Académique Documents
Professionnel Documents
Culture Documents
B and
B
H
u
= = . (2)
In addition to this, eddy currents in the lamination sheets introduce frequency dependence.
The well-known procedure [24] for deriving the effective frequency dependent complex
permeability is briefly sketched below. Faradays law
t
c
V =
c
B
E (3)
and Amperes law
t
c
V = +
c
D
H J , (4)
in combination with the constitutive relations
o = J E , c = D E ,
h
= B H , (5)
and time-harmonic assumption lead to
( )
2
h
j j o e o ec V = = +
2
H H H. (6)
For lower frequencies when wave propagation can be ignored (i.e., o ec >> ), one
has
2
h
j o e o ~ .
Fig. 1: Laminate infinite in z direction, with a width in y direction much larger than its thickness 2b, exposed to
a H field in z direction.
For analysis of the magnetic field in a laminate, the simple geometry illustrated in Fig. 1 is
appropriate. The magnetic field is applied in the z direction, hence the only component of the
magnetic field strength is H
z
which varies only in the x direction, H
z
= H
z
(x). In one
dimension, Eq. (6) reduces to
x
y
( )
z
H x
2b
9
2
2
2
z
z
H
H
x
o
c
=
c
, (7)
which has the general solution
1 2
( ) e e
x x
z
H x A A
o o
= + . (8)
The field strength on the both sides of the laminate is assumed to be H
0
. For the reason of
symmetry the following condition is obtained
0
( ) ( )
z z
H b H b H = = . (9)
The final expression for the magnetic field strength then becomes
( )
( )
0
cosh
( )
cosh
z
x
H x H
b
o
o
= . (10)
The effective, complex permeability of a lamination is given as the average magnetic flux
density B in the laminate normalized to the surface magnetic field strength H
0
,
eff eff eff
j ' '' =
h h
0 0
1 tanh( )
( )d
2
b
z
b
B b
H x x
H H b b
o
o
= = =
}
. (11)
This expression accounts for the effect of hysteresis without saturation, and the effect of eddy
currents. It is assumed here that additional (or excess) losses are either negligible or have a
similar frequency dependence so that they can be incorporated in the expression (11) for
eff
.
3.3 Detailed Hysteresis Model
Later in this Chapter, some results obtained with a more detailed model of the magnetic
hysteresis, eddy current and excess losses will be reported, so therefore, this more detailed
model, which has been developed by Dr. David Ribbenfjrd in [21], is described here in short
(for a thorough description and explanation, see [21]).
The total hysteresis is a combination of three different phenomena, namely, static
hysteresis, eddy current effects and excess eddy currents. For the detailed hysteresis model,
the following approach has been used. The static hysteresis is modeled using Bergqvists lag
model [25, 26], the classical eddy currents are modeled using Cauer circuits [22, 27], and the
excess losses are modeled using an approach by Bertotti [20].
3.3.1 Static Hysteresis
The Bergqvists lag model of static hysteresis starts from the idea that the magnetic material
consists of a finite number of pseudo particles n
p
, i.e., volume fractions with different
magnetization. The total magnetization is then a weighted sum of the individual
magnetization of all pseudo particles.
10
Fig. 2: An-hysteretic curve (left), play operator (middle), and resulting hysteresis curve (right); figure taken
from [21].
The hysteresis curve for one particle is introduced by applying a play operator with a play
equal to the pinning strength k (which will determine the width of the hysteresis curve) on
the an-hysteretic curve, see Fig. 2, where m is the magnetization of the actual pseudo particle,
and is the back field i.e. the field that will give the magnetization m if no hysteresis is
present.
Using a population of pseudo particles with different pinning strengths allows constructing
minor loops. An individual pinning strength
i
k is assigned to every pseudo particle, where k
is the mean pinning strength, and
i
is a dimensionless number for particle i. The total
magnetization is then given by a weighted superposition of the contributions from all pseudo
particles (Fig. 3).
Fig. 3: Weighted superposition of the contributions from pseudo particles describes a minor loop; figure taken
from [21].
The expression
an s
s
2
( ) arctan
2
H
M H M
M
t _
t
| |
=
|
\ .
(12)
is used for the an-hysteretic magnetization, where M
s
is the magnetization saturation and is
the susceptibility at H = 0. For infinite number of pseudo particles, the total magnetization of
the material is then given by
( )
an an
0
( ) ( ) ( )d
k
M cM H M P H
= +
}
, (13)
where c is a constant that governs the degree of reversibility, and the integral describes the
hysteretic behaviour (irreversible part). P
k
is a play-operator with the pinning strength k, and
() is a density function describing the distribution of the pseudo particles. Finally, the
magnetic flux density is obtained from B =
0
(H+M).
m m
q
q
H H
k
k
2k
2
1
k 2
2
k 2
3
k
H
H H H
11
3.3.2 Excess Losses
Excess losses are caused by microscopic eddy currents induced by local changes in flux
density due to domain wall movements. For the detailed model, an approach described by
Bertotti [20] is used. In this approach, a number of active correlation regions are assumed
randomly distributed in the material. The correlation regions are connected to the micro-
structure of the material like grain size, crystallographic textures and residual stresses. In
Bertottis model, the resulting contribution to the magnetic field strength is given by
0 0
excess 2
0 0
4 2 d d
1 1 sign
2 d d
n V G bw B B
H
n V t t
o
| |
| |
= + |
|
|
\ .
\ .
, (14)
where w is the width of the laminate and 2b, as before, its thickness. G is a parameter
depending on the structure of the magnetic domains. n
0
is a phenomenological parameter
related to the number of active correlation regions when the frequency approaches zero,
whereas V
0
determines to which extent micro-structural features affect the number of active
correlation regions.
The parameters n
0
and V
0
are by definition frequency independent, but they are expected in
reality to depend on the amplitude of the B field [28]. Since the precise form of this
dependence is unknown, their values are usually adjusted empirically for a given amplitude.
In the simulations reported here one set of (empirically determined) values is used, although
the amplitude of the B field varies slightly in the measurements.
3.4 Measurements and Model Adjustments
The magnetic measurements were carried out using a Single Sheet Tester. It consists of two
equal U-shaped yokes placed face-to-face to each other (Fig. 4). The magnetic sheet to be
tested is placed between the yokes and most of the flux is forced through it due to its high
permeability. For the measurement of the flux in the test material, a coil is surrounding the
strip which is connected to a flux meter. The magnetic field strength is measured with a Hall
probe placed close to the surface of the sample and connected to a Tesla meter. A sinusoidal
H field was applied to the sample; the H and B field values were measured for 100 periods
and numerically filtered. Thereafter, the mean values at different phase angles of the B and H
fields were calculated and used for the experimental verification of the complex- model.
Fig. 4: Cross sectional view of the Single Sheet Tester.
12
The measured (mean-value) H-B curve is then approximated with a complex- ellipse
characterized by the permeability
meas
. This is done by matching both the peak values H
p
, B
p
and the area A of the ellipse to the measured results (i.e. the measured H-B curve and the
complex- ellipse will have the same area and peak values). This is of course appropriate as
long as the shape of the measured H-B curve is close to an ellipse, i.e., if saturation effects are
not too pronounced. The power loss per cycle, the area A within the H-B curve, is given by the
integral
meas
meas meas meas
0
d
d d
d
T
H
A B H B t
t
= =
} }
, (15)
where B
meas
and H
meas
are time dependent fields (correlated trough
meas
) defined as
approximations of the time dependent measured B and H fields, respectively, and T is the
duration of a period. If the measured H field is assumed to vary sinusoidally, then one can use
the approximation
( )
j
meas p p
( ) Re e cos( )
t
H t H H t
e
e = = , (16)
with the derivative
meas
p
d ( )
sin( )
d
H t
H t
t
e e = . (17)
The approximation of the measured B field becomes then
( ) ( ) ( ) ( )
j j
meas meas p meas meas p p meas meas
( ) Re e Re j e cos( ) sin( ) .
t t
B t H H H t t
e e
e e ' '' ' '' = = = +
(18)
By inserting Eq. (18) and (17) into Eq. (15) one gets
meas 2
p
A
H
t
'' = . (19)
Furthermore, from the relation
meas p p
H B = one obtains
( ) ( )
2
2 2 2
p
meas meas meas
p
B
H
| |
' '' = + =
|
|
\ .
, (20)
which implies
( )
2
2
p
meas meas
p
B
H
| |
' '' =
|
|
\ .
. (21)
Both
meas
' and
meas
'' are functions of frequency.
13
-400 -300 -200 -100 0 100 200 300 400
-1.5
-1
-0.5
0
0.5
1
1.5
H
B
(a)
-400 -300 -200 -100 0 100 200 300 400
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
H
B
(b)
Fig. 5: H-B curves from measurements (blue) and complex- model (green) with
meas meas
meas
j ' '' =
, for
(a) f = 50 Hz and (b) f = 400 Hz.
Fig. 5 compares the measured H-B curves with complex- ellipses, generated with the
adapted
meas
at frequencies f = 50 Hz and 400 Hz.
eff
as defined in Eq. (11) is a function of frequency and of a vector x = (
r
,
h
, o b
2
)
containing the model parameters. It is adjusted to measured data by numerically minimizing
the expression
eff meas
2
1
( ) ( ) ,
i i
N
i
e e
=
x (22)
with respect to x.
meas
( )
i
e are the measured complex permeability values, defined by
Eq. (19) and (21), at N different frequencies
i
= 2f
i
, i = 1, , N. Measurements at N = 9
different frequencies ranging from 50 Hz to 2 kHz (see Fig. 6 and Fig. 7 below) were
performed on a 100 mm 3.2 mm strip of the non-oriented magnetic material M600 with a
thickness of 2b = 0.5 mm.
3.5 Comparison between Model and Measurements
Since the measurement setup was quite sensitive to noise, the measurements had to be
numerically filtered. Adjusting
eff
to the filtered data using Eq. (22), the following model
parameter values were obtained:
r
= 3366,
h
= 0.477 rad, and ob
2
= 0.243 Sm, i.e.,
o = 3.8910
6
S/m which is somewhat larger than the true DC conductivity
o
dc
= 3.3310
6
S/m since excess losses were included in the classical phenomenological
form (11). In Fig. 6, the real and imaginary parts of the measured complex permeability are
compared at different frequencies with the adjusted
eff
.
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
14
10
0
10
2
10
4
10
6
10
8
0
500
1000
1500
2000
2500
3000
Frequency [Hz]
'
ef f
/
0
''
ef f
/
0
'
meas
/
0
''
meas
/
0
Fig. 6: Real and imaginary parts of the measured complex permeability (symbols) and of the fitted
permeability function (curves), normalized by
0
.
The agreement is quite satisfactory considering the simplicity of the model, especially at
higher frequencies. The deviation between
meas
'' and
eff
'' at the lowest frequencies is
probably due to saturation effects which are not properly taken into account by the
expression (11) for
eff
, see for instance the measurement at 50 Hz (Fig. 5(a)), where the
amplitude had to be chosen large enough for the signal not to be covered by noise.
Below, the H-B hysteresis curves are shown for all measured frequencies. Measurement,
simple model, and detailed model are represented by solid green lines, dashed blue lines and
dotted red lines, respectively.
-400 -300 -200 -100 0 100 200 300 400
-1.5
-1
-0.5
0
0.5
1
1.5
f = 50 Hz
H
B
-400 -300 -200 -100 0 100 200 300 400
-1.5
-1
-0.5
0
0.5
1
1.5
f=100
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
15
-400 -300 -200 -100 0 100 200 300 400
-1.5
-1
-0.5
0
0.5
1
1.5
f = 200 Hz
H
B
-400 -300 -200 -100 0 100 200 300 400
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
f = 400 Hz
B
-400 -300 -200 -100 0 100 200 300 400
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
f = 500 Hz
B
-400 -300 -200 -100 0 100 200 300 400
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
f = 800 Hz
B
-400 -300 -200 -100 0 100 200 300 400
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
f = 1000 Hz
B
-400 -300 -200 -100 0 100 200 300 400
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
f = 1250 Hz
H
-500 -400 -300 -200 -100 0 100 200 300 400 500
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
f = 2000 Hz
B
Fig. 7: H-B curves from measurements (solid green line), detailed model (dotted red line) and complex-
model (dashed blue line) with calculated from expression (11), at different frequencies ranging from 50 Hz to
2 kHz.
