IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 9, NO.

I , JANUARY 1994 127

Improved Analytical Modeling of

Conductive Losses in Magnetic Components
Jan A. Ferreira, Member, IEEE

Abstract- The one well-known one-dimensional method for At dc, the resistances are equal, although at high fre-
calculating the ac resistance of multilayer windings contains a quencies, the square representation of round conductors
built-in orthogonality which has not been reported previously. becomes inaccurate.
Orthogonality between skin effect and proximity effect makes a
more generalized approach for the analytical solution of ac resis- 4) The skin depth is made a function of the axial packing
tance in windings possible. This includes a method to calculate factor. Obviously, something as fundamental as skin
the ac resistance of round conductor windings which is not only depth cannot be a function of geometry.
convenient to use, but gives more accurate ansers than the basic Despite these shortcomings, this method received much
one-dimensional method because the exact analytical equations
for round conductors can be used. attention during recent years to such an extent that an overview
paper [6] was presented. In this paper, it will be shown that
it is possible to overcome the above shortcomings within the
I. ANALYTICALCALCULATION OF COPPER LOSSES scope of an exact analytical approach, and a more generalized

D UE to a trend towards higher frequencies and the emer-

gence of many resonant converter configurations which
all have inductors or transformers, the analysis of copper
method of modeling conductive losses during the design of
magnetic components is presented.

losses has been receiving renewed attention during the past

11. NOTATION
few years. These papers (for example, 131-161) are almost
exclusively based on the one-dimensional solution developed The original one-dimensional analysis is applicable to paral-
by Bennet [ 11 and Dowell 121. Perrys approach 131 is similar lel bus bars of infinite width, and since this paper examines the
to that of Bennet [ l ] , and both involve a multilayer infinite fundamentals of the one-dimensional approach, we return to
solenoid wound with foil conductors. Dowell [2] has been this notation. Provided the curvature is significantly smaller
the first to adapt this strictly one-dimensional solution for that the thickness of the conductor, the analysis can be
practical transformers, recognizing that the leakage flux lines applied to a tall solenoid. Both Bennet [l] and Perry [3]
run parallel to the surface of foil windings. He introduced two indicated that a Bessel function version of the solution can be
convenient geometrical adaptations which have also been used derived, which takes account of the curvature. If the method
in subsequent papers, for example, [4), 151. of images is applied to model the leakage field of a foil
This particular approach, being one dimensional in rectan- winding in a transformer, the highly permeable core projects
gular coordinates, is in principal applicable to foil conductors an infinite series of windings along the axial direction [13]. It
having a magnetic field parallel to the conductor surface, and effectively gives rise to an infinitely long solenoid, provided
is therefore subject to certain restrictions: that the windings extend over the complete length of the
Magnetizing current of transformers cannot be included, winding window. This is the reason why the other papers [ 2 ] ,
as it results in a magnetic field component that is not [4], [5] successfully applied the one-dimensional analysis to
parallel to the foil conductor surface. For the same transformers.
reason. this approach cannot be used to analyze the The one-dimensional analysis in this paper uses the bus bar
conductive losses in most practical inductors. configurations shown in Figs. 1 and 2. The following should
Even when considering transformers with negligible be observed:
magnetizing current, it should be realized that, strictly 1) 1, = current per unit width of a bus bar.
speaking, the analysis is valid for infinitely long solenoid 2) The average magnetic field across the single foil con-
windings. When the windings fill the window length ductor (Fig. 1) is equal to zero.
completely, or if the distance between primary and 3) The average magnetic field across the mth foil conductor
secondary is small, the resultant field approaches that (Fig. 2) is equal to ( m - + ) I f Dowell
. [ 2 ] extended the
of infinite solenoid windings scope ofthe analysisto include round conductors (wire)by
Error is introduced by replacing round conductors by replacing the round conductors with square conductors
square-shaped conductors of equal cross-sectional area. of equal cross-sectional area. This inevitability leads to
gaps between adjacent square conductor representations,
Manuwipt revi5ed September 29, 1993. and a packing factor 7 as shown in Fig. 3 is defined.
The author is with the Energy Laboratory, Rand Afrikaans University,
Aucklandpark 2006. South Africa. This results in the following:
IEEE Log Number 9214823. 4) Current per square conductor = I f / t .
0885-8993/94$04,00 0 1994 IEEE
I28 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 9, NO. I , JANUARY 1994

10

RAC
Rdc

Fig. 1. Magnetic field intensity at the boundary of a single foil conductor

1
0.1 100
fnorm
Fig. 4. Comparison between the exact cylindrical solution (0trace) and
the square conductor, one-dimensional representation (+ trace) of a round
conductor.

