ISBN: 978-1-4799-8903-4 2015 IEEE 11th International Conference on the Properties and Applications of Dielectric Materials (ICPADM)

Highly Accurate Current Response Measurements

of Insulation Materials for HVDC Applications

Malte Tschentscher, Christian M. Franck

High Voltage Laboratory, ETH Zurich
8092 Zurich, Switzerland

Abstract— This paper presents a current measurement setup polarization processes, followed by very low and slow time-
to determine polarization and conduction processes of solid varying currents, requires current measurements in a range of
insulation materials under HVDC stress. The setup enables 10-3 A to 10-15 A. Some measurement setups described in
simultaneous polarization, depolarization, and conduction literature show noise levels above 10-12 A that compromise the
current measurements of up to four material samples. The lower end of the measurement range [3, 9, 10]. In other studies,
volume or surface current measurements are performed in a built-in low-noise voltage sources of Keithley® electrometer
pressure vessel, adjustable to the whole range of possible long- are used to reduce the noise caused by voltage harmonics,
term operating conditions present in gas-insulated equipment.
limiting the maximum voltages to 1 kV [4, 11].
Besides discussing the electrical and mechanical construction,
this contribution addresses restrictions of the measurement range In order to analyze if an insulation material qualifies for
caused by an insufficient signal-to-noise ratio. gas-insulated HVDC equipment, it is essential to investigate its
behavior in all possible operation conditions. In case of gas-
Keywords— Polarization measurements; Conductivity insulated equipment, the temperature at the solid insulation
measurement; Surface resistance; HVDC transmission may range from -20 to 90 °C and the DC electric field could
reach up to 10 kV/mm in case of polarity reversals [1].
I. INTRODUCTION
This paper introduces a measurement setup for highly
Both development and design of reliable gas-insulated accurate current response measurements to voltage step
equipment for HVDC differ significantly from that optimized excitations after a sudden DC application, removal, or reversal.
for HVAC. Phenomena such as surface charge accumulation The setup enables parallel investigations of up to four
along with temperature-, field-, and humidity-dependent insulation material samples within the entire range of possible
material properties are dominant in determining the electrical operating conditions present in gas-insulated equipment.
field gradient [1, 2, 3, 4]. In order to calculate the electrical
field stress in gas-insulated HVDC equipment, knowledge of II. EXPERIMENTAL SETUP
both conduction and polarization in the bulk and at the
insulator surface is essential [5, 6]. Voltage step excitations A. Electric Circuit
result in a time-dependent current response at the electrodes. The current measurement setup is shown in Figure 1. The
The time-dependent response is caused by charge transport and high voltage DC source (FUG HCN 140-35000),
polarization in the insulation. Both time-dependent polarization programmable for voltages up to ±35 kV, consists of a
and steady-state currents, i.e. residual currents after relaxation grainacher cascade.
of transient phenomena, are an integral part of insulation
characterization. Studies on commonly used solid insulations
such as filled/unfilled epoxy resins [7], reinforced
thermoplastics, or resin impregnated papers/synthetics have
shown transition times to steady-state currents in the order of
hours up to months. A large number of the materials response
with very slow time-varying currents following power-laws
t-n with small values of n (0<n<1) [8].
The information obtained by frequency domain (FD)
measurements of these insulation materials is typically
restricted to minimum frequencies of 10-3 Hz. Therefore,
current response measurements to a voltage step excitation in
the time-domain (TD) are a powerful technique for the
understanding of very slow polarization and conduction
processes. Nevertheless, the transient characterization of highly
resistive insulation materials under HVDC stress is challenging
Fig. 1. Measurement setup to determine polarization and conduction
in terms of measurement range, accuracy, and precision. The processes of solid insulation under HVDC stress
implication of the transient current response due to fast

978-1-4799-8903-4/15/$31.00 ©2015 IEEE

208
 ISBN: 978-1-4799-8903-4 2015 IEEE 11th International Conference on the Properties and Applications of Dielectric Materials (ICPADM)