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
H [A/m]
B
[
T
e
s
l
a
]
16
The above way of defining a best fit of ellipses to the more complicated H-B hysteresis
relations approximately preserves both H and B amplitudes and magnetic losses in the whole
frequency range. This is illustrated in the Fig. 7, where the measured H-B curves are
compared with the corresponding complex- ellipses and the detailed model at different
frequencies. As can be seen, the simple model agrees very well with the measurements as
long as saturation is not too strong, which means for low amplitude fields and/or for
frequencies higher than about 200 Hz.
17
4 High Frequency Winding Model
In this Chapter, a winding model based on a lumped element approach with three different
levels of discretization has been developed. The developed models are analyzed using state
space analysis in the frequency domain and the impedances for the three different models are
compared to each other to detect the new phenomena emerging for higher frequencies as the
model discretization is made finer and finer. Parts of this chapter are based on [29].
4.1 Transformer Winding Model for High Frequency Applications
In the field of transformer winding modeling, various approaches and tools are available.
Among the most common tools are lumped element circuits [5, 3032]. Usually all the turns
of one or two discs are lumped together into one inductive element (segment) of the model,
which leads to a decreased computation time but also to a reduction of the models upper
frequency limit, typically to values around some 100 kHz.
In this thesis, lumped element models with much higher resolution (up to 4 segments per
turn) are used, in order to increase the upper frequency limit.
Each lumped element represent a part, a section of the physical geometry with similar
quantities like magnetic flux, electric potential, resistance and etc, and these lumped elements
are connected together to represent the whole geometry. For power transformers, the windings
are divided into finite sections represented by lumped resistance, inductance, and capacitance,
where each section should be small enough so that it can be assumed that the current through
it is constant and not influenced by the displacement current which will be noticeable at
higher frequencies.
Up to a few hundreds of kilo hertz, the displacement current will not be so remarkable and
can be approximated to zero, so that a winding can merely be modeled by means of its self
and mutual inductances and resistances alone. But at higher frequencies, the aforementioned
approximation is no longer valid, and the displacement currents from a section to other
sections or to conductive bodies have to be accounted for, to make the model realistic, and this
is done by means of capacitors. The total capacitance for a particular section is then split into
two halves and located at both ends of the section. As mentioned above, the complete winding
model is made by connecting all the sections together.
In this thesis, the transformer winding is a single phase continuous disc winding (with the
order of turns as shown in Fig. 9), composed of quadratic discs as in Fig. 8. Also, since the
low voltage winding is on a much lower voltage than the high voltage winding, it (the LV-
winding) is replaced by ground in the models developed here.
4.2 Three Different Resolutions of the Model
Three different model resolutions are studied: in the models labeled 1, 2, and 3, each turn in
the discs is modeled by one, two, or four segments, respectively. This implies that there are
one, two, or four turn-to-turn capacitances between any two neighboring turns in a particular
disc of the HV winding, respectively (see Fig. 8 from left to right). Further, there are one, two,
18
or four disc-to-disc capacitances between any two turns facing each other in two neighboring
discs, respectively. Also, at the innermost and outermost turn of every disc, there are in
addition one, two, or four capacitances to ground, respectively (these capacitances are not
shown in Fig. 8 due to the lack of space).
Fig. 8: The three different levels of discretization with the nodes numbered in an increasing sequence from one
winding end to the other. The resolution increases from model 1 to model 3, and the disc-to-disc capacitances
between neighboring discs are oriented perpendicular to the plane and are not shown here.
Fig. 9: The cross section of the continuous disc winding.
Fig. 10: One segment and its electrical circuit equivalence.
It has been argued in the literature on the basis of simulations and measurements that the
impulse voltage distribution in transformer windings, and the winding impedances for higher
frequencies can be analyzed using air-core self and mutual inductances neglecting the iron
core without serious error (see for instance [3334]). This implies that the complex
permeability model developed in Chapter 3 does not need to be included in the calculations of
L
self R
seg
h
i
/ /2
ks
t
1 2 n
w
h
tot
o
d
i
t
Ground
i
d
Ground
Plane of
symmetry
2n+1
2
1
3
2n-1
mutual
inductances
model 2
2n
4
5
1
2
n+1
n
3
mutual
inductances
model 1
~ U
I
4n+1
2 3
4
1
5
6
4n
4n-2
model 3
mutual
inductances
self
inductance
resistance
4n-1
7
8
9
o
/ 2 /
19
inductances and losses. In order to further facilitate the calculation of self and mutual
inductances, the discs have been chosen to be of quadratic shape here. The whole HV winding
thus consists of straight segments which are all either parallel or perpendicular to each other,
so that the self and mutual inductances can be calculated by simple analytic formulas [35].
Every segment is modeled by one resistance in series with one self inductance, as it is
depicted in Fig. 10, and there are mutual inductances between any two parallel segments of
the whole winding. For simplicity, self inductances and resistances are shown in Fig. 8 on one
segment only, and some of the mutual inductances are indicated by arrows. In case of model
1, also the connection of the voltage source for impedance measurement for one single disc is
shown.
Finally, all the inductance and capacitance parameters (i.e., all model parameters except for
the damping resistances) are estimated from the physical winding geometry, and not fitted to
measurements, and they are discussed in Sections 4.3 and 4.4. The models are analyzed by
solving their state space equations [3032] in both the frequency- and time domain, which is
discussed in Section 4.5 of this chapter and in Chapter 9, respectively.
4.3 Calculation of the Capacitances
The capacitance depends on the geometry of the conductors and on the permittivity of the
dielectric material between them. Here, the relation for the capacitance between two planar
surfaces, capacitance = permittivity area / distance, is used [36]. The total capacitance
between two turns C
tt
in the quadratic disc is then approximately given by
i o i
tt i 0
i
2
4
2
h
C
t
c c
t
+ +
=
/ /
, (23)
where h is the height of the conductor,
i
is twice the insulation thickness,
i
is the relative
permittivity of the conductor insulation,
i
/ is the inner length of one side of the square
discs,
o
/ is the outer length of one side of the square discs, and the addition of 2
i
to h
accounts for the fringing effect [37]. The total capacitance between two discs, if they are close
enough to each other and if air is used as insulation between them, is approximately given by
2 2
o i
dd air 0
i ks
( / 2) ( / 2)
4 C c c
t t
=
+
/ /
(24)
where
air
is the relative permittivity of air (=1) and
ks
is the distance between two discs. The
total capacitance between one disc and the outer ground wall is given by
o tot
og air 0
o
4
h
C
d K
c c =
/
(25)
and the total capacitance between one disc and the inner ground wall is given by
i tot
ig air 0
i
4
h
C
d K
c c =
/
(26)
20
where K is the total number of discs in the winding, h
tot
is the total height of the winding, and
d
0
and d
i
are the outer and inner distances between the winding and the ground respectively.
4.4 Calculation of the Inductances and Resistances
For the calculation of the self and mutual inductances, formulas in [35] are used. The self
inductance L
self
of each straight segment with the length / , height h and width w in the disc is
0
self
2
ln 1
2 0.2235( )
L
w h
t
| | | |
=
| |
+
\ . \ .
/
/ . (27)
The mutual inductance M between two segments which are perpendicular to each other is
zero, whereas for two parallel segments of length / , separated by a distance x (as in Fig. 11),
it is given by
2 2
0
ln 1 1
2
x
M
x x x
t
| | | |
| | | |
| |
= + + +
| |
| |
\ . \ .
\ . \ .
/ / /
/
/
. (28)
Fig. 11: Two filaments with negligible cross section area with same lengths.
When the segments are parallel but have different lengths, as in Fig. 12, the mutual
inductance is given by
( ) ( )
2
m p m q p q
M M M M M
+ +
= + + , (29)
Fig. 12: Two filaments with negligible cross section area with different lengths.
where for example M
m+p
is the mutual inductance between two straight wires both having the
length m+p and being placed relative to each other as in Fig. 11, and which for the symmetric
case p = q reduces to
m p p
M M M
+
= . (30)
x
p
m
q
/
x
/
21
Of course, in the formulas for the mutual inductances it is assumed that the conductors have
very small cross section areas, which is just an approximation. The resistance R
seg
of each
segment is assumed to be of the form
0
seg
1 1
2( )
f
R
wh w h
t
o
o o
| |
= +
|
|
+
\ .
/ , (31)
where o is the conductivity of the conductor and f is the frequency. The first term is the DC
resistance [38] (which has a vanishing effect for the frequencies dealt with here) and the
second term accounts for the skin effect at higher frequencies [3940]. Since proximity losses
are not included in the model, a numerical factor > 1 has been introduced and adjusted so
that a realistic level of resonance damping is obtained.
4.5 State Space Model for the Winding
The circuit model for the three different winding models in Fig. 8 and Fig. 9 of the single
winding consists of Kn
i
winding sections resulting in K(n
i
+ 1) nodes and Kn
i
inductive
branches and associated capacitances and resistances, where n
i
= 2
i1
n, for i = 1, 2, 3 for
model 1, 2 and 3 respectively, and where n is the number of turns in one disc. K is the total
number of discs used in the winding and hence one will arrive at the following two matrix
equations by considering the voltage difference between the nodes of inductive branches and
the current conservation at the nodes:
d
dt
= I C V , (32)
T
d
dt
= + V L I R I . (33)
Here V and I are the vectors containing the voltages at the nodes and the currents in the
inductive branches, respectively
1
2
( 1)
( 1) 1
i
i
i
Kn
K n
K n
V
V
V
V
+
+
(
(
(
(
=
(
(
(
V
.
,
1
2
( 1)
1
i
i
i
K n
Kn
Kn
I
I
I
I
(
(
(
(
=
(
(
(
I
.
. (34)
The matrix connects the currents and voltages and consists of 1, 1 and 0, and
T
is the
transpose of
( 1)
i i
K n n + K
(
(
(
( =
(
(
(
S 0 0
0 S 0 0
0
S 0
0 0 S
. . .
. . .
. .
, where
( 1)
1 0 0
1 1 0 0
0
0 1 1
0 0 1
i i
n n +
(
(
(
( =
(
(
(
S
. .
.
. . . .
. .
. .
(35)
22
and 0 is a zero matrix. The resistance matrix for the whole winding R is a diagonal matrix
disc
disc
disc
disc
i i
Kn Kn
(
(
(
( =
(
(
(
R 0 0
0 R 0 0
R
0 0 R 0
0 0 R
. .
.
. . . .
.
. .
, (36)
composed of the resistance matrix for individual discs R
disc
, where
seg,1
seg,2
disc
seg, -1
seg,
0 0
0 0 0
0 0 0
0 0
i
i
i i
n
n
n n
R
R
R
R
(
(
(
(
=
(
(
(
R
. .
.
. . . .
.
. .
, (37)
is composed of the resistances of all segments in one disc (for example, R
seg,j
is the resistance
of segment j (see Eq. (31))). The inductance matrix for the whole winding L is composed of
smaller matrices
disc 12 13 1
21 disc 23 2
disc
1 2 disc
i i
K
K
K K
Kn Kn
(
(
(
( =
(
(
(
L L L L
L L L L
L
L
L L L
.
.
. . . .
. . . .
. .
, (38)
where the off-diagonal matrices L
kj
are n
i
n
i
matrices for the mutual inductance between disc
k and j, and the matrix in the diagonal L
disc
is the inductance matrix for a single disc being
composed of the mutual inductances M (Eqs. (28)(30)) and self inductances L
self
(Eq. (27))
of the segments in one disc
self,1 12 13 1
21 self,2 23 2
disc
self, 1
1 2 self,
i
i
i i
K
K
n
K K n
n n
L M M M
M L M M
L
M M L
(
(
(
(
=
(
(
(
L
.
.
. . . .
. . . .
. .
, (39)
(for example, M
kj
and L
self,j
are the mutual inductance between segment k and j, and the self
inductance of segment j, respectively). The total capacitance matrix C in Eq. (32) is
23
( ) ( ) ( )
DD DD
( ) ( ) ( ) ( )
DD DD DD
( ) ( ) ( ) ( )
DD DD DD
( ) ( ) ( )
DD DD
( 1) ( 1)
2
2
i i
i i i
i i i i
i i i i
i i i
K n K n + +
( +
(
+
(
(
=
(
+
(
(
+
C C C 0 0
C C C C
C 0 0
C C C C
0 0 C C C
.
.
.
.
, (40)
where, as mentioned before, i = 1, 2, 3 for model 1, 2 and 3 respectively. C
(i)
is the specific
capacitance matrix for model i, and for model 1, the roughest model, it is
(1) (1) (1)
tt ig tt
(1) (1) (1)
tt tt tt
(1) (1) (1)
tt tt tt
(1)
(1) (1) (1)
tt tt tt
(1) (1) (1)
tt tt tt
(1) (1) (1)
tt tt og
1 1
0 0
2 2
1 3
0 0
2 2
0 2 0 0
0 2 0
3 1
0
2 2
1 1
0 0
2 2
C C C
C C C
C C C
C C C
C C C
C C C
(
+
(
(
(
(
(
(
=
C
. . .
