Design and Comparison of High Frequency Transformers Using Foil and Round Windings
Design and Comparison of High Frequency Transformers Using Foil and Round Windings
Design and Comparison of High Frequency Transformers Using Foil and Round Windings
Topic number: 14
Kartik V Iyer, William P Robbins and Ned Mohan
Department of Electrical and Computer Engineering
University of Minnesota
Email: iyerx070@umn.edu, robbins@umn.edu, mohan@umn.edu
Abstract-High frequency transformers are widely used in inevitable to determine the winding design to be used for a
Switched-mode power supplies and now are being proposed to be given specification.
used with power electronic converters to replace line-frequency
transformers. This paper presents a winding design procedure
In [4] winding losses are computed by considering a 1-
for minimizing the power losses using foils and solid round wires
under sinusoidal excitation to limit the temperature rise. This
D model of a transformer with foil conductors. The optimum
paper derives the range from which the thickness of the layers thickness of foil conductors obtained by the I-D analysis,
can be chosen to obtain the minimum power loss. This thickness as a function of number of layers, were used to design the
range is a function of the number of layers and does not include foil conductors. However, the designs at optimal thicknesses
the "optimum" based on the previous literature. Using this design resulted in ac losses more than the dc losses. Losses can be
procedure, it is shown that interleaving is not necessary in foil reduced if the winding can be designed such that the ac losses
-wound transformers to obtain the minimum loss. A comparison match the dc losses. A comparison between foil and solid
of winding losses between foil windings and round conductors
round conductors for a particular design was shown in [5]
is also given. The analytical results are verified by designing six
which showed that the foil conductors are less lossy than round
different winding configurations for the same specifications using
conductors but still are more than dc losses. In [1] it was shown
2-D Ansys Maxwell finite element design package.
that the foil conductors should be extremely thin to have low
losses. This led to the use of extremely thin strips of conductor
I. INTRODUCTION
at high frequencies known as litz wires, which have ac losses
same as the dc losses. The optimal design and the cost analysis
With better availability of fast switching semiconductor of litz wires were done in [6] and [7] respectively. But the litz
switches there is an increase in trend to employ switch-mode wires are expensive and the window utilization factor is low.
power converters at high frequencies. Magnetic components The choice between foil and litz wire is difficult and it depends
like transformers are an integral part of Switch mode power on various design trade offs as shown in [8]. In literature,
supplies (SMPS). High frequency transformers determine about the comparison between different winding types are more for
25 % overall volume and more than 30 % of the overall weight specific cases than a generalized comparison [5], [8].
of SMPS [1]. Hence, in order to reduce the footprint of SMPS
the reduction in transformer volume is imperative. An increase The winding losses for both foil and solid-round conductor
in frequency allows for the volume reduction of transformers are analyzed. For foil conductors, for a given number of layers,
but at the cost of increased losses [2]. instead of an optimal thickness as stated in literature previously,
there exists a range of thickness, designing at which leads to
The losses occur in the core and in the windings of the
an ac-to-dc resistance ratio less than l.35-l.4. The paper shows
transformer. The transformer is designed to be operated at
that it is possible to design foil, so that the ac to dc resistance
low flux densities to avoid core saturation resulting in lower
is close to 1 thereby giving an alternative to the expensive
core losses. The winding losses in the transformer depend on
litz wires. The paper demonstrates a design procedure with
the number of turns and the winding dimensions. At high
foil conductors for ac-to-dc resistance ratio equal to 1.05 for
frequencies due to eddy currents the effective conduction area
low power high frequency transformers. It is also shown in the
reduces, which increases the losses due to skin effect. The
paper that if designed effectively the interleaving to minimize
increase in losses due to skin and proximity effects depend on
winding losses can be avoided.
the frequency and the winding type. Foil, solid-round wire, litz
wires are the different conductors used in transformer winding
[3]. The paper is organized in the following way: Section
II and III show the conventional and the proposed way of
A proper estimation of the winding losses is crucial for the winding design using foil conductors respectively. Section IV
design of high power density transformer. For the design of describes the step-by-step winding design procedure for foil
low/medium power high frequency transformers, area product and solid round conductor. Section V gives a comparison
is the most common method for core selection. With solid between the foil and solid-round winding in terms of power
round, foil and litz wires being the different winding types, loss. Section VI shows six different winding configuration for
a lot of different winding design schemes are available, which the same specifications and power loss for each is computed
might fit the given window specifications. Hence, a comparison and validated using 2-D FEM. Finally section VII gives the
of the different winding types is terms of winding losses is conclusion.
