factor K as described in equation 2.4. The calibration
procedure will however be applied on the processed signal(after deconvolution and attenuation
and dispersion correction). The factor in equation 2.4 will then be designated Kcal which is
the calibration factor as shown in equation 2.5.
To calculate the calibration factor Kcal a known charge distribution is needed. The signal
without space charge can provide a known charge distribution which in this case corresponds
to the earth electrode surface charge. The applied voltage is known therefore the charge
distribution at the earth electrode can be calculated.
With the known applied voltage the electric field at the earth electrode can be calculated
according to equation 2.6.
V
Ee = (2.6)
d
With the applied electric field at the earth electrode(which is the same as anywhere in the
sample) the electrode surface charge density can be calculated with equation 2.7 in which
is the permittivity of the sample.
σe = Ee (2.7)
With the electrode surface charge density the calibration factor Kcal is calculated with equa-
tion 2.8.
R x2 att
vs (x) dx
Kcal = x1 (2.8)
σe
The integral in eqaution 2.8 represents the surface of the earth electrode in the voltage signal
vsatt . The points x1 and x2 are the start and the end of the earth electrode in vsatt respectively.
Figure 2.7 contains a schematic representation of the earth electrode surface in the voltage
signal. The surface representing the integral in equation 2.8 is highlighted with a light blue
color. The surface covered by the integral is related to the surface charge density at the earth
electrode because integrating the space charge density along the length of the sample would
result in the surface charge density.
With the now known calibration factor Kcal the space charge profile can be obtained according
to equation 2.9.
vsatt (x)
ρ(x) = (2.9)
Kcal
Chapter 2. Experimental methods 12
Figure 2.7: Voltage signal with earth electrode surface highlighted in light blue
The electric field profile E(x) can be calculated from the space charge distribution with
equation 2.10.
1 d
Z
E(x) = ρ(x) dx (2.10)
0
From the electric field profile the voltage distribution across the sample V (x) can be obtained
with equation 2.11
Z d
V (x) = − E(x) dx (2.11)
0
When the calculated voltage distribution is correct the calibration procedure is also correct.
Note that this check also applies for the attenuation and dispersion correction. Furthermore
the voltage signal used for the calibration procedure should be free of space charge. The
voltage distribution in a sample with space charge is different and can not be calculated
beforehand.
As explained in section 2.1.1 each mini cable consists of a solid copper inner conductor, an
inner and outer semiconductive layer and an XLPE based insulation layer. Each cable is as-
sumed to be free of space charge. The DC electric field is applied between the inner conductor
and outer semiconductive layer by applying a high DC voltage on the inner conductor. The
electric field causes space charge to form inside the insulation layer.
The PEA setup for cables also has an earth electrode which is in contact with the outer
semicon layer of the mini cable. Two types of earth electrode can be used for cables. The
Chapter 2. Experimental methods 13
first type is a curved electrode which fits the outer semicon layer of the cable. The piezoelectric
sensor and absorber are in this case also curved. The second type is a flat earth electrode
with a flat sensor and absorber. In this thesis the flat configuration is used which is shown
schematically in figure 2.8. The cable is pressed down onto the earth electrode with the spring
Figure 2.8: Schematic representation of PEA flat earth electrode and cable cross section[1]
system and holder as shown in figure 2.8 to ensure a good acoustic contact between the earth
electrode and outer semicon. Furthermore a thin film of silicon oil is applied on the earth
electrode before pressing down the cable to further ensure the quality of the acoustic contact.
The acoustic impedance of the semicon layers is very similar to the acoustic impedance of the
insulation layer therefore no reflections of acoustic waves at the semicon-insulation interface
will occur.
The flat earth electrode is used in this setup because of the fact that a curved electrode has
two main disadvantages. It is difficult to achieve a good fit of the outer semicon layer in
the curvature of the earth electrode. A good acoustic contact is therefore difficult to obtain.
Furthermore only one cable size can be used with such an electrode, the setup could therefore
not be used on other cables without changing the entire electrode and sensor combination.
In the flat electrode configuration these disadvantages are not present.
However the flat electrode configuration also has a disadvantage. Because of the narrow
contact between electrode and semicon the piezoelectric sensor is also narrow with respect to
the curved configuration. Because of this the sensor has a low capacitance which can cause
distortion when an amplifier with low imput impedance is used(in the order of 50 Ω)[1, 5].
Of course the use of an amplifier with a high imput impedance(in this case 1.5 kΩ) can solve
this problem. The noise level of such amplifiers is however much higher[5].
Pulse voltage
The pulse voltage applied on the specimen in the thin sample PEA is of course also present
in the cable PEA method. There are however multiple possibilities for applying the pulse
voltage across the cable insulation[1]. Firstly, the pulse voltage can be applied on the inner
conductor along with the DC voltage. This method is principally the same as with the thin
plaque PEA. As with the thin plaque PEA a decoupling capacitor in series with the pulse
generator is in this case needed to apply both the pulse and DC voltage simultaneously. There
Chapter 2. Experimental methods 14
are however some constraints in using this method. The decoupling capacitor should be able
to withstand the DC voltage and have a much larger capacitance than the cable resulting
in a large component. Furthermore if the cable is longer than the wavelength of the voltage
pulse the pulse could be distorted while propagating through the cable and be reflected at
the terminations.
Secondly, the pulse voltage can be applied between the PEA cell(earth electrode) and earth.
The constraints of applying the pulse voltage on the inner conductor are not present here.
No decoupling capacitor is needed because the cable itself acts as a decoupling capacitor.
Furthermore there will be no reflections and distortion of the pulse at the measurement
location. The main disadvantage of this configuration is the fact that the PEA cell is at the
same potential as the pulse voltage. The measurement equipment can thus not be directly
connected to the PEA cell.
Thirdly, the pulse voltage can be applied on the outer semicon of the cable with the use of
a pulse electrode wrapped around the cable with copper tape. This type of configuration is
used in this thesis. Figure 2.9 shows a schematic representation of the PEA setup with a pulse
electrode wrapped around the outer semicon layer. In this configuration the same advantages
Figure 2.9: Schematic representation of PEA cable setup with wrapped copper electrode
apply as with the configuration where the pulse is applied between the PEA cell and earth.
The pulse voltage is in this configuration applied between earth and the pulse electrode. The
PEA cell is thus at earth potential. The measurement equipment can therefore be directly
connected to the PEA cell. Note that this configuration can also be used on cables with an
outer conductive screen. The screen is in that case interrupted at the location of the PEA
cell and the pulse is applied on both sides of the interrupted screen[1].
Chapter 2. Experimental methods 15
However, due to the cylindrical shape of cables the electric field inside is divergent. The
pulsed field will therefore also be divergent causing the resulting voltage signal to be distorted.
Furtermore the acoustic waves travelling through the insulation are also divergent due to the
cylindrical shape giving rise to another error. The output signal has to be corrected for both
forms of divergence to obtain the correct space charge distribution.
The calibration of the signal is basically the same as the calibration for flat samples. The
electric field calculations are however different because of the cylindrical shape. The processing
steps for cable PEA are shown in figure 2.10.
Divergence correction
The output signal has to be corrected for the divergence of both the pulse field and the
acoustic waves. The correction procedure is described in this section[1, 8].
The electric field distribution Ep (r) inside the cable due to the pulse voltage Up is described
in equation 2.12.
Up
Ep (r) = (2.12)
rout
r ln rin
As can be seen from equation 2.12 the electric field inside the cable insulation is inhomo-
geneous as opposed to the electric field in a thin plaque sample(parallel plate construction).
The amplitude of the pressure wave originating from the space charge location is a function of
the pulsed electric field(equation 2.2). Because the pulse field is a function of radial position
r in the insulation the resulting signal will also be dependent on the position of the space
charge. To correct the output signal for the divergence in the pulse field a correction factor
Kg,pulse is defined according to equation 2.13.
p(rout ) e(rout ) r
Kg,pulse = = = (2.13)
p(r) e(r) rout
Chapter 2. Experimental methods 16
The term p(rout ) corresponds to the pressure wave resulting from a fixed charge at the outer
radius of the cable insulation rout . The term p(r) corresponds the pressure wave resulting
from the same amount of charge at an arbitrary radius r in the insulation.
The cable is assumed to be very long with respect to the insulation thickness and the insulation
material is assumed to be homogeneous in the axial direction. The space charge distribution
only changes in the radial direction inside a cylindrical object. The acoustic waves are there-
fore only dependent on the position in the radial direction resulting in a one-dimensional
representation in cylindrical coordinates as shown in equation 2.14.
The term A is a constant determined by the boundary conditions of the partial differential
equation and ω is the angular velocity of the acoustic wave. According to equation 2.15
the amplitude of an acoustic wave in radial direction decreases with increasing radius. The
acoustic wave detected by the sensor is thus smaller than the acoustic wave at the space
charge location.
To correct for the divergence in the acoustic waves a correction factor Kg,wave is defined
according to equation 2.16.
r
p(r, t) p(r, t) rout
Kg,wave = = q = (2.16)
p(rout , t + ∆t) r
p(r, t) rout r
The function p(r, t) represents the acoustic wave originating from a charge at radius r. The
function p(rout , t + ∆t) represents the same acoustic wave at the outer radius ro ut after
travelling through the insulation.
Both electric field and acoustic wave divergences can be contained in one correction factor
called the geometrical correction factor Kg which is defined in equation 2.17.
r
r
Kg = Kg,pulse Kg,wave = (2.17)
rout
The radius r corresponds to the location of the space charge. The voltage signal corrected
for the electric field and acoustic divergence can be calculated from the deconvoluted signal
with equation 2.18.
Note that the attenuation and dispersion correction will in this case be applied on vsdiv (t)
in stead of vsdecon (t). The divergence correction procedure is implemented in MATLAB. The
corresponding MATLAB code is contained in appendix B.
Chapter 2. Experimental methods 17
Calibration
The calibration procedure for cable geometry PEA is basically the same as for the thin
plaque PEA described in section 2.2.1. The calculation of the surface charge density at the
earth electrode is however different. The surface charge density at the earth electrode σe is
calculated with equation 2.7. As can be seen from equation 2.7 the electric field at the earth
electrode is needed. The calculation of the electric field in cylindrical objects is of course
different from the calculation of the electric field in parallel plate structures. Equation 2.19
represents the electric field at the outer radius of a cable where the earth electrode is located.
The applied voltage corresponds to V .
V
Ee (rout ) = (2.19)
rout ln rrout
in
The calculation of the electric field distribution from the obtained radial space charge dis-
tribution ρ(r)is also different. The electric field in the cable is calculated with Equation
2.20.
1 rout
Z
E(r) = rρ(r) dr (2.20)
r rin
Furthermore, the voltage distribution can be calculated with equation 2.21 which is basically
the same as for parallel plate structures.
Z rout
V (r) = − E(r) dr (2.21)
rin
The same checks apply here as for the thin plaque PEA calibration procedure mentioned in
section 2.2.1. The calibration procedure for cable PEA is implemented in MATLAB. The
corresponding MATLAB code is contained in appendix B.
The high DC voltage Udc is applied to the conductor via a series resistor Rdc to limit the max-
imum current drawn from the DC supply and to decouple the power supply. The resistance
of Rdc is 30 MΩ.
As explained in section 2.2.2 the voltage pulse is applied on the outer semiconductive layer
via an electrode wrapped with copper tape around the cable(figure 2.9). The pulse is created
with a small HVDC power supply connected to a switching device which switches the high DC
voltage. The output of the switching device is then connected to the wrapped pulse electrode.
The amplitude of the pulse is between 3.5 and 4 kV. The pulse has a rise time of 10 ns and
the pulse width ∆t is 80 ns resulting in a spatial resulation of 165 µm.
