Documente Academic
Documente Profesional
Documente Cultură
2, February 1981
873
Toshiba Corporation
KAWASAKI, JAPAN
Abstract
In this paper, local voltage oscillation of each
turn within sections of interleaved transformer winding
Because of interlacing of turns, each
are discussed.
turn within a pair of sections of the winding has
initial voltage distributions quite different from
final voltage distributions. This difference brings
about local oscillation when a steep impulse voltage
enters the winding.
The effects of the front time of impulse voltage
and the number of turns per section of local oscillation are investigated. And the local oscillations are
elucidated by calculating the voltage oscillation occurring in the equivalent circuit of a pair of sections.
LOCAL OSCILLATION
The interleaved winding was developed and patented
Figure 1 (a) is a diagrammatic
by G. F. Stearn [21k
view showing arrangement of turns within a pair of
In the arrangement shown in
sections in the winding.
Fig. 1 (a), each section consists of 8 turns and the
numerics in each block of the sections indicate the
For the purpose of explanation in
number of the turn.
this paper, a pair of sections is termed "a coil" and
the turns numbered 1 through 8 are termed "the first
strand", the turns numbered 9 through 16 "the second
strand"c.
INTRODUCTION
Because of its excellent property against lightning surge, the interleaved winding is widely used as
the high voltage winding of a core type transformer.
high-frequency oscillation occurs between
However,
sections of the winding when such steep impulse waves
The
front-of-waves enter it [l11.
as chopped-waves or
magnitude of this local oscillation depends on the
steepness of the applied impulse voltage and/or the
number of turns per section.
All impulse voltages impinged on a transformer do
For example, switching
not have standard waveforms.
surges generated at gas insulated substations (GIS)
generally have much of high-frequency oscillatory comMeanwhile, transmission lines have a tendency
ponent.
toward higher voltage *today, and 1,000 kV-class UHV
transmission lines are also being planned in some
The interleaved winding is one of the most
countries.
suitable high-voltage windings used on UHV transformers, and the higher the voltage, the more the
Under
number of turns per section of the windings.
these circumstances, as the interleaved winding will
it would
more possibly have a large local oscillation,
be important to elucidate this phenomenon clearly.
4i1A3
j9
6 14 5 8 16
[1l3
(a) Stearn
Fig. 1
winding
24 113291
I=
[J1fJ146 J15J74168J
.--
(b)
Nuys
winding
oscillation [5].
81 IEEE
874
A report written by R. van Nuys says that, when
steep impulse waves such as chopped-waves or front-ofwaves enter a conventional interleaved winding, a high
oscillatory voltage occurs locally between sections, at
the center of its radial build.
He made various approaches to clarify the phenomenon and proposed a winding allowing little oscillation [1, [61.
Figure 1 (b) is a diagrammatic view showing arrangement of turns within a pair of sections in the
Nuys winding.
A feature of the Nuys winding is that
the two strands do not exchange radial locations in the
upper and lower sections of a coil. That is, the first
strand is located outside the second strand in either
section. On the other hand, the Stearn winding has the
first strand located outside the second strand in
section 1, and the first strand located inside the
second strand in section 2.
Considering that the major cause for local oscillation is magnetic coupling, R. van Nuys explained the
difference in oscillation between the Stearn's and his
as follows:
Simulating one coil using a transmission line
model as shown in Fig. 2 (b) and solving this circuit
gives a distribution of the fundamental harmonic component of oscillating current in Fig. 2 (a). As pointed by the arrows in Fig. 2 (b), therefore, currents in
the two strands flow in reverse directions each other
in both sections.
Then the direction of the magnetic
flux induced by the current depends on which conductor
of the first or second strand occupies outward in radiIn the Stearn winding, the magnetic
al location.
coupling in the upper and lower sections is positive,
In the Nuys winding, the
reinforcing oscillation.
magnetic coupling is negative, suppressing oscillation.