FR = II
[ Sinh(2ll) sin(2ll)
+
+
�
(p2
_
1)
sinh(ll) - sin(ll) ] Pac =
J
IN(MLT)
a
x
(FR) (5)
cosh(2ll) - cos(2ll) 3 cosh(ll) + cos(ll) Fig. 2a shows the plot of FR/ II versus II for different layers.
(1)
d The Table III shows, in the conventional design, the optimal
where II is the ratio of the layer thickness, and the skin
value of II as a function of number of layers. The FR value
depth, /j at the operating frequency.
varies in a range of 1.36-1.4 for all layers. Hence, if designed
(2) at llopt, using the conventional method, the foil conductors
for any number of layers will give approximately the same
Fig. 1 shows a transformer EE-core with primary and sec- loss, which obviates the need of interleaving to minimize
losses, provided (3) is satisfied. If not, then the thickness of
the conductor has to be increased to fit in the required Nl,
thereby compensating for increased losses. The reason being,
d p
with increasing number of layers the thickness of conductor
Q..-o I··. at.···. Q··-ol reduces which increases the width of the foil conductor, and
D··{] •...• •...• D··-D this increase in width counters the increase in losses by the
O···U •..• •...• Q ..-o Nl c proximity effect.
, , , ,
, , , , , ,
" ,
0···01··1 •...• D···OI As seen in Fig. 2b, by designing the foil conductors in a
specific range of ll, the FR value can be reduced to close to 1.
D Fig. 3 shows that, using llopt = 0.43, which is obtained from
Fig. 2a for p = 10 will give FR = 1.379, but by designing at
II < llopt will lead to FR < 1.379 and hence lower losses.
Fig. 1: Multi-turn Multi-layer foil winding of a transformer. By designing at II such that FR is close to 1, the losses can
be reduced by 37% compared to the conventional design. Even
in the proposed method, by designing the foil conductors such
that FR is close to 1, for all layers, the need of interleaving to
ondary windings.
Fig.2a shows a plot of FR/ II versus II for different layers. minimize the losses can be avoided.
The conventional foil winding design procedure is:
FR can be approximated as [9] for II < 1,
Decide the number of layers, p individually for pri
mary and secondary. FR = 1+ 5p2 - 1 ll4 (6)
2 Locate llopt corresponding to the number of layers
45
chosen from Fig. 2a. It is not possible to take FR = 1, but it is definitely possible
3 Compute dopt, the foil conductor thickness from llopt to design with foil thickness, such that FR is close to 1. For a
as the skin depth is known. particular FR value, II as a function of number of layers, can
4 Compute a, the width of the conductor from dopt and be computed using (7). Here, for the designs shown in section
J VI, FR = 1.05 is considered. Table III based on the proposed
J, the current density, using a = -d-. method provides the value of pll product for 1-20 layers for
J opt
which FR = 1.05.
x
Nl a/'rJl < C (3)
3038
The 2014 International Power Electronics Conference
0.1 10
(a) (b)
6. 6. 6.
Fig. 2: (a) Plot of FRj versus for different layers. (b) Plot of FR versus for different layers.
V-;]
p6. NI
= (11) (13)
J'T}lOC
Rearranging, (8), (9) , (10) and (2), we get, (11).
(14)
All the quantities on the right hand side of (11), are
known.
p
2 From the look up Table
6. III the number of layers, (15)
and corresponding can be chosen.
