IEEE Transactions on Dielectrics and Electrical Insulation Vol. 19, No.

3; June 2012 855

SF6 High Voltage Circuit Breaker Contact Systems under

Lightning Impulse and Very Fast Transient Voltage Stress
P. Simka, U. Straumann and C. M. Franck
Power Systems and High Voltage Laboratories, ETH Zurich, 8092 Zurich, Switzerland

ABSTRACT
In the past few years gas insulated high voltage circuit breakers have improved
considerably, in particular with respect to required drive energy for operation,
compactness and reliability. A further reduction in size implies increasing operational
field stress inside the breaker and, thus, a reduction in built-in safety margins. If the
design comes closer to its physical limits, in particular the dielectric coordination of the
contact systems becomes more and more challenging. The aim of the present work is to
identify factors that may influence the dielectric coordination of contact systems with
reduced coordination margin. The breakdown probability distribution of SF6 insulated
multi-contact systems was investigated with the focus on two aspects. First of all, the
breakdown probabilities under lightning impulse (LI) stress were measured. These
indicated that it is not feasible to extrapolate the breakdown probability curves of
single contacts to a combined full contact system. Secondly, the findings on the same
contact system stressed with very fast transient voltages (VFT) showed a considerable
reduction in breakdown voltage and suggested a significant increase in the main
contact breakdown probability, compared to the LI tests. The results and
interpretation within this work indentified the polarity and voltage shape dependent
breakdown initiation of different contacts as causes for circuit breaker contact system
failure.

Index Terms — SF6 high voltage gas circuit breakers, breaker contacts, electric
fields, dielectric breakdown, breakdown probability, switching transients, very fast
transients.

1 INTRODUCTION contacts. Later, when the arcing contacts separate, the switching
arc is formed and burns between them. In a closing operation,
HIGH voltage gas circuit breakers are the key safety devices
the two contact systems approach each other and the gas gap
in electric power systems and have to separate faulty parts of the
between them breaks down, also called pre-strike. The axial
grid quickly and reliably. The interruption capability of high
exposure of the arcing contacts with respect to the main contacts
voltage circuit breakers have been increased over the last
leads to an increased electric field stress and ensures breakdown
decades and today, SF6 insulated circuit breakers are capable of
(and subsequent arc initiation) between the arcing contacts. Arcs
switching up to 20 GVA per interrupting unit [1].
on the main contact during breaker opening must be avoided
Today’s circuit breakers have two different contact systems: under any circumstances as they cannot be extinguished and
the main (nominal) and the arcing contact system. The outer could lead to a complete breaker failure. An arc initiated by a
main contact system is optimized for minimum contact breakdown between the main contacts during breaker closing
resistance in order to carry currents with low losses in the closed would extinguish when the contacts touch, but could still lead to
position, whereas the inner arcing contact system is designed to damage of the insulator and must also be avoided.
carry switching arcs. The breaker itself and its nozzle system are
Dielectric coordination is usually a compromise between
designed to quench and interrupt an arc burning between the
strong electric field enhancement on the arcing contacts to
arcing contacts. It must be ensured that any breakdown of the
promote the breakdown initiation between them, minimum field
contact gap and all switching arcs only occur on the arcing
enhancement to increase the voltage withstand of the open
contacts. The associated design procedure is known as dielectric
breaker, and a minimum breaker size. To enable further circuit
coordination. Typically, this is achieved by exposing the arcing
breaker design optimization, the following questions are
contacts in axial direction with respect to the main contacts.
investigation here:
When the circuit breaker opens, the main contacts separate first
and the current is commutated from the main to the arcing - Which phenomena beyond the electrostatic field
distribution are important for the dielectric coordination of
Manuscript received on 12 August 2011, in final form 24 January 2012. circuit breaker contact systems?

1070-9878/12/$25.00 © 2012 IEEE

856 P. Simka et al.: SF6 High Voltage Circuit Breaker Contact Systems under Lightning Impulse and Very Fast Transient Voltage Stress

- Is it possible to calculate breakdown voltages reliably? calculated using the electron production rate and the
- Is it possible to extrapolate (lightning impulse) breakdown
probability distribution functions from single contact ionization α and recombination coefficients , with
measurements to full contact systems? 1 exp ∭ 1 dV dt . (2.2)
- Are dielectric coordination failures more frequent if very
fast transients (VFT) exist compared to LI voltages, only? The volume-time criterion accounts for the fact that different
statistical time lags occur for different electric field stresses in
These questions are investigated on a specific simplified volumes of different sizes. Generally speaking, the statistical
circuit breaker model with typical contact configurations and time lag decreases with increasing volume exceeding the critical
common coordination margins. Even though this does not field strength (Vcr).
permit to draw universally valid conclusions, insights into Consequently, the impulse breakdown voltages differ
the questions mentioned above can be given. significantly for negative and positive voltage polarity in such
This paper is an extension of the data presented in [3] and it is situations. Besides longer statistical time lags in the case of
structured as follows: In Section 2, the theoretical basis for positive polarity, the time to arc formation is also influenced by
breakdown in low inhomogeneous, SF6 insulated electrode the formative time-lag [10]. The time scales of this formative
arrangements is reviewed and the most important mathematical time-lag are larger for negative polarity [13].
tools are introduced. Section 3 describes the experimental It was observed that the breakdown voltages are smaller if the
method; that is the setup, diagnostics, and the evaluation insulating gas gap is stressed with VFT voltages compared to
procedure. Section 4 presents the results of both, the LI and LI [14, 15]. The transition of a streamer discharge to a
VFT experiments as well as the results of the breakdown conductive leader channel can be supported by high frequency
voltage calculations, which are discussed in Section 5. Section 6 voltages U(t) of sufficient amplitude. The streamer volume is
concludes this paper with a summary of the key findings. heated by the high frequency displacement currents i(t) [15]:

