Dr.

Qais Alsafasfeh

Electrical Breakdown of Gases in Uniform Fields

Introduction
At normal temperature and pressure, gases are excellent insulators but background
currents of the order of micro-amps can be measured if an electric field of several
kV/mm is applied. This current results from the electron/ion pairs produced by high-
energy particles, either cosmic rays or derived from natural radioactivity, striking an
air molecule:
high-energy particle + M ==> M
+
+ e
If the voltage is increased sufficiently, the electron is accelerated by the electric field
towards the positive electrode (or anode) and further ionisation can occur. The
electron will collide with gas molecules and most of these will be elastic collisions,
but, if it has gained enough kinetic energy (KE), it will ionise the gas molecule it hits:
e + M ==> M
+
+ e + e.
Enough energy means energy greater than the ionisation energy of the molecule.
The process of acceleration until a collision with a molecule occurs, with most of the
collisions elastic, and some inelastic (i.e., ionising) is illustrated in the AVAL-1.exe
program. The KE gained by the electron is (electric field)*(distance travelled before
the next collision) see box below.
Now there are two electrons and the process can repeat, and repeat, and repeat,
causing an exponential increase in the number of electrons. The situation after 4 such
sets of ionisations by accelerated electrons is illustrated in the AVAL-2.exe program
from which the diagram below is taken. (The original high-energy particle ionisation
occurs at A subsequent ionisations are caused by the accelerated electrons).

There are 1+1+2+4+8 = 16 positive ions
and 16 electrons here
Note that while the electron is accelerated towards the anode, the positively-charged
ion is obviously accelerated towards the cathode. However the ion, being far
heavier, is accelerated more slowly: the average velocity of the electrons is about ten
times faster than that for the ions. This causes the situation seen above with the
electrons moving swiftly to the right in a group, leaving clumps of 1, 2, 4 and 8 ions
behind. This is seen more clearly in the AVAL-4.exe program in which 10 sets of
ionisations have occurred (so there are 2
10
, or 1024, electrons and the same number of
positive ions).




1.1
Kinetic energy gained by the electron = work done on the electron
= (force on the electron) * (distance travelled by the electron)
= (e.E)*x i.e., proportional to both the electric field and the distance gone.
Dr.Qais Alsafasfeh
The electron will normally have many elastic (low-energy) collisions before the
ionising collision hence the crooked paths seen in the animations. Because of the
random nature of the number of collisions before the ionising collision, the distance
between ionisations is also variable again, as seen in the animations.
Avalanches
In the AVAL-4 diagram below there are 1024 electrons (and the same number of
positive ions), the electrons and ions being indistinguishable in black and white
reproduction. The distribution graphs for the two charged particles are shown, and
explain the comet-like shape of the avalanche, as this phenomenon is called.
Note the overlapping of the two graphs: this means that there will be a number of
electron/positive-ion collisions which may result in recombination:
e + M
+
==> M + energy
This recombination energy is usually released as a photon of light energy.













(a) (b)
Cloud chamber photography of single avalanches (a) in nitrogen (N
2
) at 0.37 bar
and (b) in carbon dioxide (C0
2
), both in a 36-mm gap.
The voltage was a DC voltage pulse lasting 0.4 ms.
[H. Raether, Electron Avalanches and Breakdown in Gases, 1964, p.5]
1.2
positive
ions
electrons
Dr.Qais Alsafasfeh
The ionisation coefficients
The ionisation coefficient, , is defined as the probability that an electron will make
an ionising collision in travelling unit distance in the direction of the anode.
In addition there is a possibility (especially for slower-moving electrons) that an
attaching collision takes place:
e + M ==> M

.
The attachment coefficient, , is defined as the probability, per unit distance travelled
in the direction of the anode, that an electron will attach to a molecule to form a
negative ion.
As mentioned earlier, the kinetic energy (KE) gained between collisions needs to
exceed the energy required to ionise the molecule. The distance between collisions is
inversely proportional to the density and hence to the pressure so it should not be
surprising (see box) that
/p = f(E/p).
For similar reasons it is found that
/p = g(E/p).
For simplicity, an effective ionisation coefficient, , is defined as
=
Many text books use the empirical equation
/p = 1100(exp{-27.4 E/p})
for air but it is very approximate. Better ones are available but are generally more
troublesome to apply (see, for example, MacAlpine & Li, IEEE Trans.D&EI, Vol.7,
pp.752-757, 2000). Here, and in general in this course, the units are assumed to be
mm, kV and bar or their combinations.









Clearly, if attachment is more likely
than ionisation when a collision
occurs, or >, avalanches cannot
develop.
In nitrogen the attachment
coefficient, , is negligible; in
oxygen it is very small. So, in this
graph of /p versus E/p for air it is
only at low fields, below 2.3 kV/mm
bar, that > (i.e., is negative).
Even then it is only just below zero.
The field at which = is called the
Critical Field.
1.3
-5
0
5
10
15
20
25
0 1 2 3 4 5 6
E/p (kV/mm.bar)
a
l
p
h
a
/
p

(
1
/
m
m
.
b
a
r
)
Li-MacA
Geballe & H
Prasad
Morruzzi & P
The energy gained between collisions = e.E. = eE/p
because (= mean free path between collisions) is proportional to 1/p (p = pressure)
The probability of a collision resulting in ionisation is a function of the energy gained between
collisions, that is, from the above, a function of E/p, say F(E/p).
The number of collisions (any kind) per unit distance = 1/ = Ap (A is a constant)
the number of ionising collisions, , is F(E/p)*Ap,
or, /p = f(E/p)
Dr.Qais Alsafasfeh
Avalanche calculations
The average size of an avalanche may be
calculated for uniform-field conditions by
considering the number of electrons n passing
through a plane at a distance x from the cathode
in the direction of the electric field (towards the
anode) in a time t. Simple integration (see
box) gives
n(x) = e
x

How big is an avalanche? Consider an electron
produced at x = 0 in a 10-mm gap (e.g. as an
electron-ion pair due to a cosmic ray, or by
emission from the electrode) in air at
atmospheric pressure.
If the applied voltage is 25 kV, E = 2.5 kV/mm,
and is found (from the graph on the previous
page) to be close to zero.
Now try 30 kV: = 1.3 mm
-1

the number of electrons in the head of the
avalanche when it strikes the anode will be
n(10mm) = exp(13) = 4.4x10
5

Now try 35 kV n(10mm) = exp(30) = 1.1x10
13
. This is a huge increase!
Clearly if the current (the sum of all the electrons in all the avalanches which occur in
a second) increases at this rate, breakdown MUST occur near this voltage.
The reverse calculation: what is the voltage which gives, say, 10
8
electrons in the
head of the avalanche when it strikes the anode?
n(10mm) = exp(10.) = 10
8

Therefore = 1.84 mm
-1
,
So, V = 32 kV.

Avalanches in non-uniform fields
In a non-uniform field, the same approach as used above gives
n(x) = exp( (E).dx )
This is useful in for example, coaxial cable or busbar systems see next lecure.

Sulphur hexafluoride, SF
6
This is a very important insulating gas and is widely used in equipment for electrical
power transmission and distribution. It is colourless, odourless and heavy.
For SF
6
the critical field is at 8.85 kV/mm as shown in the graph overleaf and
the attachment coefficient is much larger than that for oxygen.
From the right-hand graph, an empirical expression for the effective ionisation
coefficient, or (-), may be obtained as
/p = 26E - 230.
1.4
The average size of an avalanche
Consider the number of electrons n
passing through a plane at a distance x
from the cathode in the direction of the
electric field (towards the anode) in a
time t.
The number of electrons passing through
a plane at a distance x + dx may be
written as n + dn (again in a time t)
where,
dn = n.dx,
using the definition of ,
so that, integrating between limits of x
= 0 and x, and remembering that each
avalanche is started by a single electron,
n(x) x
dn/n = .dx,
1 0
or, n(x) = e
x

Unless the field is non-uniform, varying
with x, which means varies with x,
so n(x) = exp( .dx).


Dr.Qais Alsafasfeh
60
70
80
90
100
110
120
130
140
8 9 10 11
E/p, kV/mm.bar
I
o
n
i
s
a
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t
s
,

1
/
m
m
.
b
a
r
-50
0
50
100
150
200
8 10 12 14 16
E/p kV/mm.bar
E
f
f
e
c
t
i
v
e

a
l
p
h
a

/

p
,


1
/
(
m
m
.
b
a
r
)

This formula gives the critical field as E
c
= 230/26 = 8.85 kV/mm.
How big is an SF
6
avalanche? Consider an electron produced at x=0 in a 10-mm gap,
as previously, but in SF
6
at atmospheric pressure. If the applied voltage is 88.5, 89.0
and 90.0 kV, then, from the formula, = 0, 1.4 and 4.0, respectively, and the number
of electrons in the head of the avalanche is n(10mm) = 1, 1.2 x 10
6
and 2.35 x 10
17
!
Clearly the size of an SF
6
avalanche increases hugely within 1 to 2% above the
critical field. But what is the criterion for breakdown?

The Streamer Theory of electrical breakdown in gases
The electrons in the head of the avalanche increase the field ahead of it. Similarly, the
positive ions, particularly those in the high concentration zone just behind the head,
increase the field in the tail. When the avalanche is small, this is an insignificant
effect, but as the avalanche grows there must be a critical size when it is the
concentration of charge is great enough for the field to be increased to twice the
original gap field.
Both Raether (in Germany) and Meek (in England) came to the conclusion that when
there were about 10
8
electrons in the avalanche head, the field due to the avalanche
itself could equal the main field, and that this would lead to the development of a
channel of ionised conducting gas. Meek called this a streamer (Raether called it a
kanal, the German word for channel).
The mechanism they proposed was that in the region where the electrons and positive
ions overlap (see diagram on page 1.2), recombination would take place:
e + M
+
==> M + photon
This region will therefore be a source of photons which would speed off (at the
velocity of light) in all directions. When they strike molecules, photo-ionisation - the
ionisation of molecules by photon impact can occur:
photon + M ==> M
+
+ e
These photo-electrons initiate new avalanches: those in the higher-field zones ahead
and behind the mother avalanche, will grow even faster than the mother avalanche
did because the value of increases very quickly above the critical field. Between 4
and 8 kV/mm in air increases by a factor of 10; in SF
6
the situation is even more
spectacular as breakdown tends to be close (within 1 or 2% - see above) to the critical
field.
1.5

-
Dr.Qais Alsafasfeh
They would produce further avalanches ahead and behind, thus producing a column
of conducting ionisation. Consider electrons produced by photo-ionisation at A B C,
D and E in the figure below. At A the field is perhaps 5 times in the gap remote from
the avalanche a baby avalanche will develop far faster than the original or
mother avalanche owards the anode. Even at B the field is perhaps twice the gap
field so the baby avalanche will still develop much faster than the mother
avalanche. At C the field is zero nothing will happen. At D the field is again about
twice the gap field so the baby avalanche will still develop much faster than the
mother avalanche, and again in the anode direction (or, more precisely, towards the
maximum density of positive ions). At E the field is again about twice the gap field,
but in the opposite direction, so the baby avalanche will still develop much faster
than the mother avalanche, but in the cathode direction.
All this development of avalanches of similar size in electron numbers, though
occupying a smaller space, occurs immediately ahead of the head of the avalanche,
behind it in the tail region, and close besides the dipole area (in the mother
avalanche near E) thus extremely quickly, almost explosively, forming a column of
avalanches bridging of the gap between the electrodes (see the STREAMER
program). This column of free charges (positive ions and electrons) is conductive so
current flows. Positive feedback occurs as, while the field is maintained, more and
more avalanches occur, the number of charged particles increases, the current
increases until the source impedance limits the current to the short-circuit current,
thus dropping the voltage across the gap to near zero. This defines electrical
breakdown.






The electric field
around a large
avalanche







The critical size for an avalanche to transform into a streamer for air is usually taken
as 10
8
, which is therefore the breakdown criterion for uniform-filed gaps. It is easy to
demonstrate that taking the critical size as 10
8
or, say, 2x10
8
makes a change to the
calculated breakdown voltage which is usually negligible in practical terms.
For SF
6
, many workers have suggested that 10
7
gives better agreement with
experiment.
Taking N = 10
8
= e
x
= e
K
, gives K = 18.4, an alternative criterion.
For N = 10
7
, K = 16.1.
1.6
B
A
C
D
E
Dr.Qais Alsafasfeh
Paschen's Law V
bdn
is a function of pd
This was discovered experimentally by a German scientist, F. Paschen, who published
it in 1889. It may be derived as follows: breakdown occurs when
exp(d) = 10
8
, or d = K
/p = f(E/p) = K/pd
Hence, V
bdn
= F(pd) or Paschen's Law.