3 For the number of layers chosen, check if Nl is an where, 'T}2 is the distance between two consecutive
integer, in order to fit integral number of turns in each layers and 'T}l is the layer porosity factor as defined in
layer. [3].
a) If yes, then compute
6. d, the foil conductor 2 The Nln can be determined from window height, C
thickness from as 0, the skin depth, is using (13)
3039
The 2014 International Power Electronics Conference
V. COMPARISON OF FOIL AND SOLID-RoUND WINDING
(l7)
I Jrd2
Substituting, Ac 4' gives,
J
= =
1.50 d2 FR"-R
'T}l
V"4 52 x Sh = FR,,-F (18)
(19)
Yes which shows that the the round wire should be used if and only
if, FR"- < FRf:>F'
R
TABLE I: Specifications
Parameter Value
Power I kW
Voltage 200 V
Fig. 4: Winding design procedure for foil and solid-round Brnax 0.3 T
conductor. kw 0.4
a) If yes, then compute b.R using (15). mm and D 7.78 mm, where c and D are window height and
=
b) If no, then re-compute, NlR by changing 'T}l. window width respectively, as shown in Fig. 1. The number
of turns, N 36, can be computed as core area is known.
4 Compute the winding losses using, (16),
=
'T}
= 0.9 is assumed for all computations.
A 2-D Finite element analysis is done to verify the proposed
(16) design methodology using ANSYS M AXWELL 16.0. A double
EE-core with dimensions of OP44721EC is used. The same
core with 6 different winding configurations is analyzed. Four
Np(MLT)
where, Rdcld=o =
1f<52 cases are of the foil conductors with/without interleaving for
the conventional and proposed method and two cases are for
4 round conductors with/without interleaving.
The section VI will go through the detailed winding design A detailed description about the computation of analytical
procedure which will take into account the interleaving as well. winding loss for all six cases is presented.
3040
The 2014 International Power Electronics Conference
A. Case-A: Non-interleaved Foil winding designed using the If interleaving of primary and secondary winding is consid
conventional method ered, then the best possible way is to have I-layer of primary
and I-layer of secondary. Fig. 6 shows the above described
winding configuration. For such a configuration, from Fig. 2a
or the Table I the value of b. can be computed as 1.58. Using,
(3), Nl can be chosen to be 12 to fit the conductors inside
the window height. The winding arrangement will be 3 sets
of 12 turns of primary followed by 12 turns of secondary. In
this case, a = 1.35mm using (20) and the ac power loss is
44.96Wjm.
The analytical ac loss is around the same as that of the case
with non-interleaved winding. This is because, the width of the
conductor is less in case of single layer as compared to the non
interleaved case. If the design is such that the thickness is the
same and the number of layers are increased then the losses
for single layer would be less. But in general, with the increase
Fig. 5: Non-interleaved Foil winding with 9 layers designed in the number of layers, the thickness is reduced as seen by
using conventional method: (red)-primary winding, (blue) "conventional method" column in Table III, as proposed in
secondary winding previous literature, hence the losses are almost the same.
I •
method with 9 layers. As shown in the Fig. 5, it is a non
interleaved winding structure with 9 layers of primary having
4 turns/ layer followed by 9 layers of secondary. From Fig. 2a
I •
or the Table I the value of b. can be computed as 0.45. Now,
a, which is the height of foil conductor can be computed from
20,
I
a = (20) Fig. 7: Non-interleaved Foil winding with 18 layers de
Job. signed using proposed method: (red)-primary winding, (blue)
which gives a = 4.75mm. As (3) is satisfied, the ac power loss secondary winding
is computed using (4) which gives 42.393Wjm. The ac power
loss can also be computed using (5) which gives the same result
and it shows that FR = 1.366. Even if, p = 18 would have
been chosen then still the losses would have been around the Fig. 7 shows the winding configuration designed using the
same value as in that case the width of the conductor would proposed method with 18 layers. As shown in the Fig. 7, it is
be increased such that the FR is around 1.36 - 1.4. a non-interleaved winding structure with 18 layers of primary
having 2 turns/ layer followed by 18 layers of secondary. If
B. Case-B: Interleaved Foil winding designed using the con
instead of using Dowell's curve, the curve as shown in Fig.
ventional method
2b is used, then the b. can be selected such that FR is close
to 1, provided the computed width of the conductor satisfies
IIIIII (3). Here, in this analysis b. is chosen such that FR = 1.05.