. .
.
. .
. .
. . .
. . .
( 1) ( 1) n n + +
(
(
(
(
(
(
(
(
(41)
where
(1) (1) (1)
tt tt ig ig og og
, , C C C C C C = = = , (C
tt
is the turn-to-turn capacitance (Eq. (23)) and
C
og
/C
ig
are capacitances to ground (Eqs. (25)(26))). For model 2, the next finer model, it is
(2) (2) (2)
tt ig tt
(2) (2) (2)
tt ig tt
(2) (2) (2)
tt tt tt
(2) (2) (2)
tt tt tt
(2)
(2) (2) (2)
tt tt tt
(2) (2) (2)
tt tt tt
(2)
tt
1 1
0 0 0
2 2
0 0 0 0
1 3
0 0 0
2 2
0 0 2 0
0 0 2 0 0
3 1
0 0 0
2 2
0 0
C C C
C C C
C C C
C C C
C C C
C C C
C C
+
+
=
C
. . . .
. . .
. . .
. .
. .
. .
. . .
. . . .
(2) (2)
tt og
(2) (2) (2)
tt tt og
(2 1) (2 1)
0
1 1
0 0 0
2 2
n n
C
C C C
+ +
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
+
(
(
+
(
. . . .
(42)
where
(2) (2) (2)
tt tt ig ig og og
/ 2, / 2, / 2 C C C C C C = = = . For model 3, the finest model, it is
24
(3) (3) (3)
tt ig tt
(3) (3) (3)
tt ig tt
(3) (3) (3)
tt ig tt
(3) (3) (3)
tt ig tt
(3) (3) (3)
tt tt tt
(3) (3) (3)
tt tt tt
(3)
1 1
0 0 0 0 0
2 2
0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0 0
1 3
0 0 0 0 0 0 0
2 2
0 0 0 0 2 0 0 0 0
C C C
C C C
C C C
C C C
C C C
C C C
+
+
+
+
= C
. . . . . .
. . . . . .
. . . . .
. . . .
. . .
. .
.
(3) (3) (3)
tt tt tt
(3) (3) (3)
tt tt tt
(3) (3) (3)
tt tt og
(3) (3) (3)
tt tt og
(3) (3) (3)
tt tt og
(3) (3)
tt tt o
0 0 0 0 2 0 0 0 0
3 1
0 0 0 0 0 0 0
2 2
0 0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0
1 1
0 0 0 0 0
2 2
C C C
C C C
C C C
C C C
C C C
C C C
+
+
+
+
.
. .
. . .
. . . .
. . . . .
. . . . . .
. . . . . .
(3)
g
(4 1) (4 1) n n + +
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(43)
where
(3) (3) (3)
tt tt ig ig og og
/ 4, / 4, / 4 C C C C C C = = = . The matrix
( )
DD
i
C in Eq. (40) accounts for the
capacitive coupling between two neighbouring discs and it has the form
( )
dd
( )
dd
( )
DD
( )
dd
( )
dd
( 1) ( 1)
1
0 0
2
0 0
0 0
1
0 0
2
i i
i
i
i
i
i
n n
C
C
C
C
+ +
(
(
(
(
(
=
(
(
(
(
C
. .
. .
. .
. .
. .
(44)
where
( )
dd dd
/
i
i
C C n = for all three models i.e. i = 1, 2, 3 (C
dd
is the disc-to-disc capacitance
according to Eq. (24)).
The pre-factors 1/2 in the capacitance matrices are due to the fact that the capacitances
connected to the first and the last nodes in a disc account only for a half segment, and the pre-
factors 1, 2 and 3/2 in the diagonals are due to the fact that, if there is no capacitance to
ground, the sum of the elements in a row/column must be equal to zero.
When an external voltage source is connected to a node (k), its node voltage is no longer
unknown (k = 1 here). The voltage at that node V
k
and its time derivative dV
k
/dt should
therefore be separately inserted in (32) and (33) as inputs accompanied by the corresponding
columns of matrices C and
T
. Eq. (32) and (33) are then transformed to
d d
d d
k
V
t t
= + I C V O (45)
T
d
d
k
V
t
= + P V L I R I . (46)
Here, O consists of one column taken out from the C matrix corresponding to index k, and P
consists of the k:th column taken out from
T
(transpose of ). In addition, appropriate
changes should be made according to the connection of the terminals of the discs and node
grounding [41]. By rearranging the terms in these equations and putting them in one matrix
equation, Multi Input Multi Output (MIMO) state space model of the lumped parameter
circuit can be formulated as:
25
d
d
k
V
t
' = + X A X B (47)
where
(
=
(
V
X
I
,
1
1 T 1
(
=
(
0 C
A
L L R
and
1
1
d
dt
(
(
' =
(
(
C O
B
L P
. (48)
The state vector X consists of all the nodal voltages (except the applied one) and inductor
currents of the lumped circuit. By taking the Fourier transformation of the equation system
and selecting all state variables as outputs, one will arrive at
( )
1 ( )
( ) j
( )
k
V
e
e e
e
'' = =
X
TF II A B , (49)
where e is the angular frequency, II is the identity matrix with the same size as A, and
1
1
je
(
'' =
(
C O
B
L P
. (50)
TF(e) contains all the transfer functions of the nodal voltages (except the applied one) and
inductor currents with respect to the applied voltage V
k
.
4.6 Comparison between the Three Models
The first comparison between the three models is made for a single disc winding, and the new
phenomena that emerge with increasing model resolution are studied. The disc in the models
consist of n = 10 turns of varnished copper wire with the conductivity = 5.810
7
S/m, the
conductor height h = 7 mm and width w = 3 mm. The inner sides of the square disc have a
length of
i
/ = 1.2 m, and the gap between any two neighboring conductors (turns) i.e. twice
the insulation thickness is
i
= 0.4 mm, and there is no ground wall which means that d
o
and d
i
are set to infinity in the calculations (see Fig. 9 for a geometrical illustration of the
parameters). These are also the dimensions of the experimental setup which will be described
in Chapter 5. The reason for this choice of dimensions will be explained in the same Chapter.
The magnitudes and phases of the calculated in-impedances Z(f)=U(f)/I(f) (see Fig. 8) are
compared to each other in Fig. 13(a) and (b), respectively. It can be seen that several
resonances occur: the first, pronounced impedance maximum is the fundamental resonance of
the winding, due to the total inductance and series capacitance of the whole disc. As it will be
argued in Section 7.1, the three following resonances can be interpreted as radial resonance
modes, and the two after that, which form a pronounced impedance minimum above 10 MHz
and do not appear in the lowest-resolution model 1, as azimuthal resonance modes.
As can be seen in Fig. 13, for model 1 the impedance becomes purely capacitive i.e. it
becomes of the form Z = (jC
hf
)
-1
after the radial modes i.e. for frequencies higher than about
6 MHz (C
hf
is the winding model capacitance at high frequencies). This is not a physical
reality and it means that model 1 is for sure not valid for that part of the frequency spectrum.
For model 2 and 3, the impedance becomes purely capacitive in the end of the frequency
spectrum after the azimuthal modes, which as expected would mean that with finer
discretization the model becomes valid for higher frequencies.
26
10
6
10
7
10
0
10
2
10
4
Frequency [Hz]
I
m
p
e
d
a
n
c
e
a
m
p
l
i
t
u
d
e
[
O
]
model 1
model 2
model 3
10
6
10
7
-100
-50
0
50
100
Frequency [Hz]
I
m
p
e
d
a
n
c
e
p
h
a
s
e
[
d
e
g
r
e
e
s
]
model 1
model 2
model 3
Fig. 13: (a) Impedance amplitude and (b) impedance phase for the different models.
radial modes
azimuthal modes
fundamental coil
resonance
(a)
(b)
27
5 Frequency Response Measurements
In this Chapter, the measurement device and the experimental setup are introduced and the
choice of dimensions for the setup is explained. Also, the different measurements performed
are described.
5.1 Impedance Measurement Device and Dimensioning of the
Experimental Setup
A network analyzer Bode 100 from Omicron Electronics [42] (frequency range 1 Hz
40 MHz) was used for the impedance measurements (see Fig. 14).
Fig. 14: Impedance measurement device Bode 100 from Omicron Electronics.
As mentioned before, the shape of the discs is chosen quadratic so that all the self and mutual
inductances can be calculated by simple analytic formulas from [35]. The location of the
resonances of the winding in the frequency spectrum depends on the physical geometry and
material properties of the winding, and generally, larger dimensions of the winding leads to
larger inductance and capacitance values, which in turn leads to the resonances occurring for
lower frequencies.
The measurement device can measure up to the frequency of 40 MHz, but the
measurements will be more sensitive to the effects of environmental noise and measurement
cables for the higher part of this frequency range. Consequently, the measurements will be
more disturbed and unreliable for this high-frequency part of the frequency range, and due to
this fact, the geometrical size of the discs had to be chosen so that all the interesting
phenomena and resonances occur for frequencies below approximately 20 MHz.
The most effective way to satisfy this requirement is to design the square discs with large
side lengths. So the constructed discs consist of n = 10 turns of varnished copper wire with
rectangular cross section (7 mm 3 mm). The inner sides of the square discs have a length of
28
1.2 m, and the gap between any two neighboring conductors (turns) is varying between
0.4 mm (= twice the insulation thickness) and about 1 mm because of manufacturing
irregularities. The cross section dimensions of the varnished copper wires (7 mm 3 mm) are
regular wire dimensions used in power transformers, and the number of turns can not be
chosen too high (the discs will be to heavy and impractical to handle), or too low (there will
be too few resonances), so n = 10 turns seemed reasonable and was chosen.
10 units of these discs were manufactured by the transformer manufacturing company
Nordtrafo AB [43] (see Fig. 15).
Fig. 15: Separate disc units manufactured by Nordtrafo AB.
5.2 Impedance Measurement Results
The first measurements were performed on each of the ten different single disc units
separately (Fig. 16).
Fig. 16: Frequency response measurement on a single disc.
By comparison between the impedances of the different discs, it could be affirmed that due to
the manufacturing irregularities mentioned above, the impedances of the discs differ from
29
each other more and more as the frequency is increased. This seems consistent, because for
higher frequencies, the capacitances between the turns in the discs play a larger part in
shaping the impedance. Since a capacitance is per definition sensitive to small distance
changes between two conductors, and since (as mentioned earlier) the gap between any two
neighboring conductors (turns) is varying between 0.4 mm and about 1 mm (because of
manufacturing irregularities), hence the discs have different capacitive features for higher
frequencies, leading to different high frequency impedances. In Fig. 17, the measured
impedance for four of the discs is plotted to illustrate the differences.
10
6
10
7
10
0
10
2
10
4
Frequency [Hz]
I
m
p
e
d
a
n
c
e
a
m
p
l
i
t
u
d
e
[
O
]
Fig. 17: Impedance measurement for four different discs.
It can from Fig. 17 be seen that for frequencies up to approximately 300 kHz, the discs have
the same impedance value. This is the so called inductive regime where the impedance
behaves as Z = jL
lf
(where L
lf
is the inductance of a disc for low frequencies), and since the
inductance is not so sensitive to small irregularities in the physical geometry, hence the discs
have all the same impedance. But as the frequency is approached to the fundamental coil
resonance frequency, the capacitance starts to play in, and it can be seen that from now on the
impedances do not coincide with each other. The radial and azimuthal modes are still
there for all of the discs, but they do not have exactly the same location, shape and amplitude.
Measurements were then performed on units with a varying number of discs connected
together. The distance between two neighbouring discs
ks
was changed between
approximately 1 mm to 5 cm, and different types of connections between the discs were tried,
and all these for comparison between models and measurements. It was found that as the
distance between the discs is decreased, the models deviate more and more from the
measurements, and the reason for this will be discussed and explained in Chapters 6 and 8.
Next, the model simulations will be compared to measurements for verification.
fundamental coil
resonance
radial modes
azimuthal modes
30
31
6 Model Verification
In this Chapter, model 3 is verified by comparison with measurements. The model is
compared with measurements for a winding with one, three, six and nine discs respectively,
and it is shown that the resonances predicted by the model also occur in the measurements.