A log/log graph of the
breakdown voltage of
nitrogen against pd.
Gallagher & Piermain,
page 49




Note: 1 atm.cm = 1.013 bar.cm = 10.13 bar.mm = 1013 kPa.mm
The graph shown is for nitrogen, but similar shapes are found for all gases.
Note that for pd > 100 bar.mm (= 10 atm.cm in the figure) Paschens Law fails. This
is because of the field distortions around small imperfections, and dust particles, on
the electrode surface. This will be covered in Unit 2 on breakdown in compressed
gases.
Similarly, for a near vacuum, pd < 10
-3
bar.mm (= 10
-4
atm.cm in the figure),
Paschens Law also fails. This is because the number of mean free paths in the
distance from cathode to anode becomes too low for it to be possible for avalanches to
develop. This will also be covered in Unit 2.
The so-called Paschen minimum can be of importance for electrical circuits on, for
example, printed circuit boards in space or near-space conditions. For power
engineering it is seldom of direct interest. Vacuum circuit-breakers operate on the
left-hand part of the curve. Atmospheric-pressure air as insulation involves
dimensions of the order of metres (e.g., overhead lines). SF
6
is normally used at a
pressure of 4 to 5 bar and dimensions in the region of 10s of mm. Do the
calculations yourself.



The breakdown voltages at the
Paschen minima for various gases
Gallagher & Piermain, page 49






1.7
Dr.Qais Alsafasfeh


1. Electrical Breakdown of Gases in Quasi-Uniform Fields
Quasi-Uniform Fields
By this meant an electrode gap where the ratio of the maximum and minimum fields
is less than about 10. Common examples are electrode systems where one electrode is
a rod and the other is a tube and both are coaxial.
The avalanche size is given by
n(x) = exp( (E).dx )
(See page 1.4). If is known as a function of E ,and E is known as a function of x,
the avalanche size can be determined. If analytical expressions are not known for
these functions, or cannot be integrated, numerical integration can be used.
See Example quasi-uniform fields.pdf.

2. Electrical Breakdown of Compressed Gases in Uniform Fields
Introduction
The electrical breakdown strength of gases increases with pressure, as may be seen
from the Paschen's Law graph (page 1.6). Why? As the pressure increases the mfp
decreases and so a higher electric field is required in order that the electrons may gain
sufficient kinetic energy (k.e.) between collisions to cause ionisation. However, it is
found that Paschen's Law no longer applies for pressures above about 5 bar, unless
great care is taken
to limit the emission current from the cathode and
to avoid the presence of dust, particularly conducting particles.
The higher the pressure the harder it becomes to avoid the effects of cathode emission
currents and dust.

Electron Emission Currents
The graph at the right shows the very high
currents caused by field-enhanced electron
emission. The 3 graphs refer to plane-
parallel electrodes:
the electrodes just after being
manufactured and polished (top
graph),
the same after several breakdowns
(middle graph), and
the same with a thin (0.1mm) layer of
polythene stuck to the surfaces (lowest
graph).


2.1
Dr.Qais Alsafasfeh
There are two mechanisms by which such emission can occur: (a) Schottky and (b)
Fowler-Nordheim emission.
Simply explained, Schottky emission is the effect of a high field in reducing the work
function, that is, the energy which electrons need in order to leave the metal; and
Fowler-Nordheim emission is a quantum mechanical effect by which electrons can
tunnel through a thin potential wall.
The latter requires virtually perfectly clean surfaces and is therefore only possible in
near vacuum conditions. In all practical cases, therefore, only Schottky emission or
field-enhanced thermal emission occurs.
Current density, J = AT
2
. exp(-BE/T)
where E is the electric field at the surface, T is the temperature in degrees Kelvin and
A & B are constants. It requires relatively high electric fields at the surface, before
significant emission occurs, and so does not greatly affect breakdown at atmospheric
pressure.


The
breakdown of
nitrogen at 15
bar and a 2mm
electrode
separation
0
1
2
3
4
2
8
2
9
3
0
3
1
3
2
3
3
3
4
3
5
3
6
3
7
3
8
3
9
4
0
Impulse breakdown field, kV/mm
N
u
m
b
e
r

o
f

b
r
e
a
k
d
o
w
n
s
Clean
Dusty


The two histogram figures demonstrate the effect of dust on the impulse strength of
nitrogen (above) and SF
6
(next page) at 15 atm - a lowering of the breakdown strength
by some 30 to 50%, even with polythene-covered electrodes stopping electron
emission.
Clearly imperfections such as emission currents and dust are the cause of the
breakdown voltages for high pressure gases being lower than expected from
Paschens Law.
Why does dust cause problems, and increasingly so as the pressure increases? Why
do emission currents cause problems, and what causes them?

The
breakdown of
SF
6
at 15 bar
and a 2mm
electrode
separation
0
0.5
1
1.5
2
2.5
3
3.5
7
0
7
5
8
0
8
5
9
0
9
5
1
0
0
1
0
5
1
1
0
1
1
5
1
2
0
1
2
5
1
3
0
1
3
5
1
4
0
Impulse breakdown field, kV/mm
N
u
m
b
e
r

o
f

b
r
e
a
k
d
o
w
n
s
Clean
Dusty


2.2
Dr.Qais Alsafasfeh
Why does dust cause a problem?
This illustration is merely for interest and for those of a mathematical bent.
The answer lies in the exponential relationship between and E, as the following example will show.
Consider a conducting particle of 10-m size attached to one electrode: suppose, for simplicity, it
enhances the field by a factor of 5 for the first 5 m, 3 for the next 5 m and 2 for a further 5 m. First,
the breakdown field for the case without a particle is calculated, using simple equations for (E) and
assuming breakdown occurs for d=18.4 since this is the criteria for streamer formation across the
gap.
Then, for 4 different pressures and some value of E, the values of .x for these 3 regions and the rest
of the gap are calculated. By trial and error the value of E is found which makes (.x) = 18.4 (I
used Excel).
The results are tabulated below. The value of E with a particle present is compared with the value for a
clean gap without any particles present and it is seen that the effect of particles is very small at
atmospheric pressure but rapidly becomes very significant.

Air pressure (in bar) 1 5 10 15
/p = 18.4/(p*10*1100) = 1.84 0.368 0.184 0.1227
ln(/p1100) = -27.4*p/E = -6.393 -8.003 -8.696 -9.101
without particle, E (in kV/mm) = 4.29 17.12 31.51 45.16
By trial and error: main gap field, with
particle =

4.20

16.15

28.61

39.54
First 5 m: .x = 1.49 5.04 8.100 10.32
Second 5 m: .x = 0.62 1.63 2.26 2.58
Third 5 m: .x = 0.21 0.40 0.46 0.46
Rest of gap, .10 = 16.07 11.37 7.61 5.04
Sum across gap of all these .x = 18.40 18.43 18.43 18.40

E(with particle)/E(without) = 0.980 0.943 0.908 0.876
Percentage strength reduction = 2.1 5.7 9.2 12.4

In case this is not clear, the full calculation is given overleaf for the 5 bar case. (N.B., as usual, all
quantities are in units of kV, mm, bar and their combinations.
For the uniform-field case (no particle), at 5 bar,
exp(d) = 10
8
, i.e., d = 18.4, so = 1.84 /mm
But /p = 1100 exp(-27.4 p/E)
So 1.84/5 = 1100 exp(-27.4 * 5/E)
E = 17.12 kV/mm
For the uniform-field case with a particle, at 5 bar, using E = 16.15 kV/mm,
exp(.dx) = 10
8
, i.e., d = 18.4 .x
Since = 5*1100*exp(-27.4 p/E)

1
.x = 5*1100*exp(-27.4 p/5E) *5*10
-3
= 5.04

2
.x = 5*1100*exp(-27.4 p/3E) *5*10
-3
= 1.63

3
.x = 5*1100*exp(-27.4 p/2E) *5*10
-3
= 0.40

4
.x = 5*1100*exp(-27.4 p/E) *9.975

= 11.37
Sum, .x = 18.4


2.3
Dr.Qais Alsafasfeh
Repeat calculation but for SF
6

Consider the same conducting particle of 10-m size attached to one electrode and again enhancing the
field by a factor of 5 for the first 5 m, 3 for the next 5 m and 2 for a further 5 m. The breakdown
field for the cases with and without a particle are calculated in the same way, but assuming breakdown
occurs for d=16.1, this being the criteria for streamer formation across the gap for SF
6
.
Again it is seen that the effect of particles is very small at atmospheric pressure but rapidly becomes
very significant, indeed disastrous, for SF
6
.

SF
6
pressure (in bar) 1 5 10 15
without particle, E (in kV/mm)
= (230*p +1.61) / 26 =

8.91

44.29

88.52

132.75
By trial and error: main gap field, with
particle =

8.88

25.66

37.60

48.65
First 5 m: .x = 4.62 10.93 12.94 14.37
Second 5 m: .x = 2.31 4.26 3.16 1.72
Third 5 m: .x = 1.16 0.92 (-1.72) (-4.60)
Rest of gap, .10 = 8.01 (-4821) (-13204) (-21818)
Sum of positive .x values = 16.10 16.11 16.10 16.10

E(with particle) / E(without) = 0.996 0.579363 0.425 0.366
Percentage strength reduction = 0.4 42.1 57.5 63.4

Note: for most cases in SF
6
the value of is negative for the main gap, so K is determined for the
enhanced area only.

The Jumping Particle Mechanism
Another possible breakdown mechanism involving dust particles is the jumping
particle mechanism: dust particles on the electrodes will become charged to the same
potential and therefore be attracted to the opposite electrode. If light enough, they
will accelerate towards the other electrode and, it is suspected, discharge by a minute
spark when a few micrometres away, just before contact is made. It has been shown
that at high pressures electrical breakdown can easily be triggered by small
injections of plasma ionised gas into the gap.

Electron Emission
When the bare-electrode graphs of the
emission figure on page 2.1 are re-
plotted for the Shottky relationship, as
log(emission current) versus (electric
field), a good straight line was obtained
as seen on the right. This shows that
the emission was indeed the Shottky
field-enhanced thermal emission
process and not the Fowler-Nordheim
quantum-mechanical barrier-tunnelling
one.


2.4
Dr.Qais Alsafasfeh
As stated above, in a good vacuum it is likely that the mechanism for electron
emission is the Fowler-Nordheim one, but for compressed gases the necessary thin
energy barrier could not exist as the surfaces of metals are covered in thin layers of
tarnish such as moisture, oxides and surface-adhering gas molecules there is no such
thing as a clean surface at sub-microscopic levels. Consequently Schottky emission
has to be assumed.
Virtually all the electron emission takes places at tiny protuberances little hills of
micron dimensions on the surface of conductors where the field there may be
magnified several times.
These can cause the electron
emission to increase by several
orders of magnitude. Typically
90% of the emission from a surface
will come from a few of these hot
spots. The figure shows the field
enhancement caused at the tip of a
hill shaped as half an ellipsoid of
revolution.
Note that even a hemispherical
hill has a field enhancement by a
factor of 3 at its top.
The very high number of electrons
emitted from the top of the hill
are more likely to form avalanches
which can transform into streamers
in the high-field area near the
emission point.
In addition, the avalanches do not need to become so large, in order to transform into
streamers, as there are many developing in parallel, all adding to the field distortion
which is what causes streamer development.

Conditioning of Electrodes
The lowering of breakdown strengths by the very high electron emission currents
from bare electrodes at the high electric fields which can be reached at high pressures
also need explaining. The electron emission tends to be from very small areas where
a pointed protuberance (=a microscopic hill) enhances the field. However the very
high current densities further distort the field locally leading to streamer-type
breakdown mechanisms. Consequently gas gaps between bare electrodes will break
down at very low voltages initially.
After several breakdowns the breakdown voltage will stabilise at a higher level: it is
believed that this reflects the burning off, of areas of high emission and of particles
of dust on the electrodes. The energy released by the spark burns, vaporises or melts
the surface protuberances and dust particles. The electrodes are then said to be
conditioned.

2.5

Dr.Qais Alsafasfeh
This agrees well with the graphs on page 2.1 and 2.4 where it is seen that the emission
from the unconditioned electrodes greatly exceeds that of the conditioned
electrodes. Also near the top of the conditioned-electrodes graph there seem to be a
break where the emission decreases by what at first sight may appear a small amount,
but is in fact a decrease by a factor of two. A reasonable explanation would be that
the emitter of the highest current was melted by the current and ceased emitting.

Consequences and cautions
The effect of dust and electron emission in reducing the breakdown strength of gases
from the values calculated using the streamer criterion has been demonstrated, but we
cannot estimate the actual breakdown voltage. The field above the irregularity (dust
or micro-hill) IS increased, so avalanches can form there easily. Electrons ARE
emitted, so multiple avalanches can form simultaneously presumably 100
avalanches of 106 electrons distort the field much as an avalanche of 108 electrons
would. But we obviously cannot know what irregularities are present their sizes and
numbers.
Furthermore, even if we could know that the conditions are such as to cause a
streamer to form in the enhanced-field region above a particular irregularity this
does not mean breakdown will occur.
For example, in SF6, the field away from the enhanced-field region will probably be
below the critical field and therefore the streamer will only develop in the high-field
region. This localised streamer will act as a conducting extension of the irregularity -
causing a new enhanced-field region at the streamers tip. But whether this will be
sufficient to cause a further streamer, and thus be a self-propagating phenomena,
leading to complete bridging of the gap and complete breakdown, is obviously
unknown.
In general, practical applications of compressed gases are in general restricted to a
maximum of about 5 bar, as a result of the many uncertainties and problems with the
use of highly compressed gases, that is, the difficulties in achieving both a dust-free
system and low electron emission rates.