2 BREAKDOWN IN SF6 (2.3)

A breakdown in the gas gap between two contacts with a In summary, VFT voltages may have three different effects
homogeneous (or slightly inhomogeneous) electric field relevant to the present investigation: Firstly, an increase of
distribution can occur (necessary condition) if the critical voltage stress through constructive reflection of travelling waves
electric field strength (E/p)crit is exceeded in some places and if within the gas insulated system. Secondly, the potential field
the streamer criterion is fulfilled [4]. For typical filling pressures distribution may differ from the electrostatic one and there could
of gas insulated switchgear (pfill = 0.4-0.6 MPa (4-6 bar)), the be a voltage difference between the main and arcing contacts.
streamer criterion is fulfilled on very short lengths of a few Thirdly, the breakdown voltage may be reduced by the high-
100 m or lower in the frequent case, where electrode frequency mechanism described above.
roughness determines the breakdown voltage [4, 5]. Therefore, a
simpler and more practical criterion is often used for SF6 gas
circuit breaker design: the maximal macroscopic field strength 3 EXPERIMENTAL APPROACH
Emax on an electrode surface. In order to investigate a possible failure of the dielectric
The streamer criterion is sufficient with slowly rising voltages coordination, a contact system similar to that of a circuit breaker
(ac and dc) in homogeneous or technically relevant (weakly in- was designed and constructed. In the following this specimen is
homogeneous) configurations, but it is observed that the referred to simply as the `contact system`. The macroscopic
breakdown voltages are typically higher with impulse voltage electrostatic field distributions were calculated and the
stress. In addition, a first electron must be available in the breakdown probability of single contacts and the complete
critical region to start the avalanche. For impulse voltage contact system was measured using lightning impulse
stresses of short duration this may not apply and the statistical voltages [16] and very fast transient voltages superimposed on
time-lag of the first electron creation becomes relevant [6, 7]. lightning impulse.
If the polarity of the electrode with higher electric field stress 3.1 GEOMETRY OF CONTACT SYSTEM
is negative, such first electrons are likely to be produced by field Five different electrodes were used to investigate the
emission [8]. A microscopic electric field of 1 MV/cm [4] up to breakdown behavior as depicted in Figure 1. A plane
10 MV/cm [9] has to be present in order to enable this process. electrode and four contacts whose geometry is similar in size
Such field stress can be reached at the electrode surface even in and curvature to those used in high voltage circuit breakers
low inhomogeneous field configurations due to small surface were used for all experiments. The electrodes are made from
defects or electrode roughness. The rate of electrons leaving the stainless steel (NiCr) and the tips are spherically rounded.
electrode due to field emission can be calculated by the Fowler- The contacts are only used to study the breakdown behavior,
Nordheim equation as in [10]. i.e. they are not designed to carry current or withstand
At positive polarity, the dominant mechanism to generate arcing. The tulip contact, for example, is thus designed as a
first electrons is collisional electron detachment from SF hollow cylinder with rounded front sides. The contacts do
ions [11]. To judge the breakdown probability P of gas gaps not support a nozzle as no arc quenching capability is
with low inhomogeneous electric field distribution and different needed. The contacts are fixed during the experiments but
sizes, the volume-time criterion can be applied [7, 12]. It is the axial position of each of the four contacts can be adjusted
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 19, No. 3; June 2012 857

individually. By this, there is the flexibility to expose the parallel, with the stray capacitance of the connecting GIS-
contacts to different electric fields. Thus, the dielectric circuit). The resistor is inserted in order to minimize the
coordination in specific situations without the influence of energy impact on the electrodes in the case of a breakdown.
the movement of the contacts can be investigated. The
surfaces of the electrodes were sandblasted (80 μm grain
size) in order to minimize the relative change in surface
condition during the experiments. This treatment resulted in
a mean surface roughness of 1.85 μm and a maximum
surface roughness of 15.22 μm (measured according
to ISO 4288:1996E).
In analogy to a real circuit breaker, the contact system
with the plug is hereafter called fixed contact system and the
one with the tulip moving contact system.

a.)
impulse voltage source
capacitor bushings limiting resistor

disconnector,
switch earthing
switch
b.) all measures [mm]
specimen

5
134
170

79
42
22

300 ~
voltage measurement
photo
AC voltage source current measurement
Figure 1. Photographs a.) and schematic drawing and dimensions b.) of the
contacts used in the experiments. The sandblasted stainless steel electrodes Figure 3. Picture and schematic drawing of the experimental setup used to
represent a high voltage circuit breaker contact system. From left to right: evaluate breakdown probabilities of SF6 insulated high voltage circuit
Fixed contact system with main contact (ring), arcing contact (plug), moving breaker contacts used in the investigations.
contact system with main contact (ring) and arcing contact (tulip) and plane
electrode. An ac voltage source was used to sweep excessive ions in
between the individual breakdown experiments in order to
Figure 2 shows the macroscopic electrostatic field reach a reproducible condition of the gas and to ensure
distribution of the contact system, calculated with Comsol statistical independence of the experimental series [17].
Multiphysics 3.4. The tips of the arcing and nominal contacts Consequently, no long waiting time in between the
are aligned in axial direction. It can be seen that the electric field measurements was needed.
strength for this configuration is largest at the fixed arcing
contact (plug); it is almost twice as high compared to the other 3.3 DIAGNOSTICS
electrodes, which is to assure arcing to happen on this contact. The applied voltage was measured using a series damped
capacitive voltage divider (cf. also Figure 3). The current across
140 fixed contact side 130 the electrodes was measured with a Pearson current monitor,
electric field - norm [V/m]

arcing contacts
electric field - norm [V/m]

120
contact separation [mm]

120
main contacts attached in the ground return path of the GIS encapsulation.
side

100 110
contactside

Both measurements were recorded with a digital storage

side

100
contactside

80
movingcontact

90 oscilloscope (LeCroy, Wavepro7000, 1 GHz, 10 Gs/s).