The cable which connects the pulse generator (HVDC supply and switch box) to the pulse
electrode has a characteristic impedance of 50 Ω. The cable is therefore terminated with a
50 Ω resistor Rpulse between the conductor and earth. Furthermore, stray inductances Ls are
Chapter 2. Experimental methods 18
present at the test cable (every wire has a stray inductance). The short rise time and narrow
pulse width in combination with the stray inductances may cause the generation of fly-back
voltages[1](like ignition coils used with petrol engines). These fly-back voltages will distort
the shape of the pulse(oscillations).
The outer semiconductive layer also has a finite resistance between the pulse electrode and
the earth electrode which can cause a mismatch in the termination of the pulse cable(creating
more distortion). These two effects can be dampened by using a series resistor Rd between
the termination of the pulse cable and the pulse electrode. An equivalent circuit drawing of
the connection of the pulse cable to the pulse electrode and test cable is shown in figure 2.11.
The amplifier used in the PEA setup to amplify the output signal of the sensor has an
input impedance of 1.5 kΩ because of the low sensor capacitance mentioned in section 2.2.2.
The gain of the amplifier is 70 dB and the bandwidth ranges from 0.1 to 100 MHz. The
piezoelectric sensor used in this setup consists of a PVDF foil of 0.025 mm thickness. PVDF
is an abbreviation of polyvinylidene fluoride which is a piezoelectric polymer.
Heating
The space charge measurements in this work are conducted at different temperatures. In this
case temperatures of 20 ◦ C, 40 ◦ C and 60 ◦ C are desired. The cable thus has to be heated with
a heating system. The cable could be heated inside an oven, creating isothermal conditions
within the cable insulation.
However in this work normal operating conditions are simulated. In normal operation a
cable is heated by resistive losses in the conductor caused by the current flowing through the
conductor. Because the cable is heated from the inside a temperature gradient will arise in
the insulation. The temperature at the outer semicon layer is thus lower than the conductor
temperature.
To simulate these temperature conditions the cable is heated by an inductive heating system.
A large current transformer is used in which the secondary consists of the test cable connected
in a loop(figure 2.12). Both ends of the cable are therefore connected to each other and
the HVDC supply. A current can then flow through the loop. The primary of the current
transformer is connected to a variable autotransformer to regulate the current flowing through
Chapter 2. Experimental methods 19
the primary. The current through the test cable can thus also be regulated which results in
temperature control of the conductor.
Figure 2.12: Schematic representation of PEA cable setup including heating system
The actual temperature of the cable has to be measured at different locations because of the
temperature gradient. Temperature sensors should therefore be placed on the inner conductor
and on the outside of the cable. However the conductor of the test cable is at high voltage
during a test. Temperature sensors can thus not be placed at the inner conductor.
To overcome this problem a dummy cable is used which is layed in a loop like the test cable
(creating a second secondary winding). The dummy cable is not connected to high voltage,
thus temperature sensors can be placed at the inner conductor of the dummy cable. Because
the temperatures are measured at the inner conductor and the outside of the dummy cable the
current flowing through both the dummy and test cable should be equal. The temperature
at the outside of the test cable is also measured to compare both the dummy and test cable
temperatures.
The currents flowing through both the dummy and test loops are also measured to make
sure that both currents are equal. To measure the currents two rogowski coils are used in
combination with two ammeters. A rogowski coil is shown in figure 2.12. Note that figure
2.12 represents only the test cable. The current flowing through a specific loop can also be
adjusted by changing the size of the loop. In this way the two currents can be equalized. The
devices used in the test setup are listed in table 2.2.
Measurement procedure
Space charge measurements can be performed in two ways. A voltage-on and a voltage-off
measurement. In a voltage-on measurement the high DC voltage is applied on the cable
conductor during the entire test. This method is taken as an example in section 2.2.
Chapter 2. Experimental methods 20
As explained in section 2.2.1 the DC electric field, which is continuously applied, causes
accumulation of space charge in the insulation. The process of charge accumulation can
therefore be recorded with a voltage-on measurement. The charge accumulation process is
also called polarization. In this work voltage-on tests are used to record the accumulation
of space charge at different DC field strengths and temperatures. Section 3.1 describes the
theoretical background on space charge accumulation.
Each voltage-on test in this thesis has a duration of 2 hours. Thus, space charge accumu-
lates during 2 hours. The space charge profile of a voltage-on measurement obtained after
processing of the voltage signal is displayed on the right side of figure 2.13.
Figure 2.13: Left: Space charge profile of voltage-off test, Right: Space charge profile of voltage-on
test
Chapter 2. Experimental methods 21
When the DC voltage on the conductor is removed and the conductor is connected to earth
the space charge present in the insulation will start to deplete. This process is called the
depolarization of the test object. When the space charge signal is recorded without the high
DC voltage on the conductor a voltage-off measurement is performed. The electrode surface
charges are in this case not present in the signal. Only the space charge and the induced
mirror charges at the electrodes are present as shown on the left side of figure 2.13.
The voltage-off tests are used in this work to record the depolarization process of the cables.
The duration of the voltage-off tests is 3 hours. The space charge thus depletes during 3
hours. Note that before a depolarization characteristic can be measured the cable first has to
be polarized which also takes 3 hours. After polarization the conductor is connected to earth.
The traces shown in figure 2.13 correspond to the space charge profile at a given point in
time. A full poling or depoling test consists of hundreds of these traces depending on the size
of the time step between two traces. Furthermore in this research only the accumulation or
depletion of the average space charge in the insulation is desired. From each trace the average
space charge density is calculated with equation 2.22.
Z rout
1
ρavg = |ρ(r)|dr (2.22)
rout − rin rin
The calculation of the average space charge density in the cables from the space charge profiles
is implemented in MATLAB. The corresponding matlab code is contained in appendix C. An
example of a polarization characteristic with average space charge density values is given in
figure 2.14
The voltage pulse firing at the pulse electrode is a source of electromagnetic disturbance in
the output signal. This disturbance is however constant in time. By saving the space charge
profile before applying the high DC voltage and subtracting this signal from the results this
disturbance can be removed. The total uncertainty of the measurements is thus about 15%.
The small polarization and conduction currents are measured with an electrometer which is
actually a very sensitive ammeter(in the order of 1 fA). The electrometer is connected to a
Chapter 2. Experimental methods 23
computer with the GPIB(IEEE-488) interface to store the measured data. To protect the
electrometer from overcurrents and overvoltages caused by for example a breakdown in the
specimen a large resistance Rdc is connected in series with the HVDC supply. The value of
Rdc is in this case 660 MΩ.
When performing a conduction current test only the current through the specimen is to
be measured. However currents can also flow from the HV electrode to the measurement
electrode over the surface of the sample due to the surface conductivity of the sample. These
currents will distort the measurement. Guarding electrodes which are connected to earth are
used to remove these surface currents. The surface currents will then flow to earth in stead
of the measurement electrode.
The entire setup is placed inside an EMC shielded cage because of the sensitivity of the
measurements. Any electromagnetic disturbance could distort the measurement.
Measurement procedure
The conduction current measurements are performed on the semiconductive plaques described
in section 2.1.2. The purpose of conducting these measurements is to determine the conduc-
tivity σ of the semiconductive materials. The conductivity of a material can be calculated
according to equation 2.23.
Jss
σ= (2.23)
E0
The term Jss corresponds to the steady state conduction current density flowing through the
sample. The applied electric field is designated E0 . The steady state conduction current which
is reached after the polarization processes are finished is measured with the electrometer at
a known applied field strength. The current density can then be calculated by dividing the
measured current by the area of the measurement electrode Ael (equation 2.24).
Iss
Jss = (2.24)
Ael
Iss = lim i(t)
t→∞
Because the conductivity of the semiconductive materials is relatively high(in the order of
10−12 S/m) the steady state value of the current is reached in a few minutes. When an
insulator like XLPE(σ = 10−17 S/m) would be tested the time until the steady state is reached
would be approximately 24 hours at field strengths higher than 10 kV/mm. The equipment
used in this test setup is listed in table 2.3.
polarization current will then start to flow from the inner conductor to the measurement
electrode at the location of the measurement electrode. The polarization current and conse-
quently the conduction current are measured with an electrometer which is connected to the
measurement electrode. The measurement electrode is in this setup a copper tape wrapped
tightly around the outer semicon layer.
The electrometer is in this setup also connected to a computer with the GPIB interface(IEEE
488) to store the measurement data. The protection resistance Rdc is connected between the
HVDC supply and the cable conductor. The value of the resistance is 660 MΩ.
The guarding electrodes are also present in this setup to prevent surface currents from reaching
the measurement electrode and distorting the measurement. The guarding electrodes are
wrapped around the insulation layer with the same copper tape used for the measurement
electrode.
Measurement procedure
The cable type conduction current measurements are conducted on the mini cables described
in section 2.1.1. The main purpose of these measurements is to obtain the conduction current
density versus applied electric field(J-E) characteristic of each mini cable. The steady state
conduction current is reached, as explained in section 2.3.1, after the polarization processes
are finished.
The time in which the steady state is reached varies between 24 and 48 hours depending on
the applied field strength. Each conduction current measurement thus takes up to 48 hours.
When a J vs. E characteristic containing 8 data points is required a total of 312 hours or
roughly three workweeks would be needed to complete such a characteristic(3 data points at
low field strengths and 5 data points at high fields are assumed).
Because the J-E characteristics are required at different temperatures the cable has to be
heated up. Furthermore the conductivity of a cable at different temperatures is required in
other research. The use of an oven for heating purposes is therefore required to ensure an
isotropic temperature distribution(a temperature gradient implies a conductivity gradient).
Furthermore using the same heating method as used in the cable PEA setup is unfeasible
because the large cable loop and current transformer then need to be placed inside the EMC
shielded cage which are normally fairly small. The equipment used in this test setup is listed
in table 2.4.
The frequency characteristic of the permittivity of a material can be measured with a dielectric
spectrometer. The relative complex permittivity of a material can be calculated from the
complex impedance which is measured with a dielectric spectrometer. Figure 2.17 represents
the measuring circuit for complex impedance measurements.
The AC voltage U with frequency f is applied to the test specimen. The test specimen is
placed between the electrodes(parallel plate). The resistor R is included in the circuit to
convert the current flowing through the sample to a voltage. Both the amplitude and the
phase of the voltage across and the current through the sample are measured with the two
phase sensitive voltmeters. The complex impedance Z can be calculated from these measured
values with equation 2.26.
U
Z= (2.26)
I
From the sample capacitance C calculated from the complex impedance the relative complex
permittivity (ω) can then be calculated with equation 2.27. Edge effects at the ends of the
parallel plates are neglected. The standard parallel plate capacitance formula can then be
used.
d
(ω) = 0 − i00 = C (2.27)
0 A
The automatic temperature control system is depicted schematically in figure 2.18. Liquid
nitrogen is contained in the Dewar vessel. An evaporator is also present inside the vessel to
regulate the nitrogen gas flow to the sample cell in the cryostat depending on the desired
Chapter 2. Experimental methods 27
Device Specifications
Impedance Analyzer Frequency range from 3 · 10−6 Hz up to 20 · 106 Hz,
impedance range from 10−2 Ω up to 1014 Ω and sen-
sitivity tan δ = 10−5
Cryostat Temperature range from −160 ◦ C up to 400 ◦ C, the
sensitivity is limited to 0.01 K
temperature. The evaporator regulates the nitrogen flow by building up a certain amount of
pressure inside the Dewar. Both temperature and pressure inside the Dewar are measured
by channel 1 and 2 respectively. An additional gas heater heats up the gas flow to further
increase the temperature of the nitrogen gas again depending on the temperature setting.
Channels 3 and 4 measure the temperatures of the gas and sample respectively. The vacuum
pump and gauge are present to maintain a low vacuum (< 1 · 10−5 bar). The low vacuum
maintains the thermal isolation of the cryostat.