2.6
Dr.Qais Alsafasfeh
3. Electrical Breakdown in Vacuum
Surprisingly, vacuum breakdown is a topic which has much in common with
breakdown in compressed gases in that it is largely controlled by the imperfections.
These are of two kinds: small areas of very high electron emission from small
protuberances (very small pointed hills) on the electrode surfaces; and the presence
of free particles.
In a vacuum of 1 bar (=10
-3
torr) the average distance between collisions is about 50
mm so no avalanche or streamer breakdown can take place.
Small conducting particles will increase the field at their own surface and there may
then be high levels of electron emission from their surface.
However, high-field emission of electrons from small sites (hills or conducting
particles) on the cathode surface causes local heating of the anode surface opposite,
where the jet of electrons impacts. This impact area will be very small since there
is no spreading out of the beam of electrons since there are no collisions with gas
molecules. Furthermore the current density at the emission source may be sufficient
to cause melting or even explosion of the cathode protuberances, if it is pointed
enough.


The extra scale shows the number of mean free paths in the gap width,
i.e., the number of times an electron will on average collide
(elastically or inelastically) with molecules in crossing the gap

The heating at the point where the electrons impact causes the release of metal vapour
and absorbed gases from the metal surface. It is this resulting vapour which breaks
down.
Thus, in a vacuum of better than 1 bar, the breakdown strength is related to the
electric field and to the condition and material of the electrodes, and not to the actual
level of vacuum. There is therefore no difference between a l-bar, or a 0.0l-bar,
vacuum as far as breakdown strength is concerned.
The small protuberances causing the high-density electron emission can be removed
by medium-energy conditioning of the electrodes, again, as is the case with
compressed gases.


2.7
Dr.Qais Alsafasfeh
Corona and Stepped Breakdown

Corona Introduction
The field is very high close to a metallic point or a knife-edge, if they are at a high
voltage. It decreases very quickly, most of it being gone within a distance equal to
the radius of the point or edge.
The graph shows the field between a 1-mm radius hemispherical point (at the end of a
1-mm radius rod) 30 mm from an earthed cup electrode, when 10kV is applied
between the two electrodes.
0
2
4
6
8
10
12
14
0 1 2 3 4 5
Distance from the surface, mm
E
-
f
i
e
l
d

a
n
d

(
o
v
e
r
a
l
l

a
l
p
h
a
)
/
1
0
E-field, kV/mm
(Overall alpha)/10, per mm

1 mm from the surface, the field has dropped by some 75%, 2 mm from the surface,
the field has dropped by some 90%.
Given the fast rise of with E (see the graphs on pages 1.3 and 1.4), it should not be
surprising that has dropped to zero by 1mm (and then goes negative), so that
nothing much happens outside the first 1 mm of gas. This is generally true provided
the radius of the pointed electrode is small compared with the gap size.
If the field is strong enough streamers will occur in the high-field area, this is called
corona.
Each streamer is a pulse of current and therefore can be measured in the circuit
supplying the high voltage. Since they last for a time of the order of a microsecond
and occur at intervals of 10 to 100 s, they cause radio interference and can also be
detected audibly (a sizzling noise). This can be a nuisance but can also be used to
detect incipient trouble in, for example, high-voltage machines.
All points or sharp edges when at a high voltage can exhibit corona. The point/cup
electrode system shown above is merely to make explanation and calculation easier.

Negative corona - calculations
The electric field at any point in the space between a point electrode and a plane may
be determined by a combination of computer field plotting and calculation. In the
case of the hemispherically-tipped point and cup system described above, the field
near the tip is given by E = 1.0345 V/r
2
.
3.1
Dr.Qais Alsafasfeh
The effective ionisation coefficient for air, = -, was shown in a graph on page 1.3.
The best-fit polynomial for the data shown is found to be
/p = 0.0726(E/p)
3
+ 0.6551(E/p)
2
2.972(E/p) + 2.2805
Using this equation, the value of /p was also plotted on the graph.
Now, the size of an avalanche starting from one electron at the surface of the
electrode is
n(x) = exp( (E).dx ) = exp(K) (see page 1.4)
K is the area under the graph and can be found analytically by integration or, using
Excel, by adding up the very small areas between (e.g.) two radial distances 0.01 mm
apart. The area under the curve is K=23, so it would transform into a streamer well
before reaching the boundary of the ionisation area.
Why? The critical size of an avalanche to transform into a streamer is 10
8
electrons,
or e
18.4
electrons. Another way of writing this is that the criterion for transformation
to a streamer is K = 18.4.
In fact careful calculation shows that an avalanche starting from one electron at 0.045
mm from the surface of the electrode will reach a size of 10
8
, or e
18.4
, electrons at the
ionisation boundary. This is where E, has dropped to 2.3 kV/mm, and so has
dropped to zero. Further out E is less than this critical value and will be negative,
so the avalanche will not only cease to grow, but will decrease in size as the free
electrons have attaching collisions. (In this case the ionisation boundary is at 1.1 mm
from the surface).
Calculation shows that if the applied voltage is reduced to 9.4 kV, then the size of an
avalanche starting from one electron at the surface of the electrode is the critical size,
10
8
, or e
18.4
, electrons by the time it reaches the ionisation boundary and stops
increasing in size.
If the voltage is now increased to
12 kV, the value of K reaches 18.4
(integrating from the surface of the
point electrode) at 0.12 mm from
the surface i.e., the avalanche
reaches critical size within a
distance of 0.12 mm.
Furthermore, an avalanche starting
from one electron at 0.135 mm
from the surface of the electrode
will reach a size of 10
8
, or e
18.4
,
electrons at the ionisation
boundary.
0
2
4
6
8
10
12
14
16
18
20
0 1 2 3 4 5
Dist from the surface, mm
E
-
f
i
e
l
d

a
n
d

(
o
v
e
r
a
l
l

a
l
p
h
a
)
/
1
0
E-field, kV/mm
(Overall alpha)/10, per mm
Summarising, as the voltage is increased, streamers will start to occur at 9.4 kV this
is called the Corona Onset Voltage. At 10 kV any electron starting between the
surface and 0.045 mm away from the surface will develop into an avalanche which
will transform into a streamer. (An electron starting say 0.05 mm away from the
surface will develop into an avalanche which will never get big enough to transform
into a streamer). And at 12 kV any electron starting between the surface and 0.135
mm away from the surface will develop into an avalanche which will transform into a
streamer.
Another worked example is downloadable for a case where the radius of the tip of the
point electrode is r = 1.5 mm.
3.2
Dr.Qais Alsafasfeh
Negative corona - qualitative
An electron produced at A (see diagram)
will cause a larger avalanche to develop
than an electron produced at B. So there is
a region 1 (see diagram) near the
electrode where electrons will develop into
large enough avalanches (about 10
8

electrons) to transform into streamers.
There must also be a region 2 outside of
region 1 where electrons will develop
into avalanches which never get large
enough to transform into streamers.
Streamers consist of the column of ionisation formed by many (30 to 100?) baby
avalanches developing ahead and behind a main avalanche. Outside of the high-field
regions (1 and 2) where > (or > 0) that is, in region 3 attachment
predominates, so avalanches cannot form. It follows that streamers can only form in
the high-field region where > .
Note:
In both regions 1 and 2, > .
In region 3, > .
The boundary between region 2 and region 3, is at =

Negative corona - measurements
By inserting a resistor in the earth lead and applying the voltage to an oscilloscope,
these streamers may be seen as pulses (see next page), whose frequency increases
with the applied voltage. They are called Trichel pulses after the first person to
observe them (or, at least, the first to publish his observations).
Generally speaking, no
streamer can occur until the
previous one has decayed and
disappeared because its
conductivity lowers the electric
field nearby. This happens (a)
because the electrons and
positive ions collide and
recombine (to form molecules)
and (b) because the electrons
and positive ions drift away
towards the electrodes. Then
there is the further delay,
waiting for a free electron to
appear in the volume near the
pointed electrode from which
avalanches can develop and transform into streamers (i.e., region 1 in the figure
above). At higher voltages, both these processes happen faster, so the streamers
occur more frequently.

3.3


= line
0
20
40
60
80
100
0 5 10 15 20 25
Applied voltage, kV
P
u
l
s
e

f
r
e
q
u
e
n
c
y
,

k
H
z
0.2 bar 0.4 bar
0.6 bar 0.8 bar
1.0 bar
Dr.Qais Alsafasfeh
The following graphs from saved DSO screens are for concentric electrodes as in
the diagram on page 3.1 with the earthed electrode having a 30-mm internal diameter
and the pointed electrodes hemispherical tip having a radius of 1.5 mm, as in the
worked example (Example neg corona.pdf).


Positive Corona
It would be expected that the situation when a positive voltage is applied to the
pointed electrode would be similar. The calculations do give similar results except
that the avalanches are starting near the ionisation boundary and coming inwards
towards the pointed electrode. The onset voltage is the same for the same electrode
geometry. However, although streamers appear at the onset voltage, as soon as the
voltage is raised by about a kV they cease! See the graphs overleaf.
Why? The reason is that there are many more avalanches than in the negative case
coming in towards the pointed electrode (think about the increased volume for
initiatory electrons to appear) and as soon as they turn into streamers, they clog up
the area near the surface of the pointed electrode with positive ions. These lower the
field there until streamers can no longer appear negative feedback control!

3.4
Dr.Qais Alsafasfeh


The 13-kV graph shows a pulse, due to a
streamer, just above the onset voltage. The
graphs below it show a noisy DC current,
larger for 21 kV than for 15 kV. This DC
current is in fact the sum effect of many small
pulses due to the many avalanches arriving at
the pointed electrode.
The cloud of positive charges (called space
charge) around the tip shown in the figure on
the right lowers the field locally: graph with
and without space charge are shown.


3.5
Dr.Qais Alsafasfeh
AC Corona
As all these events take place on a microsecond scale, and AC voltage changes so
slowly on a millisecond scale the positive and negative half-cycles are almost
independent.
Thus, in the positive half-cycle, there may be one pulse as the voltage passes through
the onset voltage in the rising part of the half-cycle but not on the falling part, as the
cloud of positive ions has not yet dispersed.
In the negative half-cycle, there
are pulses from the moment the
voltage passes through the
onset voltage in the rising part
of the half-cycle until it passes
through the onset voltage in the
falling part of the half-cycle.
As the voltage increases, the
frequency increases, as
expected. And vice versa.
This may be seen on the
diagram on the right, where the
pulses due to streamers are
shown on the lower graph.

-1
0
1
0 90 180 270 360

0 90 180 270 360
An actual oscilloscope reading of the negative half-cycle is shown below, firstly the
full half cycle, and then below that an expansion of the part during which pulses are
occurring (270
o
+ 45
o
). The increased pulse frequency near a phase angle of 270
o
is
clearly seen..
6000 7000 8000 9000 10000 11000 12000 13000 14000 15000 16000


9400 9600 9800 10000 10200 10400 10600 10800 11000 11200 11400 11600 11800 12000 12200 12400 12600 12800 13000


3.6
Dr.Qais Alsafasfeh
The 3-dimensional graph above shows pulses counted during several complete
cycles, plotted against phase angle, and separated according to size.


Stepped Breakdown Long Sparks and Lightning
The breakdown of a point/plane gap is not as simple as for a parallel plane/plane gap: in the latter,
when the streamer bridged the gap, it continually increased in conductivity until it formed a short-
circuit a spark, a breakdown. As shown above, in a point/plane gap, the streamers only develop
within the ionisation boundary and their length is typically of the same order as the radius of the
pointed electrode.
The exact situation is argued over but the following is your lecturers opinion! It appears that if the
voltage is high enough much more than the onset voltage for corona the streamer becomes so
highly conducting that it becomes self-propagating for a much greater distance than out to the ionising
boundary, because it distorts the field it is then called a leader. It stops propagating, perhaps because
it decreases the charge and hence voltage of the electrode system (a capacitor). Once recharged to full
voltage from the external circuit, the leader can advance again, as soon as a new streamer forms and
increases in conductivity, width and length, to be a continuation of the leader. The direction of the
streamer depends on where, near the tip of the previous leader, the first initiatory electron occurs.
Hence the twisting and turning of a long spark and of lightning as every step of the spark sends the
leader in a slightly different direction.












3.7
Dr.Qais Alsafasfeh
Electrical Breakdown of Solids & Liquids and across Surfaces

1. Solids as Electrical Insulators

Introduction
Now for something completely different. The breakdown of perfect solids is almost
impossible, and in practice the breakdown process always starts at imperfections,
either within the solid or at the conductors bonded to the solid insulation. The
mechanism by which breakdown occurs is called treeing.
The imperfections at which breakdowns are initiated in solid insulation are (a)
protuberances (micro hills) on the electrodes, or (b) metallic inclusions, (c) water
droplets or (d) gas bubbles (or voids) within the insulator.

Treeing
The field at the tip of a pointed
protuberance on the surface of an
electrode will be very much higher than
the main field elsewhere and will lead to
localised high electron emission currents
as for the compressed gas case. Because
of the small emission area these currents
can have very high densities and cause
local overheating of the insulator: this can
cause damage by either melting or
decomposition (as appropriate) and result
in a small void around the tip of the
protuberance. This effectively extends the
point because the void will break down
and the discharge is of course a conducting
plasma - and the process then repeats itself.
Thus, a tubular channel (or micro tunnel)
is formed which extends away from the
electrode towards the other one.
Sometimes the channel splits and forms
two branches, so that the channels begin
to look like a bush or tree. Not
surprisingly, this process is called treeing.
The simulation shown on the right is from
the program treeing.exe.