is given by the following equation:

Rd,( sinh ( sin [ +
&kin = ~

2 <
cosh - COS (
where
Fig. 2. Magnetic field intensity at the boundaries of two adjacent sets of foil
conductors (according to Amptres circuital law). E = - -f, i d
2 6
0 0 Although (1) and ( 2 ) give the same dc resistance, the approx-
imation with rectangular conductors (2) becomes optimistic at
high frequencies, as can be seen from the graph in Fig. 4.
For example, even at f n o r m = 10 (i.e., the frequency where
the skin depth is one tenth the wire diameter), the rectangular
approximation overpredicts by 16%. It is evident from this
graph that even for a single conductor configuration, significant
* improvement is obtained by opting for the cylindrical solution
+ U=- t instead of the equal area one-dimensional solution.
a) b) S+t C ) d)
Fig. 3. Representation of round conductors in terms of the one-dimensional BETWEENSKINEFFECT
IV. ORTHOGONALITY
model: a) foil conductor, b) foil consisting of square conductors, c) spaced
square conductors, d) equal area round conductors. AND PROXIMITY EFFECTIN CONDUCTORS

5 ) Average magnetic field across mth square conductor Orthogonality exists between skin effect and proximity
layer = ~ ( -m+)If. effect when the applied magnetic field due to other conductors
The frequency response and conductor dimensions (proximity field) is assumed to be uniform over the conductor
have been normalized in the paper by introducing a cross section. This permits decoupling of the two effects
normalized frequency as follows: and simplifies calculation significantly; it is also described in
6) fnorm = 1 is defined as the frequency at which the [8] and [9], but for the purpose of this paper, it has been
conductor diameter is equal to the skin depth. summarized and placed in Appendix A.

OF SINGLE ROUNDWIRE
111. AC RESISTANCE V. ONE-DIMENSIONAL
ANALYSIS
USINGORTHOGONALITY
APPROACH
A single round wire, far removed from the retum conductor,
will only be subjected to skin effect. The exact analytical References [ 11 and [31 derived identical expressions for the
solution is readily found in the literature [7]: ac resistance of the with layer (see Fig. 2). Rewriting the
expression in terms of the notation adopted in this paper,
+
where
d
R,, = Rd,< (2rn
[
- (4m2 - 4m)
-
sirili 2< sin 2(
2rn + 1)
cosh 2[ - COS 2[
<
cos sirih sin [cosh + <
I
(3)
y=- cosh 2 [ - cos 2[
6Jz
Upon substitution, identities (B2) and (B3) in the above
and
expression, the following equation is obtained:
6 = skindepth.
+ sin E + ( ~ r r c
If the round conductor is replaced with a square-shaped con-
ductor of equal cross-sectional area, the skin effect resistance
sirih
- 112
sirih
cos11
E
(
- sin (
+ cos ( I
(4)
.

1 7
FERREIRA: IMPROVED ANALYTICAL MODELING OF CONDUCTIVE LOSSES 129

The first term in the above equation is identical to (2) and

describes the skin effect. The second term gives the proximity
effect and is directly compatible with (A7) and an applied
magnetic field of strength (m - (kextstyle { 1 \over 2)])I-(f]
(see point 3 in Notation section).
It seems that over a period of almost 50 years, the one-
dimensional analysis has been applied to multilayer windings,
many publications have appeared (of which six are listed
as references), and the inherent orthogonality between skin
effect and proximity effect for the particular analysis has not
been recognized. The orthogonality opens the way to new
opportunities concerning analytical solution of eddy currents
in magnetic components. It indeed becomes possible to apply
the analysis to two-dimensional problems, as has been done
0.1 f m
with the CAD programs described in [13].
(a)

VI. AC RESISTANCE
OF ROUNDWIRE WINDINGS
The orthogonality principle makes theoretically more accu-
rate prediction of the ac resistance in multilayer round wire
windings possible. The assumption is made that the applied
field due to proximity effect is uniform, in which case skin
effect and proximity effect are orthogonal as described in
Appendix A, and a similar expression to (4) can be obtained
by applying skin effect resistance [(l) and proximity effect,
factor G (A8)], and the expression for the ac resistance of the
mth layer becomes
ber y bei' y - bei y ber' y
= Rdc - - 27r(2m - 112
+
berl2 y beil2 y
+
ber2 y ber' y bei2 y bei' y
+
ber2 y bei2 y
(5)