Measurement procedure
The dielectric spectroscopy tests were performed on two thin plaques consisting of the two
insulation materials used in the mini cables(insulations I and II). Each dielectric spectroscopy
test has a duration of approximately 5 hours. Frequencies between 10−2 Hz and 106 Hz were
used in these measurements. Furthermore the temperature ranged from −40 ◦ C to 120 ◦ C.
The applied voltage was set at 1 V.
Chapter 2. Experimental methods 28
Note that the maximum frequency which can be used in the measurement is limited at
approximately 3 · 106 Hz. This limit is caused by the fact that the impedance of the sample
may reach a value in the same order as the inductive impedance of the BNC cables connecting
the sample cell to the analyzer[10].
Chapter 3
Theoretical Background
In this chapter the background theory of the measurements is explained. Section 3.1 contains
a description of space charge accumulation in general. The threshold electric field present in
space charge accumulation is described in section 3.2. Section 3.3 describes the theory behind
the estimation of apparent trap-controlled mobility and trap depth values and distribution
with the use of depolarization characteristics.
Space charge behavior at high electric field strengths related to the generation and movement
of charge packets is described in section 3.4. Section 3.5 contains the background theory on
conduction current measurements. Both flat sample and cable type conduction current mea-
surements are included. The theory on frequency domain dielectric spectroscopy is described
in section 3.6.
A divergence in the current density J can be found at several locations and conditions: at an
electrode-dielectric interface, at a dielectric-dielectric interface, in the case of a temperature
gradient and in the case of an inhomogeneity.
29
Chapter 3. Theoretical Background 30
Electrode-dielectric interface
The buildup of space charge at an electrode-dielectric interface is defined by the difference
∆Jint between two charge flows. The flow through the interface Jinj (E, T ) which is defined by
the charge injection and extraction processes and the flow through the dielectric Jtrans (E, T )
which is defined by the conduction or transport processes in the bulk material. Note that both
flows are dependent on electric field and temperature. Equation 3.1 represents the difference
mentioned above.
Jinj (E, T ) − Jtrans (E, T ) = ∆Jint (E, T ) (3.1)
Three different situations can occur at electrode-dielectric interfaces which will be described
in the next part of this section.
Firstly, the injection current Jinj could be equal to the transport current Jtrans . In this case
the difference is zero hence there is no buildup of space charge. The injection current is in
this case just enough to replace the electrons transported away from the interface.
Secondly, the injection current could be larger than the transport current. Negative (homo)
charges will in this case accumulate at the cathode interface. However the electric field
strength will drop as the amount of negative is building up. Due to the drop in electric field
strength the injection current density will also decrease until a steady state is reached in
which again the injection and transport current densities are equal. Note that at the anode
the same process can take place. In this case the extraction current of electrons is larger than
the transport current supplying the elctrons to the interface resulting in a positive (homo)
charge.
Thirdly, the injection current could be smaller than the transport current. In this case more
electrons are transported away from the cathode interface than the injection current can
supply. A positive (hetero) charge will build up at the electrode interface causing a local
increase in electric field strength. Due to this increase the injection current will rise until a
steady state is reached again in which both currents are equal. Note that this process can
also exist at the anode where the extraction current of electrons is smaller than the transport
current supplying electrons to the interface resulting in a negative (hetero) charge.
Dielectric-dielectric interface
When crossing an interface between two different dielectrics both the permittivity and
the conductivity σ jump to another value. The difference in conductivity between the two
dielectrics results in a change in current density (J = σE. Because of the different current
densities charge accumulates at the interface(total current density is divergent). Furthermore
the difference in permittivity results in a difference in field strength E across the boundary
according to equation 3.2. The difference in field strength then results in again a difference
in current density and thus charge accumulates at the boundary.
1
En2 = En1 (3.2)
2
In general space charge accumulates at every boundary where the quotient σ changes. Note
that there is a situation in which both permittivity and conductivity change across a boundary
Chapter 3. Theoretical Background 31
but the quotient does not. In that case no charge should accumulate.(equation 3.3)
1 2
= (3.3)
σ1 σ2
σ1 2
1 =
σ2
J1 = σ1 En1
σ2 1
J2 = σ2 En2 = En1 = σ1 En1
2
J1 = J2
The analysis given above can also be represented by the Maxwell capacitor and corresponding
Maxwell-Wagner theory[13].
Temperature gradient
The transport current density in the bulk material is temperature dependent. A temperature
gradient in the material therefore causes a difference in transport current density leading to
space charge accumulation. Note that this effect can also be explained by the fact that the
conductivity is temperature dependent. The conductivity thus changes along the temperature
gradient and the quotient σ also changes resulting in space charge accumulation[13].
Inhomogeneity
Material inhomogeneity is found in most types of insulation materials. Many insulation
materials contain fillers which have a different conductivity and permittivity than the host
material resulting in a large amount of boundaries in the material. Space charge will accu-
mulate at those boundaries. Furthermore polymeric insulation materials like polyethylene
consist of parts which are nicely ordered also called crystalline parts and parts which are
disordered also called amorphous parts. The conductivity of the amorphous parts is higher
than the crystalline parts. Space charge will therefore accumulate at the boundaries between
the crystalline and amorphous parts in polymers.
Trapping
In the explanation of charge trapping it is assumed that the band model of Niels Bohr is
known to the reader. A charge trap is a potential well in which a charge carrier(electron, hole
or ion) can be captured.
Charge traps are caused by defects in a material. Every insulating material contains de-
fects because no insulating material is perfectly homogeneous. Most traps are formed in the
amorphous parts of insulating polymers.
Chapter 3. Theoretical Background 32
Incompletely bound atoms present in defects in the crystal lattice, which are mostly present
in amorphous parts, can cause the formation of dangling bonds[10]. A dangling bond is an
unsatisfied valence on an atom within a crystal lattice which means that there are not enough
electrons available in the covalent bonds to completely fill the valence(outermost) shell of the
atom. Also the existance of an excess electron in the valence shell of an atom in a crystal
lattice is considered a dangling bond.
Dangling bonds can be satisfied by the acceptance of an electron from the lattice or the
donation of an electron to the lattice. Dangling bonds therefore behave as allowed energy
states within the band gap(band model). These allowed energy states in the band gap are
called trapping sites which is the same as traps.
Traps for electrons are called acceptors and traps for holes are called donors. Electrons or
holes travelling from the valence band to the conduction band may get trapped in such an
energy state.
A charge carrier which is trapped in a potential well can leave the trap when a sufficient
amount of thermal energy is acquired from lattice vibrations. The amount of energy needed
to escape from a certain trap is called the trap depth. The amount of time a charge carrier
resides in a charge trap is determined by the trap depth: the deeper the trap the longer the
time a charge carrier will spend inside the trap(the chance of acquiring enough thermal energy
is smaller for deeper traps).
Another type of charge trap called a self-trap[13] is also present in polymers. The electric field
of a free electron can locally change the structure of the polymer molecular chain creating
a local drop in potential(potential well). The electron thus traps itself. The main difference
between this type of trap and donors or acceptors is the absence of a counter charge. Due
to this absence of a counter charge space charge is formed inside the material. Self-traps are
usually deep. Charge remains in such traps for many hours or even days.
Charge traps can be located in the vicinity of physical or chemical defects. Physical defects
are mostly present at the ends of molecular chains. These defects are responsible for shal-
low charge traps. Chemical defects are caused by additives like anti-oxidants(to counteract
thermal aging), residues like crosslinking by-products and other impurities created during
production. Chemical defects are usually responsible for deep traps. Additives can thus have
a large effect on the space charge behavior of a material.
Injection
In polyethylene materials the injection and conduction of charges is mainly performed by
electrons[13]. Charge emission from the electrodes into the dielectric is dictated by the injec-
tion of electrons at the cathode and the extraction of electrons at the anode.
Before being injected from the electrode into the dielectric electrons first have to overcome a
potential barrier(figure 3.1). The energy W required to pass the potential barrier is defined
to be φ − χ according to figure 3.1. The terms φ and χ are the work function of the metal
and the electron affinity of the insulator respectively.
Chapter 3. Theoretical Background 33
Two injection processes can be defined: the Schottky process and the Fowler-Nordheim
process[12–14]. In the Schottky process, which is valid for field strengths up to 100 kV/mm,
an applied electric field reduces the potential barrier making it possible for electrons to travel
across the barrier. The injection current density dictated by the Schottky process is dependent
on both temperature T and electric field E(equation 3.4).
" p #
φ − e eE/4π 0 r
J = AT 2 exp − (3.4)
kT
The Fowler-Nordheim process is valid for field strengths above a few 100 kV/mm. In this case
electrons will tunnel through the barrier which is very narrow because of the high electric
field strength. Because these field strengths are not reached in this work the Fowler-Nordheim
process is not considered.
Conduction
The band gap of an insulator is very wide (> 2eV) according to the band model. Due to such
a wide band gap very few electrons gain enough energy to travel directly from the valence
band to the conduction band. The conduction band of an insulator thus contains almost no
electrons resulting in a very low conductivity.
The crystalline regions in PE are perfect insulators with a conductivity of less than 10−20 S/m[13].
The conduction present in polyethylene thus takes place in the amorphous regions where most
material defects are present. Most charge carriers travelling through the amorphous parts are
trapped in a charge trap and released again after a certain amount of time. Charge carriers
thus move from trap to trap.
The time a charge carrier spends inside a trap is as already explained in this section dependent
on trap depth. For example an electron which is present in a trap with a depth of 1eV will
stay trapped for approximately 1 hour. The velocity at which charge carriers move from trap
to trap is about 105 m/s. Charge carriers thus spend most of the time inside traps in stead
of moving from trap to trap. The conductivity of PE is therefore dependent on the density
Chapter 3. Theoretical Background 34
and depth of the traps present in the material. This conduction mechanism is also called trap
limited conduction which is shown schematically in figure 3.2.
Although an applied electric field cannot move trapped charge carriers directly, the application
of an electric field can decrease the potential barrier present at a potential well. The energy
needed to escape from the trap is in this case smaller which results in increased conductivity.
This process is called the Poole-Frenkel effect[12, 14]. The conduction current density dictated
by this effect is presented in equation 3.5.
" √ #
eφ − β E
J = σ0 E exp − (3.5)
2kT
Another conduction mechanism called resonance tunneling can be defined[12, 13]. According
to quantum mechanics a finite probability exists that a charge carrier is present at the other
side of the potential barrier between two traps. This probability is of course very small.
However when the distance between two traps is very small (< 1nm) this probability is not
zero and an electron can sometimes appear on the other side of the barrier. This process is
called tunneling because it seems that the electron digs a ”tunnel” through the barrier.
Finding the electrical degradation threshold of a material would thus be advantageous for the
design of insulation systems. However two main problems arise in finding such a threshold[17].
Firstly, determining the threshold with life tests at different electric field strengths would be
virtually impossible because the lifetime of a material subjected to DC fields close to threshold
is extremely long(> 50 years). A life test for threshold estimation would therefore take too
much time. Short-term alternatives thus need to be found for indirect threshold estimation.
Secondly, multiple electrical aging processes with distinct threshold fields can be detected
under dc stress(threshold for tree formation, threshold for partial discharge inception etc.).
Knowledge on which electrical aging process is relevant for the estimation of the degradation
threshold would thus be needed to develop a proper short-term test.
Space Charge
Space charge accumulation has shown to only occur above a certain value of applied electric
field strength. This value is called the space charge accumulation threshold. The electrical
degradation of an insulator under DC stress is found to begin above the space charge accu-
mulation threshold[16, 17]. An indirect estimation of the electrical degradation threshold can
therefore be acquired by finding the electric field above which space charge starts to accumu-
late. The space charge accumulation threshold field is also called the ”lowest” degradation
threshold because it is the lowest field strength above which aging processes are initiated.
The space charge accumulation threshold can be derived with two measurement methods:
conduction current measurements and space charge measurements. The derivation of the
space charge accumulation threshold of a material with conduction current measurements is
described in section 3.5 whereas the derivation of the threshold with space charge measure-
ments is described in this section.