4.1
Dr.Qais Alsafasfeh
The overheating mechanism described above is not the only method of tree-formation.
The branches may also grow, due to the electrostatic repulsion between high
concentrations of electrons on the inside surface at the tip of the channel: these are
deposited after discharges inside the channel. The strong forces of repulsion can
cause cracking or splitting of the material. These cracks will tend to open up where
there is a weakness or fault in the insulating material and account for the randomness
in the direction of channel growth.
A third possible mechanism is that of electrons being accelerated along the channel to
such high energies that, when they hit the end of the channel, they can break chemical
bonds in the solid there. In the case of the long-chain polymers (plastics) this results
in shorter molecules: these will probably be liquid or gases, especially when the
channel tip is likely to be hot due to the discharges.
Three possible mechanisms have been described. Probably all three occur
simultaneously.
The acceleration of electrons to the extent that molecular bonds may be broken.
The impact of high-energy electrons causing local over-heating and hence melting
or decomposition.
The repulsive force between electrons causing splitting of the material.
Note: the structure of flexible plastics is a spaghetti-like mass of long molecules of
the form M M M M M M M M M M M
(In polythene the monomer indicated by M is CH
2
. In PTFE it is CF
2
. In
polystyrene M is much more complicated).
If the molecule is broken into shorter
lengths it may be liquid or even
gaseous. Thus in the case of the CH
2

family, molecules with more than 17
CH
2
units will be solids (waxes,
polythene), molecules of less than 17
CH
2
units will be liquids (paraffin,
petrol, oil) at room temperature, less
than 5 CH
2
units will be gases
(methane, ethane, propane, butane).

Water trees and electrical trees
It is found that water greatly increases
the rate of growth of trees, perhaps
because the polar nature of the water
molecule allows it to penetrate deeper
into small faults and cracks in the
material. Water trees may be found in
bow-tie formations inside the
insulation and vented tree formations
starting at the interface between the
insulation and the conducting screen.
Electrical trees grow in the absence of water, and, after the water is used up, may
grow from a tree which started as a water tree. Thus it may continue the damage
until eventually the insulation is bridged and complete breakdown occurs.
4.2
-200
-100
0
100
200
300
400
0 10 20 30 40
n, number of C-atoms per chain
T
e
m
p
e
r
a
t
u
r
e
,

d
e
g
r
e
e
s

C
Gases Liquids Solids
BP


MP
Dr.Qais Alsafasfeh
Dr.Qais Alsafasfeh
2. Liquids as Electrical Insulators
Introduction
Transformers, power capacitors and cablesare the main users of insulating liquids and
petroleum mineral oils are the most common and the cheapest of them.
Theoretically, in pure, clean liquids the mean free paths are too short for ionisation to
occur. How then do liquids break down electrically? The theoretical intrinsic
strength as with solids is seldom if ever achieved in practical engineering
situations, as illustrated by the table below.

Power-frequency design fields for insulation systems compared with breakdown fields
for highly purified liquids

Insulation
systems
Design field
(kV/mm)
Very pure liquids
Breakdown field
(kV/mm)
Transformer 2-6 Transformer oil 100
Cables 12-20 n-hexane 130
Capacitors 10-25 Polybutene 110

As solids, the answer is not simple but certainly the following are factors affecting the
breakdown strength and reducing it below the intrinsic strength:
the electrode material and its condition
the presence of water droplets, gas/vapour bubbles and free particles
(particularly if long rather than round).
Then trees will start to grow, in the locally enhanced field, but cannot remain once the
voltage is removed surface tension turns the branches into strings of small bubbles
which float off towards the surface.
Water droplets become elongated in a large enough field as shown in the sequence
of photographs shown overleaf. When very long the end becomes rather pointed and
breakdown channels develop until the gap is bridged and breakdown initially of the
water vapour occurs.

Breakdown of a liquid dielectric (a very viscous silicone fluid) due to the instability of
a water globule at a very high electric field. Tenfold enlargements from 16-mm film
(after Krasucki).

4.4

Dr.Qais Alsafasfeh
Solid particles in the oil will experience a force tending to move them into the region
of highest field if the particle's permittivity is greater than that of the oil. This can
cause the setting up of a bridge of particles across the gap if there are sufficient of
them, particularly if they are long particles, like fibres. If they are metallic this is a
short-circuit and will probably lead to breakdown; if they are of high permittivity this
can cause significant field distortion and assist breakdown.
The field at the surface of an individual particle will be greater than elsewhere (in the
case of a spherical metallic particle, 3 times the average field). This may cause local
breakdown leading to vapourisation and breakdown as described above (bubbles
mechanism).
Weak link effects: as shown in the graphs below, there is also a distinct volume
effect, probably due to an increase in the probability of finding a particle in the
volume or an asperity on the electrode.

The electric strength of transformer oil under uniform-field conditions at 90
o
C
for (a) power-frequency alternating voltage and (b) impulse voltage
(from Palmer & Sharpley, Proc IEE, 116, p.2029 (1969)


2. Surface Breakdown across Solid Electrical Insulators

Introduction
Flashover is the breakdown of a gap between two conductors, one at high voltage, the
other earthed, where the gap is bridged by a solid (or liquid) insulator: the breakdown
occurs in the gas across the surface of the solid insulator, and much lower than if the
insulator was not there.
Tracking is a slow process, taking typically months or years in the field, by which
meandering and branching carbon tracks develop across the surface of a solid
insulator, eventually short-circuiting most of the gap, when flashover occurs.
[Note: tracking is completely different from treeing; treeing occurs inside the
insulator].
Tracking
This is a complex process which depends not only upon the insulation itself but on the
condition of the surface. By definition, tracking is the formation of a permanent
conducting path across a surface of the insulation, with the conduction due to contain
carbon. However, not all carbon-based materials can track: neither polythene nor
perspex do so, but epoxy resin does.

4.5
Dr.Qais Alsafasfeh
Tracking occurs when there is a conducting film of moisture across the surface of the
insulation, which has condensed from the atmosphere. This will usually contain
contamination such as salt in coastal areas, carbonaceous dust from burning fuel or
brush gear in industrial areas, or cellulose fibres in (for example) a textile mill. Salt is
particularly important because it increases the conductivity of the film of moisture.
The usual mechanism by which the leakage current through the film of water is
interrupted is by its drying out. This occurs because of (a) the normal evaporation
process, particularly if the ambient temperature rises, and (b) the heating effect of the
leakage current itself. This will eventually cause a break in the film which virtually
stops the current. All the voltage applied across the insulator is then across the dried-
out gap (often called a dry band in books), which will then probably exceed the
breakdown voltage of air: a spark occurs, bridging the dry band. After a period of
sparking the insulator surface will be damaged.
Degradation of the insulation by the sparks is the result of heat from the sparks, and
the consequent bombardment by high-energy electrons, which carbonises the
insulation. In some materials (like polythene or perspex) the surface is merely melted
or eroded, but in most organic materials and plastics, the long polymer chains are
repeatedly broken until they are reduced to carbon (with the other materials being
vaporised). It follows that only organic insulation which can be degraded to carbon
by heat, can track.
The carbonised area is now (permanently) conducting and so current again flows.
The current will assist other parts of the wetted area to dry out until another dry band
is formed, breaking the circuit and, again, causing a spark. The cycle repeats until the
carbon tracks begin to link up and eventually form a continuous line from electrode to
electrode when complete failure of the insulation occurs. Note that the carbon bridge
is narrow so that the current approaching it through the wet film will he of
increasingly high density, so that the evaporation will take place preferentially near
the ends of the track so that the tracks tend to grow rather than start forming in
isolation from each other.
In practice it is likely that, once a substantial proportion of the gap has been bridged
by carbon tracks, the voltage gradient across the unbridged parts will be high enough
to initiate flashover which is the next topic.
The rate of degradation of the surface may also be accelerated by other processes,
such as physical weathering, ultra-violet radiation and chemical attack. (For example,
ozone and oxides of nitrogen generated by discharges may further degrade the
insulation).

Methods of minimizing tracking
Tracking may be prevented by ensuring that all surfaces are clean, dry and
undamaged and the materials used are highly track-resistant. Cleaning is not often
practicable; but may be used in critical cses.
Using non-tracking or track-resistant materials such as porcelain and silicone rubbers
will avoid tracking but will not be suitable for all applications. Porcelain does not
contain carbon and so does not track. Silicone rubbers exude low-molecular-weight
silicones which are oily and break up the moisture into droplets, thus preventing a
continuous moisture layer from forming so no conduction.

4.6
Dr.Qais Alsafasfeh
Good design helps by (a) limiting the access of dirt, (b) avoiding its accumulation in
areas between conductors, (c) increasing creepage paths as much as is practical and (d)
for indoor installations, ensuring either good ventilation or using air-conditioning
when the humidity is high.
The tendency to track may be tested by various tests, two of which are now briefly
described. Both will be used in a simplified form in the laboratory sessions.

The Comparative Tracking Index Test (CTI Test) is a relatively simple test which
only uses 100 to 600 volts and a very small sample. Drops of a solution of 0.1%
ammonium chloride are allowed to fall on the surface of the insulator between two
electrodes 4mm apart, at 30-second intervals. This continues until failure (high
current). The higher the voltage, the fewer drops required. The voltage at which
failure occurs on average after 50 drops is called the CTI. This limits its accuracy
and usefulness. It provides
good resolution between
materials with relatively
poor tracking resistance,
but lacks resolution for
materials normally required
for outdoor use for which
the test method in IEC
Standard Publication 587 ...
should be used. It is
described in IEC Standard
Publication No.112.



Some results of a typical test
are shown in the graph on the
right. For the weaker sample
(i.e., the leftmost of the two),
below 200 volts, it does not
break down, for 250 volts less
than 50 drops are sufficient and
for 300 volts less than 10 drops.








4.7
Fine tube Droplet of NH
4
Cl solution
2 platinum
electrodes
Dr.Qais Alsafasfeh
The High-Voltage or Inclined-plane Test (IEC 587 Test) uses test voltages up to 6
kV and a larger surface area of the sample. Two stainless-steel electrodes are
clamped to the specimen which is set up with its longer axis at 45
o
to the vertical, and
the electrodes attached to the underside. A solution of 0.1% ammonium chloride with
a little wetting agent is fed into a filter paper pad clamped under the top electrode and
flows down the under-surface to the lower electrode (see the diagram from the
IEC587 standard on page 7.3). The minimum voltage to cause uniform sparking over
the wetted surface is applied (this must be a multiple of 250 V). At the end of each
hour the voltage is increased 250 V until failure or flashover occurs. In this way
materials are graded in terms of the voltage at which they fail.
A material which has a CTI of 500 on the low voltage test will fail at about 2 kV on
the inclined plane test; thus a range of 4 to 5 kV is available in which to evaluation the
more tracking-resistant materials.


4.8
Dr.Qais Alsafasfeh
Surface flashover
While tracking may take hours, days or even years to develop to complete breakdown,
flashover is an instantaneous occurrence, being a breakdown via a streamer
travelling through the air close to the surface of the insulator. The breakdown is at a
lower voltage than would be expected from the apparent values of the electric fields
and the normal streamer breakdown theory.
Thus in the three gaps shown below, cases (a) and (c) will break down at similar
voltages (which may be calculated from the streamer criterion). However in case (b)
the breakdown will be across the surface of the insulating spacer and at a significantly
lower voltage than (a) or (c). Why? Note that the field should be uniform and should
be the same in cases (a) and (b) and for most of (c).

Surface effects: there will be impurities on the surface of the insulating rod and these
provide a source of secondary electrons which are produced by the impact of photons
and electrons from the avalanche head. The surface appears to produce these
secondary electrons more efficiently than the air molecules (which are very far apart).
Thus, the surface provides in effect a greatly increased value of , and this causes
flashover on the spacer surface, at a voltage well below the breakdown voltage of the
gap if without the spacer present.
Dirty surfaces appear to produce these secondary electrons (i.e., have a high effective
value of ) more efficiently than clean surfaces. So the breakdown strength of a gap
with a spacer will vary greatly with the cleanliness of the spacer.

3-material interface. In case (b)
it is inevitable that the machining of
the spacer will not give a perfect
right angle at the ends and so, in the
sketch shown below there will be an
irregular gas gap.
This will have a higher electric field
than the rest of the gap (about twice
the main field) and will therefore
break down locally at about a half
of the no-spacer breakdown field
for the gap.
As the voltage across the whole gap increases, the discharge will get more intense and
feed electrons and ions into the main gap. These electrons will be multiplied through
the normal ionisation process, and could cause breakdown at a lower-than-expected
voltage.

4.9
Dr.Qais Alsafasfeh
The Design of High-voltage Equipment
The previous units have indicated why and how gases, liquids and solids break down
electrically, and why and how breakdowns occur across surfaces. In this chapter, the
design of some items of HV equipment are examined to see how this knowledge is
applied.