Residual AC components with amplitudes of 200 mVpp, a measurement device while earthing the others via resistors.
minimum frequency of 50 Hz, and harmonics of a multiple of The earthing resistors are equal to the DC input impedance of
40 kHz superimpose the high voltage output signal. Due to the the measurement device, i.e. ranging from 1 Ω to 1 MΩ. The
setup capacity these disturbances may cause displacement performance specification of the switch is determined by its
currents reaching up to 10-6 A in the measurement signal. system impact. Firstly, the control circuit of the switch must
Therefore, the HV signal is filtered by a passive first order not disturb the measurement signal. Therefore, either the coil
low-pass filter consisting of a 10 MΩ or 50 MΩ resistor and a voltage of the relays must be negligible during measurements
5 µF capacitor (rated voltage: 20 kVDC). The filter limits the or the coil-electrode resistance of the switch must be high
voltage harmonics to a maximum amplitude of 5 μVpp and enough to limit the interference current to orders below the
damps the displacement currents to a negligible level (< fA). measurement signal. Secondly, no significant leakage current
An alternative solution is the installation of an active HV low- from the signal path to earth must occur. Therefore, the
pass filter. The implementation had been tested but was contact to ground resistance of the switching relays must be
rejected as it caused further high- frequency interferences. high compared to the DC input impedance of the measurement
Figure 2 shows the guard electrode system used for device. Thirdly, interferences between the signal paths must
polarization and conduction measurements of solid insulation not take place during measurements. Hence, the contact to
materials. The aluminium electrodes can be used for both contact resistance of the switching relays must be large
surface and volume-current measurements. For surface current compared to the DC input impedance of the measurement
measurements, the outer ring electrode (dre= 110 mm or device. Lastly, the connection state, “open” or “closed”, must
108 mm) is connected to HV potential and currents, flowing be uniquely defined for each switching relay at any time. A
along the insulator surface to the measurement electrode toggle relay, for instance, would not satisfy this requirement
(dme= 106 mm), are analysed. For investigations of the volume and could lead to overvoltages within the switch. In
current, the counter electrode (dce= 140 mm) is connected to accordance with these requirements bi-stable reed relays
HV potential, and the outer ring electrode is grounded (GND). produced by Meder (BE05-1E66-M) are used, providing a
The assembly is centered by a hollow cylinder and an switching coil-contact resistance of 1 TΩ, minimum coil to
insulating cap of PTFE. ground isolation voltage of 2 kV, and a maximum switching
current of 0.5 A. The four inputs to one output switching
board consists of eight relays and their respective drives. It is
controlled by means of four TTL signals. The signals are
decoded on the board and define the “open” or “closed”
operation of each relay, giving access to 16 switching states.
Any disturbance from the power supply is prevented by
running the board with an insulated supply and by exclusively
applying the supply voltage during switching operations.
Resistors ensure the discharging of every board capacity and
de-energize the relay drives during current measurements.
The current signals collected at the measurement
electrodes are low-pass filtered, converted, and amplified with
a transimpedance amplifier (Femto DDPCA-300). The
amplifier provides a ±1 % accurate switchable gain from 104
to 1013 V/A and an active low-pass filtering with a variable
bandwidth (-3 dB) of 400 Hz down to 0.1 Hz. It should be
Fig. 2. Contacting method of the cylindrically electrode system for volume
noted that the usability of high gains mainly depends on the
current (red) or surface current (green) measurements maximum input current of the transimpedance amplifier, in
full scale including noise. The maximum input current is
To ensure proper contact, 0.5 mm thick conductive rubber dependent on the gain and equals 1 pA for a gain of 1013 V/A.
plates are placed between the electrodes and the sample. The Therefore, a strong limitation of the displacement currents is
rubber compensates surface inhomogeneities and prevents the required and achieved by the previously introduced HV low-
formation of voids that could cause partial discharges between pass filter between the electrodes and the HV source. The AD-
the sample and the electrodes. The plates are filled with silver conversion of the voltage output signal of the transimpedance
plated aluminum particles and add a negligible volume amplifier is done with a digital multimeter (Fluke 8845A).
resistivity of 80 µΩm. Digital filtering of the multimeter output further reduces noise
A mechanical earthing switch is placed between the by averaging readings.
electrode assemblies and the HV filter. The switch enables to B. Mechanical Construction and Temperature Control
short-circuit the samples and to measure both polarization and
depolarization currents sequentially. The earthing switch is Four electrode assemblies are placed in a pressure-
separated from the filter by a resistor of 10 MΩ, limiting the controlled vessel with nominal working pressures up to
discharge currents of the filter capacitor in case of earthing 0.58 MPa. A pressure sensor (Wika D-10-P) and a humidity
operation or faults within the system. and temperature transmitter (Vaisala HMT318) continuously
log the pressure (p), temperature (ϑ), and relative humidity
Furthermore, a switching board has been constructed,
(%RH) during current measurements. The pressure sensor is
connecting electively one of the four samples with the