To find the charge accumulation threshold with space charge measurements an electric field-
space charge density(E-ρ) characteristic is needed. An example of such a characteristic is
shown in figure 3.3.
Note that it is not possible to start a new poling test directly after the last one because
the accumulated charge might still be present in the cable insulation. A relaxation time of
another 2 hours is used before starting a new measurement effectively prolonging the test
duration to 4 hours.
The threshold electric field Eth,sc is the point in the plot above which the average accumulated
space charge density starts to rise with increasing field strength. Also the slope of the graph
above threshold can be an important parameter giving information on the rate of charge
accumulation and thus the behavior of a material at fields above threshold.
Chapter 3. Theoretical Background 36
Temperature dependency
The E-ρ characteristics are determined at three different temperatures(20 ◦ C, 40 ◦ C and 60 ◦ C)
to investigate the temperature dependancy of the charge accumulation threshold. In prior
research the threshold field has shown to decrease when the temperature is raised[18]. Fur-
thermore the space charge density in the insulation after 2 hours of poling at a fixed poling
field is temperature dependent according to measurements conducted in [18]. From the space
charge density values at different temperatures for a fixed applied field an Arrhenius type
temperature dependency can be found with corresponding activation energy[18].
Thorough investigation of charge transport quantities in a dielectric can thus give a good
insight in the charge dynamics and field enhancement of the dielectric. Two particularly
interesting quantities with respect to transport mechanisms are the trap depth distribution
and charge carrier mobilities inside a material[19, 20]. Note that these quantities also seem
to be sensitive to material degradation: a decrease in mobility and an increase in trap depth
has been found for electrically and/or thermally aged PE based materials[19].
Chapter 3. Theoretical Background 37
Physical quantities such as charge carrier mobility and trap depth distribution cannot be
calculated directly. However an estimate of the apparent mobility and trap depth can be
obtained with the use of PEA depolarization characteristics[19, 20]. Section 3.3.1 describes
the estimation of apparent mobility while the trap depth estimation is described in section
3.3.2.
The data points each correspond to the average charge density in the cable insulation at a
point in time calculated with equation 2.22. As can be seen from figure 3.4 the charge present
in the insulation decreases over time during depolarization. The charge is either depleted from
the material via the electrodes giving rise to a current in the measurement circuit or the charge
is recombined with opposite charges in the material. If it is assumed that recombination can
be neglected the charge depletes from the specimen within a time period related to the depths
of the charge traps, where charge was trapped during polarization, and to the transit time
between the traps.
Chapter 3. Theoretical Background 38
It is assumed that the transit time is negligible with respect to the time spent within traps(trap
limited conduction, section 3.1.2). The apparent mobility and trap depth calculations(section
3.3.2) can therefore be performed with the appropriate processing of the depolarization char-
acteristics. Because the apparent mobility calculated with the depolarization characteristic is
assumed to be dependent only on the time charges spend within traps this type of mobility
is called apparent trap-controlled mobility.
Two main considerations should be taken into account when the trap-controlled mobility is
derived from depolarization characteristics. Firstly, space charge values are only recorded
after a delay of 0.1 to 2 seconds. This time lag is caused by the fact that the HVDC supply
still has to discharge after turning down the voltage. Furthermore the oscilloscope also has an
initialization time because the scope is averaging over 1000 sweeps to suppress noise. Because
the first part from 0 to 0.1 seconds of the depolarization process is not recorded the derivation
of the mobility of the free or shallow trapped charges is not possible.
Secondly, the depolarization characteristics contain average values of space charge density.
No distinction is made between negative and positive charges. It is therefore assumed that
the space charge distribution in the insulation is essentially unipolar.
The relation derived for the calculation of apparent trap controlled mobility is given in equa-
tion 3.6. The derivation of this equation can be found in[19].
2 dq(t)
µ(t) = (3.6)
q 2 (t)
dt
The mobility equation is heavily approximated[19]. However, the equation is quite easy to
implement in software. Once the depolarization characteristic is acquired the mobility can be
calculated easily. Furthermore the goal of this work is to acquire information on the depletion
of charge from the insulation which is a useful quantity for material characterization not the
acquisition of the exact mobility values. Equation 3.6 is implemented in MATLAB. The
corresponding MATLAB code can be found in appendix D.
Equilibrium
In the model an arbitrary number m of trap levels is assumed to exist in a material. Each
trap level has a fixed number density N (i) of trapping sites. In an uncharged material not
all trapping sites are unoccupied. Charges are continuously moving between trap levels. The
number density of occupied traps in an arbitrary level i is denoted n(i).
The movement of charges will be explained with the following example: A charge which is
trapped in a trap at level i is promoted to the top level 1 by thermal excitation from which
Chapter 3. Theoretical Background 39
it falls back down again to any of the lower levels. The charge exchange rate at the top level
can be calculated with equation 3.7.
dn1 N1 − n1 (t) U2 − U1 N2 − n2 (t)
= n2 (t) ν exp − n1 (t) ν (3.7)
dt N1 kT N2
+Promotion from level i to 1
−Degradation from level 1 to i
The term ν defined as ν = kT /h is called the attempt frequency, h is the Planck constant,
U2 − U1 is the activation energy between levels 2 and 1 and (Ni − ni (t))/Ni is the relative
available trapping site density at level i. The first part of equation 3.7 corresponds to the rate
of promotion from level 2 to level 1 whereas the second part corresponds to the degradation
rate from level 1 to level 2.
In the equilibrium situation, where no charge is building up, the promotion and degradation
rates to and from level 1 are equal. The exchange rate dn1 /dt is thus equal to zero in
equilibrium. Note that level 1 corresponds to the energy state of the free charges.
Charging
At applied electric field charges are injected at the electrodes and will occupy empty trapping
sites in the material. An equilibrium situation is again reached but in this case with a higher
amount of charge due to the extra occupied traps. Due to the applied electric field a current
will flow through the external circuit. This charge flow can be described by the equation n1 G
where G is equal to µeE and µ is the mobility of the charges in level 1. Note that in this
equation the charge flow is assumed to consist only of drift currents. Diffusion is thus not
taken into account.
The charge exchange rate at level 1 is still determined by equation 3.7 with n1 G subtracted
from the left term dn1 /dt assuming that the trap depth is not influenced by the electric
field. At high electric fields this assumption is not valid and the energy levels have to be
corrected(poole-frenkel effect, section 3.1.2).
Discharging
When the electrodes are short-circuited and the sample is discharging a discharge current will
flow through the external circuit. The discharge current can be described in a way similar
to equation 3.7. In this case the charges which are promoted to level 1 are assumed to be
extracted from the material immediately. The charges in level 1 are thus not able to degrade
to a lower level.
The number density of trapped charges in level 1 is now assumed to be zero n1 (t) = 0 because
level 1 will be completely empty according to the assumption above. The degradation terms
and the term corresponding to the relative available trapping sites in level 1 are therefore
Chapter 3. Theoretical Background 40
From this equation the total number density of charges trapped in the material at time t can
be described by equation 3.9.
m
X m
X
ntot (t) = ni (t) = ai exp(−bi t) (3.9)
i=2 i=2
ai = ni (0)
∆Ui0
bi = ν exp
kT
∆Ui = Ui − U10
0 0
Note that the energies U and ∆Ui are in this case denoted U 0 and ∆Ui0 . The reason for this is
that the local electric field present in the materials decreases the energy barrier for detrapping
effectively resulting in lower measured values for the trap levels. ∆Ui0 represents the energy
barrier for charge detrapping which is decreased by the electric field.
The corresponding unaffected trap depth ∆Ui could be derived by subtracting a term based
on the Poole-Frenkel law from ∆Ui0 [20]. However, according to [20] the effectiveness of such
corrections is questionable. Therefore, the analysis is limited to ∆Ui0 values.
Equation 3.9 can be applied directly to the depolarization characteristic. The characteristic
is divided in an appropriate amount of segments each corresponding to a trap level. The seg-
mentation is based on the difference in steepness between different parts of the characteristic.
Note that although equation 3.9 defines the total number density of charges ntot (t) it can be
directly applied on the average charge density qavg (t). The calculation of the trap depths is
implemented in MATLAB. The corresponding code is contained in appendix D.
The mobility of these packets is typically in the range of 10−14 to 10−12 m2 V−1 s−1 which
is comparable to typical values of apparent trap-controlled mobility[19, 20]. Because of the
relatively low mobility these packets are called “slow” charge packets.
Chapter 3. Theoretical Background 41
Charge packets are considered as an important aging factor. The relatively large amounts
of charge moving through the material can locally cause high electric field enhancements.
Furthermore observations show that charge packets can even accelerate breakdown processes
in the insulation[21]. The process of “slow” charge packet generation has been described in
several proposed theories which consist of different viewpoints. At this time no theory is able
to describe all possible situations with respect to “slow” charge packets.
Another effect which is described in [21] is the fast occurence of hetero charge formation
at the electrodes(faster than a fraction of a second to a few seconds). In [21] space charge
measurements on mini cables reveal that large amounts of hetero charge accumulates at both
semicon layers.
In the work of Fabiani faster and more accurate equipment is used than is available in this
work. These fast measurements show that charge accumulates within a few seconds or even a
fraction of a second after voltage application. The process of hetero charge accumulation can
be correlated to injection[21]. Hetero charge accumulates at the counter electrode when the
charge trapping rate is low with respect to the transit time between traps and the extraction
rate is low with respect to the rate of arrival of charge carriers at the counter electrode. Thus
a region of hetero charge will be building up at the counter electrode which will grow wider
over time while no significant charge is present in the bulk.
The measurements conducted in the work of Fabiani show that both negative and positive
charge packets travel between the anode and cathode as shown in figure 3.5. Each charge
packet contains only a small amount of charge (100 - 130 pC). However, because of the high
repetition rate of charge packet injection a reasonably large amount of hetero charge is already
present at the electrodes after a few seconds.
Figure 3.5: Charge packets moving between the electrodes in the first second after the start of
polarization[21]
The accumulation of hetero charges at the electrodes only seems to occur at test specimens
which have a semicon layer between insulation and metal electrode. Furthermore only in
relatively thick specimens this accumulation is present. When measuring on a thin sample
Chapter 3. Theoretical Background 42
with a semiconductive layer present at only one electrode hetero charge accumulates only
at the electrode with semicon layer. It seems that semiconductive material somehow blocks
the extraction of injected hetero charge at the counter electrode. Accumulation of hetero
charge in a material under high dc electric field can thus be observed at the partially-blocking
electrode.
More information on charge packet dynamics can be obtained from the time in which a charge
packet travels across the insulation thickness. As can be seen in figure 3.5 the time in which
a packet travels from the anode to the electrode(or vice versa) is very short(less than 0.5 s).
This is surprisingly fast when it is considered that the overall conductivity of XLPE is very
low. Furthermore, the negative charge packets always seem to be faster than the positive
charged packets[21].
The charge packet dynamics can be represented by the average mobility µ of the positive and
negative charge packets. The average mobility can be roughly calculated with equation 3.10.
ν
µ= (3.10)
Emean
The term ν represents the velocity of the packet and Emean corresponds to the mean applied
electric field in the specimen. The mobilities of the fast charge packets calculated with
equation 3.10 in the work of Fabiani are in the order of 10−10 to 10−11 V2 /ms. The mobilities
of the fast charge packets are clearly much larger(2-3 orders of magnitude) than the mobilities
of the slow charge packets. Furthermore, the mobility of negative charge packets is larger than
the mobility of positive charge packets. The mobility also increases with rising temperature.
Including the mobility of charge packets described by equation 3.10 two main definitions of
mobility are mentioned in this work thusfar. The trap-controlled mobility corresponds to the
conduction of charges trapped in deep trapping sites and the charge packet mobility describes
the conduction of fast charge packets travelling across the insulation. A third representation of
charge carrier mobility can be defined which represents the shallow trapped and free charges.