1. Cables
XLPE Distribution Cables
Distribution cables (<50 kV) may be single- or 3-core and are usually stranded copper
conductors with extruded cross-linked polyethylene (XLPE) insulation. This is
polyethylene (polythene to laymen) where chemical bonding has been caused
between some of the long polyethylene molecules making a stiffer material.
Because of the tendency of thermal cycling (caused by load cycling) to cause some
de-bonding of the plastic from the conductor, due to the much greater expansion rate
of the polyethylene, there is usually a thin slightly-conducting layer placed between
them, consisting of a carbon-loaded polyethylene. This will have the same expansion
rate as the main insulation, and be at the same potential as the conductor.
Consequently it doesnt matter if there are gaps in between the plastic and the
conductor.
This also improves the uniformity of the field (lessening the chance of treeing) which
otherwise would be increased locally due to the stranding of the conductor and
microscopic irregularities on its surface.
Similarly, an outer layer of conducting plastic is
extruded with, and bonded to, the insulation under
the copper screen. Armouring with steel wires or
corrugated aluminium may be needed and then a
PVC covering will, typically, be applied overall.
The materials outside of the copper screen depend
on the ruggedness deemed necessary for the
particular application.
For 20kV and below the cable usually has 3 cores
(example shown on next page).
In the photograph a high-voltage cable is shown
(in this case, a DC one), probably rated at about
50kV, but clearly showing the stranded conductor
(in this case, of aluminium), the carbon-loaded
XLPE screen, the main XLPE insulation, the outer
carbon-loaded XLPE screen, the copper wire
screen, the aluminium sheath and the PVC
protective covering.


5.1

Dr.Qais Alsafasfeh
Oil-paper Transmission Cables
At transmission voltages (>50 kV) the insulation must be much thicker and may be
either oil-paper or extruded XLPE. Oil-paper cables use many layers of thin paper
tapes as insulation. After adding the sheath, almost all the air trapped between the
layers of paper is sucked out under vacuum, and then the cable is impregnated with
a mineral oil under pressure. This double process almost completely eliminates voids.
However, the oil filling results in the inconvenience of having oil reservoirs at
intervals along the route and stop joints to avoid undue hydrostatic pressure where
the cable goes up a hill. Stop joints are joints where the flow of oil is blocked: oil
cannot move from one section to the other. The good aspect is that if a cable is
damaged, it will leak oil only slowly and may be run for a considerable time (if the
pressure is kept up and the oil reservoirs are kept full) until it is convenient to repair it.

1. Oil-filled AC transmission cable
2. Mass-impregnated DC transmission cable (no oil duct)
3. XLPE AC transmission cable
4. 3-core XLPE distribution cable
5+6. XLPE distribution cables for DC
In the photograph it will be seen that oil-paper cables also use slightly-conducting
layers between the conductors and main XLPE insulation, in this case carbon-loaded
paper.

XLPE Transmission Cables
Polyethylene cable does not have these troublesome and costly oil pressurising
stations but is not yet commonly used above 300 kV except for short lengths (e.g.
within hydro electric. The problem is the increasing difficulty of avoiding the
presence of voids in larger cables and the progressive deterioration of the cable due to
the growth of trees in them. Continuous research into manufacturing techniques has
hugely improved the situation in respect of voids and hence the long-term reliability
of polyethylene cables. For greater stiffness cross-linked polyethylene (XLPE) is
used rather than ordinary polyethylene.

5.2
Dr.Qais Alsafasfeh
The photograph shows a 500-kV XLPE cable as
used in a pumped storage station in China.
The permittivity of polyethylene is lower (2.3)
than that for oiled paper (4.5), and therefore the
capacitance and charging current are lower on
XLPE cables than on oil-paper cables. This
means less reactive compensation is necessary.
The loss factor is also less: this means less heat is
generated in the insulation and thus allows an
increased current rating; it also means lower
losses.

2. Bushings
The insulators used where an hv
conductor passes through an
earthed metal wall (or similar) are
known as bushings. In air the
highest field would be at the edge
of the hole in the wall: this can be
reduced by placing a collar of metal
there. This might be sufficient if
there were no requirement to
separate one side of the wall from
the other side by a solid barrier.
Given that surface flashover occurs
at much lower fields than
breakdown, the air/conductor surface must be several times longer than the
radius of the hole in the wall.
The field will still be greater on the surface near the edge of the hole in the
wall so foil grading is used to control the voltage distribution in the
bushing: cylindrical foils are included during the winding on of the
bushing insulation and these will form a number of capacitors in series,
which, if equal in value, will ensure the voltage between each adjacent pair of foils is
equal. Their relative size can also be adjusted to have a lower gradient near the wall
than elsewhere.
5.3






Dr.Qais Alsafasfeh
A porcelain covering may be added to the external half of the bushing with shedding
to increase the creepage distance and break up the flashover path. The word
shedding refers to the strangely-shaped stack of discs or cones around the bushing
Perhaps because it causes it to shed water easily, that is, to allow the rainwater to
run together and drip off it at the edge. If the bushing is vertical, the shedding will be
smooth and sloping on top for this reason, and ridged underneath, again to break up
flashover paths.

9.3 Circuit Breakers
Sulphur hexafluoride circuit breakers
SF
6
circuit breakers are widely used at transmission voltages and increasingly at
distribution voltages, too. SF
6
s high attachment coefficient removes free electrons
from the arc and its high thermal capacity removes heat from the arc.
For transmission voltages, the circuit breaker is a single sealed tank containing the
contacts and insulated with SF
6
at a pressure of about 5 bars. As the contacts start to
move, the design is such that a quantity of SF
6
gas is compressed by the movement
and is released by the contacts separating which separation also causes an arc. The
blast of SF
6
gas blows out the arc. This puffer mechanism is unnecessary at
distribution voltages. A simplified explanatory diagram is shown below (animation:
puffer.exe) and below that is the real thing.




5.4
Dr.Qais Alsafasfeh
A major advantage of SF
6
circuit breakers over older types (oil-blast and air-blast) is
their lack of maintenance and smaller size. Almost all the transmission-voltage
switchgear in urban areas is SF
6
-insulated.

It is extremely important for manufacturers to ensure that neither contaminating
particles nor moisture (must be less than 500 ppm) are present in the gas. If
contaminating particles are present, lower-than-
expected breakdowns may occur see Lecture 2.
If moisture is present (more than 500 ppm) this
will react with the chemical dissociation products
produced in the arc, to form highly corrosive
chemicals which will etch and erode the surfaces
of the insulators.

Vacuum circuit breakers
Vacuum circuit breakers (VCBs) are the norm at
distribution voltages. They are small because of
the good insulation and the very small movement
of one contact the moving electrode of about
0.5mm. They have a long life with virtually
zero maintenance, a maximum arc duration of
only 1 or 1.5 cycles at all levels of current, cause
no additional fire risk and produce no noise or
gas emissions when operated. The construction
is completely sealed, gives very consistent arcing
times and allows very fast automatic reclosure.





5.5
Dr.Qais Alsafasfeh



One of the two contact discs A 5 m x 5 m sample of the surface

To obtain such long-lifetimes and consistent results, it is necessary to ensure
the metal contacts are not damaged by local heating at the arc root;
the metal contacts are outgassed before sealing the vacuum; and
the metal vapour from the arc does not condense on the insulating walls of the
breaker and eventually cause a short-circuit;
These are ensured by
causing the arc to move across the surface of the contacts by cutting slots in them
(see photograph above) such that magnetic fields are produced by the current
through the contact to the shaft (F=BiL);
careful heating of the contact material by
a high-frequency discharge, with a
vacuum pump running to remove the
absorbed gases before sealing
(outgassing);
removing electron emission points by
the same high-frequency discharge; and
placing a thin cylinder surrounding the
contacts so that any metal vapour will
condense there, thus shielding the
insulating walls from metal deposits.
It is common for enclosures around the VCB to be
filled with SF
6
gas to avoid external flashover ( i.e.,
vacuum inside the bottle, SF
6
outside).

9.4 Overhead Line Insulators
Post insulators support conductors and need to be rigid
(see the pin-type insulator on the right), but insulator
strings for overhead lines at voltages above 100 kV
need to be flexible to absorb the forces on the lines
from the wind. They are therefore made up of cap and
pin insulators (on right, lower picture) which are
connected together to form strings.

5.6
Dr.Qais Alsafasfeh
For 400 kV, a string of typically about 20
cap and pin insulators are used. The voltage
drop across each insulator tends to be uneven
due to their different surface resistances
because of the differing amounts of pollution
which they have collected, and to be much
higher near the HV conductor end and, to a
much lesser extent near the earthed arm of
the tower. This is due to the stray capacitance
between the metal of the cap (and the pin
which is plugged into it) and the tower. To
decrease this effect, a metal loop is attached to
the HV end, to produce stray capacitance
between the metal cap and this HV loop (see
the bottom left photo).
All insulator designs - whether of the pin-type,
for distribution voltages, cap-and-pin strings
for higher voltages, or post insulators for
outdoor switchgear have this in common:
they have long creepage paths between HV and earth. These are to make it more
difficult for surface flashover to occur but unfortunately also make it easier to trap
pollution. As always, designs must be a compromise.
When the pollution becomes moist it tends to conduct and then dry out to form dry
bands, with local arcing or flashover. Materials used for outdoor insulators are
usually chosen to resist tracking but surface erosion may occur and this encourages
the collection of moisture and pollution.

5.7
Strings of cap & pin insulators
on a 400-kV transmission tower

Dr.Qais Alsafasfeh











Composite insulators with a fibre-glass rod for tensile strength and a shedded cover
are becoming more common as a replacement for the cap-and-pin strings. This
shedded cover is commonly made from silicone rubber, a material which has the
advantage of eroding rather than tracking when a surface discharge occurs
The problem of contamination by pollution can be overcome by regular washing of
the insulators or by coating them with grease which breaks up the water film into
small separate droplets. The former can be done when the insulators are live but is
not usually cost-effective on high transmission towers unless they are situated in very
polluting conditions. Also the automatic washing equipment can be unreliable
because of corrosion and the multitude of valves and joints involved. Very wet
insulators, particularly the post-type, tend to flash over very easily, so live washing is
not always advantageous.
Greasing is effective but is labour-intensive and only practical in switchgear
enclosures not on transmission towers. Also, after a while, the grease tends to trap
dirt and is eventually no longer effective. Also in hot weather the grease tends to melt
and drip off. The removal of old and dirty coats of grease is very unpleasant work.

A power utility
helicopter
washing
insulators









5.8
Composite insulators
60 to 150 kV
Dr.Qais Alsafasfeh

Maintenance work on a 400-kV
transmission line the worker is sitting
in a seat suspended by poly-propylene
chains.

Remote-controlled jet-washing on the
190m-high towers of the 400-kV line
crossing the Thames River. Servo-
controlled nozzles are used.
















5.9

Dr.Qais Alsafasfeh

High-voltage AC & DC generators
To check that the design is within the specification, the equipment must be tested, so
voltage sources at levels higher than the rated voltages must be available.

1 AC High-Voltage Sources
1.1 Cascaded Transformers
Alternating voltages above a few hundred kV are often achieved by cascading two or
more transformers: an example is shown in the diagram where two 240V/200kV
transformers are cascaded with the top 240 volts being tapped off from the first
transformers secondary winding (c
1
-e
1
) as input to the second transformers primary
winding (a
2
-b
2
).
The first transformer has its tank earthed, but
the second has its tank insulated from earth
and at a potential of 200kV above earth: the
lower ends of the second transformers
primary (b
2
) and secondary (d
2
) windings are
connected to the tank. The advantage of
this arrangement is that each transformer
only needs to have an insulation level of
200kV.
A Haefely cascaded 1800 kV ac supply is
shown below: three 600 kV transformers are
cascaded and have outputs from the high end
of their windings at 600, 1200 and 1800 kV
respectively.

6.1
Dr.Qais Alsafasfeh
1.2 Series Resonant Testers
These are employed to give high voltages at relatively low cost: a high-voltage
transformer feeds a series circuit comprising a variable reactor and the test object.
The variable reactor is achieved either by using a low-voltage variable reactor on the
LV side of a transformer, as shown in the diagram below; or by using an HV reactor
with a variable air gap. The test object (show as a cable in the diagram below) is
effectively a capacitance.
The reactor is tuned until the circuit is at resonance at the supply frequency. The
voltage across the test object is then the supply voltage multiplied by the circuit Q-
factor (typically 20 to 50).

If the test object capacitance is low, extra capacitors may be connected in parallel to
allow tuning to the resonant frequency.
Alternately, a higher frequency may be used by replacing the mains-frequency supply
(50 Hz) by a variable frequency supply from a solid-state converter.
There are many advantages to the Resonant Tester method:
The voltage waveshape is improved by removing harmonics from the power
supply;
The power required from the source is much lower than the kVA in the series
resonant circuit (typically around 5%) and is supplied at unity power factor;
If the test object fails, the energy available is small (just the stored energy in
the capacitor) and the voltage immediately collapses (no longer in resonance), so less
damage occurs and there is less danger.

2 DC High-Voltage Sources
2.1 The Voltage Doubler
A simple way to achieve high direct
voltages is of course by half or full-wave
rectification, followed by a smoothing
capacitor. A more efficient design,
however, is the voltage doubler circuit.
In your figure, the voltage at X varies from
+V
rms
2 to - V
rms
2 and the voltage at y
between 0 and 2V
rms
2, so the output
capacitor will be charged to 2V
rms
2.