I I I The " proposed method" column in Table III is constructed
I I I considering FR = 1.05. But the Table can be designed for any
I I I other value of FR using (7). Depending on the chosen value
I I I of FR the analytical winding losses can be reduced compared
I III to the conventional method.
I I I For the example considered, the pb. = 3.537. From the
I III "proposed method" column in Table III, the number of layers,
IIIIII p, for which pb. is close to 3.537 is p = 18 which also makes
II I I Nl = 2, which is an integer. From Table III the value of b.
I III
can be computed, which is 0.19. Now, a, which is the height
I I I
of foil conductor can be computed from 20, which gives a =
Fig. 6: Interleaved Foil winding with 1 layer designed using 1l.2629mm. This value of a will not satisfy (3). Hence, b.,
conventional method: (red)-primary winding, (blue)-secondary should be chosen so as to allow 2 turns/layer. Considering,
winding b. = 0.1965 which is obtained by substituting p = 18 in pb. =
3.537 will allow 2 turns/layers but, FR = 1.0536 > 1.05 but
3041
The 2014 International Power Electronics Conference
still less than 1.37. In this case, a = 1O.89mm. The ac power Fig. 9 shows the winding configuration designed using the
loss is computed using (4) which gives 32.698Wjm. The ac round conductors without interleaving the primary and sec
power loss can also be computed using (5) which gives the ondary winding. The flowchart as shown in Fig. 4 is followed
same result and it shows that FR = 1.0536. for round conductors. As the given case is a non-interleaved
structure, both layers of primary are wound together. The
D. Case-D: Interleaved Foil winding using the proposed diameter dR can be computed using (12), which is, dR = 1.128
method mm. AWG-17 is the nearest solid-round wire. The number
of turns per layer, NIR can be computed using (13) and as
the windings satisfy (14), 6.R and hence, the losses can be
computed.
�I
Fig. 8: Interleaved Foil winding with 1 layer designed us
ing proposed method: (red)-primary winding, (blue)-secondary
L8:81
winding
tf· ! Wi
Fig. 8 shows the winding configuration designed using the
\lIDJ
proposed method with interleaving of primary and secondary
winding. As shown in the Fig. 8, it is an inter-leaved winding
structure with 4 sets of 9 turns of primary followed by 9 turns 0.0.
of secondary. For such a configuration, number of turns in 1- �=--�.... o.o.
layer of primary winding can be selected from the "proposed 0.0.
method" column in Table I using p6. = 0.88 in (21) .
0.0.
0.0.
0.0.
0.0.
(21) 0.0.
Fig. 10: Interleaved Round conductors with 1 layer: (green)
Using, the above data, number of turns in I-layer of primary
primary winding, (red)-secondary winding
winding can be selected from, (21) which gives N = 8.9. In
order to fit in, 9 turns per layer, 6. is chosen to be 0.9. Using,
6. = 0.9 gives FR = 1.058 slightly more than 1.05. In this
case, the ac power loss is 32.586Wjm. The Table II compares the losses for all cases. The an
alytical and simulated losses are computed per unit length.
E. Case-E: Non-interleaved Round conductors The average mean length of turns can be chosen as 0.125
m according to [10]. The winding thickness using proposed
8§11
method due to window dimension restrictions for 1 layer was
0.9 instead of 0.88 and 0.1965 instead of 0.19 for 18 layers.
The results match closely with the analytical calculation for
./1 8811 foil conductors but are different for round winding due to 2-
� g811 D effects. Also, the losses for solid-round conductor for both
§§ii
interleaving and non-interleaving cases are more than the foil
� 'II
conductors. In case of foil winding, interleaving was not a
concern but for solid round conductors as diameter is fixed,
6. is fixed and hence, interleaving has to be done to reduce
00 ••
losses.