6.1 Comparison of Model with Measurements
The impedance magnitudes of model 3 and measurement for one single disc are compared in
Fig. 18, where it can be seen that they have an overall satisfactory agreement. Also, the
comparison shows that the radial modes and the high-frequency azimuthal modes, which
model 1 is unable to produce and which model 2 produces partially, are no model artifacts but
real physical phenomena.
10
5
10
6
10
7
10
2
10
4
Frequency [Hz]
I
m
p
e
d
a
n
c
e
a
m
p
l
i
t
u
d
e
[
O
]
simulation
measurement
Fig. 18: Comparison between measurement and model 3 for a single disc.
In Chapter 7, the physical meaning of the radial and azimuthal resonances will be explained
and discussed.
In the measurements, the radial modes are shifted toward somewhat higher frequencies
compared to those in the simulations, but they are fully recognizable. Such a shift is expected
to occur due to the proximity effect which has not been taken into account in the model
calculations reported here, and which is going to be discussed in Chapter 8.
f
az
32
Next, the impedance magnitudes of model 3 and measurements for a three, six and nine
disc winding, with the distance between two discs
ks
being approximately 6 mm, are
compared in Fig. 19 to Fig. 21, respectively.
10
5
10
6
10
7
10
2
10
4
Frequency [Hz]
I
m
p
e
d
a
n
c
e
a
m
p
l
i
t
u
d
e
[
O
]
simulation
measurement
Fig. 19: Comparison between measurement and model 3 for a three disc winding.
10
5
10
6
10
7
10
2
10
4
10
6
Frequency [Hz]
I
m
p
e
d
a
n
c
e
a
m
p
l
i
t
u
d
e
[
O
]
simulation
measurement
Fig. 20: Comparison between measurement and model 3 for a six disc winding.
33
10
5
10
6
10
7
10
2
10
4
10
6
Frequency [Hz]
I
m
p
e
d
a
n
c
e
a
m
p
l
i
t
u
d
e
[
O
]
simulation
measurement
Fig. 21: Comparison between measurement and model 3 for a nine disc winding.
Even here in Fig. 19Fig. 21, the comparisons show an overall satisfactory agreement
between model 3 and measurements, confirming the general accuracy of the model. But, as
was the case in Fig. 18, there are shifts between some of the resonances reproduced by the
model and the measured resonances. The reason for this, as for the resonance shifts for one
single disc, is expected to be the proximity effect which is neglected in the models and which
is going to be discussed in Chapter 8.
34
35
7 Interpretation
This Chapter, which is mainly based on [29], deals with the radial and azimuthal resonances.
The features of these resonances are explained, and it is shown that they are present also in a
winding with several discs.
7.1 Explanation of the Different Oscillation Modes
As it was mentioned in Section 4.6, the three resonances somewhere between 1 MHz to
8 MHz in Fig. 13 for a single disc are called radial resonances and the ones somewhere
between 10 MHz and 20 MHz are called azimuthal resonances, and it could be seen in Fig. 18
that these resonances also occur in the measurement, which means that these are physical
realities and not model artifacts, and now the physical meaning of these resonances will be
investigated and explained.
7.1.1 Radial Resonance Modes
Those resonances whose node voltages vary rapidly in the radial direction, but slowly in the
azimuthal direction are called radial. By radial and azimuthal directions, the and
directions in polar coordinates are meant, respectively (see Fig. 22 for model 3).
Fig. 22: Geometry of the disc, showing the definitions of coordinates and .
In Fig. 23 Fig. 25, instantaneous node voltages for a sinusoidal excitation voltage U are
depicted for different resonance frequencies, each at two different instants of time during an
oscillation period, obtained from simulations of model 3.
1
2
4n
4n-2
4n+1
4n-1
3
4 5
= 0
= /2 =
= 3/2
6
1
36
A linear voltage profile along the whole winding (which is the low-frequency limiting
behavior) is subtracted, so that the values at both end nodes of the winding are equal to zero.
The green line depicts the geometry of the winding disc and the location of the nodes, and the
thin horizontal red line shows the zero level of the voltage as a reference. Black lines connect
voltage levels in radial direction, and vertical blue lines indicate the correspondence between
voltage levels and nodes.
The voltage distribution in the disc for the first, second and third radial resonance of Fig.
13 is depicted in Fig. 23, Fig. 24, and Fig. 25, respectively. f
k
(k = 1, 2, or 3 ) denotes the
frequency for that particular resonance, and 1/f
k
is the corresponding period time. The radial
resonances appear as standing voltage waves which can approximately be described by the
formula
0 0 1
rad,
1 0
( , , ) cos(2 ) sin 2
2
k k k k
V t f t A k B
t t
( | | + | |
~ +
( | |
\ .
\ .
(51)
for
0
< <
1
, where
0
and
1
are the inner and outer radii of the disc, respectively (see
Fig. 22). The resonance voltage amplitudes A
k
and B
k
are damping dependent. Note that the
approximate expression (51) is independent of . It can be seen in Fig. 23Fig. 25 that the
approximation (51) is best for low resonance order k. The amplitude B
k
is close to zero for
k = 1 and increases with increasing resonance order k.
Fig. 23: Voltage profile of the first radial resonance (k = 1), at times t = 0.7/f
1
(left) and t = 1.4/f
1
(right).
Fig. 24: Voltage profile of the second radial resonance (k = 2), at times t = 0.4/f
2
(left) and t = 0.8/f
2
(right).
V
rad,k
V
rad,k
37
Fig. 25: Voltage profile of the third radial resonance (k = 3), at times t = 0.3/f
3
(left) and t = 0.6/f
3
(right).
7.1.2 Azimuthal Resonance Modes
For azimuthal resonances, just like the radial ones, the node voltages vary rapidly in the
direction, but the difference is that there are also significant node-voltage variations in the
direction. This pattern can be seen in Fig. 26 and Fig. 27 which depict the instantaneous node
voltages for the two dominant azimuthal resonances, appearing in Fig. 13 and Fig. 18 as
pronounced minima close to f
az
, at two different instants of time. Again, model 3 has been
employed and a linear voltage profile has been subtracted.
In contrast to the radial resonances which are spread out in frequency, the azimuthal
resonances are clustered (at least when viewed on a logarithmic frequency scale) around a
characteristic frequency f
az
slightly above 10 MHz. They cannot be described by a simple
formula like that for the radial resonances (51), but their common characteristics is
approximated by the expression
( )
0 1
az az az
( , , ) cos(2 ) 1 cos( )
2
V t A f t
t
+ | |
~
|
\ .
. (52)
This fundamental behavior is indicated in Fig. 26 and Fig. 27 by dotted lines. Individual
azimuthal resonance modes differ from it by additionally superposed short-wavelength
modulations. The mode in Fig. 26 resembles more closely to the fundamental expression (52)
than the one in Fig. 27.
Fig. 26: Voltage profile of the first azimuthal resonance, at times t = 0.25/f
az
(left) and t = 0.7/f
az
(right).
V
az
V
rad,k
38
Fig. 27: Voltage profile of the second azimuthal resonance, at times t = 0.4/f
az
(left) and t = 0.65/f
az
(right).
7.2 Radial and Azimuthal Resonances for a Winding with More Than One
Disc
As it was reported in Chapter 6 and illustrated in Fig. 19, Fig. 20 and Fig. 21, measurements
were also performed for a winding with more than one disc. In Fig. 28, measurements are
shown (again) for a winding consisting of one, three, six and nine discs respectively,
connected together in a continuous way as in Fig. 9.
10
5
10
6
10
7
10
2
10
4
10
6
Frequency [Hz]
I
m
p
e
d
a
n
c
e
a
m
p
l
i
t
u
d
e
[
O
]
1 disc
3 discs
6 discs
9 discs
Fig. 28: Measurements on windings with one, three, six, and nine discs respectively.
It can be seen that even for a several-disc-winding (and irrespective of the number of discs in
it), the radial and azimuthal resonances exist, and occur around the same frequencies as for a
single disc (see the dotted rings in Fig. 28). This supports the picture that these resonances are
internal oscillations in every individual disc.
Also, it can be noticed that the radial resonances (at least the first two) and the azimuthal
resonances tend to get smoothed out as the number of discs in the winding is increased. This
is expected to be due to the small manufacturing differences leading to the positions, shapes
and amplitudes of the resonances for different discs being different (see Fig. 17), and when
V
az
Radial
modes
Azimuthal
modes
39
the number of discs in the winding is increased, these relatively different resonances
superimpose on each other, leading to a smoothening-out. Another reason for this
smoothening-out could be the proximity losses which enter the picture when several current
carrying conductors are placed close to each other (see Section 8.1).
Furthermore, measurements and simulations show that especially the azimuthal resonances
are very sensitive to small changes in the winding geometry (e.g. mechanical winding
deformations). For instance, in the measurements on different single discs of identical design
but with small manufacturing differences (see Fig. 17), the relative strength and shape of the
two dominating azimuthal resonances varied noticeably, each disc thus having its individual
finger print.
It is also worth to mention that the new resonances which emerge in the frequency range
210
5
210
6
for a winding with more than one disc are global oscillations along the total
length of the winding between different discs, called axial resonances here . A hallmark of
the global resonances is this observation that they shift considerably in frequency when the
number of discs in the winding is increased.
40
41
8 Reluctance Network Method for Inclusion of
the Skin and Proximity Effects
In this chapter, the impact of the skin and proximity effects on the inductances and losses for
a winding is explained, and a method based on the reluctance network approach is described
to compute the frequency dependent inductance and resistance matrices. This method (which
has been developed in [4445]) employs a network of complex reluctances describing the
flux paths around the conductors. The use of this method in this chapter is limited to the
conductor arrangements of multi-conductor transmission lines surrounded by a shield, since
these are easier to start with than a winding, and its extension to windings is left for future
work. The chapter is concluded with verification of the results reproduced by this method by
comparison with finite-element method calculations. This chapter is to a large extent a
description of what has been stated in [4445].
8.1 Skin and Proximity Effects
As it was mentioned earlier in Section 4.4, the formulas for the inductances are for thin
filaments and are moreover frequency independent. This is a good approximation as long as
the distance between two conductors is much larger than their largest cross sectional
dimension, and as long as the frequency is low enough. But when the frequency is high and
the two conductors are close to each other, the so called skin effect and proximity effect will be
present respectively.
Skin effect when a time-varying current flows in a conductor it creates a time-varying
magnetic field which in turn induces eddy currents i.e. induced currents that counteract the
original current. The consequence is that the total current tends to be confined to the surface
of the conductor. This effect becomes stronger as the frequency is increased and the effective
current carrying area of the conductor becomes restricted to a thin layer below the surface,
which is called skin depth and is defined by
0 r
1
f
o
t o
= (53)
The simulation in Fig. 29 (a) shows the influence of skin effect on the distribution of the
current- and magnetic flux densities. It can be noticed that due to the skin effect, the current
density becomes non-uniform in the radial direction , but it is still uniform in the azimuthal
direction when the conductor has cylindrical symmetry.
Proximity effect in contrast to the skin effect, proximity effect is about eddy currents which
are induced in a conductor due to a time-varying magnetic field produced by the currents in
the other conductors in the vicinity. Due to the proximity effect, the current- and magnetic
flux density distributions become unsymmetrical in both and directions (see Fig. 29 (b)).
42
(a) (b)
Fig. 29: Distribution of magnetic flux density (arrow plot) and current density (surface plot) for circular
conductors; in (a) skin effect is present only, while in (b) both the skin- and proximity effects are present (for the
currents in the conductors flowing in the same direction). This figure is borrowed from [46].
The impact of skin effect on the self inductance is that the internal inductance of the
conductor decreases, but since the internal inductance is a very small part of the total self
inductance of a single conductor, the influence of skin effect on the self inductance can be
neglected. The skin effect introduces of course additional, frequency dependent losses. These
losses have already been incorporated in Eq. (31)
But the proximity effect (in combination with the skin effect) is more serious, and this is
due to the fact that when the frequency is very high and the conductors are really close to each
other, the current is not only confined to the surface of the wires but it is distributed around
the axis in conformity with the law of distribution of the charges in the corresponding
electrostatic problem [35]. This means that if current flows in opposite/parallel directions in
two parallel conductors, the current density in each conductor is a maximum at the
nearest/farthest points of the cross sections of the conductors. This has the effect of a
reduction/expansion of the effective spacing of the conductors, which means that the mean
distance between the effective current carrying areas will be smaller/larger than the distance
between the centers of the two conductors which is the distance used in the thin filament
approach (see Fig. 11 and Fig. 12). This in turn means that the actual mutual inductances for
turns close to each other will deviate from the ones calculated in expressions (28) (30). The
proximity effect will also introduce additional, frequency dependent losses, which would be
equivalent with the resistance matrices in Eqs. (36) and (37) gaining off-diagonal frequency
dependent components. It is argued here that the deviations between model and measurements
in Fig. 18 to Fig. 21 are most probably due the absence of these high-frequency phenomena in
the formulas used to calculate the inductances and resistances in Sections 4.4 and 4.5.
