Full resolution (JPEG) - On this page / på denna sida - 1958, H. 5 - Flashover Characteristics of Non-uniform Spark Gaps with Standard 1×50 Impulses, by Jörgen Fryxell
<< prev. page << föreg. sida << >> nästa sida >> next page >>
Below is the raw OCR text
from the above scanned image.
Do you see an error? Proofread the page now!
Här nedan syns maskintolkade texten från faksimilbilden ovan.
Ser du något fel? Korrekturläs sidan nu!
This page has never been proofread. / Denna sida har aldrig korrekturlästs.
Fig. 4 a—h. The flashover distribution of horizontal 450
mm rod-rod gap with 900 impulses (1 X 50) per
voltage level (uncorrected values),
a—b. Front capacitor 200 pF■ Free arrangement.
c—d. Front capacitor 200 pF■ Earthed plates near
the test object.
e—/. Front capacitor 800 pF. Free arrangement.
<g—h. Front capacitor 800 pF. Earthed plates near
the test object.
presented in a more conventional way. The result
from each test series consisting of 100 impulses per
voltage level has been plotted on probability paper
and the 50 % value has been determined in the usual
manner without reference to the S-curve. The nine
50 % flashover values obtained in this way have
been dealt with as a statistical variable and plotted
on probability paper in fig. 5. For comparison the
values for similar gaps measured by 14 European
laboratories and presented in a similar way by
Berger1 are shown in the same figure.
In the Gaussian distribution case the conventional
way of treatment in fig. 5 can be expected to give
the same numerical results as the values presented
in table 2 and inspection shows this to be so for
positive polarity. For a non-Gaussian distribution
with negative polarity, however, the conventional
treatment of fig. 5 gives a more peculiar picture. For
instance earthed plates increase the 50 % value
by 6 % with a front capacitor of 800 pF while they
decrease the value by 11 % with a front capacitor
of 200 pF. This can be explained from fig. 4 through
the influence of different factors, such as
surroundings and component values of the circuit, moving
the most horizontal part of the S-region up or down
to higher or lower probability values.
Discussion
If it is assumed that the standard deviation of the
flashover distribution of the test object is the only
factor influencing the results and if all flashover
curves are Gaussian distributed it can be shown2
that all 50 % values in a number of series of 10
impulses per voltage level or more would fall within
a range less than ± 0.75 s (95 % confidence band).
The value of s seems to be about 3—4 % in the
Gaussian range, at least for spacing between 50—
600 mm. Therefore all 50 % values should fall within
± 2—3 %. A reasonable addition of uncertainty
owing to minor inaccuracies in the measurement of
test voltage or inaccuracies due to the correction for
humidity, etc. would extend the possible range of
50 % values to say ± 4 to 5 % as a maximum. The
presence of the S-curve may extend the range further.
Reference is now made to table 3 which is a
summary of different determinations of the flashover
value of a 450 mm rod-rod gap for positive and
negative polarity. In the first horizontal row the values
ELTEKNIK 1958 1 1 9
<< prev. page << föreg. sida << >> nästa sida >> next page >>