WCA May 2017

56 / 64

54 Wire & Cable ASIA – May/June 2017
Thereby channel 1 (Ch1, blue) shows the signal reflections
when the spark gap is connected at the far end of
both cables and channel 2 (Ch2, red) shows the signal
reflections when the spark gap is connected to the
connection point between the cables.
The upper diagram is the complete signal recording over
about 300µs. In the middle diagram the first and the
second reflection are zoomed out. In the lower diagram the
differentiated curves are shown with Ch11 related to Ch1
and Ch12 related to Ch2.
From this measurement the propagation velocity is
determined to v = 172.5m/µs based on T = 17.0µs of Ch1
and according to Equation 2. Now the Tx = 8.79µs of Ch2
indicates exactly the length of the cable sample of 758m.
Assuming an uncertainty of ±0.2µs of the time evaluation
for both full length and partial length, the following cable
lengths to failure can be estimated.
Based on the determined cable length of 758m the
maximum deviation is 11m, which is 0.75 per cent of the
full cable length.
Furthermore, the measured signal shows a significant
decline. This comes from the damping of the cable itself
and from its dispersion.
Comparison of the waveforms in Ch1 and Ch2 show that
the reflection losses are also a substantial part of the cable
losses, because the decrease of the voltage as a function
of the number of reflections is more or less constant.
After this initial test the same measurements with an
undamped capacitive divider were carried out. The goal
was to find out if it is possible to get usable results of fault
location even with a voltage divider with a lower bandwidth
( Figure 6).
Figure 8 shows the results of a measurement with a
divider type WCF normally used in resonant test systems
for cable tests. It is clear to see that such a divider is
actually not suitable for such fast transient measurements.
Nevertheless, there is still a possibility to evaluate a fault
position.
In the lower diagram of Figure 8 the curves are filtered
with a numerical low-pass Bessel filter to find the
transition points of the reflection. Assuming a well-known
propagation speed (172.5m/µs) the fault can be located at
759m. But it is clear that the uncertainty of determination is
much higher than before.
A second test with the same divider was performed, but
this time the divider type WCF was damped with a resistor
of 150Ω.
❍
❍ Figure 10: DC cable, detail spark gap and attenuator
It is shown that the damping resistor eliminates the
majority of the oscillations after the transition in the
waveform. Therefore, a further filtering is not necessary
for the evaluation. As before, the fault can be located with
the well-known propagation velocity: the result of the
calculation is 758m.
DC cable (PE (for DC), > 100kV)
The test configuration consisted of one cable on a
turntable. The cable was connected to an adjustable DC
source.
The breakdown test was performed by using a spark gap
at the far end of the cable (Figure 10).
The voltage was increased until the spark gap got fired.
The resulting travelling waves were recorded.
❍
❍ Figure 9: Measurement with divider type WCF, damped with
150Ω
❍
❍ Figure 8: Measurement with divider type WCF, undamped
❍
❍ Table 2: Calculated cable lengths for different signal
propagation times
T partial length [µs]
8.77 8.79 8.81
T full length [µs] v [m/µs] calculated length [m]
16.8 170.5 748 749 751
17 172.5 756 758 760
17.2 174.5 765 767 769