The 20th International Symposium on High Voltage Engineering, Buenos Aires, Argentina, August 27 – September 01, 2017

SIMULATION AND MEASUREMENT OF PD IN THE UHF FREQUENCY

RANGE - STUDIES IN TIME AND FREQUENCY DOMAIN
M. Hartje*, M. Huntke, S. Peik and
1
2
M. Akbari Azirani, P. Werle
1
Hochschule Bremen, Neustadtwall 30, Bremen, Germany
2Leibniz Universität Hannover, Institute of Electric Power Systems,

Division of High Voltage Engineering and Asset Management, Schering-Institute,

Callinstr. 25A, 30167 Hannover, Germany
*Email: [email protected]

Abstract: UHF PD-measurements can be used in a wide frequency range up to some

GHz. Many studies have been published in the past showing different maxima in the
frequency spectrum depending on the setup. From the beginning of the UHF PD-
measurement, there is a discussion, how to find a relation between apparent charge at
the measurement terminals and the characteristics in the frequency spectrum.
This paper deals with HF-simulations and measurements of different PD sources such as
different small needles and a potential free floating metal fragments in gas and
submerged in oil. The paper investigates the influence on the frequency spectrum from
different length and thickness of the needle. This evolves an influence on the discussion if
an UHF PD-measurement can be calibrated.

1 INTRODUCTION 2 EXPERIMENTAL SETUP

Partial discharge measurements have been The experimental setup consists of one PD
effective in assessing the condition of insulation intensity measuring circuit according to IEC 60270,
systems in the past. Measurement is in general as shown in Figure 1.
defined in IEC 60270. A PD calibration of the setup
is needed for comparison with other
measurements. Calibrated measurements are
typically performed with frequencies below several
MHz. The IEC 60270 currently does not provide a
recommendation for a calibration of the UHF
diagnosis [1, 2]. Various suggestions on the
assumed influences on the amplitude density
spectrum have been applied in the past, like:
 the nature of the impurity [3]
 the embedding medium [3,4]
 the discharge length [5]

Since various publications have reached different

conclusions, this contribution examines the
influence of the measurement setup on the
amplitude density spectrum.

For this purpose, a measurement setup is selected

and also a simulation with CST MICROWAVE
STUDIO is carried out, in order to investigate the
Figure 1: PD Measurement setup in an anechoic
effect of geometrical changes to the test setup. A
chamber. (up: schematic, down: photo)
selection of these setups are then simulated and
also measured with an experimental setup.
The HV setup was placed into a RF screened
cabin to omit the influence of typical noise sources
The objective is to determine the differences
like mobile phone and wireless LAN networks
between measurement and simulation, and
during the measurement. As seen in the right part
whether the measured amplitude spectra are
in Figure 1, there is a white colored logarithmic
consistent with the simulation. It is to be expected
periodic antenna Electro-Metrics EM-6950 (0.2 to
that there are deviations.
Figure 2: Screenshot of the CST graphical editor created in CST MICROWAVE Studio. The model is
window with the simulation model for depicted in Figure 2.
the UHF electrical field study
In Figure 2, the cage around the components
1 GHz) was placed in two meter distance to the shows the border of the simulated field room. The
setup. The antenna was connected via a coaxial field probe represents the antenna from the
cable to the measurement instruments in a experimental measurement. The fault model is
screened neighbor cabin. The measurements were depicted in the right part of the cage at the bottom.
performed by using the oscilloscope R&S The bottom is a (conducting) metal plate. The
RTO1004 with a bandwidth of 600 MHz and the insulated support, left of the fault model can be
spectrum analyzer R&S FSC 3 with a bandwidth of neglected for the simulations, since it has little to
9 MHz to 3 GHz. no effect on the propagating electromagnetic
signals. Right of the fault model, there is a 1.2 nF
A basic test setup was utilized as shown in Figure decoupling capacitor, which is modeled as a thick
3, where the geometrical dimensions can be varied conductor. The upper bar has a length of 680 mm.
in order to examine their effects on the received
amplitude density spectrum. Hence, the distance
between the upper and lower plate, as well as the
length of the rod and the needle fixture can be
altered for simulation purposes. A few distances for
the practical realization were chosen.

Figure 4: Simulated amplitude density spectrum

with different plate to plate distances

For the simulation results shown in Figure 4

Figure 3: Shape of the PD fault model various plate to plate distances of the PD fault
model have been used. The distances covered are
3. SIMULATION 170, 100 and 70 mm as depicted in Figure 4.