fixed contact

60
80
40 In order to determine whether a breakdown happened on
moving

70
fixed

20
0
moving contact side
60 the arcing or the main contact, each shot was photographed
50
0 2 4 6 8 10 12 14 16 18 by a digital single-lens reflex camera (Nikon D50). In
line length [mm]
addition to the direct image, the camera also recorded the
Figure 2. Electrostatic field of the contact system, calculated with 1 V
image from two mirrors installed inside the GIS. By means
potential difference and 10 mm contact separation between fixed and
moving contact side. Electric field distribution (left) and electric field along of triangulation methods the exact location of the arc root on
the shortest distance (right). the contacts was determined [3, 18].
3.2 EXPERIMENTAL SETUP 3.4 EXPERIMENTAL PROCEDURE
The experimental setup used to determine the breakdown During the assembly the specimen and the test vessel were
probability of the specimen is shown in Figure 3. The voltage kept free of dust, and contact with skin was omitted to
source was a Marx Generator with maximum charging voltage prevent pollution of the electrodes such as deposition of
of 1 MV. The test vessel was a standard SF6-filled GIS grease in order to minimize relative change of surface
element attached via disconnector, earthing switch, and gas condition during the experiments. The test vessel was then
insulated bushing to the air insulated parts on the source side. evacuated (70 Pa=0.7 mbar) and filled with SF6 at 0.6 MPa
The load side was grounded with a 2.2 MΩ resistor (and, in (6 bar) absolute pressure.
858 P. Simka et al.: SF6 High Voltage Circuit Breaker Contact Systems under Lightning Impulse and Very Fast Transient Voltage Stress

After each application of a voltage impulse, irrespective of amplitude in Figure 5 was 600 kV and thus considerably
whether a breakdown had occurred or not, the ac voltage higher than the prospective lightning impulse voltage of
source was connected for 30 s with voltage amplitude of 5 – 520 kV. Furthermore, the measurement system was not
10 % of the peak impulse voltage. By this, all ions between optimized for measuring high frequency. Hence, the peak
the contacts were removed. After this, it took 60 s to amplitude stressing the contact system is assumed to be
disconnect the ac source and to charge the impulse voltage higher than the measured one. However, some information
generator. Applying this procedure in exactly the same way on such peak amplitudes is given by calculation. According
for each measurement ensures a reproducible ion distribution to the simulation guidelines presented in [20, 21] time
and thus also statistical independence of the experiment. resolved simulations of the experimental geometry were
Changing of the contacts due to arcing foot points could made in addition to the experiments [3]. The frequency
impair statistical independence as well. However, content of the excited oscillation is similar to the one
minimizing the impact of arcing with the resistor seemed measured and shown in Figure 5. From the simulations it can
sufficient, as mathematical tests revealed statistical be seen that the peak voltage stress at the contact system is
independence of the experimental series. These tests 80% higher compared to the pure LI stress.
consisted of the iteration-test and the test of comparison of The disconnector was closed during the application of the
breakdown probabilities [17]. ac voltage (cf. section 2.4) and reopened before impulse
The breakdown probability of the contacts was determined voltage application.
by using the constant voltage method according to [16].
600 600
3.5 EVALUATION PROCEDURE
400
A typical waveform of a breakdown measured under applied 500

voltage [kV]
lightning impulse is depicted in Figure 4. A breakdown across 200
400
the electrodes happened at approximately 1.3 μs. The time to
0
breakdown was calculated between 10% of the prospective peak 300
0 50 100 150 200 1 1.5 2 2.5 3 3.5
voltage amplitude and the actual breakdown. The peak voltage time [s] time [s]

of the measured signal, the occurrence of and the time to Figure 5. Voltage measurement of VFT voltage superimposed on lightning
breakdown was determined and stored for post processing. impulse voltage shape stressing the contact system (left) and its detail in the
range of breakdown at 1.8 μs (right).
Noteworthy, the breakdown in Figure 4 did not lead to an
instantaneous discharge of the Marx Generator, indicating that
the resistor attached on the load side provided an effective The empirical breakdown probability curve of the
current limitation. This extinguished the spark and the source measured peak voltages was approximated by a normal
capacitors of the impulse generators discharged through the distribution and the 95% confidence intervals were
source resistors. Hence, the energy input and thus the damage to calculated as well as the standard deviation  and the mean
the electrodes is effectively reduced. This supports the statistical value of the distribution U50.
independence with respect to the electrode surface of the To compare these values with the theoretical breakdown
measurement series. voltage, the latter was calculated by using the ideal streamer
criterion (2.1) as well as the volume-time criterion (2.2).
900
Both were evaluated for different voltages, scaling the
800
electric field values obtained from the FEM-calculations.
voltage [kV]

600 800 Approximations for α and  were taken from [5]. The
400
700 streamer criterion (2.1) was evaluated for kst = 18.5
200
(according to [5]).
0 600
0 50 100 150 1 2 3 4 5 6
In order to calculate the volume-time criterion, the
time [s] time [s]
normalized measured voltage shapes from Figures 4 and 5
Figure 4. Voltage measurement of lightning impulse voltage application on were used to scale the electric field values. The temporal
the model contact system with an amplitude of approximately 900 kV (left)
and its detail in the range of breakdown (right). integration boundaries were set between t = 0 and 300 s,
consequently integrating the whole voltage shape. The
volume Vcr for the space integration consisted of the region
The generation of VFT stressing the contact system was
with (E/p) ≥ (E/p)crit and with d being the minimum
reached by opening the disconnector of the experimental
distance from the electrode surface. This calculation was
setup (cf. Figure 3). The voltage application resulted in a pre-
repeated for the necessary voltage range in order to obtain
ignition across the open disconnector which induced
the breakdown probability curve dependent on the voltage.
travelling waves in the compartment between disconnector
The 50%-value of this curve will be given from now on. The
and specimen. The corresponding measurement is shown in
Figure 5. Two subsequent oscillations were recorded. The electron production rate was calculated using the
first one, starting at 1.2 μs, represented the pre-ignition electron detachment ratio from [19] and calculated ion
across the disconnector contacts, whereas the second density of 1905 IP/cm3 - corresponding to 60 s of ion pair
oscillation at about 1.8 μs was the breakdown across the (IP) generation by radiation at a rate of 32 IP/(cm3 s) at 0.6
circuit breaker contacts. Notably, the recorded peak MPa (6 bar). Latter has been measured in the laboratory,
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 19, No. 3; June 2012 859