6.2
Dr.Qais Alsafasfeh
In practice the leakage resistance (tan) of the capacitors, the finite reverse resistance
of the rectifiers and other leakage paths will cause both capacitors to self-discharge
slightly and so there will be some ripple.

2.2 The Multistage Rectifier
Multistage rectifier circuits are a further development of the doubler circuit and are
able to produce output of 1 MV or more. In the drawing of a 3-stage circuit below, the
left-hand column of capacitors remain charged to a potential difference of
approximately 2V
max
(= 2V
rms
2) at steady state but the potentials at the nodes
oscillate up and down by V. The capacitors in the right-hand column also are
charged to a pd of approximately 2V
max
, but the node voltages remain constant.
The word, approximately is used above because there are ripple effects due to the
leakage currents as well as the load currents (if the generator is supplying a load).
The charging process from the initial state (all capacitors uncharged) is explained
below.




The use of a high-frequency supply will decrease the ripple and the voltage
regulation but may not be convenient.
Note that polarity reversal is readily possible by swapping the two diodes attached to
each left-hand node. The figure opposite a 4-stage rectifier shows how the design
may be adjusted to make this a simple operation.



6.3

Dr.Qais Alsafasfeh
2.3 Van der Graaff Generators are
HV DC electrostatic generators with an
output of up to (currently) 25 MV, although
relatively little current is available. The belt
is charged by a set of sharp points which are
typically set at 10kV and produce a corona
discharge.
The charged belt delivers charge to the
collector which is a set of sharp points where
corona will occur due to the charge on the belt:
these spray the opposite charge onto the belt
until it is discharged, which increases the
charge on the hv electrode. Further spray
points can be added to charge the belt in the
opposite polarity.
The voltage achieved will be limited by the
level at which a balance occurs between the
belt current and the sum of the output current,
leakage current and corona loss current. For
the higher voltages the whole unit may be
enclosed and pressurised with SF
6
.


3 High-Voltage Impulse Sources
3.1 Introduction
Impulse testing is done with a fast rise-time pulse which rises to its maximum in
about 2 s and falls asymptotically to zero much more slowly: it falls to 50% of its
peak value in 50 s. Such pulses are produced by discharging a capacitor through an
R-C circuit. There are two methods used, the single-stage impulse generator (which
will now be described) and the Marx impulse generator a multi-stage version of the
single-stage.


6.4





Dr.Qais Alsafasfeh
3.2 The IEC 60-2 (1973) Standard
The IEC 60-2 (1973) documents define the shape and tolerances as follows: a front
time T
2
, of 1.2 s (30%) and a time to half value T
2
of 50 s (20%). The front
time is defined as 1.67T, where T is the time to rise from 30% to 90% of the peak
value as shown in the figure on the previous page. The time to half value is defined
as the time from the 'virtual origin' at 0
1
to the point where the voltage has dropped to
half of its peak value. The 'virtual origin' is where BA produced intersects with the
time-axis.

3.3 Single-stage Impulse Generators comprise a wave-shaping circuit added
on to a direct-voltage generator: this circuit allows a charged capacitor to discharge
through an R-C circuit and the impulse voltage is the potential difference across one
of the capacitors as shown in the circuit overleaf.

In the circuit above, C
1
>> C
2
, so the overall situation is that C
1
discharges via R
1
+R
2

with C2 charging up initially through R
1
and then staying at the same voltage as R
2
.
By Laplace we find that the voltage across C
1
is given by
v
imp
(t) = V
o
[R
2
/(R
1
+R
2
) [exp(-t) - exp(-t)]
where = 1/C
1
(R
1
+R
2
) and = (R
1
+R
2
)/C
2
R
1
R
2
.
This may be explained more usefully by considering that, in the long term, C
1
is just
discharging through R
1
and R
2
in series, so that the voltage across C
1
may be written
after the peak as
v
C1
(t) = V
o
exp(-t / C
1
(R
1
+R
2
)) = V
o
exp(-t).
Therefore, bearing in mind the voltage-divider effect of R
1
and R
2
, the voltage across
C
2
after the maximum is given by
v
C2
(t) = V
o
{R
2
/(R
1
+R
2
)} exp(-t) for t >> 1/
Initially C
2
is uncharged, so v
C2
(0) = 0. It is charged very quickly from C
1
, but to
determine the charging rate we must consider the Thevenin equivalent of the rest of
the circuit: it is a source of magnitude V
T
= V
o
R
2
/(R
1
+R
2
), with a series resistance
of R
T
= R
1
R
2
/ (R
1
+R
2
). It follows that, before the peak
v
C2
(t) = V
o
{R
2
/(R
1
+R
2
)}{1 - exp(-t)} for t << 1/
Clearly these two equations comply with the first (Laplace) equation.




6.5



Dr.Qais Alsafasfeh

The two graphs above illustrate the last two equations.
To comply with the international standard shown on the previous page, it is quite
simple to show that the characteristic times must be about 1/ = 70s and 1/=0.4s.
The switch in a single-stage impulse generator can be
the mechanical approach of two spheres (the spark breakdown will occur before
mechanical contact occurs, so avoiding contact bounce);
by fixed-gap spheres (breakdown will occur at the appropriate voltage for the gap
setting); or
by the use of a triggertron (a small ancilliary spark-gap on one of the spheres can
be used to 'trigger' the main-gap breakdown).


3.4 The Marx impulse generator
This multi-stage impulse generator avoids
the difficulty in switching very high
voltages by connecting many capacitors in
parallel for charging and then connecting
them in series in order to discharge them
through an R-C circuit, as before.
Thus, if there are 20 capacitors all charged
to +10kV, in parallel, and they are then
connected in series, their total voltage will
be +200kV this can be discharged through
the usual impulse-generator R
1
-R
2
-C
2

network to give an impulse of over 100
kV.










6.6

0
20
40
60
80
100
0 1 2 3 4
Time, microseconds
V
o
l
t
a
g
e
,

k
V
0
25
50
75
100
0 25 50 75 100 125 150
Time, microseconds
V
o
l
t
a
g
e
,

k
V
Dr.Qais Alsafasfeh
The actual circuit of a 4-stage
Marx generator is shown at
right.





The diagram on the right shows that
the effective circuit of a four-stage
impulse generator is in effect, a
parallel arrangement when charging
each capacitor to a p.d. of, say, V
o
,
providing that R >> R
2
and R >> R
3
, so
the capacitors are charging through the
R resistors (and may take several
seconds).





6.7

Dr.Qais Alsafasfeh
The diagram on the right shows the
effective circuit of a four-stage impulse
generator when discharging the
capacitors, which are now effectively
in series and therefore have a total p.d.
of 4V
o
.
R >> R
2
and R >> R
3
so each capacitor
C discharges through the parallel
resistor R
2
with a decay time of CR
2

[about 70 s] and C
2
is charged
initially to 4V in a time C
2
(4R
3
+R
1
)
[about 1 s] and follows the decay of
the four series capacitors so that the
voltage across C
2
is the usual form for
an impulse voltage (see below).
The R
3
resistors are distributed as
shown as this assists to damp
oscillations during the discharge due to
the stray inductance of the circuit.

The voltage across C
2
is
V
C2
= -4V [ exp(-t/CR
2
) - exp{-t/C
2
(4R
3
+R
1
)}]
























6.8


Dr.Qais Alsafasfeh
Partial Discharges in Electrical Insulators

What are partial discharges (PDs)?
They are localised breakdowns which do not bridge the electrodes completely. They therefore
cause voltage pulses in the external circuit rather than complete short circuit causing voltage
collapse (as in a breakdown). However they usually also cause progressive deterioration of the
insulation which may lead eventually to breakdown and this is why they are important.
Examples are the development of a tree from a void in solid insulation, or the development of
tracks across an insulating surface.
It is therefore important to monitor partial discharges as a means of determining the condition of
the insulation and its rate of deterioration. This is achieved by detecting and observing the
electrical pulses by appropriate circuitry, or their physical side effects, such as pressure pulses
(ultrasonic), light pulses or chemical impurities.
Partial discharges can occur at many types of faults or imperfections in the insulating material:
(a) Breakdowns in voids (small bubbles of gas) within solid insulation;
(b) Breakdowns in voids between the insulating material and the electrode;
(c) Breakdowns in trees developing from (a) or (b);
(d) Breakdowns in gaps between the insulation and the electrode which are open to
the main gas gap (as discussed under flashover, page 7.4);
(e) The small sparks which are part of the tracking process (see page 7.1);
(f) Localised breakdowns across surfaces; and
(g) Corona at small-radius corners, edges and points which are surrounded by air or
other gas.


7.1
Dr.Qais Alsafasfeh
Voids in insulation
These will experience a potential difference V
v
which is a fraction of the applied voltage V (less
any voltage due to free charges on its surface):
V
v
= V (l/C
c
) / ( l/C
b1
+ 1/C
c
+ 1/C
b2
)
= V
r
x/d = V
vo
cos t (if V is alternating)
for a flat void, where x is the depth of the void and d is the width of the gap. (For other shapes
the void voltage V
v
could be larger or smaller by a factor of 2 or more).
If there is any voltage, V
fc
, due to free charges on its surface, the void voltage will be
V
v
= V
r
x/d - V
fc

The void will discharge when the applied voltage, V
v
, reaches the breakdown voltage, V
bd
, of the
void but will soon stop as the source of electrons in an insulator is obviously limited. The field in
the void after the discharge will be almost zero because the backwards field due to the charge
deposited by the discharge will cancel the applied voltage. However the externally applied
voltage, V, is still rising, so the breakdown voltage, V
bd
, of the void will be reached again and
another breakdown will occur, causing another pulse of current. Thus there will be a series of
pulses until the voltage maximum.
When the external voltage reaches zero, or shortly before, the negative breakdown voltage, V
bd
,
of the void will be reached and a negative pulse of current will occur, followed by a few more
as shown in the diagram below.

This can all be expressed mathematically as follows. The electrons and ions forming the
discharge will be deposited at opposite sides of the void and produce a back voltage V
back
so that
after the first void breakdown the instantaneous void voltage will be
V
v
= V
vo
cos t V
back
And, after the second void breakdown,
V
v
= V
vo
cos t 2.V
back

7.2
Dr.Qais Alsafasfeh
When the negative voltage exceeds V
bd
the void will break down in the opposite direction and
the new voltages after each breakdown will be, successively,
V
v
= V
vo
cos t V
back

V
v
= V
vo
cos t
V
v
= V
vo
cos t + V
back

V
v
= V
vo
cos t + 2.V
back

The void enclosed in insulation is thus seen to give a discharge pattern which is symmetrical
(same for both half-cycles), with the pulses tending to occur while the voltage is increasing in
magnitude. By contrast, a void at the earthed electrode will give much stronger pulses as the
voltage enters the positive half cycle (and vice versa).

Discharge patterns
A corona discharge will be characterised by discharge pulses at the peak voltages, the pulses at
the positive maximum being much greater in number if the point where the corona occurs is
earthed (and vice versa) since negative point-plane corona has much more frequent pulses than
positive point-plane corona.
Because different types of pd give different patterns, their detection and measurement may be
used as a diagnostic tool to determine the condition of the insulation system and its rate of
deterioration.

7.3
Dr.Qais Alsafasfeh
Partial discharge detection
The detection of PDs under ac conditions is basically filtration to remove the low frequency (50-
Hz) component. However this is not simple as the charge movements are of the order of pico-
coulombs in systems of pico-farad to micro-farad capacitance; the pulses are of milli-volt size
with test voltages of 1 to 100 kV. Furthermore, there are many sources of possible interference:
outside sources of electromagnetic radiation such as inverters, sparking from brushes on
commutators and partial discharges in the HV supply transformer and corona from the HV leads
to the specimen.
A common detection circuit is shown below, with Z being either an RCL parallel circuit to
resonate in the 50kHz to 100 kHz region, or simply a resistor followed by a high-pass or band-
pass filter. Clearly the circuit used will have considerable effect on the shape and magnitude of
the pulses: they are sharp, unidirectional pulses of a few microseconds duration so after filtering
they may be oscillatory.

This is a large subject which will not be covered in the present course, but if you are interested,
much further information may be found in Partial discharge detection in high voltage equipment
by Prof K.H. Kreuger (Engineering Library: 621.31937 20)

The classical display format
The classical method of displaying the signal is by use of an oscilloscope with a small 50-Hz
signal applied to the x-plates as the time base. The signal applied to the y-plates is the sum of a
smaller 50-Hz signal and the pulse signal from the detection circuit.
This gives an elliptical display with the pulses superimposed and makes their relative position in
respect of the supply waveform as described in the graphs on page 8.3 - clearly visible and easy
to interpret. Some photographs of such displays are reproduced overleaf (from K & Z, page 445).
The top one shows a corona discharge and agrees well with the description of ac corona given
earlier (see PPT for corona, last two slides): no or little corona visible during the positive half-
cycle, Trichel pulses on the negative half-cycle from onset voltage to extinction voltage. Also as
the voltage increases (in magnitude) to the peak value and then decreases, the pulse frequency
increases and then decreases.
The discharges in a void within the insulation occur at similar points on both the positive and
negative half-cycle it is symmetrical and the pulses occur before and over the peak voltage
(magnitude).