3.1 COMPUTER SIMULATION Figure 4 shows the normalized to zero dB

amplitude of the E-field probe is shown with a PD
To investigate which parameters influence the source of approx. 2 pC and the length of the
amplitude spectrum of the received UHF signals, simulated PD pulse is in total 4 ns.
an exact RF model of the used test setup was
The PD fault model with a plate to plate distance of Table 1 shows the roughly estimated frequencies
170 mm shows a maximum at 440 MHz. of the 3 selected plate to plate distances showing
Decreasing the distance to 100 mm, the maximum the obtained resonant frequencies.
slides to 612 MHz and with 70 mm distance one
small maximum at 610 MHz and the leading Table 1: Estimation of the resonance with the 2
maximum at 1.1 GHz occurs. As it can be seen in explained models
Figure 4, there are significant differences in the
amplitude minimum of the three different test
setups. Frequency [MHz]
Stick length Antenna with
3.2 MODEL BASED ESTIMATION LC resonator
[mm] top capacity
The test setup can also be seen as a quarter 170 406.44 372.29
wavelength ground plane antenna. A simple 100 529.93 571.30
modeling would assume the needle and its fixture 70 633.39 741.65
as a quarter wavelength ground plane. The
extension and structure is assumed as a large
inductance (Figure 5). The upper plate will be The LC resonator consists of the plate to plate
considered as a loading capacitance. capacity and the inductance of the needle fixture
between the both plates. The antenna model with
top capacity is calculated in the last column.

Table 1 shows as expected differences between

both calculation methods. There are also
differences to the simulation results in Figure 4,
although in both cases a clear shift of the main
resonance frequencies can be seen. The absolute
differences might be traced back to inaccuracies of
the simulations as well as to the used
simplifications for the simulation models. But it is
clearly shown, what variance of resonance
frequencies can be expected during a
measurement and it could explain dependent
spectrum density functions, which have to be
Figure 5: PD test setup decomposed into an verified by measurements.
antenna
4. MEASUREMENTS OF SPECTRUM
A first method to estimate this „antenna“ calculates
a simple quarter wavelength antenna with the With a spectrum analyzer in average mode with 64
correction of the conductor radius a and the length repetitions and a sweep time of 30 s the measured
h. The wavelength is λ. Considering the correction spectrum of the test setup with 170 mm distance of
of the thickness of the conductor the resultant both plates is shown in Figure 6 and compared to
wavelength res is calculated according to [6] by the simulation from Figure 4.

h
c
res   a .
f  4 1 h
a

But with the roof or top capacitance from a

conducting upper plate, there is a further correction
needed with the radius of the roof plate h [6].

res
h  h 
4

The second method of estimation is done by the Figure 6: Measured and simulated amplitude
capacity of the distant plates of the test setup and density of a 170 mm plate to plate fault
the inductance of the needle fixture and the needle. model with a spectrum analyzer
The measured data in Figure 6 show a very small reduce certain frequency shares of the received
difference between noise level and amplitude signals, even if only the first 10 ns of it are
peaks due to the measurement method with the considered for FFT calculations.
spectrum analyzer. But the maximum of the
simulation compared to the measurement shows
the expected maximums at 0.3 to 0.6 GHz. The
intensity of the measured PD was approx. 25 pC.