which is 10 m below ground level (compared to 26- main contact and then compared to the measured breakdown
55 IP/(cm3 s) at 0.4 MPa (4 bar) in [7]). probability for the combined contact system with either the
To calculate the streamer inception triggered by field same distance between contacts and the plane electrode or
emission on negative polarity the analogue procedure was approximately the same electric field stress as in the single
executed. The relevant changes are that in the volume-time contact experiments. This can be seen in Figure 6 showing
criterion (2.2) the volume was replaced by the electrode schematic drawings of these electrode arrangements together
surface and the electron production rate is given by field with the electric field along the shortest distance between the
emission. For latter the formula from [5] has been used. electrodes. The setups (a)–(c) in Figure 6 have the same
contact distance of 30 mm whereas in setup (d), the arcing
Even though it was measured, the surface roughness is not
contact tip is exposed by 6 mm with respect to the nominal
well defined and choosing protrusions to represent the
contact. Notably, the electric field between the main contact
roughness in the quantitative evaluation of the breakdown
and the plane electrode remain unchanged for setups (b)–(d),
criteria certainly offers some variety. But, the measurement
whereas the field between arcing contact and plane electrode
gives some limits for such protrusions. Further, not all
is significantly different comparing setup (a) with (c). In the
experimental behavior can be reproduced by the appropriate
combined contact setup (d) (equal field arrangement) the
matching choice of such protrusions; with the choice of the
distance was chosen so that the maximum electric fields at
protrusions, the breakdown probability distribution of one
both contacts were comparable with the single contact setups
situation can be reproduced, the consistence with the
(a) and (b).
breakdown behavior of the other situation is then only given
by the criteria. Therefore, such including of surface
roughness, reveals, whether the breakdown criteria and the
30
underlying physics is the appropriate and the decisive for the
breakdown process. This is why the surface roughness is
taken into account in the following by choosing some type of a.) b.) c.) d.)
protrusions.
100
To account for the surface roughness the field strength at arcing contact a.)
E-field, norm [V/m]

short distances r from the electrode is multiplied with the 80 main contact b.) - d.)
arcing contact c.)
factor 60
arcing contact d.)

∙ 1, (2.3) 40

20
matching the field at the tip of ellipsoids (are at the tip of
half of an ellipsoid sitting on a infinite surface) quite well, 0 5 10 15 20 25 30
arc length [mm]
where a is the smaller semi-axis of the ellipsoid and c given
Figure 6. Electrode arrangements and corresponding electric field along
by the relative of both semi-axis. For the following shortest distance between the electrodes, calculated with 1 V potential
calculations (and in agreement with the surface roughness difference: a.) arcing contact and b.) main contact vs. plane electrode. c.)
measurement), a is set to 5 m and c = 8 (representing a and d.) are combined contact systems vs. plane electrode with equal contact
distance as in case of the single contacts c.) and with equal maximum
second semi-axis b ~ 2.5). Further it is assumed that their
electric field d.).
frequency of occurrence is about 0.25%. In the case of
negative polarity, field emission is not triggered at this field
strengths, whereby a further field enhancement by a factor 3 The breakdown probability curves with standard LI
by microprotrusions at the surface of the ellipsoids is voltages are shown for both polarities in Figure 7. Further,
considered in the calculation. the measured 50% breakdown voltages U50 are summarized
in Table 1 together with the calculated breakdown voltages
Ucalc using the described streamer criterion for negative
4 RESULTS polarity and the volume-time criterion for positive polarity.
In the following sections, the results of the breakdown The arrangement a.), arcing contact vs. plane electrode
experiments are presented. Each breakdown probability showed the lowest breakdown voltage with U50 = 485 kV for
curve is the result of 150 - 600 shots. Indicated polarity in negative polarity and U50 = 598 kV for positive polarity. The
the description of the results always refers to the contact side theoretical 50% breakdown voltage calculated with the
with higher field stress (higher field inhomogeneity). streamer criterion and field emission is U50 = 498 kV. The
calculation of the volume-time criterion results in
4.1 FIXED CONTACT SYSTEM WITH LI STRESS U50 = 654 kV.
In order to investigate the breakdown behavior of an The highest breakdown voltages were measured with the
individual contact system, breakdown experiments were arrangement b.), main contact vs. plane electrode having a
carried out using the fixed contact system and a plane 50% breakdown voltage of U50 = 724 kV (negative) and
electrode. The goal was to investigate whether the U50 = 792 kV (positive), respectively. For this configuration,
breakdown behavior (or probability) of both contacts could the theoretical breakdown voltage is calculated to be
be considered individually or not. For this, the breakdown U50 = 731 kV (field emission, i.e. negative) and 817 kV
probability was measured separately for the single arcing and (volume-time, i.e. positive).
860 P. Simka et al.: SF6 High Voltage Circuit Breaker Contact Systems under Lightning Impulse and Very Fast Transient Voltage Stress