This type of mobility is called the Space Charge Limited Conduction(SCLC) mobility which
will be described in section 3.5.
The charge of a pulse can be seen as some kind of quantum of charge which is dependent
on the type of material. The value of this quantum of charge is about 3 · 10−9 C for positive
pulses in XLPE, 1.5 · 10−9 C for negative pulses in XLPE, and approximately 0.6 · 10−9 C to
0.3 · 10−9 C for nanostructured epoxy. Furthermore, the pulse charge varies with an increase
in the percentage of nano-additive which indicates that mechanical properties of materials
may play a fundamental role in the formation of charge pulses.
It has also been observed that the shape of each charge pulse does not change while travelling
Chapter 3. Theoretical Background 43
through the material(no high frequency components are lost), even for a distance larger than
1 mm.
From the properties mentioned above it can be speculated that charge pulses propagate like
charged solitary waves(meaning that they behave like solitons) instead of independent charge
packets. (A soliton is a wave packet or pulse which maintains it shape while travelling at
constand speed[23])
Although space charge is not measured directly in this method the threshold shown in figure
3.6 is related to the space charge accumulation threshold[16]. When no space charge is present
in the insulation the electric field will be uniform and the J-E characteristic will follow Ohm’s
law (J = σE) which is a linear relation. This can be seen from the J-E characteristic which
has a slope of approximately one below the threshold in a double logarithmic scale.
At field strengths above the space charge accumulation threshold trapped and mobile charges
will influence the electric field and thus also the J-E relation. In the double logarithmic scale
the characteristic will follow a power law above threshold resulting in a straight line with
slope ≥ 2. This situation is called space charge limited conduction(SCLC).
Chapter 3. Theoretical Background 44
Note that the supra-linear relationship above threshold can also be described by the fact that
carrier mobility is enhanced by the space charges occupying trapping sites. This enhancement
is caused by the fact that there are less traps available for free carriers to be trapped in.
Mobility
With the J-E characteristics a third type of mobility called SCLC mobility can be derived.
The equation for the calculation of the SCLC mobility is shown in equation 3.11 and can be
applied to the steady-state conduction current measurements[19].
8Jd3
µ= (3.11)
9V 2
Because this type of mobility is applied to the steady-state conduction current measurements
the mobility will give information on the free and shallow trapped charges. The SCLC mobility
is thus not time dependent as opposed to the trap controlled mobility which is related to the
progressive emptying of increasingly deeper charge traps.
Temperature dependency
Just as with the space charge measurements the threshold field decreases with increasing
temperature. The temperature dependency of the threshold can again be investigated. Fur-
thermore, the conductivity of a material is expected to show an Arrhenius type temperature
dependancy which is shown in equation 3.12 [11].
B
σ(T ) = A exp − (3.12)
T
The terms A and B are arbitrary constants. When the Arrhenius type temperature depen-
dancy has been found the activation energy of the underlying process can be derived.
Chapter 3. Theoretical Background 45
The term ε0 (ω) is the real part of the complex permittivity and represents the energy stored in
each period of polarization(assuming periodic voltage waveform). The term ε00 (ω) represents
the energy dissipated in each period which is known as dielectric absorption.
Another term which represents the relative losses in a dielectric is tan δ. This term is defined
by the ratio of the real and imaginary parts of the complex permittivity tan δ = ε00 /ε0 . Note
that for the testing of insulation materials on dielectric losses(related to thermal breakdown)
tan δ can also be measured with a schering bridge. However, the frequency of the applied
voltage is in that case limited to 50 Hz.
The complex permittivity ε̄(ω) is not only dependent on frequency, it is also dependent
on temperature. Several direct relations between the complex permittivity,temperature and
frequency are proposed in literature[10]. Debye was the first to find such a relation which is
shown in equation 3.14.
εs − ε∞
ε(ω, T ) − ε∞ = (3.14)
1 + iωτ (T )
The term ε∞ represents the permittivity at ω → ∞, εs represents the permittivity at ω = 0
and τ (T ) is the relaxation time. The real and imaginary parts of the complex permittivity
can be derived from Debye’s equation resulting in equation 3.15.
εs − ε∞
ε0 (ω, T ) − ε∞ = (3.15)
1 + ω 2 τ 2 (T )
iω
ε00 (ω, T ) = (εs − ε∞ )
1 + ω 2 τ 2 (T )
The main problem involving Debye’s law is the fact that it represents only one relaxation
time constant. This means that the law can only be applied on gases and liquids with a low
molecular content which can be seen as one polar group. When such a relation is to be found
for polymers some modifications to Debye’s law are desired.
As already mentioned the permittivity shows several different relaxation processes with cor-
responding relaxation parameters. The most pronounced relaxation process is α−relaxation.
Chapter 3. Theoretical Background 46
α−relaxation processes are generally associated with temperatures above the glass transi-
tion temperature Tg [24]. The glass transition temperature is the temperature above which a
polymer changes from a rigid state to a more rubber-like state.
α−relaxation is characterized by a large narrow loss peak which will shift to higher frequencies
at increasing temperature. This type of relaxation is thought to be related to the cooperative
motion of polar segments of the main molecular chain.
Both relaxation processes are not seperated by a fixed distance in the frequency spectrum.
In some situations these two relaxation peaks may overlap making it very difficult to differ-
entiate between the two seperate relaxation peaks. Figure 3.7 shows the effect of overlapping
relaxation peaks.
Figure 3.7: Schematic representation of ε00 (ω) and the contributions of both relaxation peaks in the
frequency domain[10]
Chapter 4
In this chapter the results of the measurements are described. Section 4.1 contains the re-
sults of the space charge polarization measurements including the values of the accumulation
thresholds. In section 4.2 the results of the apparent trap-controlled mobility and trap depth
evaluation are described. The results of the conduction current measuremenst on both semi-
conductive thin plaques and mini cables are presented in section 4.3. Section 4.4 describes the
results of the dielectric spectroscopy measurements conducted on the thin insulation samples.
The ρ-E characteristic at 20 ◦ C was obtained without using the heating system because the
ambient temperature in the lab is on average 20 ◦ C(a maximum deviation of +2 or -2 degrees
has been noticed). Thus no temperature gradient is present in the cable insulation.
The threshold field is found to be 8 kV/mm. From this point on the average space charge
starts to increase. Extra measurements at electric fields close to the threshold were conducted
to confirm the threshold value. The value of the threshold field is found to be comparable to
values presented in literature[25].
The data points at field strengths of 30 kV/mm and 60 kV/mm are obtained from measure-
ments conducted on this cable in an earlier project. The space charge density at 30 kV/mm
47
Chapter 4. Experimental results and analysis 48
seems to correspond nicely with the measurements conducted in this work. However, the space
charge density at 60 kV/mm is much higher than expected from the fitting line. Furthermore,
the end value of the average space charge density was present almost instantaneously and a
large amount of positive charge was noticed at the earth electrode after voltage turn-off(figure
2.13).
Further investigation in literature revealed that this kind of space charge behavior could be
caused by the existance of fast moving charge packets. The proposed theory on charge packets
is contained in section 3.4. Because of the fact that the underlying process corresponding to
the space charge density value at 60 kV/mm is apparently different, the data point is not
included in the fitting line.
The ρ-E characteristics at 40 ◦ C and 60 ◦ C are shown in figure 4.2 and 4.3 respectively. To
obtain the threshold characteristics at temperatures above ambient the heating system de-
scribed in section 2.2.3 is used. Due to the use of this type of heating the temperature
distribution inside the insulation is non-isothermal resulting in a conductivity gradient. As
explained in section 3.1.1 this conductivity gradient leads to space charge formation. Two
different processes are thus present at these conditions: space charge formation due to a con-
ductivity gradient and increased space charge formation due to the increased thermal energy
of the charge carriers.
Chapter 4. Experimental results and analysis 49
The electric field thresholds at 40 ◦ C and 60 ◦ C are found to be 6 kV/mm and 4.5 kV/mm re-
spectively. The threshold value clearly drops with increasing temperature as expected(Section
3.2).
In the ρ-E characteristic at 60 ◦ C the lines above threshold at 20 ◦ C and 40 ◦ C are also
indicated by the green and magenta lines respectively. From this it can be seen that the
slope of the line above threshold is decreasing at rising temperatures. This means that the
rise in space charge accumulation rate for increasing field strengths also decreases at higher
temperatures. The lines of the three different ρ-E characteristics could therefore intersect
resulting in a situation in which, for a certain field strength above this intersection, the total
space charge density is higher at low temperatures than at high temperatures which is the
opposite of what is expected. The values of the slope b at 20 ◦ C, 40 ◦ C and 60 ◦ C are 1.7 · 10−3 ,
1.1 · 10−3 and 8.7 · 10−4 respectively.
The cause of this behavior could be found in the temperature dependancy of the charge
detrapping and extraction rates. The charge injection current at the electrodes has a positive
temperature dependancy according to Schottky’s law(section 3.1.2) causing a decrease in the
space charge accumulation threshold at increasing temperatures. However, charge detrapping
is also known to have a positive temperature dependancy because of the increase in the thermal
energies of the trapped charges resulting in a higher number of charges which are able to leave
the trapping sites(section 3.1.2).
If it is assumed that the charge extraction rate at the counter electrode also increases with an
increase in temperature or that the extraction rate is much higher than the charge detrapping
Chapter 4. Experimental results and analysis 50
Figure 4.3: ρ vs. E characteristic of cable 11544-4 at 60 ◦ C including the lines above threshold at
20 ◦ C(green) and 40 ◦ C(magenta)
rate the combination of the extraction and detrapping rates also has a positive temperature
dependancy.
With both injection and extraction of charges from the material(including detrapping) having
a positive temperature dependancy it can be stated that a possible cause of the decreasing
charge accumulation rate is the fact that the total charge extraction rate including detrapping
increases faster with increasing temperature than the injection rate.
Note that in the ρ-E characteristic at 60 ◦ C the value of the average space charge density
at 30 kV/mm is much higher than expected. This could also be explained by the existance
of fast moving charge packets which furthermore implies that the inception field of charge
packet generation is also temperature dependent.
As explained in section 3.2 the value of the average space charge density at a fixed poling
voltage is temperature dependent. In literature this value at different temperatures is used to
obtain an Arrhenius type temperature dependancy with corresponding activation energy[18].
The activation energy can give more information on the underlying thermally activated pro-
cess. However, as already explained the value of the average space charge density for a fixed
poling voltage does not always show a positive temperature dependancy. The usefulness of
trying to obtain an Arrhenius type temperature dependancy with this method is questionable.
The ρ vs. E characteristics at 40 ◦ C and 60 ◦ C are shown in figures 4.6 and 4.7 respectively.
Also in the results of this cable it can be seen that the slope of the graphs above threshold
decreases with increasing temperature. The values of the slope b at 20 ◦ C, 40 ◦ C and 60 ◦ C are
in this case 1.0 · 10−3 , 8.8 · 10−4 and 7.7 · 10−4 respectively. The slope at 20 ◦ C is in this case
clearly smaller than for cable 11544-4. The difference in slope at 40 ◦ C and 60 ◦ C is however
much smaller. Furthermore the decrease in slope between 40 ◦ C and 60 ◦ C is comparable for
both cable types.
The space charge accumulation thresholds at 20 ◦ C, 40 ◦ C and 60 ◦ C are 7.3 kV/mm, 6 kV/mm
and 4.5 kV/mm respectively. The threshold values are the same as the threshold values found
for cable 11544-4 except for the threshold at 20 ◦ C which is somewhat lower. This is a
surprising result as it was expected that the values would be entirely different due to the
difference in insulation material.