209
 ISBN: 978-1-4799-8903-4 2015 IEEE 11th International Conference on the Properties and Applications of Dielectric Materials (ICPADM)

sized for a working pressure up to 1.6 MPa, guaranteeing a essential. It decreases the offset currents as the circuit for
total measurement error <0.1 %. Over the used range the thermally induced currents is closed by the low voltage
percentage error of the humidity transmitter is <2 %RH, and insulation resistances and the DC input impedance of the
the Pt100 temperature sensor shows an accuracy of ±0.35 K. measurement device. Furthermore, junctions should be
avoided or their number at least minimized. Especially,
joining of oxidized copper conductors can produce strong DC
disturbances as the Seebeck coefficient for such junctions (Cu-
CuO) is approximately 1 mV/°C, three orders of magnitude
higher than for Cu-Cu. Nevertheless, a voltage drift of the
source can cause low frequent displacement currents.
Therefore, the voltage drift (dU/dt) times the sample capacity
must be small compared to the measurement signal.
IV. MEASUREMENT PERFORMANCE
In order to assess the setup performance, three
representative current measurements are shown in figure 4. The
vessel was pressurized with 0.45 MPa SF6 to avoid an electrical
breakdown of the gaseous insulation.
Two electrode units were connected for surface current
Fig. 3. System temperatures for a control step from 40°C to 60°C measurements, as shown in figure 2. One electrode was filled
with a material sample (MS) of epoxy resin and one was left
The pressure vessel is temperature-controlled by a liquid empty to investigate if charge injection (CI) occurs at metal-gas
filled temperature circuit. A chiller (Huber CC150W), placed interfaces of the assembly. Furthermore, one measurement
in distance to the measurement setup, brings the vessel to a channel was high resistive capped (HRC) directly at the input
temperature between -20 and 90 °C. This approach avoids of the switch, to measure how much interference coupling of
mechanical or electrical disturbances of the current the three channels occurs on the switching board. Figure 4
measurements. Double layer caps with milled spiral notch shows the raw measurement data and a filtered curve that has
between the layers are used to provide a good thermal contact been collected by filtering the observed data (20 °C and
between the liquid circuit and the current measurement setup. 2 kV/mm) with a 200 data points moving average (MA) filter.
One of the caps is placed at the top and one at the bottom of The measurement was performed sequentially on each
the pressure vessel (see Fig. 1). Figure 3 shows the time- electrode, with a sample rate of 2 Hz and collection time of 60
dependent temperature measured at different setup parts. The data points.
sample surface reaches stable conditions following an
exponential change with a heating time constant (τheat) of
7200s or a cooling time constant (τcool) of 14000s.
III. SOURCES OF INTERFERENCE
Noise and temporary disturbances in the low frequency
current signal can be due to electrical, mechanical, and
thermal interferences. Oscillating disturbances, such as
voltage harmonics or permanent mechanical vibrations, can be
handled by low-pass filtering of the HV and measurement
signal. Furthermore, any cable connection should be made of
low noise cables that shield the signal path. In this setup,
cables with carbon-impregnated conductive PTFE layers
surrounding the PTFE dielectric are used. The layer dissipates
static charges and reduces mechanically generated electrical
noise when the cables are bent. Rather challenging is the
handling of disturbances caused by thermal interferences. DC
offset currents can be introduced by thermoelectric effects. Fig. 4. Surface current measurements of three representative contact
Temperature differences across junctions of dissimilar topologies at 20 °C, 2 kV/mm, and a measurement gain of 1011 V/A
materials or oxidized contacts in the measurement line cause
The observed measurement behavior can be divided into
disturbances in the millivolt range, resulting in pA DC-offset
three time windows. Firstly, the background noise validation
currents. The DC offset currents may lead to the conclusion
takes place when no HV signal is applied to the sample.
that the current saturation is due to the end of transient
Furthermore, this time window shows whether thermal
relaxation phenomena. Therefore, the measured currents
interferences take place and have to be considered or not.
should be at least one order of magnitude higher than the
Secondly, the voltage rise time of the setup, starting from the
oscillating disturbances. A good electrical insulation to
moment that HV is applied until the measurement voltage (Um)
ground, of the low voltage setup components and connections
reaches stable conditions, strongly depends on the charging
between the electrodes and the measurement device is

210
 ISBN: 978-1-4799-8903-4 2015 IEEE 11th International Conference on the Properties and Applications of Dielectric Materials (ICPADM)