In the equal distance arrangement (c), the resulting 686 kV and is thus comparable to configuration (a). At
electric field between the main contact and plane remains negative polarity, all breakdowns occurred on the plug. For
unchanged, whereas the arcing contact is partly shielded by positive polarity, out of the 146 breakdowns (in a series of
the main contact and the maximum electric field on the 208 shots), one occurred on the main contact having a peak
contact tip is reduced by ~ 15 %. Due to the decreased amplitude of 758 kV and a time to breakdown of 4.46 µs.
maximum electric field on the arcing contact in
arrangement (c) compared to (a), a higher breakdown contact system c.) (312 measurements) contact system d.) (208 measurements)
voltage resulted with U50 = 587 kV (negative) and arcing contact (77% - 180 breakdowns) arcing contact (99.3% - 145 breakdowns)
U50 = 774 kV (positive). The theoretical breakdown voltage main contact (23% - 54 breakdowns) main contact (0.7% - 1 breakdown)
is U50 = 596 kV for the plug (field emission) and 734 kV
(volume-time) and 809 kV and 820 kV for the plug and the 80 80

ring respectively. For negative polarity, the breakdown 40 40

probability curves of the individual contact (b) does not

y-axis [mm]

y-axis [mm]
0 0
overlap with that of the combined contact system (c) and all
breakdowns occurred on the plug. This is different for -40 -40

positive polarity. Here, breakdown probability curves (b) -80 -80

and (c) overlap and of the 234 breakdowns (in a series of -100 -50 0 50 100 -100 -50 0 50 100
x-axis [mm] x-axis [mm]
312 shots), 54 breakdowns occurred also on the main
Figure 8. Arc root point distribution on main and arcing contact when tested
contact. Figure 8 (left) shows the location of the arc roots with positive lightning impulse voltage. In the equal distance arrangement
reconstructed from the photographs of the breakdown. It (c), almost a quarter of all breakdowns occurred on the main contact. In the
can be clearly seen that the arc roots are homogeneously equal field arrangement (d), main contact breakdown probability was
distributed across the surface. significantly reduced but not eliminated.

1
Table 1. Measured (U50) and calculated (Ucalc) breakdown voltages of
a.) d.) c.) b.) arrangement (a) – (d) (see Figure 6) for positive (+) and negative (–)
breakdown probability

0.8
polarity. Breakdown voltages are calculated using the ideal streamer
0.6
criterion (negative) or the volume-time criterion (positive). Additionally the
0.4 ratio between calculated and measured values is given.
(–) (+)
0.2
[kV] U50 Ucalc ratio U50 Ucalc ratio
0
1
(a) 485 498 0.97 598 645 0.93
breakdown probability

0.8
a.) d.) (b) 724 731 0.99 792 817 0.97
0.6
b.) (c) arc 596 0.98 809 0.96
0.4 587 774
400 c.) main 734 no BD 820 0.94
0.2
(d) arc 507 1.09 686 1.06
0 557 725
1 main 735 no BD 821 one BD
breakdown probability

0.8

0.6 a.) d.) plug Figure 9 shows the measured time to breakdown values for
0.4
arrangement (a) – (d) and both polarities. It can be seen that
c.) ring
400 c.) plug the time to breakdown is always higher than the time to peak
0.2
of the impulse voltage form, indicating that all breakdowns
0
400 500 600 700 800 900
happened after the peak voltage. The breakdown voltage is
voltage [kV] generally lower for negative polarity, whereas its spread of
Figure 7. Breakdown probability curves for the four electrode arrangements time to breakdown is larger. Further, negative polarity
a.) – d.) according to Figure 6. The electrode system was stressed with generally shows a stronger increase of breakdown voltage
lightning impulse voltage of negative polarity (top) and positive polarity
(middle) on the inhomogeneous contact side. The bottom picture shows the towards smaller times.
calculated breakdown probability curves (positive polarity) for indicated
contacts. 4.2 FULL CONTACT SYSTEM COMPARING
LI WITH VFT VOLTAGE STRESS
As mentioned above, the equal field arrangement (d) was In order to investigate the influence of the voltage shape
chosen to have comparable maximum electric field stresses on the breakdown location, the contact system according to
at the main and arcing contact compared to their single the schematic drawing in Figure 10 was tested. The relative
arrangements. Nonetheless, the measured breakdown distance of the four contacts was chosen to provoke an
probability curves from (d) did not overlap with those of (a) electric field distribution comparable to real high voltage gas
(the lower breakdown voltage of the two single contacts). circuit breakers. On the moving contact side, the arcing
The resulting breakdown voltages were measured to be contact (tulip) is exposed in axial direction by 3 mm
U50 = 557 kV (negative) and U50 = 725 kV (positive), while compared to the main contact. On the fixed contact side, in
the theoretical breakdown voltage is U50 = 507 kV and turn, the plug is withdrawn by 5 mm. Despite this, the
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 19, No. 3; June 2012 861

maximum electric field strength on the plug is ~ 75 % higher polarity showed lower breakdown voltages having
than on the tulip. The fixed contact side is thus the more U50 = 426 kV for lightning impulse and
inhomogeneous one and decisive for breakdown initiation. U50 = 340 kV (prospective) and 394 kV (maximum) if VFT
The theoretical breakdown voltages calculated by the was superimposed. At positive polarity the 50% breakdown
streamer criterion are U50 = 393 kV for the arcing contact voltages were U50 = 588 kV, 530 kV and 574 kV,
and U50 = 647 kV for the main contact. The ratio of the respectively. The calculated breakdown voltage using the
maximum electric field between arcing and main contact streamer integral respecting field emission is 393 kV and
(coordination factor C) was ~1.8 for the fixed contact side 536 kV using the volume-time criterion, assuming the plug is
and ~1 for the moving contact side. Comparing the the decisive electrode.
maximum electric fields along the shortest distance between
1
the electrodes (as in Figure 10, right), the ratio was ~1.66

breakdown probability
and ~0.87 respectively. 0.8
VVFT-peak
0.6
a.) b.)
900 VVFT-prospective
0.4
VLI-peak
800
0.2
700
0
600 300 400 500 600 700
voltage [kV]
500
Figure 11. Breakdown probability curves for complete contact system and
voltage [kV]

400 both polarities, when stressed with lightning impulse voltage shape (VLI-peak)
0 5 10 15 20 25 0 5 10 15 20 25
c.) d.) and with VFT voltage superimposed on lightning impulse (VVFT-peak) see
900 Figure 4 and Figure 5. The amplitude of the prospective LI impulse voltage
800 shape, which excited the VFT voltage, is referred to as
VVFT-prospective.
700