The type of semiconductive material used for the semicon layers in the cables designated
11544-4 and 11543-3 is both type C. This correspondence suggests that the space charge
accumulation threshold is only dependent on the type of semicon used in the semiconduc-
tive layers. Note that the lower threshold value at 20 ◦ C could be caused by measurement
inaccuracy. Further discussion on this observation is contained in section 4.1.5.
Another interesting property of the ρ vs. E characteristics is the fact that the slope of the
graphs above threshold is much less steep than the slope of the graphs of cables 11544-4 and
11544-1 at 40 ◦ C and 60 ◦ C. The values of the slope b at 20 ◦ C, 40 ◦ C and 60 ◦ C are 2 · 10−3 ,
5.3 · 10−4 and 3.3 · 10−4 respectively. Furthermore, the decrease in the slope of the graph
at increasing temperature is larger for this cable than for cables 11544-4 and 11544-1 which
Chapter 4. Experimental results and analysis 54
Figure 4.7: ρ vs. E characteristic of cable 11544-1 at 60 ◦ C including the lines above threshold at
20 ◦ C(green) and 40 ◦ C(magenta)
suggests a stronger temperature dependency of the charge detrapping and extraction rates
for this cable.
The temperature dependancy of the threshold field is approximately the same as for cable
11544-4 assuming that the threshold at 20 ◦ C is inaccurate.
The space charge accumulation threshold at 20 ◦ C is found to be 4 kV/mm while the thresholds
at 40 ◦ C and 60 ◦ C are both equal to 3 kV/mm. It should be noted that the average space
charge values at very low field strengths are less accurate because of the fact that the signal
is relative small with respect to the distortion. The combination of this with the fact that the
average space charge rises relatively slowly with increasing field strength at 60 ◦ C makes the
determination of the threshold difficult. A very small increase of average space charge at low
field strengths could be diminished by the limited accuracy resulting in a higher threshold in
the ρ-E characteristic. However, results in literature also show equal threshold fields at 40 ◦ C
and 60 ◦ C[18]. Thus the cause of this effect can be either technical or physical.
The threshold values found for this cable are equal to the threshold values found for cable
Chapter 4. Experimental results and analysis 55
11544-1 except the value at 60 ◦ C. Again the type of semiconductive material used in both
cables is the same(type D). This fact supports the hypothesis that the threshold field is
influenced only by the semiconductive layer. This topic is further discussed in section 4.1.5.
The values of the slope b at 20 ◦ C, 40 ◦ C and 60 ◦ C are 8.6 · 10−4 , 7.5 · 10−4 and 6.9 · 10−4
respectively. These values are higher than the slopes of cable 11543-3(except for the slope at
20 ◦ C) but are lower than the values of the cables 11544-4 and 11544-1.
The temperature dependancy of the threshold values of this cable is equal to the temperature
dependancy of cable 11544-1 at 20 ◦ C and 40 ◦ C. Note that if it would be assumed that the
threshold value at 60 ◦ C is correct, the fitting of the temperature dependancy of the threshold
field to an exponential function would be highly inaccurate.
Figure 4.11: ρ vs. E characteristic of cable 11543-3 at 60 including the lines above threshold at
20 ◦ C(green) and 40 ◦ C(magenta)◦ C
important factor in the determination of the space charge accumulation threshold it can be
stated that the threshold fields are determined by charge injection processes.
The main difference between both types of semicon(type C and D) is the concentration of
carbon black. From the results it can be concluded that a high concentration of carbon black
in the semicon layers results in a low threshold field and vice versa. If the semicon material
with a high concentration of carbon black shows to have a high conductivity and the semicon
with a low concentration shows a low conductivity the previous statement also holds for the
conductivity in stead of the concentrations. The results of conductivity measurements on
both semicon types, contained in section 4.3, should reveal if this is correct.
The difference in the temperature dependancy of the threshold field between cables 11544-
1 and 11544-4 is minimal according to the results. The temperature dependancy of the
threshold field of the cables 11543-3 and 11543-4 are equal to the temperature dependancy
of the cables 11544-4 and 11544-1 respectively(neglecting the aberrent threshold values). In
total it can be said that the temperature dependancy is not influenced by either the carbon
content in the semicon layers or the type and concentration of additives in the insulation.
A more general conclusion on this is that the temperature dependancy of the threshold field
could be determined only by the basic molecular structure of the materials which is in this
work the same for both the semicon materials and the insulation materials.
The slope of the ρ-E characteristics above threshold is shown to be decreasing at increasing
temperature. This could be caused by the fact that the charge detrapping and extraction
Chapter 4. Experimental results and analysis 58
Figure 4.14: ρ vs. E characteristic of cable 11543-4 at 60 ◦ C including the lines above threshold at
20 ◦ C(green) and 40 ◦ C(magenta)
rate under applied field shows a larger positive temperature dependancy than the injection
rate. Of course the rate of charge detrapping and extraction is lower than the injection rate
above threshold because otherwise there would be no charge build up.
The slope of the ρ-E characteristics at 40 ◦ C and 60 ◦ C is less steep for cables 11543-3 and
11543-4 than for the cables with designation 11544-4 and 11544-1. This could mean that
the electric field dependancy of the charge extraction and detrapping rate of the cables with
designation 11543-3 and 11543-4 is larger than for the cables with designation 11544-4 and
11544-1 at 40 ◦ C and 60 ◦ C. A more general proposition is that the charge detrapping rate
of insulation material I exhibits a larger electric field dependancy than insulation material II
at 40 ◦ C and 60 ◦ C. The results of the apparent mobility and trap depth could support or
invalidate this proposition. These results are contained in section 4.2.
In figures 4.15 and 4.16 the influence of the semiconductive material and the insulation ma-
terial on both the threshold field and the slope at 40 ◦ C is displayed.
Note that at 20 ◦ C the slopes of the ρ-E characteristics of the cables with designations 11544-4
and 11543-3 are quite steep (1.7 · 10−3 and 2.0 · 10−3 ). These cables both have semicon type
C as semiconductive layer. The cables with semicon type D show much smaller values for
the slope b, which are much closer to the values of the slope at higher temperatures. The
reason for this could be that at room temperature the charge extraction rate at the electrodes
is the limiting factor while at higher temperatures charge detrapping is the most important
mechanism.
Chapter 4. Experimental results and analysis 60
Figure 4.15: The influence of the semicon and insulation on the threshold field
Figure 4.16: The influence of the semicon and insulation on the slope
Chapter 4. Experimental results and analysis 61
Note that the conclusions and propositions contained in this section only apply to these four
types of mini cables.
Secondly, the charge present in the cables did not show any significant decrease at room
temperature within the 3 hours depolarization time. When a good comparison of the ap-
parent trap-controlled mobility and trap depths is desired far longer depolarization times are
required(i.e. weeks).
The apparent trap-controlled mobilities of the four test cables are shown in figure 4.17. The
value of the trap-controlled mobility at the end of the 3 hour depolarization time for cable
Chapter 4. Experimental results and analysis 62
11543-3 is 3.6 · 10−15 m2 /V s. Cables 11543-3, 11544-4 and 11544-1 show a trap-controlled
mobility of 2.6 · 10−15 m2 /V s, 5.8 · 10−16 m2 /V s and 4.4 · 10−16 m2 /V s respectively. These
values are comparable to values found in literature[19, 20].
From 200 s after the start of the depolarization process the mobility of the cables with insu-
lation type I is higher than the mobility of the cables with insulation type II. This fact can
also be deduced from the depolarization characteristics in which the cables with insulation I
deplete charge faster than the cables with insulation II.
Insulation material I clearly shows an advantage over insulation material II. Especially when
field inversion or voltage turn-off would occur in operation conditions. The addition of charge
trapping agents in the XLPE based insulation thus gives better results than the addition
of carbon black with respect to the mobility. It should be noted that the charge trapping
agent was added to create more shallow traps not deep traps. The carbon black was added
to increase the overal conductivity of XLPE.
The difference in mobility between the cables with the same insulation material is very small
with respect to the difference in mobility between the two insulation materials. This sug-
gests that the apparent trap-controlled mobility and thus the charge detrapping rate is only
determined by the type of insulation material. The detrapping of charges can thus only be
influenced by changing the type and concentration of additives in the material.
The trap depth values of the deepest traps of the cables with designations 11544-4 and
11544-1 are both 1.20 eV. The cables with designations 11543-3 and 11543-4 show to have
a maximum trap depth of 1.16 eV and 1.17 eV respectively. The traps depth values in this
work are however higher than values found in literature[20].
The values of the deepest traps of the cables with the same semiconductive material are
equal or almost equal which corresponds with the results of the apparent mobility evaluation.
Insulation material II thus seems to contain deeper traps than insulation I resulting in the
confirmation of the fact that insulation I should contain more shallow traps. The trap depth
thus also seems to be determined only by the the type and concentration of additives present
insulation material.
in section 4.3.1. Unfortunately it was not possible to run conduction current measurements
on the mini cables before the end of this work. Thus there are no measurement results. The
reason for this and the work that has been done on the setup is described in section 4.3.2.
The conductivity σc1 of the plaque consisting of semicon type C(“low” carbon content) at
an applied field strength of 1 kV/mm was found to be 1.48 · 10−12 S/m. At 2 kV/mm the
conductivity σc2 is 1.45 · 10−12 S/m. The conductivities σd1 and σd2 of the plaque consisting
of semicon D(“high” carbon content) at field strengths of 1 kV/mm and 2 kV/mm are found
to be 1.62 · 10−12 S/m and 1.53 · 10−12 S/m respectively.
The measurements were conducted at two different field strengths because the conductivity is
not always constant with respect to the applied electric field. Furthermore, some fluctuations
may occur in the conduction current measurements due to external influences or movement of
the measurement cables. The average conductivity is thus calculated from the conductivity
values at both field strengths. The average conductivities of semicon type C and D are
1.47 · 10−12 S/m and 1.58 · 10−12 S/m respectively.
The conductivity of semicon type D is clearly higher than the conductivity of semicon type
C. This implies that the semicon with a high carbon content has a higher conductivity than
Chapter 4. Experimental results and analysis 64
the semicon with a low carbon content. With these results it can be stated that the space
charge accumulation threshold in the 4 tested mini cables could be dependent only on the
conductivity of the semiconductive layer. Note that this can only apply for semiconductive
materials with the same basic molecular structure as used in these mini cables.
The knowledge available to build a new setup was limited. A schematic representation of a
similar setup including some test results was found in literature[11]. Furthermore, the con-
duction current setup for thin plaques also gave valuable information. However no technical
details were available. The setup, including the oven, HVDC supply, electrometer, protection
circuitry, GPIB interface and the computer is working at the moment of writing this section.
A few test measurements were performed to correct the settings of the electrometer and to
check the connections to the electrometer.
For each sample the real part of the complex permittivity ε0 (ω), the imaginary part of the
complex permittivity ε00 (ω) and the loss factor tan(δ) are measured. The ε0 (ω), ε00 (ω) and
tan(δ) of insulation type I are displayed in figure 4.19, 4.20 and 4.21 respectively. Furthermore,
the ε0 (ω), ε00 (ω) and tan(δ) of insulation type II are shown in figure 4.22, 4.23 and 4.24
respectively.
Although the frequency domain spectroscopy uses an AC test voltage the results are still useful
for materials used in DC applications. The frequency range of interest is in this case the low
frequency spectrum below approximately 1 Hz. In this part of the spectrum contributions of
both dc conductivity and in some cases low frequency dispersion(LFD) can be found[26].
From the graphs of the tan(δ) and the imaginary component of the complex permittivity ε00
of both insulation types it can be seen that the losses increase at lower frequencies. This
is caused by the finite dc conductivity of the insulation materials. The comparison of both
materials on the basis of these graphs reveals that the losses in insulation I are larger than
in insulation II which suggests that the dc conductivity of insulation I is larger than the dc
conductivity of insulation II. This fact only holds for temperatures of 40 ◦ C and higher.