8.2 Using Reluctance Network Method (RNM) to Calculate the Frequency
Dependent Inductance and Resistance Matrices
The solution proposed here for calculation of the frequency dependent inductance and
resistance matrices is the so called Reluctance Network Method (RNM); this method is started
by considering the cross section of one side of a single quadratic disc embedded in a
reluctance network, as depicted in Fig. 30.
43
Fig. 30: The cross section view of one side of a single disc, showing the conductors, conductor-insulation,
exaggerated air gaps and different reluctances.
The dimension of the air gaps (due to the manufacturing uncertainty and irregularities)
between the conductors is exaggerated in the figure above to make is possible to illustrate the
problem picture.
As it will be shown below, it is possible to find an easy analytical expression for the
reluctance in a gap in which the length of the channel is much larger than its width, i.e. ' 9 in
between the turns in Fig. 30. But this is not the case for the reluctances outside the turns
( '' 9 and ''' 9 ) since there is no easy way to define an area which the magnetic flux density
passes through and a length which it travels along. This means that it is easier to start the
development of this method with a configuration containing narrow gaps only. This would be
a configuration in which the conductors are surrounded by a shield; a multi-conductor
transmission line. What follows below in the rest of the chapter is a description of the
derivations and results in [4445]: the method is first developed for higher frequencies, and it
is then extended to lower frequencies. After that, it is applied to some simple example
geometries, and its accuracy is verified by comparisons with Finite-Element Method (FEM)
field calculations.
8.3 RNM for High Frequencies
Consider an arrangement of parallel conductors of length / aligned in x direction, with
arbitrary cross section in the y-z plane. On these AC currents of frequency are imposed,
denoted by I
k
with k = 1,,n , where n is the total number of conductors. A very simple
example with n = 2 conductors is pictured in Fig. 31. A slightly more complex and general
example is discussed further below (Fig. 32). The convention that positive current directions
always point out of the plane (red symbols) is adopted here. For clarity, / is kept as a
parameter in the equations below; in order to obtain per-length expressions, one simply has to
divide by / .
1
R
2
R
3
R
3
u
2
u
1
u
1
I
2
I
(a)
m,1
u
m, 2
u
1
I
2
I
(b)
Fig. 31: Cross-sectional view of n = 2 parallel conductors, oriented perpendicularly to the plane. (a) Reluctance
network with chosen branch orientations, (b) circulating mesh fluxes.
''' 9 ''' 9
''' 9 ''' 9 ''' 9 ''' 9
''' 9 ''' 9
conductor
' 9 ' 9 ' 9 ' 9 '' 9
I
1
I
2
I
n-1
I
n
h
i
magnetic flux
'' 9
44
8.3.1 Principle of the Reluctance Network Description
Assume for the moment that the conductors are perfectly conducting (or, alternatively, that
the frequency tends to infinity), so that the skin depth vanishes and magnetic flux exists only
outside the conductors. The magnetic coupling between these conductors is described by a
network of reluctances
i
R (i = 1,,s), characterizing flux channels formed by the gaps
between the conductors as well as the open space around them. In other words, the
i
R
characterize all the available distinct magnetic flux paths [47]. The
i
R of the flux channels
are determined by the geometry and material parameters of individual gaps between
conductors, whereas those of the open space also depend on the availability and nature of
some external current return path. Each network branch corresponds to one of these flux paths
and reluctances, and each network mesh surrounds one of the conductors.
For each reluctance branch an (arbitrary but fixed) orientation is chosen, defining the
direction of positive flux u
i
in the branch.
At any given moment, the currents I
k
flowing through all the conductors and the branch
fluxes u
i
in the reluctance network can be combined into vectors,
1
n
I
I
| |
|
=
|
|
\ .
I . and
1
s
u
u
| |
|
=
|
|
\ .
. u . (54)
Since magnetic flux is conserved locally (i.e., at each node of the network, the total inflowing
flux is zero), any possible flux state u can be built up by superposition of mesh fluxes u
m,k
, k = 1,,n , circulating counter-clock-wise in the meshes around the n conductors (Fig. 31b,
Fig. 32b).They are combined here into a new vector u
m
with n elements, and a s n matrix c
is defined as follows:
Definition of the matrix c : (55)
For any given branch index i and conductor index k, c has an element
ik
= +1 (or 1) if
the branch i belongs to the mesh around conductor k and is oriented in parallel (or
opposite) to the circulation direction, and
ik
= 0 if it does not belong to mesh k.
c can be interpreted as the connectivity matrix of the graph which is dual to the reluctance
network. In the simple example of Fig. 31, the total number of reluctance branches is s = 3,
and the c matrix is given by:
1 2
1
2
3
1 0
1 1 .
0 1
k i = =
| |
=
|
|
\ .
c
(56)
Using c , the relation between the branch and mesh fluxes can be written as a simple matrix
equation,
m
= u c u . (57)
The line integral of the magnetic field H along any branch i is proportional to the flux
i
along that branch, with
i
R as proportionality constant. Summing up these branch
contributions around any mesh k in the network yields the line integral of H on a closed loop
45
around conductor k, which according to Ampres law is equal to the enclosed current I
k
.
Expressing the reluctances
i
R as a diagonal matrix
1
0
0
s
| |
|
|
\ .
R
R
R = , (58)
this can again be written as a matrix relation,
T
= I c u R , where
T
c denotes the matrix
transpose of c . Together with (57) this results in
T
m
= I c c u R . (59)
Finally, by virtue of Faradays law the induced voltage drops between opposite ends of the
conductors are given by the time derivatives of the mesh fluxes,
m
d dt U = u , which in
frequency domain leads to
m
je = U ZI = u . (60)
with
( )
1
T
je
= Z c c R . (61)
Finally, when losses in the conductors are included the reluctances become complex functions
of frequency (see next section), and the total series impedance matrix of the conductor
arrangement acquires a resistive contribution and can simply be expressed as
( )
1
T
j j e e
= + = Z R L c c R . (62)
In the second example shown in Fig. 32 where n = 5 and s = 8, the outermost conductor
(k = 5) forms a shield surrounding the other conductors. Its cross section is multiply
connected, so that the contour describing its mesh flux
m,5
(dashed line in Fig. 32b) consists
of two parts.
k = 1
3
2
4
5
i = 1
2
3
4
5
6
7
8
(a)
k = 1
3
2
4
5
(b)
Fig. 32: Cross-sectional view of a more general arrangement of n = 5 parallel conductors. (a) Reluctance
network with chosen branch orientations, (b) circulating mesh fluxes.
46
For this example, the c matrix is given by:
1 2 3 4 5
1
2
3
4
.
5
6
7
8
1 0 0 0 1
1 1 0 0 0
1 0 1 0 0
0 1 0 0 1
0 1 1 0 0
0 0 1 1 0
0 0 0 1 1
0 0 0 0 1
k i = =
| |
|
=
|
|
\ .
c
(63)
Note that the sum over each row except the last is zero, such that any common mode
voltage (i.e., U
1
= = U
n
) corresponds to current flow in the shield only (i.e., I
k
= 0 for
k n), and the shield current obeys the relation j
n s n
I U e = R where s is the index of the
outermost flux path surrounding the shield (in the present case, s = 8).
On the other hand, if the shield is short-circuited (i.e., if U
n
= 0 is imposed) all flux outside
the shield is suppressed at nonzero frequencies, and so flux path s can be disregarded
altogether. For the calculation of reluctances of the inner flux paths the shield is treated in
the same way as the other conductors. Under these conditions, the remaining conductor
voltages U
1
,,U
n1
and currents I
1
,,I
n1
are related to each other by a reduced Z matrix
which does not depend on
s
R , and the shield acts as common current return path. The
numerical examples in Section 8.6 below will be limited to this situation (it may be called the
differential mode situation).
8.3.2 Calculation of the Reluctances for High Frequencies
At a finite frequency which is still so high that the skin depth in all conductors is small
compared to their thickness, the flux penetrates only slightly into the conductors, and the
above picture of separate flux channels which do not interfere with each other is still valid.
Eq. (62) has provided an explicit expression of the impedance matrix in terms of the gap
reluctances
i
R , but one still needs to calculate the latter for a given geometry. In the limit of
high frequencies, one can focus on a single gap and model its geometry locally as shown in
Fig. 33. Orienting the channel i in such a way than the positive flux direction is to the right,
all quantities related to the conductor above the gap have been labelled here with a subscript
+,i and those below the gap with ,i. Later on, conductor specific quantities will instead
sometimes be labelled by the original conductor index k, which obeys the condition
ik
= +1
or
ik
= 1 for the conductor above or below the gap, respectively, according to definition (55)
of the matrix c .
It is assumed that the medium in the gap is perfectly insulating (o
g,i
= 0) and has a finite
permeability
g,i
> 0. The conductors above and below the gap are given finite conductivities
o
,i
> 0 and permeabilities
,i
> 0, respectively, which leads to flux penetration into the
conductors characterized by finite skin depths
,i
(see further below).
47
conductor (
, , ,
, ,
i i i
o k
)
insulator (
g,i
)
i
a
i
w
,i
o
+
,i
o
,i
d
+
,i
d
conductor (
, , ,
, ,
i i i
o k
+ + +
)
flux
channel
g,i
u
,i
u
+
,i
u
x
z
y
skin layer
skin layer
Fig. 33: Local geometry of the gap between two conductors.
x
z
y
g,i
R
,i
R
,i +
R
g,i
u
,i
u
+
,i
u
i
u
Fig. 34: Parallel reluctances of insulating gap and skin layers in the adjacent conductors.
Assume for the moment that the gap is straight and that its width a
i
is constant along its whole
length w
i
. The case of varying gap width will be discussed below in Section 8.5. For a long
and/or narrow gap (a
i
<w
i
) and for strong skin effect (
,i
<d
,i
, where d
,i
are the thicknesses
of the two conductors in direction perpendicular to the channel, see Fig. 33), the conductor
cross sections can be approximated locally as infinite half planes, and it can be assumed that
the current densities in the conductors and all the fields are oriented parallel to the surfaces
everywhere (in the x-y plane) and are varying in z direction only. The current density is
chosen to point in x direction and the magnetic field to point in y direction. Since the quasi-
static magnetic coupling between the conductors is of interest here, solutions of Maxwells
equations where the charge density is zero everywhere are looked for. The only non-
vanishing components of the electric and magnetic fields are then E
x
and H
y
, and Maxwells
equations reduce to the one-dimensional problem given by
d
j
d
x
y
E
H
z
e = , (64)
d
j
d
y
x x
H
E J
z
ec = . (65)
Since there are no surface currents, both E
x
and H
y
must be continuous in z direction. Current
density and electric field are assumed to be related by J
x
= oE
x
, so (64) and (65) can be
combined into a single equation for H
y
. Neglecting the displacement current jE
x
(which has
48
a noticeable effect only at much higher frequencies than the ones of interest here), one ends
up with the well-known simple diffusion equation [48] for H
y
inside a conductor,
2
2
2
d
j
d
y
y y
H
H H
z
eo k = = , (66)
where
1 j
j k eo
o
+
= = , with
2
o
eo
= . (67)
For simplicity, denotes the total permeability here and in the following, including the
permeability of free space
0
, and it may have a different value in each conductor. The
conductivity o may also have a different value in each conductor. is the complex wave
number for magnetic field penetration into the conductor at frequency and the
corresponding real skin depth. Since the gap is insulating, H
y
is constant in this region,
according to Ampres equation (65). The fundamental solutions of Eq. (66) within the
conductors are exponential functions e
z k
.
In order to calculate the reluctance of the channel, one has to relate the total flux u
i
through it to the magnetic field H
y
in the gap. One can start with the contribution of the upper
conductor, assumed for the moment to occupy the region {z > 0}, so that the solution to (66)
has the simple exponential form
,
,0
( ) e
i
z
y y
H z H
k
+
= . (68)
Here H
y,0
is the magnetic field at the conductor surface {z = 0} and inside the gap i (where it
is independent of z), and ( )
, ,
1 j
i i
k o
= + are the complex wave numbers in the conductors
above and below gap i, as defined in (67). As a consequence of (68), the magnetic flux u
+,i
inside the upper conductor (which is roughly confined within a distance
+,i
from the surface)
is given by
, ,0
, ,
, 0
d
i y
i i y
i
H
z H
u
k
+
+ +
+
= =
}
/
/ . (69)
The corresponding expression for the magnetic flux u
,i
inside the lower conductor is
analogous. Inside the gap the magnetic field is constant and equal to H
y,0
, so that the flux in
the gap is given by
g, g, ,0 i i i y
a H u = / . Together with the fluxes (69) in the adjacent skin layers
it adds up to the total flux in the channel,
, ,
, g, , g, ,0
, ,
i i
i i i i i i y
i i
a H
u u u u
k k
+
+
+
| |
= + + = + +
|
|
\ .