time constant of the HV filter. Lastly, the measurement zone 10 kV/mm, this setup enables investigations of the conduction,
enables conclusive identification of sample characteristics. polarization, and injection currents down to a temperature of
Prior to every measurement, a background noise validation is approximately -3 °C.
performed showing a typically filtered noise level of ±1 fA. In
the moment that HV is applied, the observed currents jump by V. CONCLUSION
orders of magnitude. Capacitive charging of the setup takes A measurement setup has been designed for highly accurate
place and can even be noticed in the capped channel due to and precise current measurements under controlled conditions.
capacitive couplings on the switching board. In the discussed By means of interference suppression, the configuration
case, we fixed the gain of the transimpedance amplifier to enables current measurements down to amplitudes of ±3 fA.
1011 V/A. This led to an exceeding of the amplification limit Measurements can be performed at all possible long-term
during charging, and a saturation of the current amplifier could operating conditions present in gas-insulated equipment. The
be seen. The voltage rise time of Um applied to the electrodes temperature range extends from -20 to 90 °C. Furthermore,
may be significantly reduced by pre-charging the HV filter and voltages up to ±20 kV can be investigated. The setup gives
closing a switch between the HV filter and electrodes access to up to four simultaneous current measurements, using
afterwards. The switch enables to investigate relaxation times only one measurement device. Particularly for measurements
down to 10 s. Furthermore, it is possible to adjust the gain of slow time-varying currents, time can be a limiting factor.
during operation to avoid saturation of the transimpedance The signal switch enables to increase the number of
amplifier. investigations by a factor of 4 and consequently reduces the
time for analyses equally. Furthermore, simultaneous
measurements of volume, surface, and charge injection
currents are possible. This approach enables to exclude
misinterpretations of current variations caused by
environmental influences.

ACKNOWLEDGMENT
The authors gratefully acknowledge the technical help and
financial support of ABB Switzerland Ltd.

REFERENCES
[1] H. Fujinami, T. Takuma, M. Yashima and T. Kawamoto, “Mechanism
and effect of dc charge accumulation on SF6 gas insulated spacers”,
IEEE PES Winter Meeting, 89 WM 076-1 PWRD, 1989.
[2] B. Weedy, “DC Conductivity of Voltalit Epoxy Spacers in SF6,“, IEE
hc., Vol. 132, pt. A, pp. 450-454, 1985.
[3] H. Yahyaoui, P. Notingher, S. Agnel, Y. Kieffel, A. Girodet, “Analysis
of Conduction Mechanisms in Alumina-Filled Epoxy Resin under dc
Field and Temperature”, IEEE Conf. Electr. Insul. Dielectr. Phenomena
Fig. 5. Evaluation of the accessible measurement range (NL1= 1 pA and (CEIDP), pp. 667-670, 2013.
NL2= 3 fA) for a unfilled epoxy resin [11] [4] L. Zavattoni, O. Lesaint, O. Gallot-Lavallée, “Surface resistance
measurements on epoxy composites: Influence of surrounding gas,
The accessible temperature range of solid material pressure, temperature, and analysis of the origin of measured currents”,
investigations is strongly influenced by noise, disturbing the IEEE Conf. on Solid Dielectrics (ICSD), pp. 370-373, 2013.
measurement signal. An insufficient signal-to-noise ratio will [5] B. Lutz, „Einflussfaktoren auf die elektrische Feldverteilung in
compromise the lower end of the measurement range and in Isoliersystemen mit polymeren Isolierstoffen bei
turn restrict the significance of the investigation. Figure 5 Gleichspannungsbelastung“, Ph.D. thesis, Technical University Munich
TUM, 2011.
shows the calculated conduction currents of a solid material
[6] R. Lovell, “Decaying and steady currents in an epoxy polymer at high
plate sample, flowing between the outer ring electrode and the electric fields“, J. Phys. D.: Appl. Phys., Vol. 7, 1974.
measurement electrode, as a function of the volume [7] C. Guillermin, P. Rain and S. Rowe, “Transient and steady state currents
conductivity (σ). The field dependence of the conduction in epoxy resin”, J. Appl. Phys. D: Vol.39, pp. 515-524, 2006.
current is neglected and surface conductivity is set equal to the [8] A.K. Jonscher, “Dielectric relaxation in solids”, Chelsea Dielectrics
volume conductivity. By following the grid lines, from Press, London, 1983.
intersection of the noise levels and the calculated [9] B. Lutz, J. Kindersberger, "Influence of absorbed water on volume
measurement currents to the exponential Arrhenius fit of an resistivity of epoxy resin insulators", Conference on solid dielectrics,
example unfilled epoxy resin [11], the minimum accessible Germany, 2010.
temperature for current observations can be found. In this [10] C. Zou, J.C. Fothergill, S. Zhang, X. Zhou “DC conduction mechanisms
example, the minimum temperature, allowing current in epoxy nanocomposites under the humid environment”, International
Conference on Solid Dielectrics, Potsdam, Germany, 2010.
observations for a noise level of 1 pA (NL1) and a field stress
[11] B. Lutz and J. Kindersberger, “Surface Charge Accumulation on
of 1 kV/mm, is approximately 75 °C. The same material could Cylindrical Polymeric Model Insulators in Air: Simulation and
be investigated for an improved noise level of 3 fA (NL2) Measurement”, IEEE Trans. Dielectr. Electr. Insul, Vol. 18, pp. 2040-
down to a temperature of 18 °C. For a field stress of 2048, 2011.

211