However, the above explanation involves a question. If both Nuys and Stearn windings have the current
distribution shown in Fig. 2 (a) and the current flowing into the first strand is reverse to that flowing
into the second strand, the magnetic flux between t-urns
is also reverse every other turn.
Thus, the difference in the magnetic flux within the coil between the
two windings is made only by a one-turn radial deviaThis might be insuffition between sections 1 and 2.
cient when one explains the significant difference in
oscillatory voltage between the Stearn and Nuys windings.
MEASUREMENTS
Local oscillation of the six types of windings as
listed in Table 1 was measured using a low-voltage impulse generator in order to compare the conventional
Stearn winding with the Nuys winding and to investigate
the effects of the number of turns per section.
Each
of the models had an inside diameter of 900 mm, a spacing of 5.6 mm between sections, and a 1 mm thick insulation between turns.
Table 1
Type of models
IIIIIIE]I
I I I I 1 m
#900 mm
1 mm
5
4
3
0
--
Section 1
Section 2
1~~ ~ ~ ~
Fig. 2
of current flow
directions
20
No. of section
Fig. 3
(b) Directions
875
0D
Section 1
35%
Section 2
60%
64%
Section 3
_~~
Section 4
J
/
0
Section 5
Section 6
0D
35'
98%
Fig. 4
Figure 4 gives oscillograms showing voltage waveforms that appear in the windings A and B when 0.5/50s
The voltages were
full-waves are applied to them.
outside the radial build
measured at four positions:
inside the radiin Fig. 4 (c)
between sections
,
al
build
coils
between
at
the
center
the
and
of
the
radiof
center
the
Each oscillogram
.
radial build between sections
is accompanied by the crest value of voltage developed.
In the case of winding A, a large oscillatory voltage
An oscillatory comas indicated by Nuys is observed.
ponent having a frequency of approx. 800 kHz which is
attenuated rapidly in
-X
voltage
level
Maximum
voltage
times
the
few
decided
by
value
appears
calculated
cycles
the
between
from
is
superposed
on
which
is
the
initial distribution.
the
coils,
final
voltage
2.7
dis-
distributed
0.2
0.8
0.6
0.4
Front
time
1.0
1.2
(ks)
the
Fig. 5
Wdg.
-_0
-x -
:
:
(Steamn,
30
turns
per
section)
linearly
in the coil.
between sections contains a larger oscillatory com3.9 times the value calculated from
ponent, which is distribution.
final
voltage
Figure 5
of the applied
pearing
sections.
at
the
center
The data
were
of
the
obtained
radial build
on
between
windings A, B and
876
C (Table i). As is obvious, the oscillatory voltage of
winding A reaches 2.9 times the value calculated from
the final voltage distribution when Tf = 1.0 is but
that of winding C remains at 1.85 times even when Tf =
It thus is found that the oscillatory voltage
0.2 jis.
is affected considerably by the number of turns per
section. With shorter front time of the applied voltage
compared with the time required for surge to propagate
As can
in a coil, larger oscillatory voltages result.
be understood from the curve on winding A, however, the
oscillatory voltage tends toward saturation if the front
The voltage
time is reduced to shorter than 0.5 isec.
between sections of winding B remains nearly constant
even if the front time is reduced.
Measured
Calculated
Calculated
o
b_
1o, I
a:
.'
G-
U,_
00
oL
Don
-0 - -
I;
6 '..00
4.00
(C00
4.00
j0u00
0002d 00
00
J0
4.00
O0
00
4.00
(> 00
uo
00
0o
4.
.1 0)0)
'
!E0
J<
2 00
T WME
',
4.00
SEC i
1.00
to
..