/ . (70)
The reluctance
i
R of the channel is now defined as the proportionality constant between u
i
and the line integral of the magnetic field along the channel w
i
H
y,0
, and thus
1
, ,
g,
, ,
i i
i
i i i
i i
w
a
k k
+
+
| |
= + +
|
|
\ .
/
R . (71)
49
Note that this result can be interpreted as a parallel connection
( )
1
1 1 1
, g, , i i i i
+
= + + R R R R of
individual reluctances
,
,
,
i
i
i
i
w
k
+
+
+
=
/
R ,
g,
g,
1
i
i
i i
w
a
=
/
R ,
,
,
,
i
i
i
i
w
k
=
/
R , (72)
which refer to insulating gap and adjacent skin layers, as indicated in Fig. 34.
8.4 Extension of RNM to Low Frequencies
8.4.1 Summary of the Generalized Approach
With decreasing frequency the skin depth increases until it exceeds half the conductor
thickness, i.e., skin layers of opposite sides of the conductor start overlapping and the above
picture of individual flux channels is not applicable anymore. However, it will be argued
below that even for low frequencies a reasonable approximation can still be obtained by
generalizing Eqs. (62) and (71) in the following way:
( )
1
T
int int
j j e e
= + = + Z R L Z Z c c
R
R , (73)
where R is still a diagonal matrix, but its elements are modified by additional factors
,i
to
the inverse skin layer reluctances
1
,i
R (compare (72)):
1
, ,
, g, ,
, ,
i i
i
i i i i i
i i
w
a
u u
k k
+
+
+
| |
= + +
|
|
\ .
/
R . (74)
It will be shown below that these factors have the form
, ,
,
tanh( )
2
i i
i
d k
u
= , (75)
coming from the usual skin effect formula for slabs [48] with finite thicknesses d
,i
. At high
frequencies (i.e., for
,i
<d
,i
),
,i
tend to 1 so that the previous expression (71) is recovered.
An alternative way to write (75), which will be more convenient further below, uses
conductor indices k instead of subscripts :
,
,
tanh( )
2
k k i
k i
d k
u = , (76)
where d
k,i
denotes the thickness of conductor k perpendicular to the adjacent channel i (i.e.,
d
k,i
= d
+,i
or d
,i
, and k
k
= k
+,i
or k
,i
, etc. for
ik
= 1 or 1, respectively).
int
Z and
int
Z
R
are diagonal matrices, which have been added since the original expression
(62) in combination with (74) leads to a wrong low-frequency limit for the internal
impedances of individual conductors. The element
int ,k
Z
R
is the internal impedance of
conductor k, resulting under the assumption that (62) together with (74) are valid at all
50
frequencies. It is calculated from these equations by setting all permeabilities in the system
equal to zero, except for
k
(which physically means that conductor k is completely immersed
in a perfectly diamagnetic medium, so that all flux is confined to the interior of that
conductor). In that limit, all matrix elements of
T 1
( )
c c = R
1
1
,
with 0
ik
i k
k i
i
k
w k
u
c =
| |
|
\ .
/
(77)
(which is the inverse of the sum of all the skin layer reluctances around conductor k), so that
int, 1
,
with 0
j
ik
k
k
k i k i
i
Z
w
e
k u
c =
=
/
R
. (78)
The summation is around the mesh surrounding the conductor k. (Note that the sum over the
w
i
without the
1
, k i
u
, (79)
where only
1
, k i
u
. (80)
At high frequencies (
,i
<d
,i
) where both
1
, k i
u
and
, k i
u
tend to 1,
int,k
Z and
int ,k
Z
R
cancel each
other. In the limit 0, expressions (79), (80) lead to the correct DC resistance for
conductors with rectangular or circular cross section, and therefore they are adopted here as a
general rule. For instance, for a rectangular cross section with side lengths p and q, numbering
the channels around it by i = 1,,4, one has d
k,1
= d
k,3
= w
2
= w
4
= p and
d
k,2
= d
k,4
= w
1
= w
3
= q, as well as
1
, , 4 k i k k i
d u k
= 0
= 0
,1
( ) o o
+
=
,2
( ) o o
=
1
a
d
2
a
skin layer
skin layer
2 d
2 d
0
z
Fig. 35: Simple single-slab geometry.
x
z
y
= 0
= 0
I
g,2
R
,2
R
,1 +
R
g,1
R
m
u
g,2
u
,2
u
,1
u
+
g,1
u
Fig. 36: Reluctance network of the single-slab geometry.
The reluctances are calculated in a similar fashion as in Section 8.3.2 above. Now, two
surfaces of the slab have to be take into account, so the origin of z is chosen in the middle of
the slab and the functions
( )
1
2
sinh e e
z z
z
k k
k
= and
( )
1
2
cosh e e
z z
z
k k
k
= + are used as
52
fundamental solutions of (64), (65) (instead of e
z
). The general solution in a slab of
thickness d extending from z = d/2 to z = d/2 can then be written in the form
cosh( ) sinh( )
( )
cosh( 2) sinh( 2)
y
z z
H z C C
d d
k k
k k
+
= + , (82)
sinh( ) cosh( )
( )
cosh( 2) sinh( 2)
x
z z
E z C C
d d
k k k
o k k
+
| |
= +
|
\ .
, (83)
with arbitrary coefficients C
= = and ( 2)
x x
E E z d
+
| |
= +
|
\ .
, (85)
with
sinh( 2)
tanh( 2)
cosh( 2)
d
d
d
k
u k
k
= = . (86)
At low frequencies, when the skin depth o (
,1 ,2
o o
+
= = , see Fig. 35) is larger than half the
conductor thickness d, the two skin layers overlap and
+,1
,
,2
can no longer be calculated
separately. However, the total magnetic flux
c
=
+,1
+
,2
in y direction inside the
conducting slab can be obtained by integration of (82),
2
c
2
2
d
d
y
d
C
z H
u
u
k
+
= =
}
/
/ . (87)
Similarly, the total current I flowing in x direction through the slab is obtained by integration
of (83) (or, alternatively, using Ampres law (65)):
( )
2
2
d 2
d
x y y
d
I w z E w H H wC o
+
= = =
}
. (88)
The total fluxes
g,1
,
g,2
in the insulating layers below (z < d/2) and above (z > d/2) the
slab, respectively, are given by
( )
g,1 1 g,1 1 g,1 y
a H a C C u
+
= = / / and
( )
g,2 2 g,2 2 g,2 y
a H a C C u
+ +
= = + / / . (89a,b)
The voltage drop U in the slab can be obtained by integration of the electric field along a
closed contour which runs in x direction along the conductor (for instance, at one of its
surfaces) and returns in the field free region below (z < d/2 a
1
) or above (z > d/2 +a
2
) the
system. According to Faradays law, both choices of contour must lead to the same result,
53
since the total flux in y direction (both inside and outside of the slab) must vanish due to flux
conservation,
g,1 c g,2
0 u u u + + = . (90)
Here, the contour below is closed and the following result is obtained
g,1
j
x
U E eu
= + /
( )
1 g,1
j
C
a C C C
k
e u
o u
+ +
| |
= +
|
\ .
/
/
( )
1 g,1
1
j j
C
a C C
e u e u
k k u
+
| | | |
= + +
| |
\ . \ .
/ / . (91)
Now new quantities are defined
1
1 1 g,1
w
a
u
k
| |
= +
|
\ .
/
R ,
1
2 2 g,2
w
a
u
k
| |
= +
|
\ .
/
R . (92)
Since these expressions are special cases of (74) for the present geometry, they are interpreted
as generalized channel reluctances. Re-writing (91) as
( )
1
1
1
j
U C w
C C
u
e k u
+
| |
+ =
|
\ .
/ /
R , (93)
and inserting (87), (89a,b) into (90), the following is obtained
( ) ( )
1 1
1 2
0 C C C C
+ +
= + + R R
( )( )
1 1 1
1 2 2
2 C C C
+
= + + R R R . (94)
The second line in (94) makes it possible to express C
+
C
in (93) by C
Fig. 37: Some basic geometries of the gap (white space) between neighbouring conductors (shaded). The
channels are vertically oriented here. The current direction (x-direction) is perpendicular to the plane, as before.
(a) Planar, parallel conductor surfaces: This is the basic case (74) discussed earlier where
a = const and so
1
g
w
a
u u
k k
+ +
+
| |
= + +
|
\ .
/
R . (97)
55
(b) Planar but inclined conductor surfaces: Here a varies along the gap as
a(y) = a
min
+ (a
max
a
min
)y/w for 0 < y < w, and so
1
2
g
2 1 0
1
d ( ) ln 1
w
w K
y a y
K K
u u
k k
+ +
+
| | | |
= + + = +
| |
\ . \ .
}
/ /
R
with
1 min g
K a
u u
k k
+ +
+
= + + ,
( )
2 g max min
K a a = . (98ac)
When a
min
, a
max
are equal, this reduces to case (a).
(c) Locally cylindrical conductor surfaces with curvature radii r
+
, r
: The largest
contribution to the channel reluctance comes from its narrowest region, which is assumed to
be located approximately in the middle. The origin of y is chosen at that point, so that
2
min
1 1
( )
2
y
a y a
r r
+
| |
~ + +
|
\ .
in the vicinity and
2
1
1
2
2
g
2 1 2
2arctan( )
1
d ( )
K w
w
K
w
y a y
K K
u u
k k
+ +
+
| |
~ + + ~
|
\ .
}
/ /
R
with
1 min g
K a
u u
k k
+ +
+
= + + ,
g
2
1 1
2
K
r r
+
| |
= +
|
\ .
. (99ac)
One or both of the radii r
+
, r
and
, x k
E
, the fluxes
c,k
and
g,i
, as well as the total currents I
k
can again be expressed in terms of
coefficients
k
C
+
and
k
C
.
x
z
y
= 0
2
I
g,2
R
,2
R
,1 +
R
g,1
R
g,2
u
,2
u
,1
u
+
g,1
u
1
I
,1
R
,0 +
R
g,0
R
,1
u
,0
u
+
g,0
u
,2 +
R
,2
u
+
3
I
1
a
2
d
2
a
0
a
1
d
c,2
u
c,1
u
w
Fig. 54: Geometry of a stack of slabs, generalizing Fig. 35, Fig. 36. Also shown are the different flux paths in
the insulating gaps and skin layers, together with their respective reluctances.
86
Assuming as before that the connections to the conductors are made from below the stack, the
following are obtained for the voltages U
k
:
1 g,0 ,1
j
x
U E eu
= + / , (A1)
( )
2 g,0 c,1 g,1 ,2
j
x
U E e u u u
= + + + / , (A2)
etc., so that in general
( )
1
g, 1 c, g, 1 ,
1
j j
k
k i i k x k
i
U E e u u eu
=
= + + +
/ . (A3)
(By definition, the empty sum for k = 1 is equal to zero.) Using (85), (87), (89a,b) this can be
re-expressed as
( )
1
1 g, 1
1
j 2
k
i i
k i i i i i
i
i
U a C C C
u
e
k
+ +
=
(
= +
(
/
( ) ( )
1 g, 1
j j
k k
k k k k k k
k
a C C C C
u
e e
k
+ +
+ + / /
1
j
k
k k
k k
C
e u
k u
| |
+
|
\ .
/ . (A4)
In the same way, the condition of zero total flux can be re-expressed as follows:
( )
g, 1 c, g,
1
0
n
i i n
i
u u u
=
= + +
( )
1 g, 1
1
2
n
i i
i i i i i
i
i
a C C C
u
k
+ +
=
(
= +
(
/
( )
g, n n n n
a C C
+
+ + / . (A5)
Since the field H
y
is constant across any given gap k and equal to the adjacent surface fields
, y k
H
+
and
, 1 y k
H
+
, (84) implies the following matching condition for
k
C
+
and
k
C
:
, , 1 k k y k y k
C C H H
+ +
+
+ = =
1 1 k k
C C
+
+ +
= , for 1,..., 1 k n = . (A6)
Eq. (A6) allows in particular to express the terms 2
i
C
+
in the sums of (A4), (A5) in terms of
differences
i i
C C
+
, since
( ) ( )
2
i i i i i
C C C C C
+ + +
= + +
( ) ( )
1 1 i i i i
C C C C
+ +
+ +
= + , (A7)
so that the sums of (A4), (A5) can be re-arranged into the following forms:
87
( )
1
1
1
j 1
j
k
k
k i i i k k
i
k k
U w C C C
e
e u
k u
+
=
| |
= +
|
\ .