8811
8811 VII. CONCLUSION
3042
The 2014 International Power Electronics Conference
ratio close to 1 can be achieved using foil conductors provided [2] W.-J. Gu and R. Liu, "A study of volume and weight vs. frequency for
the winding can fit inside the window. Hence, it can replace high-frequency transformers," in Power Electronics Specialists Confer
ence, 1993. PESC '93 Record., 24th Annual IEEE, 1993, pp. 1123-1129.
the expensive Iitz wires which also have low window utilization
[3] R. Wojda and M. Kazimierczuk, "Analytical optimization of solid
factor. The paper also demonstrates that a proper design can
x2013;round-wire windings," Industrial Electronics, IEEE Transactions
help to avoid interleaving in case of foil conductors for min on, vol. 60, no. 3, pp. 1033-1041, 2013.
imizing winding loss. The paper also presents a comparison [4] P. Dowell, "Effects of eddy currents in transformer windings," Electrical
between foil and round conductors in terms of winding losses. Engineers, Proceedings of the Institution of, vol. 113, no. 8, pp. 1387
Six different winding configurations with/without interleaving, -1394, august 1966.
foil/round conductors for the same specification and on the [5] W. Hurley, W. Woltle, and J. Breslin, "Optimized transformer design:
same EE-core are analyzed using 2-D FEM simulations. For inclusive of high-frequency effects," Power Electronics, IEEE Transac
tions on, vol. 13, no. 4, pp. 651-659, 1998.
all six cases losses are computed and compared with analytical
[6] c. R. Sullivan, "Optimal choice for number of strands in a litz
results. The results show that interleaving of winding is a must
wire transformer winding," in Power Electronics Specialists Conference,
in case of solid- round conductors whereas interleaving can be 1997. PESC'97 Record., 28th Annual IEEE, vol. 1. IEEE, 1997, pp.
avoided in case of foil winding if designed properly. 28-35.
[7] C. Sullivan, "Cost-constrained selection of strand diameter and number
in a 1Hz-wire transformer winding," Power Electronics, IEEE Transac
TABLE III: Parameters to compute p6 with the conven tions on, vol. 16, no. 2, pp. 281-288, 2001.
tional/proposed method . [8] M. Kheraluwala, D. Novotny, and D. Divan, "Design considerations
for high power high frequency transformers," in Power Electronics
p conventional method proposed method Specialists Conference, 1990. PESC '90 Record., 21st Annual IEEE,
I 1.58 1.58 1.449 0.88 0.88 1.052 [9] W. Hurley, E. Gath, and J. Breslin, "Optimizing the ac resistance
of multilayer transformer windings with arbitrary current waveforms,"
2 0.97 1.94 1.361 0.59 1.18 1.051
Power Electronics, IEEE Transactions on, vol. 15, no. 2, pp. 369 -376,
3 0.78 2.34 1.356 0.48 1.44 1.052 mar 2000.
4 0.67 2.68 1.351 0.41 1.64 1.05 [10] N. Mohan, T. M. Undeland, and W. P. Robbins, "Power electronics:
5 0.6 3 1.355 0.37 1.85 1.052 converters, applications, and design," 1989.
6 0.55 3.3 1.362 0.33 1.98 1.047
7 0.51 3.57 1.365 0.31 2.17 1.05
8 0.48 3.84 1.374 0.29 2.32 1.05
9 0.45 4.05 1.366 0.27 2.43 1.048
10 0.43 4.3 1.376 0.26 2.6 1.05
II 0.41 4.51 1.377 0.25 2.75 1.052
12 0.39 4.68 1.368 0.24 2.88 1.053
13 0.38 4.94 1.387 0.23 2.99 1.052
14 0.36 5.04 1.364 0.22 3.08 1.051
15 0.35 5.25 1.373 0.21 3.15 1.049
16 0.34 5.44 1.377 0.21 3.36 1.055
17 0.33 5.61 1.378 0.2 3.4 1.051
18 0.32 5.76 1.375 0.19 3.42 1.047
19 0.31 5.89 1.368 0.19 3.61 1.052
20 0.3 6 1.359 0.18 3.6 1.046
REFERENCES
3043