The losses are larger at higher temperatures because of the rise in conductivity. At very low
temperature (−40 ◦ C and 0 ◦ C) almost no conductivity is present in both materials. Note that
Chapter 4. Experimental results and analysis 65
the conductivity at 40 ◦ C is also very low for insulation type II resulting in a large difference
in conductivity between insulation I and II.
When looking at the graphs of the real part of the complex permittivity ε0 (ω) of both insula-
tion types another important effect can be noticed. The permittivity’ of insulation I shows a
steep rise below 1 Hz. This effect is called Low Frequency Dispersion (LFD) and is believed
to be related to the reversible storage of a finite amount of charge within the material[26].
This effect is different from pure dc conductivity because in that case no storage of charge
can be present.
The amount of charge stored in the material is related to the ratio ε0 (ω)/ε00 (ω) which usually
has a value between 0.01 and 0.1[26]. The permittivity’ of insulation type II remains constant
which indicates that there is no or no distinguishable amount of charge stored in the material.
Therefore only dc conductivity seems to be present in the material at frequencies below 1 Hz.
However at temperatures of 110 ◦ C and 120 ◦ C it can be seen that the permittivity’ is lower.
This can be caused by the fact that the specimen is placed between two electrodes which
tightly hold the specimen. At high temperatures the material expands causing the pressure,
exerted by the electrodes, on the specimen to increase. Due to this effect the permittivity of
a material can be changed. Furthermore, the permittivity’ at 110 ◦ C and 120 ◦ C does rise at
low frequencies which suggests that there is a minor LFD effect present at temperatures of
110 ◦ C and higher.
Note that in the graphs of ε00 (ω) and tan(δ) below 1 Hz no loss peak can be found because no
distinguishable dipolar processes are involved at these frequencies. Furthermore at the high
frequency range (up to 2 MHz) there is also no loss peak. It is also important to note that
some relaxation processes may go unnoticed at low frequencies because of the strong influence
of the dc conductivity and LFD.
When comparing these results with the results of the depolarization characteristics it should
be noted that in this case the measurement is performed at low voltages. The use of time
domain characteristics at high electric fields obtained with the use of the conduction current
setup could give more insight into the frequency response at high fields. A few examples of
those characteristics and information on how to read those characteristics can found in [26].
The LFD effect is only visibly present at temperatures of 80 ◦ C and higher which suggests
that at the temperatures used in the PEA measurements these effects are not present. Fur-
thermore, the higher losses measured at 40 ◦ C for insulation type I could mean that the dc
conductivity of insulation I is higher than the dc conductivity of insulation II or that the
effect of LFD is more pronounced in insulation I. If it is assumed that the influence of dc
conductivity is much larger than LFD, the difference in dc conductivity would fit well with
the fact that at 65 ◦ C the apparent trap controlled mobility is also larger for insulation I. The
notes presented in this section should however be kept in mind.
The permittivity” of both insulation materials was compared to the permittivity” of a sample
of unfilled XLPE(no additives) found in literature[27]. It was noticed that the permittivity”
of insulations I and II at low fequencies is much larger than the permittivity” of the unfilled
XLPE sample(0.5 · 102 opposed to 0.5 · 10−1 ). This suggests that additives can affect the dc
conductivity and the storage of charges(LFD) in XLPE significantly.
Chapter 4. Experimental results and analysis 66
The aim of this thesis was to investigate in depth the dynamics of space charge accumula-
tion of a selected set of mini cables. This is accomplished by the determination of the space
charge accumulation threshold field, the apparent mobility of the charge carriers and the trap
depths in the insulation and the frequency domain characteristics of the insulation materials.
Several test techniques were used in the determination of these parameters: space charge
measurements with the use of the PEA method , frequency domain dielectric spectroscopy
and conduction current measurements. Section 5.1 contains the main findings on the ex-
perimental observations summarized in a number of conclusions. Section 5.2 contains some
recommendations for future research.
5.1 Conclusions
Space charge accumulation threshold
The space charge accumulation threshold measurements have shown that the value of the
threshold field of the four minicables is dependent only on the type of semiconductive material
used in the outer and inner semicon layers. The space charge accumulation threshold is thus
presumably only dependent on charge injection processes.
The difference between the two semicon materials is the concentration of carbon black. The
semiconductive material which shows the lowest threshold also has the highest concentration
of carbon black and the semicon material which shows the highest threshold has the lowest
concentration of carbon black. Thus the threshold field shows a negative dependancy on the
concentration of carbon black in the semiconductive layers.
Following the statement above the conductivity of the semiconductive materials was deter-
mined in the conduction current setup for thin plaques. Results show that the conductivity
of the semiconductive materials is positively dependent on the concentration of carbon black.
Connecting this statement with the threshold results obtained from the PEA measurements
suggests that the mini cables which have semicon layers with a relatively high conductivity
exhibit a lower threshold than the cables which have semicon layers with a low conductivity.
This means that the threshold shows a negative dependancy with respect to the conductivity
of the semicon layers.
69
Chapter 5. Conclusions and Recommendations 70
The temperature dependancy of the threshold field has shown to be approximately equal for
all four mini cables. This suggests that the temperature dependancy of the threshold is not
determined by the additives used in both the semicon and the insulation materials.
Not only the threshold field was an important parameter found in the ρ-E characteristics. The
slope in the ρ-E characteristics has shown to be decreasing for an increase in temperature.
This is believed to be caused by the fact that the charge extraction and/or detrapping rate
has a larger positive temperature dependency than the charge injection rate. Furthermore,
results have shown that the cables with insulation type I show to have a slope which is less
steep than the slopes obtained for the cables with insulation type II at 40 ◦ C and 60 ◦ C. This
fact suggests that, if the slope is determined by the charge extraction and/or detrapping rate,
the detrapping rate in insulation I is larger than the detrapping rate in insulation II at 40 ◦ C
and 60 ◦ C.
Because the threshold field seems to correspond only to injection processes the influence of the
temperature gradient on the threshold field is believed to be negligible(there is no temperature
gradient at the electrodes). Note that there are no measurement results with a homogeneous
temperature distribution available for comparison. Furthermore the space charge which is
present due to the conductivity gradient could be removed by the processing.
The results also show that the mobility is almost only dependent on the type of insulation
material(there are small differences in conductivity between different semicon types). This
statement suggests that the apparent trap-controlled mobility is only dependent on charge
detrapping not on charge extraction at the electrodes.
Trap depth calculations have shown that insulation type II has deeper traps than insulation
type I. The charge trapping agents present in insulation type I seem to create more shallow
traps.
Because the depolarization characteristics are only dependent on the detrapping of charges in
the insulation the calculated trap depths are only dependent on the type of insulation material.
This suggests that calculating the trap depths with the use of depolarization characteristics
is a valid approach.
At temperatures of 80 ◦ C and higher a low frequency dispersion effect is present in the per-
mittivity’ characteristic of insulation I suggesting that a build up of charge is present in the
Chapter 5. Conclusions and Recommendations 71
material at very low frequencies. This effect is not present in the characteristic of insulation
type II. Below 80 ◦ C this effect is not present thus the significance of this observation for the
characterization of the cables is questionable.
General conclusions
In general it can be observed that for this type of mini cables the space charge accumulation
threshold can be changed by changing the amount of carbon black(and thus changing the
conductivity) in the semiconductive layers. The apparent mobility and trap depths could
be influenced by changing the type and/or the concentration of additives in the insulation.
These two properties seem to be independent resulting in the fact that the design of such
cables can be split up in two independent parts(i.e. the semicon layer and the insulation).
It is important to note that the results and conclusions presented in this work are not directly
applicable to other cable systems and base materials.
The combination of semicon C and insulation type I seems to deliver on average the best
results. The threshold values corresponding to semicon C are the highest of the two types.
Furthermore, the trap-controlled mobility of the insulation is also the highest resulting in
the highest depolarization rate and the insulation also has the lowest trap depths of the
two insulation types. The slope of the ρ-E characteristics at 40 ◦ C and 60 ◦ C are also much
less steep than the slopes of the other semicon-insulation combinations. The cable with
designation 11543-3 thus seems to be the most suitable cable for dc applications.
Furthermore, as explained in section 4.4 time domain spectroscopy can also be used to evaluate
the complex permittivity of each sample at high field strengths. This could reveal more
information on the frequency response around normal working conditions with respect to
applied field. Time domain spectroscopy can also be used on the mini cable samples removing
the need to produce special semicon-insulator plaques.
A.1 Deconvolution
The high-pass characteristic of the sensor-amplifier system can be seen as a transfer function
H(f ). The resulting signal vs (t) will then be a function of the original signal vsorg (t) and the
transfer function as shown in equation A.1.
From the measured signal vs (t) the original signal has to be obtained by the deconvolution
technique. The deconvolved signal is never exactly the same as the original signal thus the
deconvolved signal is designated vsdecon (t). To obtain the original signal the output signal
vs (t) should be divided by the system response H(f ) as shown in equation A.2.
org −1 Vs (f )
vs (t) = F (A.2)
H(f )
The system response is however unknown. The simple equation given above is thus not
directly applicable. The system response has to be approximated which is done in the de-
convolution technique. The deconvolution technique is schematically represented in figure
A.1. The signal vs−el represents the earth electrode and bulk material part of the output
signal vs (t). The signal outside that interval is zero. The signal designated vs−el−id is the
ideal signal from the electrode charge. Ideal means without the response of the sensor and
sensor-amplifier system. In this case the ideal signal is represented by a pulse which has
approximately the same width as the earth electrode in vs (t) and has a height of one(which
is usually much larger than the height of the electrode signal in vs (t)).
1
The ideal signal is divided by the output signal vs−el which should result directly in H(f ).
However the result of this division contains high frequency components which will distort the
deconvoluted signal vsdecon (t). Therefore a low pass gaussian type filter with transfer function
G(f ) is used to remove or attenuate these components. After multiplication with G(f ) the
73
Appendix A. Deconvolution and attenuation and dispersion correction 74
1
signal represents H(f ) which is an approximation. With the approximated transfer function
the deconvoluted signal vsdecon (t) can be directly calculated from the output signal vs (t) with
equation A.2.
The ideal electrode signal has height one in stead of the actual voltage because of the fact
that the signal has not yet been calibrated. Because of this the resulting deconvoluted voltage
signal will not have the correct value. To correct this difference the maximum value of the
deconvoluted signal vsdecon (t) is divided by the maximum value of the output signal vs (t) which
results in a size factor. The deconvoluted signal is then divided by this factor to acquire the
correct signal as shown in equation A.3.
max(|vsdecon (t)|)
Ksize = (A.3)
max(|vs (t)|)
v decon (t)
vsdecon (t) = s
Ksize
Note that the deconvolution can only work correctly when the signal is free of space charge.
When there is space charge present the ideal signal could not be flat outside the electrode
region. The deconvolution method described in this section is implemented in MATLAB. The
corresponding MATLAB code is contained in section A.2.