600
When tested with LI negative polarity, all breakdowns
500
occurred on the arcing contact, whereas at positive polarity a
400
0 5 10 15 20 25 0 5 10 15 20 25 distribution of breakdowns on both contacts was detected (cf.
time to BD [s] Figure 12). Main contact breakdown happened 9 times out of
Figure 9. Time to breakdown measurements for electrode arrangements 255 breakdowns (in a series of 375 shots) when tested with
(a) – (d) (see Figure 6).
lightning impulse voltage shape. At the superposition of VFT,
an increase in main contact breakdown was recorded with 63
This geometry was tested with standard LI voltage stress out of 91 breakdowns (in a series of 160 shots). In Figure 12, a
(disconnector closed) and with superimposed VFT concentration of arc root points on one quarter of the ring
(disconnector opened). Although the measurement setup was electrode is visible. A slight misalignment between the fixed
not calibrated for high frequency measurements, an FFT- contact and the plane electrode could be identified as the
Analysis of the measured signals as shown in Figures 4 and 5 cause. Nonetheless, the comparison between LI and VFT
was performed. This analysis revealed that the LI impulse experiments was still valid because no change of the
voltage is confined to a frequency up to approximately experimental setup was made for the entire measurements
500 kHz whereas most of the energy of the VFT signals is series, except for the change of applied voltage shape.
within 0 – 58 MHz with further components up to 156 MHz.
Therefore there is a clear distinction of the two voltage LI (375 measurements) VFT (160 measurements)
waveforms concerning their frequency content.
arcing contact (96.5% - 246 breakdowns) arcing contact (31% - 28 breakdowns)

moving contact side main contact (3.5% - 9 breakdowns) main contact (69% - 63 breakdowns)
120
E-Field, norm [V/m]

100 arcing contacts 80 80

moving contact side

main contacts
fixed contact side

17.1 80 40 40
3 10
y-axis [mm]

16.2
60 0 0
5

40 -40 -40

20 -80 -80
fixed contact side 4 8 12 16
arc length [mm] -100 -50 0 50 100 -100 -50 0 50 100
Figure 10. Complete contact system representing high voltage circuit x-axis [mm] x-axis [mm]

breaker contacts. Different relative contact separations were chosen, Figure 12. Arc root point distribution when testing with positive polarity on
resulting in a low coordination on the moving contact side and a high inhomogeneous contact side. The experiments with lightning impulse show
coordination on the fixed contact side. The electric field calculation was a much lower main contact breakdown probability (3.5%) compared to the
performed with 1 V potential difference between the electrodes. same arrangement tested with VFT-voltage (69%).

The corresponding measured breakdown probability The measured and calculated breakdown voltages are
curves are shown in Figure 11 for LI and VFT voltage of summarized in Table 2. Since each of the four contacts
both polarities. In line with the previous results, negative composing the contact system may have an influence on the
862 P. Simka et al.: SF6 High Voltage Circuit Breaker Contact Systems under Lightning Impulse and Very Fast Transient Voltage Stress

breakdown behavior, the breakdown voltage was calculated Although the calculated breakdown voltages are not
for each contact individually respecting the voltage polarity, precise, an overlap of the breakdown probability curves in
i.e. streamer (neg) and volume-time (pos). For the latter the arrangement (c) and (d) (cf. Figure 7) could be reproduced
corresponding voltage shape (LI or VFT) was used. through the application of the volume-time criterion (small
difference of the voltage values in Table 1). Whereas for
Table 2. Measured and calculated LI and VFT (see Figure 4 and Figure 5)
negative polarity a coordination factor of C = 1.4 and 1.63
breakdown voltages for negative a) and positive b) polarity. Ucalc denotes the for arrangement (c) and (d) was sufficient to prevent main
calculated breakdown voltage using the ideal streamer criterion (s) or the contact breakdown, it failed at positive polarity. Especially
volume-time criterion (v) dependent on whether the resulting stress on the noticeable are the measured breakdown voltages, which are
single electrode is of negative or positive polarity.
a)
larger for both polarities on the equal field arrangement (d)
Negative Lightning Impulse VFT than on the arcing contact arrangement a.). Clearly, only the
U50 U50
maximal field strengths are the same between (a) and (d), but
[kV] U50 Ucalc Ucalc because of the shorter contact separation one would expect
(prosp) (max)
fixed 393 (s) 444 (s) lower breakdown voltages in arrangement (d) at first sight.
One reason for the contrary behavior is certainly the fact
arc

moving 802 (v) 935 (v)

426 340 394 that, apart from the maximal field strengths, the fields are
fixed 647 (s) 731 (s)
generally smaller in the equal field arrangement (d) than in
main

moving 692 (v) 806 (v) arrangement (a) (e.g. due to the shorter distance from the
plug to the plane, the distribution of surface gradients is less
b) homogeneous). Therefore, the surface for field emission
Positive Lightning Impulse VFT
(negative polarity) and the critical volume Vcr (positive
U50 U50 polarity) are smaller in the equal field arrangement d.)
[kV] U50 Ucalc Ucalc
(prosp) (max)
compared to arrangement a.). Indeed, this circumstance is
fixed 536 (v) 626 (v) reflected in the higher calculated positive breakdown voltage
arc

moving 648 (s) 732 (s) from arrangement (d) of 686 kV compared to 645 kV of
588 530 574
fixed 734 (v) 853 (v) arrangement (a) (see Table 1). However, this difference does
main