7.4
Dr.Qais Alsafasfeh

Elliptical display. (a) Point plane, (b) Void breakdown at inception,
(c) Void breakdown at twice inception voltage.
7.5
Dr.Qais Alsafasfeh
Computer-based PD Detectors
If the signals can be converted to digital form then far more detailed analysis can be done; the
results of the analysis can be displayed in many different formats; and records can more easily be
kept.
On the analogue screen the eyes persistence mean that several cycles of data are seen at any
one time but the display is constantly changing, and has a flickering effect. However, with a pc-
based pd detector several thousand cycles typically 2 minutes of data can be superimposed
to give a histogram which will have lost most of the purely random noise. It can be displayed
either as the classical elliptical display, like those shown on page 8.5, or as histograms, either in
two dimensions or three, as shown below and on the next page. One such PD detector is the
Haefely TE-571, which we have available in our laboratory.
The TE-571 detector is a complete partial discharge detection and analysing system designed to
measure, record, display and analyse PD's according to the IEC 270 standard. The timing of the
PD pulses relative to the alternating voltage is recorded and displayed. The hardware and
software provides continuous transfer of PD data to the computer during the measurements as
well as real-time processing, storage and displaying of PD measuring data.
Thus the pictures below show the PD pattern for a motor before and after steam-cleaning and re-
varnishing of the end windings, showing a significant reduction in the (previously excessive)
surface discharges.

Before cleaning After cleaning

The two 3D graphs below show the results of some recent research on corona showing the
surprising effect of humidity on the discharge pattern of air.

39% RH 89% RH



7.6

Dr.Qais Alsafasfeh








Many forms of display can be generated from the basic data, relating to the maximum pulse size,
average pulse size and number of pulses at a particular time interval, and the way in which these
vary with time.


A case study: about one hundred 6-kV
motors at HECs Lamma Power Station
wre tested and all the data consolidated
into just a single figure representing the
data for the positive half-cycle and
another for the negative half-cycle. The
equipment used was a TGA Analyzer for
which the connection is shown below.

0
100
200
300
400
500
600
700
800
900
1000
0 100 200 300 400 500 600 700 800 900 1000
NQN+
N
Q
N
-
All motors
MILL1A(B)
IDF1B(R)
CEP1
MILL1D(B)
7.7
Dr.Qais Alsafasfeh
It is clear from the graph of positive half-cycle data against negative half-cycle data, that, as
expected, almost all the data pairs are quite similar, but that there are 4 motors where there is a
very large difference between the two data. This is presumably because of damage to insulation
resulting in corona which only happens on one half-cycle).
Plotting the PD level against running hours, the effect of the vibration and the torque applied to
the windings during starts is apparent. Furthermore there are certain motors that appear to be
deteriorating much faster than the others, and need to be carefully monitored.
0
200
400
600
800
1000
1200
1400
1600
0 10,000 20,000 30,000 40,000 50,000 60,000 70,000
Running hours
P
D

l
e
v
e
l
s
1A
1B
1C
2A
2B
2C
4A
4B
4C
5A
5B
5C
6A
6B
6C
7A
7B
7C


Applications
Capacitive couplers can be used as above for on-line monitoring, or the current between each of
the windings (on open-circuit) and earth can be used for off-line measurements for both motors
and transformers.
Another method commonly used is the detection of the pressure vibrations caused by PDs:
high-frequency pressure sensors piezo-electric microphones are installed all around an oil-
filled transformer on the tank (outside!). By correlation of the arrival time and magnitude of the
pulse at the microphones it is possible to determine the location and severity of the PD site.
However the method is subject to many difficulties due to the different speed of sound in the
solid insulation, the oil and the steel tank.
Directional ultrasonic microphones can be used to detect corona and surface discharges on the
insulators of overhead lines.





7.8

Dr.Qais Alsafasfeh
UHF PD Detectors
When a partial discharge occurs inside gas-insulated switchgear (GIS), it causes brief
electromagnetic oscillations within the system at the various resonant frequencies. By use of
couplers or aerials inside the GIS the instrumentation can listen at a particular resonant
frequency of the GIS. Any signals received may be displayed on an oscilloscope triggered by a
50-Hz signal, so that a point-on-wave (POW) display may be seen see, for example, Figures 2
and 16 on the next page. This will indicate what type of PD caused the original PD pulse which
caused the oscillation.


In the case of GIS, the radii are about half a metre or so and the length of a section is typically 3
or 4 metres, so for resonance, half a wavelength is of the order of 0.5 to 4m. Using c = f, f =
(3 x 10
8
m/s) / (1 to 8 m) = 40 to 300 MHz (and multiples) i.e., ultra-high frequencies (UHF),
up to a GHz. Thus a discharge pulse lasting a microsecond or so, can cause oscillations at
resonant frequencies of up to GHz. These typically last for several microseconds (s).
Alternatively, the voltage from two aerials can be shown simultaneously so that the two pulses
due to the same pd can be compared: the time interval between the arrival of the pulse at the two
aerials is proportional to the difference in the distances from the pd source each aerial hence it
may be located.
This method is particularly useful for commissioning GIS.











8.9

Dr.Qais Alsafasfeh























Dr.Qais Alsafasfeh
8.10
Dr.Qais Alsafasfeh

Condition monitoring in oil-filled transformers

1. Introduction
Condition monitoring in oil-filled transformers may be done by PD detection, either
electrically or by ultra-sonics, as described in the previous Chapter, or as described in
this Chapter by dissolved gas analysis (DGA) or furfural analysis.
Both of these methods refer to oil/paper transformers for use in distribution (roughly 10
to 100kV) and transmission networks (roughly 100kV upwards). The coils are insulated
by paper and immersed in tanks of mineral oil. They are widely used except in built-up
areas where fire regulations may require dry-type or silicone-oil transformers.
A sample is drawn off and analysed. In Dissolved Gas Analysis (DGA) the concentration
of the by-products from electrical faults in the oil is measured, because the faults cause
local heating which results in partial decomposition of the oil. In Furfural Analysis the
concentration of certain by-products from the deterioration of the paper are measured.
The faults cause local heating and so partial decomposition of the paper.

2. Dissolved Gas Analysis
When overheating occurs, whether local (due to faults) or overall (due to loads above the
rated level), a number of chemical reactions occur. The nature of the fault affects the
percentage mix of the various gases formed during the fault and dissolving in the oil.
There are several approaches advocated by different institutions, but all are based on the
relative proportions of different gases. The gases produced by the different types of fault
are indicated in the table below.
Note: H
2
= hydrogen, CO = carbon monoxide, CO
2
= carbon dioxide,
CH
4
= methane, C
2
H
6
= ethane, C
2
H
4
= ethylene, C
2
H
2
= acetylene.

Discharge type Details Gases evolved
Corona or void discharges In oil H
2

In cellulose (paper) H
2
, CO , CO
2

Pyrolysis (over-heating) In oil, < 300
o
C CH
4
, C
2
H
6

In oil, 300 to 700
o
C C
2
H
4
, H
2

In paper < 300
o
C CO
2

In paper, 300 to 700
o
C CO
Arcing (high energy) H
2
, C
2
H
2

8.1
Dr.Qais Alsafasfeh
The value of DGA is that it
Gives advance warning of developing faults
Indicates the improper use of units (e.g., overloading)
Allows checks on new and repaired units
Assists the convenient scheduling of repairs
Allows the monitoring of units under temporary overload
Much more detail can be founded, if interested, at
http://www.nttworldwide.com/tech2102.htm

3. Furfural Analysis
3.1 Method
When overheating of the paper insulation
occurs, whether local or overall, a number
of chemical reactions occur resulting in
the degradation of paper and the
production of a water, carbon dioxide,
furfural and other members of the furan
family (see diagram on right). The
concentration (in ppm) of furfural in the
oil is an indicator of the progress of the
deterioration, as will now be explained.
The formal chemical name for furfural is
2-furfuraldehyde.
Paper is made from cellulose fibres
dispersed in water and drained to form a
mat. The cellulose fibres are prepared
from wood and form a natural long-chain
polymer (the monomer is a sugar, glucose
see diagram on right). Heating breaks
the cellulose molecules, resulting in
shorter cellulose molecules, and also
fragments the furfural compounds.
The tensile strength of the paper decreases
linearly with the average length of the
molecules (known as the degree of
polymerisation of the paper) as shown in
the graph overleaf.

The furfural family of compounds

Three sections of the cellulose molecule, a
natural polymer, and its monomer,
glucose, which is a sugar
The amount of the furfural compounds present due to the break-up of the long cellulose
molecules increases as the average length of the cellulose molecules decreases and about
20% of the furfural dissolves in the oil.
8.2
Dr.Qais Alsafasfeh
It has been shown experimentally that the logarithm of the furfural concentration in the
oil decreases linearly with the average length of the molecules. Hence the tensile
strength is linearly related to log(furfural concentration).


The furfural concentration is measured by liquid chromatography. When the paper is
embrittled the tensile strength of the paper has dropped to levels hard to test any
moderate voltage surge (due to a switching fault, for example) is likely to cause
mechanical failure of the insulation and an inter-turn fault.

3.2 Furfural Analysis Case Study
This section is included mostly for interest and as it may help to demonstrate and explain
the reason for furfural testing on oil-filled transformers.
The results of measurements of the furfural content of the oil from 190 transmission
transformers (from one manufacturer and in service with one particular utility) are shown
overleaf. It will be seen that there are 10 transformers whose FC is greater than or equal
to 2 mg/l (and 4 intermediate cases, 1 to 2 mg/l); but that the majority of those over 12
years old (25 out of 39) continue in good condition, that is, with FC values less than 1
mg/l. It would seem likely that the 10 poor-condition transformers had a harder service
life with more and longer periods close to, or above, full rated load. This follows from
the paper deterioration being caused by overheating.

8.3
Dr.Qais Alsafasfeh
0
2
4
6
8
10
12
0 5 10 15 20
Years of service
F
C
,

m
g
/
l

The furfural content of the oil from 190 transmission transformers (from one
manufacturer and in service with one particular utility) versus years in operation
In 22 cases a second test had been done on particular transformers and the ratio of Test 2
to Test 1 is plotted below. It shows that over a period of almost 2 years there is no
significant ageing trend. If the two data of just over 150% and the one below 150% are
discarded, the standard deviation of the remainder is 25%: the 95% confidence limits of
+50% are indicated in the figure. The three extreme values are more than 6 standard
deviations away from zero and cannot therefore be considered as part of the data
population.
-200
-150
-100
-50
0
50
100
150
200
0 5 10 15 20 25
Time separation of T1 and T2, months
P
e
r
c
e
n
t
a
g
e

c
h
a
n
g
e

The surprisingly low result with a 7-month gap between the two measurements was in
fact a drop from 0.15 to 0.02 mg/l, the latter being the resolution limit of the instrument.
It seems likely that this was a data collection, or sample handling, error.
The tolerance range for the individual measurements is +36%. This error level is
surprisingly large: it is over twice as large as that found for tests in which 4 different
laboratories in Australia were compared by independently measuring the furfural
concentration in 5 different oil samples.
8.2
Dr.Qais Alsafasfeh
The poor repeatability of the measurements described here may well originate in the long
time inevitably elapsing between the sample taking and the measurements. The utility
in question is some considerable distance from Beijing, where the furfural analysis was
done: the samples could not all be sent out on the same day and were held until a
sizeable batch could conveniently be sent to the testing laboratory, allowing a potential
for oxidation of the samples.
More details of the case quoted above can be founded, if interested, in J.M.K. MacAlpine and
C.H. Zhang, Observations from Measurements of the Furfural Content of Transmission
Transformers, Electrical Power Systems Research, Vol.57, pp.173-179, 2001.
























8.5

Dr.Qais Alsafasfeh
IEEE International Symposium on Electrical Insulation, Vancouver, BC, Canada, June 8-11, 2008


High Voltage Insulation Failure Mechanisms
Rudolf Bruetsch
Von Roll Switzerland Ltd., 4226 Breitenbach, Switzerland
Makoto Tari
Von Roll Isola Japan, Tokyo, 150-0013, Japan
Klaus Froehlich, Tilman Weiers
1)
and Ruben Vogelsang
2)

High Voltage Laboratory , Swiss Federal Institute of Technology Zurich, 8092 Zurich, Switzerland
Actual addresses:
1)
Areva T&D Australia Ltd., Brisbane QLD, Australia;
2)
Brugg Cables, 5201 Brugg, Switzerland


Abstract- There are many statistics dealing with failure causes
of high voltage rotating machines. While these statistics prove
the relevance of the insulation as failure cause, they give little
information on the real failure mechanism. This paper gives
results of a research project which was launched to investigate
degradation and final breakdown of the mica insulation in
rotating machines.
Final failure of the insulation is usually caused by electric
breakdown through a tree channel. Formation and propagation
of trees in the mica insulation therefore reflect the ageing
process of the insulation. Dominating ageing factors of the mica
insulation are thermal and mechanical stress. Manufacturing
defects of the mica insulation - such as delaminations, cracks,
voids and wrinkled or damaged mica layers - accelerate the
formation and propagation of trees. Good workmanship and
proper selection and application of the insulation materials are
important factors for the lifetime of high voltage rotating
machines.