5 EFFECT OF THE TRANSFORMER TANK

In order to investigate the effect of the transformer

tank on the received signals with UHF probes, the
PD model with a plate to plate distance of 170 mm
was used. For capturing the electromagnetic
signals, an OMICRON UVS 610 conical monopole
antenna and a 3.5 GHz, 10 GS/s oscilloscope were
used. Hereby, the corona discharge model was
placed inside the transformer tank and the UHF
probe was inserted into the tank through its DN 80
oil valve. Since the effect of position of the PD
model inside the transformer tank on the UHF
signals is investigated in [7], in this contribution,
the PD model is placed in the center of the tank.
Subsequently, a second measurement followed
using the same test setup with both the PD model
and the UHF probe placed in free space, while (a)
maintaining the same distance and alignment as
previously existing in the transformer tank.
Simultaneously, the electrical PD measurement in
accordance to IEC 60270 was carried out. This
served the purpose of comparability, so that in both
measurements, the PD intensity could be roughly
kept at a constant level, while the signals were
captured with the UHF probe and the oscilloscope.
High voltage is applied via a 220 V / 200 kV
transformer. For the measurement in the
transformer tank, the PD model is connected to the
end side of a 123 kV bushing installed on top of
the air-filled transformer tank. The used test circuit
and test setup are depicted in Figure 7. For both
measurements, the mean value of 10 captured
signals, each with a duration of 10 ns, was formed
and the results were weighted using a Hamming (b)
window and transformed into the frequency domain
via FFT. Since according to the manufacturer, the Figure 7: Test circuit to evaluate the effect of
OMICRON UVS 610 is designed for a frequency transformer tank (a) and measurement
range up to 1 GHz, the focus of the investigation is in free space (b) with the same test
on the frequency range from 200 MHz to 1 GHz [8]. setup
The obtained results are presented in Figure 8, in
which all curves have been normalized in such a Moreover the UHF probes used for capturing the
way that the maximum was shifted to 0 dB. The electromagnetic signals have a different antenna
simulation results are shifted by 100 MHz for better characteristic in comparison to the measurements
fitting. and simulations presented in Figures 4 and 6.
Despite these facts, there seems to be a relation
A correlation between the measurements and the between the maxima and minima of the depicted
simulation results is observed. However, since the diagrams. Hence in order to reach a clear
free space measurement was not carried out in an conclusion, further investigations are necessary.
anechoic chamber, certainly surrounding
disturbances, e.g. mobile and wireless networks
affected the results and lead to overall deviations.
The measurements inside the transformer tank is
shielded from external noise. Due to the relatively
small size of the tank reflections can amplify or
 CONCLUSION

For UHF PD-measurements the electrode

configuration that creates the PD impulses seems
to be responsible for some main resonance
frequencies, which could be shown by simulations
and measurements for a needle-plate arrangement.
However, the whole frequency spectrum received
by the UHF antenna during a PD measurement is
influenced by various parameters e.g. reflection
due to the transformer tank, antenna characteristic,
amplifier, recording unit etc. Despite this fact, it
might be possible to distinguish different PD root
causes by analyzing the frequency spectrum of the
captured UHF PD signals, but more investigations
are necessary in order to find corresponding
correlations.
Figure 8: Captured PD pulse with the UHF probe
via measurements in free space and
transformer tank in comparison to
simulation results REFERENCES
[1] VDE: Hochspannungs-Prüftechnik-Teilent-
ladungsmessung,“ IEC 60 270, 2016.
[2] A. Küchler: Hochspannungstechnik, Springer
6 DISCUSSION AND CONCLUSION
Berlin Heidelberg, 2009, pp. 431-433.
[3] M. Judd. G. Cleary: UHF and current pulse
The simulation, calculation and the both
measurements of partial discharge activity in
measurement methods for the PD UHF test setup
minearl oil, IEE Proc.-Sci. Meas. Technol., Vol.
show the following results:
153, No. 2, March 2006, 2006.
[4] W. Gao, D. Zhao, D. Ding, S. Yao, Y. Zhao, W.
 The electrode configuration results in preferred Liu: Investigation of Frequency Characteristics
frequencies in the spectrum density, which is of Typical PD and the Propagation Properties
clearly shown in Figure 4.
in GIS, IEEE Transactions on Dielectrics and
 The measurement method with the spectrum Electrical Insulation Vol. 22, No.3; June 2015,
analyzer has to be set in the average mode. 2015.
This limits the sensitivity even with maximum [5] X. Li, X. Wang, M. Rong, D. Xie, Y. Fu, Q.
hold mode. Gao, N. Yin: Relationship Between Time-
frequency Representation of PD-induced UHF
 The Fourier transformation of repeated time Signal and PD Current Pulse, 2016
overlaid measurements has to be done for the International Conference on Condition
first few nano seconds to avoid overlaid Monitoring and Diagnosis - Xi'an - China, 2016.
reflections. These reflections would disturb the [6] G. Ronniger: Lineare Antennen, HFT-
measured spectrum density of the PD Praktikum, 2016.
impulses. [7] M. Akbari Azirani, P. Werle, A. Akbari, H.
 The standard antenna used outside the tank Jahangir and J. Szczechowski, "An
showed a typical low pass characteristic (see Investigation on PRPD Patterns Generated by
Figure 9, free space). the UHF Measurement Technique for Power
Transformers." 25th Iranian Conference on
 The measurements inside the tank show a Electrical Engineering (ICEE), 2017.
good relation to the simulation of the test setup [8] OMICRON electronics GmbH, MPD600
with 170 mm of the plate to plate distance (see Product Manual, 2017.
Figure 9)
 Differences between simulation and
measurement can be explained with the
simplified modeling for the simulation and due
to the properties of the receiving antennas.