moving 614 (s) 693 (s) not seem to explain the whole increase of breakdown voltage
from arrangement a.) to the equal field arrangement (d), for
which no complete explanation can yet be given.
5 DISCUSSION Therefore, the initial question – whether the experimental
The measurements presented in this paper indicated that the evaluation of single contact breakdown (with approximately the
major breakdown criterion for low inhomogeneous electrode same fields) is sufficient to predict the breakdown behavior of
arrangements under impulse voltage stress was the the full contact system and to ensure that only arcing contact
occurrence of a first electron. This condition was fulfilled breakdown occurs – can only be answered in the negative.
more easily if the high field electrode was on negative Similarly, in the full contact arrangement (Figure 10) a
polarity. The necessary electric field values to trigger coordination factor of C = 1.81 is sufficient to prevent main
electron field emission are likely to be reached in the present contact breakdown at negative polarity, but not at positive
application, taking the macroscopic electric fields (cf. polarity. Considering the calculated breakdown voltages (cf.
Figures 6 and 10) at the measured breakdown voltages and Table 2) of all four contacts, the measured main contact
the field enhancement due to surface roughness into account. breakdown can be explained as follows: The electric field is
This results in an overall lower breakdown voltage at maximal at the fixed arcing contact, which is stressed with
negative polarity. The time to breakdown curves (Figure 9) positive polarity, calculated to be 536 kV. The electric field
show a minimum required time to breakdown of at the moving (main and arcing) contacts is significantly
approximately 1 µs, but a generally wider spread compared lower, but they are stressed with negative polarity, which has
to positive polarity. It is assumed that an electron is present a lower breakdown voltage of 614 kV and 648 kV,
as soon as the necessary field value for electron emission is respectively. Considering the widths of the breakdown
reached. But the time scales of the negative formative time probability curves, the initiation of the breakdown may thus
lag are larger than those at positive polarity [13]. The lower also be on the moving contact side. This may explain the
electric fields in the case of negative polarity (due to lower measured overall breakdown probability of 3.5% on the main
breakdown voltages), may further increasing the negative contacts, although the maximum electric field at the arcing
formative time lag [10]. At positive polarity, the formative contact is ~70% percent higher.
time-lag is smaller and the time to breakdown is dominated When the same contact system was stressed with VFT
by the statistical time-lag. voltage, a dramatic decrease in breakdown voltage and an
Consequently, the breakdown voltages, i.e. breakdown increase in main contact breakdowns could be observed. It
probability curves, can be calculated by applying the seemed likely that the increased voltage stress at the contacts
streamer criterion for negative polarity and the volume-time due to constructive reflection of the incident travelling wave
criterion for positive polarity. In the present application this and the additional heating of the streamer region by the high-
could explain the difference between measurement and frequency mechanism was responsible for the decrease in
prediction. (prospective) breakdown voltage (cf. Figure 11).
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 19, No. 3; June 2012 863