The method to calculate the ac resistance of round wire

windings in past publications by some authors (for example,
[ 2 ] ,[4], and [ 5 ] )has been to replace the round conductor with a
square conductor of equal cross-sectional area, and then to use
either (3) or (4). Equation (5) is a more accurate expression,
and to illustrate the error introduced by the equivalent square
conductor substitution in the one-dimensional analysis, the (b)

ratio Rac/Rdchas been plotted in Fig. 5. Note how the Fig. 5. Comparison between the exact cylindncal solution (0trace) and the
prediction by the square conductor representation is too low. square conductor, one-dimensional representation (+ trace) of the ac resistance
in two different layers of a multilayer winding.
In fact, at fnorm = 10, the prediction error is already -30%
for m = 1 and -47% for m = 4. of each layer is reduced by the factor 7 and the magnetic field
must be compensated by same factor, as is described in point
VII. GEOMETRICAL DEPENDENCY
OF SKIN DEPTH 5) in the Notation section. Equation (4) then becomes:
Dowel1 [2] introduced a skin depth that is dependent on the
packing factor 7 (defined in Fig. 3), which was also adopted
by subsequent papers [4], [5]. The geometrical skin depth is
defined as follows: (7)
1
(6) where
F .

where 0 < 7 5 1.
Such an adaptation was necessary because 7 involves a
second dimension within a strictly one-dimensional solution. The factor q2 in the above equation is valid and makes
In effect, the gaps between the square conductor are filled up physical sense. Skin depth, on the other hand, is a physical
and the conductivity is adjusted to create infinitely wide foil material constant and cannot be dependent on geometry.
conductors. Furthermore, the average current along the width Proximity effect is caused by eddy currents induced by an
130 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 9, NO. 1. JANUARY 1994

the equivalent rectangular conductor-based solution gives good

predictions, provided magnetizing current is small, the winding
is tightly packed, and fnorm < 3.

APPENDIXA
ORTHOGONALITY IN THE POWER
Rac INTEGRAL OF EDDYCURRENT LOSSES
Rdc The current flow inside a conductor lead is taken to be in
the z direction and the density distribution is only a function
of the z,y coordinates.
J ( z ,y, t ) = J ( z , y) cos wt (41)
where J is the current density (phasor quantity) and w is the
repetition frequency of the current waveform.
1 The power dissipation per unit length (Pu)of the conductor
ai f m is, by definition,
Fig. 6. Ratio R a c / R d ,for true skin depth (0trace) and geometrical skin
depth (+ trace) when m = 4 and 11 = 0.6.

externally applied magnetic field. More appropriate would

be to temporarily remove a particular conductor from the where
configuration, compute the total field due to the other current- A cross-sectional area of the conductor,
carrying conductors, and from this determine the proximity T period of the current waveform,
effect. The difference and possible error which is introduced U conductivity of the winding material.

by using the geometrical skin depth is illustrated in Fig. 6. Since we have a sinusoidal waveform, (A2) can be simpli-
fied to
VIII. EXPERIMENTAL
MEASUREMENTS
An aspect common to the papers listed on the one-
P, = & s, J . JdA
dimensional analysis is the absence of proper experimental
Designating the skin effect and proximity effect current den-
verification. References [ l ] and [3] do not report any
sities J , and Jp. respectively, (A3) becomes
experimental work. Dowel1 [ 2 ] reports that the experimental
results have not been checked thoroughly and presents
no formal experimental results. Vandelac [5] mentions one
measurement. One reason for the lack of experimental suport
Provided the conductors have an axis of symmetry and the
is the technical difficulties involved with the measurement [ 101
applied field is uniform and parallel to the symmetry axis,
and the influence of secondary effects [ 111 such as interturn
an induced current distribution Jp displays reciprocal sym-
capacitance. Measurements were conducted by the author [8],
metry. The skin effect current distribution relative to the same
using the orthogonality approach, and good agreement was
symmetry axis is an even function, and under these conditions,
obtained between measured and predicted values for coils
using the orthogonality approach for round conductors.