/
R , (A8)
( ) ( )
1 1
1
1
0
n
i i i n n n
i
C C C C
+ +
=
= + +
R R , (A9)
with the coefficients
1
i
R defined by
1 1 1
0 0 g,0
1
a
w
u
| |
= +
|
\ .
/
R ,
1 1 1
g,
1
i i i i
i i i
i i
a
w
u u
k k
+ +
+
| |
= + +
|
\ .
/
R for 1,..., 1 i n = ,
and
1
g,
n n
n n n
n
a
w
u
k
| |
= +
|
\ .
/
R . (A10)
The
i
R have exactly the form of the channel reluctances defined in (74), adapted to the
particular geometry and labelling convention of the stack in Fig. 54. Furthermore, due to (88)
one can simply write
2
k k
I wC
= . (A11)
Finally, defining as in (96)
int ,
j
2
k
k k
k
Z
w
e
u
k
=
/
R
and
int ,
j 1
2
k
k
k k
Z
w
e
k u
=
/
, (A12)
and moving the second term on the r.h.s. in (A8) to the left, one obtains
( )
1
int , int , 1
1
j
k
k k k k k i i i
i
U Z I Z I w C C e
+
=
+ =
R
R
. (A13)
The r.h.s. of this equation is now denoted by Q
k
,
( )
1
1
1
j
k
k i i i
i
Q w C C e
+
=
=
|
=
|
|
\ .
c
(A16)
so that
0 1 1
1 1 2 2
T
2 2 3
1
1 1
0 0
0 0 .
0 0
n
n n n
+ | |
|
+
|
| = +
|
|
|
+
\ .
R R R
R R R R
R R R
R
R R R
c c R
(A17)
To facilitate the argument, one can resort to the following trick. One can choose some
(arbitrary)
1 n
C
+
and
1 n
C
+
+
with
1 1 n n
C C
+
+ +
=
n n
C C
+
+ , so that the condition of zero total flux
(A9) takes the form (A14) with Q
n+1
= 0, whereby the validity of (A14) is extended to
1 k n = + . Defining in addition Q
0
= 0, one has for all 0,..., k n = :
( )
1
1 1 1
j
k k k k k
Q Q w C C e
+
+ + +
= R . (A18)
This in turn implies that for all k = 1,,n , the k:th component of the l.h.s. of (A15) can now
be written as
( )
1 1 1 1 k k k k k k k
Q Q Q
+
+ + R R R R
( ) ( )
1 1 1 k k k k k k
Q Q Q Q
+
= R R
( ) ( )
1 1
j j
k k k k
w C C w C C e e
+ +
+ +
=
j 2 j
k k
w C I e e
= = , (A19)
where the last two equalities follow from (A6) and (A11), respectively. But the end result is
just the r.h.s. of (A15) which completes the proof of (73) for the particular stack geometry
of Fig. 54.
89
Appendix B
Wave Velocity in Transmission Lines and
Transformer Windings
In this Appendix, the relations governing the voltages and currents in a transmission line and a
transformer winding are derived and the wave velocities are obtained. These derivations are
largely borrowed from [6869]. For the transmission line ladder network in Fig. 42, if u is the
voltage in the nodes, and i is the current in the inductors, then the voltage difference between an
arbitrary node and the node before it is
i
u l x
t
c
A = A
c
, (B1)
which in the limit x 0 leads to
u i
l
x t
c c
=
c c
. (B2)
Further, the difference between the currents in one inductive branch and the one before is
g
u
i c x
t
c
A = A
c
, (B3)
leading to
g
i u
c
x t
c c
=
c c
. (B4)
By differentiating Eq. (B2) and (B4) with respect to x and t respectively, and eliminating i(x,t),
one will arrive at
2 2
g 2 2
u u
lc
x t
c c
=
c c
, (B5)
which is the wave equation for a voltage wave (a similar one can be obtained for the current
i(x,t)) propagating with the velocity
g
1
v
lc
= . (B6)
For the case of a transformer winding, consider a tightly wound coil of individual pancake turns
(or discs), which are parallel and closely spaced (Fig. 55(a)).
90
c
g
c
s
ground
(a)
LV
c
g
c
s
(b)
HV LV
Fig. 55: Tightly wound coil consisting of individual pancake turns (a), and a transformer winding in cross-
section (b).
The charging current for each element of conductor length x of the n:th turn of the winding,
which is due to the ground-capacitance and thus flows mainly to ground, is
g, g
n
n
u
i c x
t
c
= A
c
, (B7)
where c
g
is the ground-capacitance per unit length of conductor. Let c
s
be the disc-to-disc
capacitance per unit length between adjacent pancake discs. Then the charging current that
flows from the n:th turn to the (n 1):th turn is
( )
1
s, s s
n n n
n
u u u
i c x c x
t t
c ' cA
' = A = A
c c
, (B8)
and that to the (n +1):th turn is
( )
1
s, s s
n n n
n
u u u
i c x c x
t t
+
c '' cA
'' = A = A
c c
. (B9)
The sum of these two currents,
( )
2
s, s s
n
n n n
u
i c x u u c x
t t
cA c
'' ' = A A A = A
c c
, (B10)
is the total disc-to-disc capacitive current per unit conductor length. Here u is the voltage
difference between adjacent turns and
2
u is the difference of this difference between
successive turns. The voltage difference is thus that due to a length of conductor
x A = / (B11)
of one complete turn of length / . Thus the second difference of voltage can be rewritten in
differential rather than difference form
2 2
2 2 2
2 2
u u
u
x x
A c
A = =
A c
/ / , (B12)
which is valid if the coil contains many turns and if the detailed distribution of the phenomena
along a single turn length are not of interest. The total disc-to-disc and ground capacitive
91
charging current of each turn, as given by Eqs. (B7) and (B10), is then equal the decrease i
n
in the current in the n:th turn. Thus, using Eq. (B12), one obtains
3
2
g s 2
n n
n
u u
i c x c x
t t x
c c
A = A A
c c c
/ , (B13)
and hence the space derivative of the current is
3
2
g s 2
i u u
c c
x t t x
c c c
= +
c c c c
/ , (B14)
which shows that it depends on the change of voltage not only with time but also with space.
For fast travelling pulses with steep portions, and thus pronounced changes in space along the
conductor, the second term of Eq. (B14) can become very significant, particularly if the disc-to-
disc capacitance c
s
is substantially larger than the ground capacitance c
g
, which is frequently the
case for tightly wound coils.
The current as well as the voltage in a long coil varies from turn to turn. If the n:th turn, of
self inductance per unit length, were separated from the rest of the coil, the voltage induced by
the current in it would be
n
n
i
u x
t
c
= A
c
. (B15)
The two adjacent turns coupled by the mutual inductance m per unit length further add an
induced voltage
( )
1 1 n n
mn
i i
u m x
t
+
c +
= A
c
. (B16)
By subtracting i
n
from the current in each of the adjacent turns and rearranging the terms, the
following expression for the effect of the adjacent turns is deduced:
( ) ( )
2
1 1
2 2
n n
mn n n n n n
i i
u m x i i i i i m x m x
t t t
+
c cA c
= A + = A + A (
c c c
. (B17)
Thus the mutual inductance of the adjacent turns contributes an induced voltage that can be
thought of as consisting of two parts, one of which is proportional to effects in the n:th turn and
the other due to second differences between the currents in the turns. Each succeeding turn will
produce an induced voltage in the n:th turn that will have the same form as Eq. (B17), except
that m will become smaller with increasing distance between turns. All induced voltages
proportional to the current i
n
in the n:th turn can be summed from Eqs. (B15) and (B17), to give
( )
n n
ln
i i
u m x l x
t t
c c
= + A = A
c c
. (B18)
Here l is the self-inductance of the coil per unit length derived from the total inductive effects
between all turns; thus it is the self-inductance of the entire coil divided by the length of wire.
In the last term of Eq. (B17) one can introduce Eq. (B11) and write
92
2 2
2 2 2
2 2
i i
i
x x
A c
A = =
A c
/ / , (B19)
when many turns are again used as basis for transformation from the difference to the
differential form. For this term, only the influence of the immediately adjacent turns is taken
into account, because the induced effects due to the farther ones have already been included in
Eq. (B18) via the total inductance of the coil and should not be counted again. The additional
inductive influence of the two adjacent turns is effective here, like that of the inter-disc
capacitance treated earlier. The voltage induced in the n:th turn, in accordance with Eqs. (B17)
and (B18), must be equal to the decrease u
n
of the observable voltage difference. Using Eq.
(B19) one thus obtains
3
2
2
n n
n
i i
u l x m x
t t x
c c
A = A + A
c c c
/ , (B20)
And the voltage difference along the coil is
3
2
2
u i i
l m
x t t x
c c c
=
c c c c
/ . (B21)
Again it can be seen that this depends on the variation of current with time as well as with
space. For fast travelling pulses with steep portions the third-order derivative may attain
appreciable magnitudes. However, its influence here is smaller than the corresponding
capacitive one on voltage in Eq. (B14) because the mutual inductance m between adjacent turns
is small compared with the self-inductance component l per unit turn of the entire coil. This
means that the second term in Eq. (B21) can be neglected as a simplification. This
approximation is fully explained and justified in [68] and [69]. Thus one has
3
2
g s 2
i u u
c c
x t t x
c c c
= +
c c c c
/ , and
u i
l
x t
c c
=
c c
. (B22)
The validity of Eq. (B22) can be extended to other cases of importance. If the coil has not only
single turns as in Fig. 55(a), but a number of turns per disc as shown in Fig. 55(b), it is possible
to consider average values u
n
and i
n
for each disc n. Correspondingly, one can now let c
g
be the
ground-capacitance for each disc and c
s
the capacitance between adjacent discs, both per unit
length. The definition of self-inductance l is unchanged and / is now the length of conductor
per disc.
By differentiating the first and second part of Eq. (B22) with respect to t and x respectively,
and eliminating i(x,t), one can write
2 2 4
2
g s 2 2 2 2
0
u u u
lc lc
x t t x
c c c
+ =
c c c c
/ . (B23)
A similar fourth-order differential equation holds for the current i. A simple form of solution for
Eq. (B23) is
0
exp(j ( / )) u u t x v e = , (B24)
By introducing Eq. (B24) into Eq. (B23) one will arrive at
93
2 2
2 2 2
g s
0 lc lc
v v
e e
e e
| | | |
+ + =
| |
\ . \ .
/ . (B25)
This shows the relation between the frequency and the velocity v of traveling waves. Eq.