The frequency content of the voltage signal seems sufficient implement proper coordination factors on both sides of the
to trigger the high frequency breakdown mechanism [15]. breaker contacts (moving and fixed side). Coordination
However, a rough estimation of the stray capacitances of factors being implemented only on one side, make the
arcing and main contact predicted only small differences breaker prone to positive over voltages applied on the
between them, making the high-frequency mechanism an coordinated side, as the contacts on the other side of the
unlikely explanation for the increase in main contact breaker, being on negative polarity, tend to lower
breakdown probability. breakdown voltages.
If the volume-time criterion is applied on VFT-voltages, a - Further it is not possible to conclude from breakdown
strong rise in breakdown voltages of the whole contact probabilities for a certain voltage waveform to the
system results from the calculations (Table 2). Due to the breakdown probabilities of another voltage waveform.
fast oscillating voltage form (with lower voltages in the Therefore, the occurrence of VFT voltages, potentially the
average than with pure LI), the occurrence of a first electron case in circuit breakers with more than one interrupter unit
initiating a discharge becomes less likely. Again, this fact in series where pre- or re-ignition excites travelling waves
applies only to the fixed contact side (being on positive in the breaker, may further accentuate coordination
polarity). Similarly to the explanation of main contact difficulties.
breakdown for the positive LI experiments above, the - Clearly, the results cannot be directly generalized to
breakdown values have to be compared with the moving contact systems of real breakers, with problems such as
main contact on negative polarity, whereas the difference contact finger edges (with less surface for field emission or
between the two breakdown values decreased (626 kV and the volume-time-criterion), roughened surfaces after arcing
693 kV). or by a macroscopic change (i.e. shortening by arcing
In the present investigations, after each voltage application, reducing the field stress). However, it has been shown that
an ac voltage was applied on the contact systems, followed there are effects which require an analysis beyond the
by a fixed time period to the next voltage application, to macroscopic field simulation or the measurement of
ensure a comparable ion density distribution between the individual contacts only, when coordinating new designs.
contacts in all experiments. This is not equivalent to the IP The developer has to take this into account by using the
density in equilibrium, reached after break times of several recipe shown in this paper.
minutes [11, 13]. Obviously, the IP density influences (in the
case of volume-time-criterion, i.e. positive polarity) the ACKNOWLEDGMENT
breakdown probability decisively; the higher the ion
densities, the lower are the breakdown voltages. In nominal Financial support of the project from ABB Switzerland,
service of circuit breakers low IP densities are to be Ltd. (Corporate Research and High Voltage Products) is
expected. gratefully acknowledged. Some of the measurements were
supported by students in the framework of their semester or
master thesis.
6 CONCLUSION
Even though the investigated specific contact REFERENCES
arrangements does not permit conclusion to generality, the [1] F. Pinnekamp, “The circuit breaker”, ABB Review, Vol. 1, pp. 75-78,
understanding of breakdown behavior of low inhomogeneous 2007.
[2] C. Flurscheim, Power Circuit Breaker Theory and Design, 2nd ed.
electrode arrangements in SF6 could be extended through the Peter Peregrinus Ltd., Chapter 1, 1982.
results presented in this work: [3] P. Simka, Dielectric Coordination of High Voltage Gas Circuit
Breakers, Ph.D. Thesis, ETH Zurich Switzerland, 2011.
- Clearly, dielectric coordination based on defining a [4] M. Beyer, W. Boeck, K. Möller and W. Zaengl,
maximal surface gradient is not sufficient if dimensioning Hochspannungstechnik, Springer Verlag, 1986, Chapter 4.
high voltage circuit breakers toward minimal size and [5] W. Mosch and W. Hauschild, Hochspannungsisolierungen mit
coordination factor. Schwefelhexafluorid, Dr. Alfred Hüthig Verlag, Heidelberg, ISBN 3-
7785-0540-8, Chapters 2 and 4, 1979.
- The breakdown voltages calculations based on streamer [6] W. Boeck and J. Kindersberger, “Determination of the statistical time
integral and volume-time criteria yield convincing results, lag in SF6“, 4th Int’l. Sympos. High Voltage Eng. (ISH), Athens,
especially in connection with coordinating multi-electrode Greece, Paper 31.06, pp. 1-4, 1983.
arrangements. However, the calculated values are too large [7] N. Wiegart, L. Niemeyer, F. Pinnekamp, W. Boeck, J. Kinderberger,
R. Morrow, W. Zaengl, M. Zwicky, I. Gallimberti and S. Boggs,
when compared with the experiment. Further, the measured “Inhomogeneous Field Breakdown in GIS – Part II”, IEEE Trans.
difference in breakdown voltage from single contact to Power Del., Vol. 3, pp. 931-938, 1988
multi-contact measurements with approximately the same [8] W. Boeck, “SF6 insulation breakdown behavior under impulse stress”,
field strengths can only be partially reproduced by the in K. Ragaller (ed.), Surges in High-Voltage networks, Plenum Press,
New York, USA, 1979.
calculation methods. [9] E. Kuffel, W.S. Zaengl and J.Kuffel, High Voltage Engineering
- Therefore, it must be concluded that high voltage circuit Fundamentals, 2nd Ed. Elsevier, ISBN 978-0-7506-3634-6, Chapter 5,
breaker contact systems should be tested as a whole and 2000.
not through extrapolation from single contact experiments. [10] M. Seeger, L. Niemeyer and M. Bujotzek, “Partial discharges and
breakdown at protrusions in uniform background fields in SF6”, J.
- Particular attention has to be paid to the dependence of Phys. D: Appl. Phys., Vol. 41, 185204, 2008.
breakdown voltages on the polarity. With regard to high [11] J. Kindersberger, The Statistical Time-lag to Discharge Inception in
voltage circuit breaker design, it is recommended to SF6, Ph.D. Thesis, TU Munich, Germany, 1980.
864 P. Simka et al.: SF6 High Voltage Circuit Breaker Contact Systems under Lightning Impulse and Very Fast Transient Voltage Stress
[12] W. Boeck, “Volumen-Zeit-Gesetz beim Stossspannungsdurchschlag Philipp Simka was born in Chur, Switzerland in 1978.
von SF6”, ETZ-A Vol. 96, No. 7, 300-305, 1975. He received the M.Sc. degree from the Department of
[13] L. Niemeyer, L. Ullrich and N. Wiegart, ”The mechanism of leader Electrical Engineering and Information Technology,
breakdown in electronegative gases”, IEEE Trans. Electr. Insul., ETH Zurich, Switzerland in 2005 and is currently doing
Vol. 24, pp. 309-324, 1989. his Ph.D. degree at the Institute for Power Systems and
[14] H. Hiesinger, ”Leader breakdown for inhomogeneous fields in case High Voltage Technology, ETH Zurich, Switzerland.
of VFT conditions“, 7th Int’l. Sympos. High Voltage Eng. (ISH), His research interests cover dielectric phenomena, high
Dresden, Germany, pp.67-70, 1991. voltage test techniques and high frequency phenomena
[15] H. Hiesinger, Der Hochfrequenz-Durchschlagsmechanismus in SF6 associated with high voltage equipment.
bei Schwingenden Transienten Überspannungen, Ph.D. Thesis, TU
Munich, Germany, 1991. Ueli Straumann (M’10) was born in Aarau,
[16] “IEC 60060-1, High-voltage test techniques – Part 1: General Switzerland in 1975. He received a diploma in
definitions and test requirements”, International Electrical theoretical physics from the University of Zurich,
Committee IEC, Geneva, 2006. Switzerland in 2001 and the Ph.D. degree from ETH
[17] W. Hauschild and W. Mosch, Statistical Techniques for High- Zurich, Switzerland, in 2007. Since then he has been
voltage Engineering, Peter Peregrinus Ltd., Vol. 1, Chapters 2 and 5, Senior Assistant and Lecturer in the High Voltage
1992. Laboratory of ETH Zurich. His research interests
[18] P. Simka nd C. M. Franck, “Measuring breakdown probabilities of include corona of OHL, the description of ion flow
SF6 insulated, low inhomogeneous electrodes”, XVIIIth Int’l. Conf. fields and gaseous insulation in general.
Gas Discharges and Their Application, Greifswald, Germany, pp.
610-613, 2010.
[19] X. Xu, S. Jayaram and S.A. Boggs, “Prediction of Breakdown in Christian M. Franck (M’04-SM’11) received a diploma in
SF6 under Impulse Conditions”, IEEE Trans. Dielectr. Electr. Insul., physics from the University of Kiel, Germany in 1999 and
Vol. 3, pp. 836-842, 1996. the Ph.D. degree in physics from the University of
[20] P. Simka, “An Approach to Model Very Fast Transients in High Greifswald, Germany in 2003. He was with the Swiss
Voltage Circuit Breakers“, IEEE Int’l. Sympos. Electr. Insul. (ISEI), corporate research center of ABB from 2003-2009 as a
pp. 449-452, 2008. Scientist and Group Leader for gas circuit breakers and high-
[21] P. Simka, ”A Complete Circuit Breaker Model for Calculating Very voltage systems. Currently, he is an /Assistant Professor for
Fast Transient Voltages“, IEEE Int’l. Sympos. Elecr. Insul. (ISEI), High Voltage Technology at the Swiss Federal Institute of
pp. 1 – 5, 2010. Technology (ETH), Zurich, Switzerland.