IX. CONCLUSIONS
It has been shown that skin effect and proximity effect are
where Pusare the skin effect losses nd Pup are the proximity
orthogonal to each other for the one-dimensional Cartesian
effect losses.
solution for conductive losses in magnetic components, which
Skin effect losses per unit length is simply calculated using
means that the losses can be calculated separately and then
the following equation:
simply added together. This orthogonality concept can be
extended to a more exact method for calculation of the ac Pus = i R s k i n I 2 (A6)
resistance of round conductors.
It has been shown that this method of calculation is in- where I = peak value of current and &in is given by either
herently more accurate than a popular existing method using (1) or the rectangular approximation ( 2 ) .
an equivalent rectangular conductor and the one-dimensional The proximity effect losses for a round conductor can be
Cartesian coordinate system solution. Two areas of erroneous expressed per unit length as
prediction of the latter solution include the potential error due
pup= G H , ~ (A7)
to a geometrical dependency of the skin effect parameter,
and a prediction error which becomes prominent at frequencies where He = peak value of the external magnetic field caused
larger than fnorm = 3. On the positive side, it appears as if by currents in surrounding conductors.
 1 1

FERREIRA: IMPROVED ANALYTICAL MODELING OF CONDUCTIVE LOSSES 131

The exact expression for G for a round conductor is as P. L. Dowell, Effects of eddy currents in transformer windings, Proc.
follows: IEE, vol. 113, pp. 1387-1394, Aug. 1966.
M. P. Perry, Multiple layer series connected winding design for

Gz-
-21ry ber2 y ber y + be22 y ber y . (A8)
minimum losses, IEEE Trans. Power App. Syst., vol. PAS-98, pp.
0 ber2 y + bei2 y 116-123, Jan./Feb. 1979.
B. Carsten, High frequency conductor losses in switch mode magnet-
ics, in PCI Proc., Munich, Germany, 1986, pp. 161-182.
The expression for G for the equivalent rectangular conductor J. P. Vandelac and P. D. Ziogas, A novel approach for minimizing high-
presentation is given by frequency transformer copper losses, IEEE Trans. Power Electron., vol.
3, pp. 266-276, July 1988.
\/r A. M. Urling, V. A. Niemela, G. R. Skutt, and T. G. Wilson, Char-

Gr-2 dE s i n h t - s i n (
acterizing high frequency effects in transformer windings-a guide to
several significant articles, in Conf. Rec. 1989 PESC, pp. 373-385.
0 cosh< +cos< R. L. Stoll, The Analysis of Eddy Currents. Oxford: Clarendon Press,
1974.
APPENDIXB J. A. Ferreira, Electromagnetic Modeling of Power Electronic Convert-
ers. Boston/Dordrecht/London: Kluwer Academic, 1989.
HYPERBOLICAL/TRIGONOMETRICAL
IDENTITIES J. A. Ferreira and J. D. van Wyk, A new approach to analyze conduction
losses in high frequency magnetic components, Trans. SAIEE, vol. 80,
The eddy current equations contain hyperbolical functions no. 1, pp. 6 1 4 7 , 1989.
of a complex variable cr of which the real and imaginary parts V. A. Niemela, G. R. Skunt, A. Urlin, Y. N. Chang, T. G. Wilson, H.
are equal: A. Owen, and R. C. Wong, Calculating the short-circuit impedance
of a multiwinding transformer from its geometry, in Con$ Rec. 1989
PESC, pp. 607417.
a = a+ja. 031) J. A. Ferreira and J. D. van Wyk, Experimental evaluation of losses
in magnetic components for power converters, IEEE Trans. Ind. Appl.,
The following two mathematical identities arise from the vol. 27, pp. 335-339, Mar./Apr. 1991
solution of the hyperbolic functions:
sinh 2a
cosh 2a
+ sin 2a
- cos 2a
=-[ +
1 sinh a sin a
2 cosh a - cos a
sinh a - sin a
cosh a cos a+
Jan A. Ferreira (M88) was bom in Pretoria, South
Africa, in 1958. He received the B.Eng., M.Eng.,
sinh 2a + sin 2a - sinh a - sin a and D.Eng. degrees in electrical engineering from
cosh 2a - cos 2a cosh a + cos a the Rand Afrikaans University, Johannesburg, South
Africa, in 1980, 1982, and 1988, respectively.
COS a sinh a + cosh a sin a) In 1981 he worked at the Institute of Power Elec-
+

cosh 2a - cos 2a tronics and Electric Drives, Technical University of

Aachen, and worked in industry from 1982 to 1985.
(B3) Since 1986 he has been on the Faculty of Engi-
neering, Rand Afrikaans University, and currently
REFERENCES holds the Carl and Emily Fuchs Chair of Power
Electronics. He is the author and co-author of 50 papers, and his research
[1] E. Bennet and S. C. Larson, Effective resistance of altemating currents includes magnetics, power electronics, and manufacturing technology.
of multilayer windings, Trans. Amer. Znsr. Elecr. Eng., vol. 59, pp. Dr. Ferreira is currently Vice Chairman of the IEEE South Africa Section
101&1017, 1940. and Chairman of the South African Energy Chapter of the IEE.