I. INTRODUCTION

Several statistics are dealing with failure causes of high
voltage rotating machines in [1 - 6]. Some of these statistics
only specify the part of the machine which failed without
giving any deeper insight in the failure mechanism. Other
publications distinguish between the damage which caused
the machine to fail and the root cause which effected the
damage. All statistics agree that breakdown of the winding
insulation plays a major role in the failure of high voltage
rotating machines.
Several authors investigated failure mechanisms of the
mica insulation under laboratory conditions using sample
bars or other test specimen [7, 8]. Their main findings are:
1) Although electric breakdown is causing the final failure of
the electrical insulation, electrical stress is not the
dominating ageing factor. It is rather believed that the ageing
mechanism is dominated by thermal degradation of the
binder resin, mechanical stress caused by vibration and
switching pulses and stress caused by the different thermal
expansion coefficients of the materials involved.
2) Ageing under thermal, mechanical and electrical stress
shows an increase in lifetime at moderate temperatures up to
approximately 130C and a rapid decrease if the ageing
temperature is increased up to 180C. The increase in
lifetime at moderate temperatures is explained by a decrease
of internal stress and crack formation in the binder resin,
whereas the accelerated ageing at higher temperature is
caused by an increasing thermal degradation of organic
matter.
The goal of the study was to develop a failure model
dependent on the operating conditions of the high voltage
insulation in rotating machines.

II. INVESTIGATIONS AND RESULTS

A. Electrical treeing in the needle-plane arrangement
The needle-plane arrangement in epoxy cubes allows direct
observation of the phenomenon of electrical treeing. Test
samples as shown in fig. 1 with an edge length of the epoxy
cubes of about 4 cm were prepared. Electrical tree
propagation was monitored with a video camera.














Fig. 1. Needle-plane test sample

Treeing and the formation of branches started immediately
after voltage application. Propagation of the trees towards the
grounded electrode was detained by the embedded mica tape
and trees spread on its surface instead. When the tape edge
was reached treeing continued towards the grounded
electrode (Fig. 2). Measured average breakdown times were
around 50-100 hours. In all experiments trees could only
penetrate the mica tape if it was damaged, e.g. by bending or
folding.
Treeing in the needle-plane arrangement was also inves-
tigated by partial discharge measurements. Fig. 3 shows the
Dr.Qais Alsafasfeh
IEEE International Symposium on Electrical Insulation, Vancouver, BC, Canada, June 8-11, 2008









Fig. 2. Treeing in the needle-plane arrangement. The mica tape acts as
barrier for electrical trees

typical evolution of PD peaks during an experiment which
lasted about 60 hours or 1.110
7
50Hz voltage cycles.
PD activity in the tree channels started after 10 hours and
then rapidly decayed to a relatively low and constant rate
with values oscillating for 30 hours. This effect is usually
attributed to the formation of conductive degradation
products of the binder resin. During the last 20 hours of the
experiment the PD activity slightly increased until
breakdown.













Fig. 3. Evolution of partial discharge peaks during electrical ageing in the
needle-plane arrangement

B. Electrical treeing in the mica insulation
One-metre-long model bars with an insulation thickness of
2 mm prepared with VPI and resin-rich technology were
used for 3U
n
voltage endurance tests. The voltage applied
was 32 kV AC. To verify if electrical trees can be found in
the mica insulation sections of the bars were examined by X-
ray methods. Fig. 4 shows an example of a bar section; the
tree is accented graphically. The tree path could also be
traced by burning off the binder resin of the mica insulation
in an oven and carefully removing the mica tape layer by
layer.
Applying both methods it was found that treeing mostly
started at the edge of the copper conductor stack which is
expected from the increased electric field strength at this
point. While the tree was propagating it travelled to the outer
grounded electrode, by following the edges of the mica tape
(Fig. 5).
It is assumed that the tree discharge pierced the layers
coincident with locations where damages or imperfections
were observed such as voids, cracks, delaminations, resin
accumulations at tape overlaps and wrinkled or disrupted










Fig. 4. Tree path in a section of a sample bar uncovered by X-ray methods
(left) and by burning off the binder resin and removing the mica tape layer
by layer (right)










Fig. 5. Micrographs of the breakdown tree in a mica insulated sample bar.
The pictures show the tree at the tape overlap (left), at the bar edge (center)
and after breakdown. The tree tends to follow the mica tape layers

mica layers. Such defects abet the formation and propagation
of trees in the mica insulation and thus reduce its lifetime.
These assumptions could be confirmed by micrographs of the
mica insulation of bars with short life time showing many
wrinkles and voids (Fig. 6). These findings are in good
agreement with the results of the needle-plane experiments,
where it was observed that the electrical tree was not able to
penetrate the mica tape if it was undamaged (Fig. 2). The tree
could only pierce the tape if it had been bent or kinked before
embedding in the needle-plane test specimen.










Fig. 6. Mica insulation showing perfect (left) and wrinkled tape layers with
many voids (right)

It is therefore obvious that the manufacturing quality of the
mica insulation has an influence on the lifetime of high
voltage coils or bars. Voltage endurance at 3U
n
was done to
verify this. The Weibull plots in fig. 7 give a significant
difference in the lifetime of sample bars with and without
wrinkles in the mica insulation as shown in fig. 6.
Dr.Qais Alsafasfeh
IEEE International Symposium on Electrical Insulation, Vancouver, BC, Canada, June 8-11, 2008















Fig. 7. Voltage endurance of sample bars with and without wrinkled mica
tape layers. The horizontal bars at P[F] = 0.63 indicate the 95% confidence
intervals around the characteristic lifetime of each test group.

C. Thermal and mechanical ageing
It was stated in the introduction that thermal and
mechanical stress are the dominating ageing factors of the
insulation in high voltage rotating machines. Hence a large
number of voltage endurance tests was performed to
investigate the two stress factors. Fig. 8 shows the influence
of the ageing temperature on voltage endurance investigated
in a combined thermal / electrical ageing test. Results show
that the lifetime of the insulation is higher at 160C than at
20C which supports the statements cited above [7, 8]. This
can be explained by a higher flexibility of the binder resin at
elevated temperature which minimises the risk of crack
formation, but also by a reduction of internal stresses coming
from the curing reaction which usually takes places at
temperatures around 160C. However, at 180C thermal
degradation of the binder resin was the dominating ageing
factor and lifetime of the model bars was considerably
reduced. According to T. Weiers [10] the optimal ageing
temperature for maximum lifetime of epoxy based mica
insulations is about 90C.














Fig. 8. Influence of the ageing temperature on voltage endurance

Mechanical stress is exerted on the insulation of high
voltage rotating machines due to vibration, different thermal
expansion coefficients of the materials involved and rapid
changes in load. Vibration experiments were conducted on
sample bars using a newly developed vibration equipment
which allowed 100 Hz vibrations with an amplitude of 0.5
mm. Insulation life was measured applying vibrational and
electrical stress simultaneously and compared to bars
exposed to electrical stress only. Two different insulating
materials were used for this test: Insulation A, a combination
of a glass-backed mica tape and an epoxy based VPI resin
and insulation B, a combination of a polyesterfilm/-fleece
backed mica tape and a polyester based VPI resin. Results
show that the lifetime of insulation B was reduced through
vibration by a factor around 10, whereas the lifetime of
insulation A was not altered (Fig. 9).














Fig. 9. Influence of mecha nical stress (vibration) and type of insulation on
voltage endurance














Fig. 10. Influence of electrical and mechanical stress (vibration) on a
polyesterfilm-backed mica insulation. The micrograph shows an electrical
tree following a crack between adjacent tape layers

Micrographs of insulation B show delaminations parallel to
the mica layers which are supposed to trigger and accelerate
treeing; but the formation of cracks perpendicular to the mica
layers could not be observed. Polyester resins are known to
give a lower adhesion and a higher shrink during curing
compared to epoxies. This effect is enhanced by the carrier
material of the mica tape used for insulation B; polyester film
results in a lower mechanical strength of the insulation than
glass fabric.
Dr.Qais Alsafasfeh
IEEE International Symposium on Electrical Insulation, Vancouver, BC, Canada, June 8-11, 2008


The vibration experiments did not include mechanical
abrasion of the slot corona protection as the sample bars were
not placed in a slot model. Abrasion of the conductive layer
in the slot by the sharp edges of the laminated stator core is
an additional factor damaging the insulation and reducing its
lifetime, especially if the wedging of the bar in the slot
becomes loose. As a consequence partial discharge occurs,
adding chemical degradation of organic insulation materials
to the mechanical erosion caused by vibration. In the
literature, this vicious circle is considered one of the main
causes of premature failure of high voltage rotating machines
[1]. Fig. 11 shows the impact of vibration: the slot corona
protection is worn except for the positions of the air ducts
[9].












Fig. 11. Abrasion of the slot corona protection due to vibration [9]

III. CONCLUSIONS

Breakdown of the winding insulation plays a major role in
the failure of high voltage rotating machines. The final
failure of the insulation is usually caused by electric
breakdown through a tree channel. Formation and
propagation of trees in the mica insulation therefore reflect
the ageing process of the insulation. It was found that treeing
is accelerated by defects in the mica insulation such as
delaminations, cracks, voids and wrinkled or damaged mica
layers. Such defects are either manufacturing failures or they
result from ageing.
Ageing is not dominated by electrical stress but rather by
the combination of different stress factors, of which thermal
and mechanical wear are the most important. Thermal stress
of epoxy based insulation systems is relevant if it exceeds the
optimal temperature range substantially, i.e. above 135-
140C.
Mechanical stress of the insulation is due to vibration,
different thermal expansion coefficients of the materials
involved as well as rapid changes in load. It was found that
vibration alone had no negative effect on the voltage
endurance of an insulation system composed of glass backed
mica tapes and epoxy based binder resins, whereas a
reference system consisting of a combination of a
polyesterfilm/-fleece backed mica tape and a polyester based
VPI resin sustained a considerable reduction of lifetime due
to the formation of cracks parallel to the mica layers. In the
long run mechanical stress and thermal degradation of the
binder resin can provoke a loosening of the bar or coil in the
slot. The resulting vibrations will cause abrasion of the slot
corona protection by the sharp edges of the laminated stator
core and partial discharge will occur irrespective of the
insulation system used.
Insulating systems used for rotating high voltage machines
are mature products and will last for decades if they are
applied properly. It was found that early breakdown of the
insulation can very often be attributed to manufacturing
defects; thus it appears that good workmanship and correct
selection and application of the insulating materials are
important factors for the lifetime of rotating machines. The
experimental study of individual deterioration factors is
essential for future establishment of residual life estimation
of the generator insulation and further work in this field is
necessary.

ACKNOWLEDGEMENT

The project was performed at the high voltage laboratory of
the Swiss Federal Institute of Technology in Zurich in
collaboration with several industrial partners and electric
power companies. A detailed description of the test
equipment and the specimens is given by T. Weiers [10] and
R. Vogelsang [11].

REFERENCES
[1] D.L. Evans, IEEE Working Group Report of Problems with
Hydrogenerator Thermoset Stator Windings, IEEE Trans. on Power
Apparatus and Systems, PAS-100, pp. 3284-3293, (1981)
[2] W. Zwicknagl, Zum Dauerverhalten der Hochspannungsisolationen
umlaufender elektrischer Maschinen, Oesterreichischer Verband der
Elektrotechnik, Journal fr Elektrotechnik und Maschinenbau, E und
M, 98/6, pp. 221--233, (1998)
[3] R. Bomba, U. Gross, J. Kaiser, VGB PowerTech, 11/2005, pp. 87-92,
(2005)
[4] CIGRE Study Committee SC11, EG11.02, Hydrogenerator Failures
Results of the Survey, (2003)
[5] O.V. Thorsen, M. Dalva, A survey of faults on induction motors in
offshore oil industry, petrochemical industry, gas terminals and oil
refineries, IEE Trans. on Ind. Appl., 31/5, pp. 1186-1196, (1995)
[6] N. Srb, Erfahrungen mit Stossspannungsprfungen an elektrischen
Maschinen, Allianz Report, 70/2, pp. 58-62, (1997)
[7] A. Wichmann, P. Grnewald, J. Weidner, Betriebliche Einfluss-
grssen auf die elektrische Lebensdauer von Hochspannungsisolierun-
gen in Turbogeneratoren, ETG Fachberichte, 16, pp. 44-48, (1985)
[8] K. Kimura, Y Kaneda, Breakdown Voltage and SEM Observation on
Mica Insulation Systems after Thermal Aging and Mechanical
Fatigue, Proc. 7
th
Int. Symp. on High Voltage Engineering, pp. 151-
154, Dresden, (1991)
[9] A. von Streitberg, Neue Erkenntnisse aus Generatorschden
Gewinn fr die Betriebssicherheit, Allianz Report, Heft 1, pp. 29-35,
(1995)
[10] T. Weiers, Eine Methode zur aussagekrftigeren Bewertung von
Maschinenisolierungen, Dissertation Swiss Federal Institute of
Technology Zurich No. 17363, Zurich, (2007)
[11] R. Vogelsang, Time to breakdown of high voltage winding
insulations with respect to microscopic properties and manufacturing
qualities, Dissertation Swiss Federal Institute of Technology Zurich
No. 15656, Zurich, (2004)