Section
Section 2
00
o oo
00
:n
turns
per
Fig. 7
section)
per
section)
877
speaking, inductance elements have resistance which deBut, in the calculation, conpends upon frequency.
stant values at the natural frequency of the coil are
A digital computer was used to calculate the
used.
oscillation occurring in the equivalent circuits. Since
using bulky computer memory and spending much computation time were considered uneconomical, circuits
having only a pair of sections were calculated, though
more accurate data could have been gotten if calcuThe data obtained
lations were made on more sections.
might be sufficient for elucidation of the oscillating
phenomenon.
Figure 7 shows calculated data and measured data of
potentials to ground at the node in the center of the
The models used for these calculation
radial build.
and measurement were D (Stearn winding) and E (Nuys
In both amplitude and
winding) shown in Table 1.
frequency of the oscillation, the calculated data agree
relatively well with the measured data.
A comparison between the two windings reveals not
only a difference in the amplitude of oscillation but
also a difference in the phase of oscillation at the
In the Stearn
node in the center of the radial build.
winding, the first strand and the second strand within
the same section (b and f of section 1, and d and h of
section 2, Fig. 7) are oscillating in the same phase,
and section 1 (b, f) and section 2 (d, h) are oscillating in mutually opposite phases. While, the Nuys wind.ing has the first strand (b and d) in the two sections
oscillating in the same phase, but the second strand
(f and h) hardly oscillating.
The difference between the two windings becomes
more distinct if comparison is made on the voltage disFig. 8 shows the calculated
tribution along turns.
In the Stearn
voltage distributions in both windings.
winding, the anti-node of the oscillation appears at
the center of the radial build and the node appears at
the inside and outside crossovers. In the Nuys winding,
130
at T =
100
1.0 ,us
at T = 1.8 ,us
10
100
: Final distribution
r-H
.H
P-H
50
50
50
First strand
-40
Fig. 8
(b)
878
The initial voltage distribution along the first
and second strands in the Stearn winding shown in Fig.
9 (a) gives a nearly symmetric with respect to the node
If the 50 % potential
in the interleaved connection.
is assumed a grounded potential and the discussion is
focused on only the first strand, then this strand
would be oscillating in the same pattern as when a rectangular wave is applied to the L-C ladder circuit
shown in Fig. 10 (a) as long as the mutual inductance
Potential e at a distance x away from the
is ignored.
grounding terminal can be given by the following
equation:
00
Ca
e = EX + E n=l
X21 Un sin njfx'cosCnt
2H
where,9
n
30
I
a
*I
-First
6060
Number of turns
*'
Second strand
strand
a
0
30
c
First
strand
Calculated
data
of
per
initial
Second
60
4
Number of turns
stranad
section)
voltage
tributions
Ct: Capacitance between turns
(Cg
Cg: Capacitance to earth
c) n =
I/L-
and
dis-
< Ct)
00
EX + E
OTOrIoyI T
I 1 ITT T
00ri
I~iDooI
Fig. 10
E(l-x) + E
sin
coo
X7l
n=l
n7rxcos6)
sin n 7r (l-x)-coso t
U
n
5n=l
ir
(l-x)
879
Potential
Final
distribution
Current
%V
I\\\,
%'
%'
1.
e\f"f
Pd
%l
g
I
IL
PFh
.1
I
x
.o_
I-'T I
0 I
+-
Secti on 2
------
&
I. I.
ba
.4
The authors wish to thank the management of ToshiCorporation for permitting them to publish this
paper.
REFERENCES
W.
ACKNOWLEGEMENT
Section 2
Fig.
pitin i- %1
bu i
-t.,asr-
Ix x
ATFV.
.
lation
,/.
I
I
winding
Section
-%le.0
.,
Section 1
f
Voltage,
current,
and
magnetic
flux
CONCLUSIONS
The effects of the number of turns per section and
the front time of impulse voltage on the local oscil-
P.89
880
Discussion
D. H. Ryder (GEC Power Transformers Limited, Stafford, England): I
was very interested to read paper 80 SM 704-7 and to see that the
question of internal oscillations within disc coils is still of some interest
to transformer engineers. In the very short time available before the
reading of this paper I cannot give a full account of our views but we
are hoping that an account of our own work may be the subject of an
IEEE paper shortly.