(B25) gives the velocity of propagation as
2 2 2 2 2 s s
0
g g g
1 c c
v v
lc c c
e e = = / / . (B26)
94
95
List of Symbols
Symbol Quantity SI-Unit
A Matrix used for formulating the state space equation of the
winding model
A Area of the complex- ellipse [m
2
]
1
A Constant for determining the field H
z
in laminate [A/m]
2
A Constant for determining the field H
z
in laminate [A/m]
az
A Voltage amplitude for the Azimuthal resonances [V]
A
k
Voltage amplitude for radial resonance k [V]
a Gap width [m]
a' Gap width [m]
a'' Gap width [m]
a
i
Width of gap i [m]
a
i,max
Maximum width for gap i [m]
a
max
Maximum gap width [m]
a
min
Minimum gap width [m]
B Magnetic flux density vector [T]
' B Matrix used for formulating the state space equation of the
winding model
'' B Matrix used for formulating the state space equation of the
winding model
B Magnetic flux density [T]
B
k
Voltage amplitude for radial resonance k [V]
B
Complex magnetic flux density [T]
B
Average magnetic flux density in laminate [T]
B
meas
Approximaion for measured magnetic flux density [T]
B
p
Peak value of the magnetic flux density for the complex-
ellipse
[T]
b Half of the laminate thickness [m]
C Total capacitance matrix [F]
C
(i)
Specific capacitance matrix for model i [F]
( )
DD
i
C
Capacitance matrix for model i, for coupling between two
neighboring discs
[F]
C
dd
Disc-to-disc capacitance [F]
( )
dd
i
C
Disc-to-disc capacitance for model i [F]
C
hf
Winding model capacitance at high frequencies [F]
C
ig
Disc-to-ground capacitance for inner ground [F]
96
( )
ig
i
C
Capacitance to inner ground for model i [F]
C
og
Disc-to-ground capacitance for outer ground [F]
( )
og
i
C
Capacitance to outer ground for model i [F]
C
tt
Turn-to-turn capacitance [F]
( )
tt
i
C
Turn-to-turn capacitance for model i [F]
C
Constant coefficients for determining the field values in a
conductor
[A/m]
k
C
Constant coefficients for determining the field values in
conductor k
[A/m]
c Degree of reversibility [-]
c
g
Ground capacitance per unit length [F/m]
c
s
Series capacitance per unit length [F/m]
D Electric displacement field vector [C/m
2
]
d Conductor thickness [m]
d' Conductor thickness [m]
d'' Conductor thickness [m]
d
i
Distance between inner ground and winding [m]
d
k,i
Thickness of conductor k perpendicular to the adjacent channel
i
[m]
d
m
Location of damage [m]
d
o
Distance between outer ground and winding [m]
d
we
Total length of winding [m]
d
,i
Thickness of the conductor above/below gap i, in perpendicular
direction
[m]
E Electric field vector [V/m]
E
x
x-component of electric field [V/m]
x
E
x-component of electric field at slab surfaces [V/m]
, x k
E
x-component of electric field at the surfaces of slab k [V/m]
F Force vector [N]
f Frequency [Hz]
f
az
Frequency for azimuthal resonances [Hz]
f
i
Frequency for measurement i [Hz]
f
k
Frequency for radial resonance k [Hz]
G Magnetic domain structure parameter [-]
H Magnetic field vector [A/m]
H Magnetic field [A/m]
H
Complex magnetic field [A/m]
H
excess
Magnetic field due to excess effects [A/m]
H
meas
Approximation for measured magnetic field [A/m]
H
p
Peak value of the magnetic field for the complex- ellipse [A/m]
H
y
y-component of magnetic field [A/m]
y
H
y-component of magnetic field at slab surfaces [A/m]
, y k
H
y-component of magnetic field at the surfaces of slab k [A/m]
H
y,0
y-component of magnetic field at the conductor surface [A/m]
H
z
z-component of magnetic field [A/m]
H
0
Magnetic field strength at laminate surfaces [A/m]
h Conductor height [m]
97
h
tot
Total height of the winding [m]
I Current vector [A]
II Identity matrix [-]
I Current [A]
I
after
Current in injection-point after damage [A]
after
Fourier transform of I
after
[A]
I
before
Current in injection-point before damage [A]
before
Fourier transform of I
before
[A]
I
j
Current in segment j [A]
I
k
Current in conductor k in a multi-conductor transmission line [A]
i Index for pseudo particles and single sheet measurements
(Chapter 3), developed winding models (Chapter 4), gaps and
reluctance branches in a multi-conductor transmission line
(Chapter 8); Current (Chapter 9)
[-]; [A]
i
g,n
Current in the n:th turn flowing to ground [A]
i
n
Current in the n:th turn [A]
s,n
i Sum of
s,n
i' and
s,n
i'' [A]
s,n
i' Current flowing from the n:th turn to the next turn [A]
s,n
i'' Current flowing from the n:th turn to the previous turn [A]
J Current density vector [A/m
2
]
J
x
x-component of current density [A/m
2
]
j Imaginary unit [-]
K Total number of discs in the winding [-]
K
1
Reluctance factor [H]
K
2
Reluctance factor [H]
k Pinning strength (Section 3.3.1); index for an arbitrary node
(Section 4.5), the radial resonances (Section 7.1.1), the
conductors in a multi-conductor transmission line (Chapter 8)
[Am
2
/s]; [-]
L Inductance matrix [H]
L
disc
Inductance matrix for a single disc [H]
L
kj
Matrix for the mutual inductance between disc k and j [H]
L
kk
Self inductance of conductor k in multi-conductor transmission
line
[H]
L
kk
Mutual inductance between conductor k and k in multi-
conductor transmission line
[H]
L
lf
Inductance of a disc for low frequencies [H]
L
self
Self inductance of a straight conductor segment [H]
L
self,j
Self inductance of segment j [H]
l Inductance per unit length [H/m]
/ Conductor length [m]
i
/ Inner length of one side of the square disc [m]
o
/ Outer length of one side of the square disc [m]
M Magnetization (Section 3.3.1 ); mutual inductance (Chapter 4) [A/m]; [H]
M
an
An-hysteretic magnetization [A/m]
M
kj
Mutual inductance between segment k and j [H]
M
p
, M
q
,
M
m+p
, M
m+q
Mutual inductance between two segments with the length p, q,
m+p, m+q, respectively
[H]
M
s
Saturation magnetization [A/m]
m Magnetization of pseudo particle (Section 3.3.1); mutual [A/m]; [H/m]
98
inductance per unit length for adjacent turns (Appendix B)
N Number of measurements performed with single sheet tester [-]
n Number of turns in a disc (Chapters 4, 5, 7, 8), conductors in a
multi-conductor transmission line (Chapter 8)
[-]
n
i
Number of branches in a disc for winding model i [-]
n
p
Number of pseudo particles [-]
n
0
Phenomenological parameter [-]
O Matrix consisting of one column taken out from the C matrix
corresponding to index k
[F]
P Matrix consisting of the k:th column taken out from
T
[-]
P
k
Play-operator with the pinning strength k [-]
p Length [m]
Q Column vector used to prove the validity of RNM for a stack of
slabs
[radAH/s]
Q
k
k:th element of Q [radAH/s]
q Length [m]
R Resistance matrix [O]
R
disc
Resistance matrix for one disc [O]
R
dc
DC resistance [O]
R
kk
Resistance of conductor k in multi-conductor transmission line [O]
R
kk
Mutual resistance between conductor k and k in multi-
conductor transmission line
[O]
R
seg
Resistance of a straight conductor segment [O]
R
seg,j
Resistance of segment j [O]
r' Conductor radius [m]
r'' Shield inner radius [m]
r
Radii of curvature [m]
S Matrix consisting of 0,1 and 1 [-]
s Number of gaps and branches in the reluctance network [-]
TF Transfer function vector
T Duration of one period of measurement [s]
t Time [s]
t
1
Time parameter for V
app
(t) [s]
t
m
Time delay of the reflection from the modified disc [s]
t
we
Time delay of the reflection from the winding end [s]
U Vector containing the voltage drops U
k
[V]
U Voltage [V]
U
k
Induced voltage drop between opposite ends of conductor k in a
multi-conductor transmission line
[V]
u Voltage [V]
u
n
Voltage in the n:th turn [V]
mn
u Induced voltage in the n:th turn due to m [V]
u
0
Voltage amplitude [V]
n
u
Induced voltage in the n:th turn due to the current in itself [V]
V Vector containing the node voltages [V]
V
app
(t) Applied voltage [V]
V
app
() Fourier transform of V
app
(t) [V]
az
V Voltage distribution in a disc for azimuthal resonances [V]
V
j
Voltage in node j [V]
99
V
k
Voltage in node k [V]
rad,k
V Voltage distribution in a disc for radial resonance k [V]
V
0
Phenomenological parameter [A/m]
v Wave velocity [m/s]
v
0
Maximum wave velocity in a winding [m/s]
w Width of laminate (Section 3.3.2) and conductor (Chapter 4,
Section 8.4.2, and Appendix A )
[m]
w' Gap length [m]
w'' Gap length [m]
w
i
Length of gap i [m]
X State vector
x Vector containing parameters of
eff
after
In-admittance of winding after damage (measured in frequency
domain)
[S]
before
In-admittance of winding before damage (measured in
frequency domain)
[S]
Z Impedance matrix for multi-conductor transmission line [O]
Z
int
True internal impedance matrix [O]
int
Z
R
Internal impedance matrix [O]
Z Impedance [O]
Z
int
True internal impedance [O]
Z
int,k
True internal impedance of conductor k [O]
int
Z
R
Internal impedance [O]
int ,k
Z
R
Internal impedance of conductor k [O]
Numerical factor for adjusting R
seg
[-]
o Complex wave number for magnetic penetration in laminate [m
1
]
|
Alternative way to write v when it becomes imaginary [m/s]
I
Matrix connecting the currents and voltages [-]
I
T
Transpose of I [-]
o
Skin depth [m]
,i
Skin depth in the conductor above/below gap i [m]
Permittivity [F/m]
air
Relative permittivity of air [-]
i
Relative permittivity of the conductor insulation [-]
0
Permittivity of free space [F/m]
q
Back field [A/m]
Reluctance factor for skin layer [-]
h
Hysteresis angle [rad]
k
Reluctance factor for skin layer of conductor k [-]
, k i
u Alternative way to write
,i
[-]
, k i
u
Factor similar to
, k i
u [-]
u
,i
Reluctance factor for the skin layer above/below gap i [-]
Propagation constant in a conductor [m
1
]
k
k
Propagation constant in conductor k [m
1
]
k
i
Number for pseudo particle i [-]
Permeability [H/m]
eff
Effective complex permeability of lamination [H/m]
eff
' Real part of
eff
[H/m]
eff
'' Imaginary part of
eff
[H/m]
g
Permeability of the medium in gap [H/m]
g,i
Permeability of the medium in gap i [H/m]
h
Low-frequency complex permeability including hysteresis [H/m]
k
Permeability of conductor k [H/m]
meas
Complex permeability for approximating the measured H-B
curve with a complex- ellipse
[H/m]
meas
' Real part of
meas
[H/m]
meas
'' Imaginary part of
meas
[H/m]
r
Low-frequency relative permeability excluding hysteresis [-]
shield
Permeability of shield [H/m]
0
Permeability of free space [H/m]
Permeability of the conductor above/below gap [H/m]
,i
Permeability of the conductor above/below gap i [H/m]
Radial polar coordinate (Chapter 7); volume charge density
(Section 8.3.2)
[m], [C/m
3
]
0
Inner radius of a disc [m]
1
Outer radius of a disc [m]
o
Conductivity [S/m]
o
dc
DC Conductivity [S/m]
o
g,i
Conductivity of the medium in gap i [S/m]
o
k
Conductivity of conductor k [S/m]
o
shield
Conductivity of shield [S/m]
o
,i
Conductivity of the conductor above/below gap i [S/m]
Density function [-]
i
Twice the insulation thickness [m]
ks
Distance between two discs [m]
u Vector containing the fluxes u
i
[Wb]
u
m
Vector containing the mesh fluxes u
m,k
[Wb]
c
Total magnetic flux inside conducting slab [Wb]
c,k
Total magnetic flux inside conducting slab k [Wb]
g,i
u Magnetic flux in gap i [Wb]
u
i
Magnetic flux in branch i [Wb]
u
m,k
Magnetic flux in mesh k [Wb]
u
,i
Magnetic flux inside the conductor above/below gap i [Wb]
Azimuthal polar coordinate [rad]
Susceptibility at H = 0 [-]
e
Angular frequency [rad/s]
cr
Critical angular frequency [rad/s]
i
Angular frequency for measurement i [rad/s]
101
1
Inverse of t
1
[rad/s]
V
Del operator [-]
R reluctance matrix [A/Vs]
R Reluctance [A/Vs]
' 9 Reluctance between the turns for a single disc [A/Vs]
'' 9 Reluctance outside the turns for a single disc (placed
vertically)
[A/Vs]
''' 9 Reluctance outside the turns for a single disc (placed
horizontally)
[A/Vs]
g,i
R Reluctance of gap i [A/Vs]
i
R Reluctance in gap or channel i [A/Vs]
,i
R Skin layer reluctance in the conductor above/below gap i [A/Vs]
c connectivity matrix for the reluctance network [-]
T
c Transpose of c [-]
ik
Arbitrary element of c [-]
,i Label for all quantities related to the conductor above (+) or
below () gap i
[-]
0 Zero matrix [-]
I Difference between currents in injection-point [A]
() Fourier transform of I [A]
x Differential length [m]
() Difference between the in-admittances of the winding [S]
102
103
List of Acronyms
AC Alternating current
CB Conductor bending
DC Direct current
DTDR Differential time domain reflectometry
FDM Finite difference method
FEM Finite element method
FFT Fast Fourier transform
FRA Frequency response analysis
HV High voltage
LV Low voltage
LWC Loose winding conductor
MIMO Multi input multi output
ode Ordinary differential equation
RB Radial buckling
RNM Reluctance network method
TDR Time domain reflectometry