2 grid on;
3 disp('click start and end point of EARTH ELECTRODE and start point of HV ELECTRODE');
4 [qq, zz]=ginput(3);
5 aa=round(qq(1)); %start point of earth electrode
6 bb=round(qq(2)); %end point of earth electrode
7 cc=round(qq(3)); %start point of hv electrode
8
9 y1=zeros(5001,1);
10 y1(aa:cc,1)=mmatrice(aa:cc,1);
11
12 vin=zeros(5001,1);
13 vin(aa+30:bb−30,1)=1;
14 g=input('gaussian filter= ');
15 for i=1:5001
16 ex(i)=exp(−(g*(i)).ˆ2);
17 end
18 EX=fft(ex);
19
20 f=−100000:+40:100000;
21 vin1=vin;
22 Vin1=fft(vin1);
23 Y1=fft(y1);
24 for i=1:5001
25 H(i)=(Vin1(i)/Y1(i))*EX(i)*EX(i);
26 end
27 Z=fft(mmatrice(:,1));
28 for i=1:5001
29 X(i)=Z(i)*H(i);
30 end
31 x=−ifft(X);
32
33 dimfact=max(abs((real(x))))/max(abs(mmatrice(:,1)));
34 finalsignal=real(x)/dimfact;
35 plot(finalsignal,'r');
36 hold on;
37 plot(mmatrice(:,1),'b');
38 grid on;
39 hold off;
40 tt=input('OK? 1=si 0=no ');
41 %Applying deconvolution to all traces
p(x, ω)
G(x, ω) = = exp(−α(ω)x) exp(−iβ(ω)x) (A.6)
p(0, ω)
To calculate the coefficients α(ω) and β(ω) two known acoustic waves at two different locations
in the sample are needed. In this thesis the acoustic wave generated at the HV electrode p(d, ω)
and the acoustic wave at the sensor location p(0, ω) are used. Using these two acoustic waves
the function in equation A.6 will be different(equation A.7).
p(d, ω)
G(d, ω) = = exp(−α(ω)d) exp(−iβ(ω)d) (A.7)
p(0, ω)
When the function G(d, ω) has been found the coefficients α(ω) and β(ω) can be calculated.
With the calculated coefficients the function G(x, ω) can be defined. The original acoustic
wave at any location inside the sample can then be calculated from the acoustic wave at the
sensor(equation A.8).
Note that the acoustic wave p(0, t) is the acoustic wave at the sensor location after having
travelled through the sample. The acoustic wave p(x, 0) is the acoustic wave before travelling
through the sample.
To be able to implement this procedure the original signal at the HV electrode corresponding
to p(d, ω) should be known. If the sample is free of space charge the signal at the HV electrode
is directly related to the electrode charge. When a known voltage is applied across the sample
this signal can be calculated.
The recovered signal calculated from the detected signal with function G(x, ω) at the sensor
is however not exactly the same as the original signal. This is caused by the fact that some
high frequency components in the waveform are completely attenuated and can thus not be
retrieved from the detected signal[1].
Some problems may arise when using G(d, ω) to calculate the coefficients α(ω) and β(ω).
The calculation of G(d, ω) is defined as a ratio of two functions(equation A.7). When the
denominator contains zeros the division is impossible. Furthermore the method presented here
has a higher amplification for the high frequency components than for the low frequency part.
Noise present in the detected signal, which usually contains much high frequency components,
will be amplified as well.
To overcome these problems the functions α(ω) and β(ω) are approximated in stead of cal-
culated with equation A.7. The coefficients are assumed to be defined by equations A.9 and
A.10.
α(ω) = A + aω 2 (A.9)
β(ω) = bω (A.10)
Appendix A. Deconvolution and attenuation and dispersion correction 77
Equations A.9 and A.10 are based on the assumption that the original and the detected
waveform can be described as a gaussian function[1].The original and detected waveform can
be described by the gaussian functions in equations A.11 and A.12 respectively.
Equations A.13 and A.14 represent the gaussian functions y1 (t) and y2 (t) transformed to the
frequency domain.
ω2
r
π
Y1 (ω) = A1 exp −iτ1 ω − (A.13)
a1 4a1
ω2
r
π
Y2 (ω) = A2 exp −iτ2 ω − (A.14)
a2 4a2
ω2 1
r
Y1 A1 a2 1
G(d, ω) = = exp −iω(τ1 − τ2 ) − − = exp(−d(A+aω 2 )) exp(−ibdω)
Y2 A2 a1 4 a1 a2
(A.15)
When the terms A, a and b are defined according to equations A.16, A.17 and A.18 the
coefficients α(ω) and β(ω) can be written as equation A.9 and A.10.
r
1 A1 a 2
A = − ln (A.16)
d A2 a 1
1 1 1
a= − (A.17)
4d a1 a2
1
b = (τ1 − τ2 ) (A.18)
d
The values of the terms a, A and b are dependent on the type of bulk material used in the
specimen. Because the values of these terms are not exactly known an approximation of the
values is used. With these values the coefficients α(ω) and β(ω) are determined according to
equations A.9 and A.10.
The output signal obtained from the deconvolution technique is a voltage signal vsdecon (t).
Therefore the function G(ω) is calculated with the coefficients according to equation A.19 in
which ∆t is the time difference between the peak of the earth electrode and the peak of the
HV electrode in the signal. This time difference corresponds to the distance d between the
electrodes.(which is the same as the thickness of the sample)
The voltage signal corrected for attenuation and dispersion is then obtained using equation
A.20.
h i
vsatt (t) = F −1 G(ω)Vsdecon (ω) (A.20)
Appendix A. Deconvolution and attenuation and dispersion correction 78
Because of the fact that the terms a, A and b are approximated it should be checked if the
resulting signal is correct. This check can be done by performing a double integration on the
corrected signal after the calibration procedure described in the next section. The result of
this double integration should represent the voltage distribution across the sample. Note that
the signal without space charge is required because space charge will distort the expected
voltage distribution.
The voltage across the sample should be the same as the applied voltage at the HV electrode
and zero at the earth electrode. When this is not the case the terms a, A and b may not
be correct and should be changed after which the attenuation and dispersion correction is
repeated with the new values. The correction procedure for attenuation and dispersion is
implemented in MATLAB. The corresponding MATLAB code is contained in section A.4.
1 HVelectrodeMax=round(ginput(1));
2 HVelectrodeEnd=round(ginput(1));
3 EarthelectrodeStart=round(ginput(1));
4 EarthelectrodeMax=round(ginput(1));
5
6 % Sample milimeter distance per data point in trace matrice
7 mmperpoint insulation = (Ro − Ri) / (EarthelectrodeMax(1)−HVelectrodeMax(1));
8
9 % acoustic correction
10 matrice(1:HVelectrodeEnd,:) = 1.13 * matrice(1:HVelectrodeEnd,:);
11
12 %divergence correction%
13 for i = 1:size(matrice,2)
14 for j = 1:length(matrice)
15 matrice(j,i) = matrice(j,i) / (sqrt(Ro/(Ri−(HVelectrodeMax(1)−j)*
16 mmperpoint insulation)));
17 end
18 end
B.2 Calibration
80
Appendix B. Divergence correction and Calibration 81
12
13 % Calculate Electric Field
14 for i = 1:size(SpaceCharge,2)
15 Field(:,i)=(1/(epsilon*8.85e−12))*cumsum(SpaceCharge(:,i))
16 *mmperpoint insulation*0.001;
17 end
18
19 assignin('base', 'EField', Field);
20
21 voltage = Field;
22 % Voltage
23 for i = 1:size(Field,2)
24 voltage(:,i) = mmperpoint insulation*0.001*(cumsum(Field(:,i)));
25 end
26 voltage = V*1000−voltage;
27
28 voltage data = voltage;
29
30 space charge data = SpaceCharge;
31
32 electric field data = Field;
Appendix C
82
Appendix C. Calculation of average space charge density 83
84
Appendix D. Calculation of apparent mobility and trap depths 85
95 end
96 clear klaar
97 %extract the coefficient values of the fits
98 coeff1 = coeffvalues(trapmodel1);
99 coeff2 = coeffvalues(trapmodel2);
100 coeff3 = coeffvalues(trapmodel3);
101 coeff4 = coeffvalues(trapmodel4);
102 coeff5 = coeffvalues(trapmodel5);
103
104 b1 = −1*coeff1(2);
105 b2 = −1*coeff2(2);
106 b3 = −1*coeff3(2);
107 b4 = −1*coeff4(2);
108 b5 = −1*coeff5(2);
109
110 % calculation of trap depth
111 T = input('Temperature of test object: ');
112 k = 1.38e−23;
113 h = 6.63e−34;
114 nu = k*T/h;
115
116 depth1 = abs((log(b1/nu)*k*T)/1.6e−19);
117 depth2 = abs((log(b2/nu)*k*T)/1.6e−19);
118 depth3 = abs((log(b3/nu)*k*T)/1.6e−19);
119 depth4 = abs((log(b4/nu)*k*T)/1.6e−19);
120 depth5 = abs((log(b5/nu)*k*T)/1.6e−19);
121
122 trace = input('Name of output file: ', 's');
123 save(trace, 'depth1', 'depth2', 'depth3', 'depth4', 'depth5',
124 'bound1', 'bound2', 'bound3', 'bound4', 'bound5', 'mu', 't100', 'q')
Bibliography
[1] R. Bodega, “Space charge accumulation in polymeric high voltage dc cable systems,”
Ph.D. dissertation, Delft University of Technology, 2006.
[2] M. J. P. Jeroense, “Charges and discharges in hvdc cables,” Ph.D. dissertation, Delft
University of Technology, 1997.
[4] Y. Li, M. Yasuda, and T. Takada, “Pulsed electroacoustic method for measurement of
charge accumulation in solid dielectrics,” IEEE Transactions on Dielectrics and Electrical
Insulation, vol. 1, no. 2, pp. 188–195, 1994.
[6] Y. Li, K. Murata, Y. Tanaka, T. Takada, and M. Aihara, “Space charge distribution mea-
surement in lossy dielectric materials by pulsed electroacoustic method,” Proceedings of
the 4th International Conference on Properties and Applications of Dielectric Materials,
vol. 2, pp. 725–728, 1994.
[8] M. Fu and G. Chen, “Space charge measurement in polymer insulated power cables
using flat ground electrode pea system,” Science, Measurement and Technology, IEE
Proceedings, vol. 150, no. 2, pp. 89 – 96, 2003.
[10] B. Alijagic-Jonuz, “Dielectric properties and space charge dynamics of polymeric high
voltage dc insulating materials,” Ph.D. dissertation, Delft University of Technology, 2007.
87
Bibliography 88
[12] P. H. F. Morshuis, “Lecture notes: High voltage dc,” 2010, delft University of Technology.
[20] G. Mazzanti, G. C. Montanari, and J. M. Alison, “A space-charge based method for the
estimation of apparent mobility and trap depth as markers for insulation degradation-
theoretical basis and experimental validation,” IEEE Transactions on Dielectrics and
Electrical Insulation, vol. 10, no. 2, pp. 187–197, 2003.
[21] D. Fabiani, G. C. Montanari, L. A. Dissado, C. Laurent, and G. Teyssedre, “Fast and slow
charge packets in polymeric materials under dc stress,” IEEE Transactions on Dielectrics
and Electrical Insulation, vol. 16, no. 1, pp. 241–250, 2009.
[22] G. C. Montanari, “Bringing an insulation to failure: the role of space charge,” IEEE
Transactions on Dielectrics and Electrical Insulation, vol. 18, no. 2, pp. 339–364, 2011.
[23] G. I. Stegeman and M. Segev, “Optical spatial solitons and their interactions: Univer-
sality and diversity,” Science, vol. 286, pp. 1518–1523, 1999.
[24] B. Lennon, “Dielectric relaxation analysis to assess the integrity of high voltage dc cables
of the mass impregnated type,” Master’s thesis, Delft University of Technology, 2009.
Bibliography 89
I would like to express my gratitude to Barry Lennon for helping me with the measurements
and the processing of the data in the beginning of this project. He was always available to
help me with problems or the interpretation of the results.
Furthermore I would like to thank dr.ir. Morshuis for reading my thesis chapters and for
giving me inspiration.
Also, I would like to thank the staff of the HV-lab including mr. Termorshuizen and ing. van
Nes for helping me with the construction of the new conduction current setup and for helping
me with the problems I encountered on the PEA setup.
I also like to thank Mr. van der Graaf for creating a pleasant atmosphere in the lab and for
transporting the new Siemens cage to the HV-lab. Finally, my thanks goes to mr. Andritsch
for helping me with the conduction current measurements on the semicon plaques and the
dielectric spectroscopy measurements. To all mentioned here and to the rest of the staff: I
wish you good luck in your future adventures.
90