Meanwhile it should be stated that George Steam, myself and the
co-authors of Reference [ 1] below, started the use of the interleaved
winding in 1950. It was then an English Electric design, but following
mergers in the UK Manufacturing Industry, it has now been used for
10 years by General Electric Company (England). Since 1950 we have
used the design, together with later improvements, on perhaps 1000
transformers, from 132kV to 765kV, and only one or two of these
transformers have suffered lightning impulse failure in service.
A number of the above transformers were subjected to Front of
Wave chopped voltages according to the NEMA rules, and no difficulty
was encountered during tests. In fact our investigations showed that
such is the time delay to breakdown characteristic of insulation, designs
for front of wave testing could use the normal calculations relevant to
the full wave impulse voltage distribution.
This is not to say that internal oscillations do not occur. They do:
and they can occur in non-interleaved, and in intershielded windings.
But the first conclusion of the present paper is important and is justified by our own work. The number of turns per section is indeed significant, and we find that for almost all practical designs the effect of
oscillatory internal coil voltages is less relevant than the intercoil
voltages.
I would be interested to know whether the authors have measure-
REFERENCE
[1]
30 x
MEASUREMENT AXIS OF
MAX.
POINT
OSCILLATION AMPLITUDE
Vo
STEARNS
FIGURE 4
1.
2.
3.
4.
MAX.
AMPLITUDE
VAN NUYS
FIGURE 7
b.
d.
f.
h.
875+30
1.75 SECTIONS
.875 + .125
2 SECTIONS - 865 +* 13
1.25
SECIONS
1.25 SECTIONS
+
.625 + .125
SECTIONS
.75SECTIONS
.375 + .10
- 375
2 SECTIONS
.375 +
.25
.25 SECTIONS
2 SECTIONS
=.125 +.37
.125 + .1
The initial voltage distributions shown in Figures 9a and 9b indicate the maximum stress on the turn insulation (between turns 0 and
30 in the first section and between turns 30 and 60 in the second section) is essentially the same for the Stearns and Van Nuys windings.
Comparison of the oscillograms in Figure 7 also indicate the stress on
the turn insulation between b and f and between d and h is also similar
(approximately .5). While the amount of turn insulation required is
therefore similar, the Steams winding may require larger ducts depending on the front time, tums per section, number of sections, and the
diameter of the coil.
Although the peak oscillating voltages between coils and sections
in Stearns windings may be quite large they are of short duration (a
few tenths of a microsecond). Also with longer fronts, fewer turns per
section and more sections, the actual stress across the duct will decrease.
The minimum duct size may also be dictated by thermal considerations.
881
surge may enter the transformer windings and the magnitude of the
local oscillation at that time.
Manuscript received October 14, 1980.
Tsuneharu Teranishi was bom in Nagoya, Japan,
on February 10, 1946. He received the B.S. and
the M.S. degree in electrical engineering from
Nagoya University, in 1968, and 1970.
_ From 1970 to 1973 he continued to research the dielectric characteristics of liquid. In
1973, he joined the transformer design department in Hamakawasaki Work of Toshiba Corporation. In 1979, he joined the Heavy Apparatus Laboratory of Toshiba Corporation, and has
engaged in the development of the insulation
technique for a power transformer.
Mr. Teranishi is a member of JIEE.
Masami Ikeda was born in Yokohama, Japan, on
September 14, 1943. He graduated from Toshiba
Techniical College in 1966.
In the same year, he joined the Heavy Apparatus Engineering Laboratory of Toshiba Corporation, Kawasaki, Japan, where he engaged in
the development of the insulation technique for
a power transformer. He has been devoted to
studies conceming the breakdown of transformer oil. Since 1975 he has been Specialist
of High voltage Laboratory.