J.

Electromagnetic Analysis & Applications, 2010, 2, 627-632

doi:10.4236/jemaa.2010.211082 Published Online November 2010 (http://www.scirp.org/journal/jemaa)

Analysis of Thermal Behavior of High Frequency

Transformers Using Finite Element Method
Hossein Babaie, Hassan Feshki Farahani

Ashtian Branch, Islamic Azad University, Tehran, Iran.

Email: {hbabaei2002, hfeshki}@yahoo.com

Received August 15th, 2010; revised September 17th, 2010; accepted September 20th, 2010.

ABSTRACT
High frequency transformer is used in many applications among the Switch Mode Power Supply (SMPS), high voltage
pulse power and etc can be mentioned. Regarding that the core of these transformers is often the ferrite core; their
functions partly depend on this core characteristic. One of the characteristics of the ferrite core is thermal behavior
that should be paid attention to because it affects the transformer function and causes heat generation. In this paper, a
typical high frequency transformer with ferrite core is designed and simulated in ANSYS software. Temperature rise
due to winding current (Joule-heat) is considered as heat generation source for thermal behavior analysis of the trans-
former. In this simulation, the temperature rise and heat distribution are studied and the effects of parameters such as
flux density, winding loss value, using a fan to cool the winding and core and thermal conductivity are investigated.

Keywords: High Frequency Transformers, Thermal Behavior, Ferrite Core and Finite Element Analysis

1. Introduction conductor losses for power electronics applications have

been considered by a number of authors [8,9]. There are
Magnetic components design plays a key role in achiev-
models provided by different manufacturers [10-12], that
ing high efficiency, low volume and reasonable price of
are synthesized in a simple expression for the calculation of
power electronic equipment. With the advent of higher
the temperature rise. They provide the average temperature
switching frequencies and power densities in power elec-
in the outer surface of the device.
tronic circuits, it is important to ensure that magnetic Most of thermal models for magnetic components are
components, such as transformers and inductors, operate analytical and assume 1D heat transfer [13-15]. These
within the limits defined by the thermal specifications of models are based on thermal networks and commonly
the circuit. Temperature rise depends on the power losses consider only steady state, constant thermal properties,
through the laws of the heat transfer theory. In order to concentrated and uniform power losses and no thermal
obtain an accurate value of the maximum temperature of feedback with the electric properties. However, the ther-
the device during the design process, it is necessary to mal distribution is 2D/3D for most of cases, even if the
apply an accurate thermal model. It is necessary of a magnetic field exhibits a 1D distribution.
trade-off between thermal distribution accuracy (direc- The thermal design is usually somewhat neglected as
tions through heat transfer is considered, constant ther- it is often not clear which theory and coefficients should
mal properties, feedback with the magnetic model, be used and experiments are time consuming. In this
steady state models, etc) and the complexity of the ther- article, by designing a typical high-frequency trans-
mal model that is obtained. former with ferrite core, its thermal behavior is investi-
Losses in the magnetic components are important design gated. Many parameters such as flux density, winding
parameters. In many high-frequency designs, magnets are loss value, using a fan to cool the winding and core and
limited by their losses. Thus, it is extremely important for thermal conductivity have effects on the ferrite tempera-
the designer to have a good practical model for estimating ture which is simulated with ANSYS software.
the losses under various excitations that are frequently en- 2. Calculating the Heat Generation Caused
countered in power electronics design. There are two major
by Winding and Investigating Its
components of the losses [1-9] in a magnetic component:
the core loss (i.e., the losses in the magnetic material which
Effective Parameters
is used as core) and the winding losses. High-frequency Heat generating sources are transformer windings which

Copyright © 2010 SciRes. JEMAA

628 Analysis of Thermal Behavior of High Frequency Transformers Using Finite Element Method

generate heat by passing the current and this heat is v1(t)

transmitted to other parts. For a winding with resistance Area λ1
Rj, the power loss can be written as:
Vg
Pcu , j  I 2j R j (1)
0 0
Which the jth winding resistance is as follow: t1 t2

lj
Rj   (2) -Vg
Aw, j
Figure 1. Transformer primary voltage.
Where:
l j  n j  MLT  (3) Equation (10) consists of three parts in which part A is
electrical characteristics and B and C include core and
WA K u  j
Aw, j  (4) magnetic characteristics respectively.
nj The core loss is obtained from the following equation:
In Equation (3), MLT is mean length per turn (cm). By Pfe  k fe B  max Ac lm (12)
considering Equation (2) to Equation (4), the power loss
of jth winding is as follow: Kfe is core loss coefficient which is different for various
frequencies. Ac is core cross-sectional area and lm is
n 2j i 2j   MLT  magnetic path length. The typical value of β for ferrite
Pcu , j  (5)
WA K u  j core is 2.6 < β < 2.8. Kfe is drastically increased by in-
creasing the frequency. In addition, Kfe depends on core
And the total power dissipation is the sum of the power temperature and Bmax. The dependency of Kfe on fre-
losses for each winding: quency, Bmax and temperature can be obtained from core
characteristics. In choosing the core when combined with
  MLT  k  n 2j I 2j 
Pcu ,tot    

(6) different alloys, there has been always trade-off between
WA K u j 1   j  saturated magnetic flux density and the core loss. Using
materials with high Bsat leads to volume, size and price
The loss in winding is optimal providing that [16]: reduction. But these materials cause more core loss.
  MLT  k
Pcu ,tot 
WA K u
 nj I j  (7) 3. Transformer Characteristic for
j 1 Full-Bridge Converter
According to Equation (7), it can be found out that the To study the temperature rise in ferrite core and its effec-
power loss or generated heat depends on parameters such tive parameters, a sample transformer for full bridge con-
as ρ (wire effective resistivity), Ku (winding fill factor), verter with two secondary windings and one primary
turns, winding current and MLT (mean length per turn). winding is chosen which can be seen in Figure 2 [16].
So, these are effective parameters on generation heat
source. In a transformer, if the primary voltage is like
Figure 1, then the following equations will be written:
1  t v1  t dt
t2
(8)
1

1 1
Bmax   n1  (9)
2n1 Ac 2 Bmax Ac
By substituting Equation (9) in Equation (7), the result
can be written as:
  MLT  I tot
2
  2 I 2    MLT    1 
Pcu  n12   1 tot     2  (10)
WA K u ku   WA Ac2   Bmax
    
A B C

where:
k nj
I tot   Ij (11) Figure 2. Transformer for full bridge converter with two
j 1 n1 outputs and one input.

Copyright © 2010 SciRes. JEMAA

 Analysis of Thermal Behavior of High Frequency Transformers Using Finite Element Method 629

Table 1. Parameters of designed transformer for full bridge

converter
Transformer frequency 75kHz Air
Turn ratio 110 :5 :15
Duty cycle 0.75
Selected core EE40
kfe 7.6 W/Tβcm3
Winding
β 2.6 Core
Ku 0.25
ρ 1.724×10-6 Ω-cm
λ1 800 V-µsec
I1 5.7 A
I2 66.1 A Air
I3 9.9A
Itot 14.4 A
Bmax 80 mT
n1 22
Figure 3. Modeled transformer in ANSYS.
n2 1
n3 3
Pfe 230 mW Table.2 Electrical and thermal characteristics for different
Pcu 3.89 W parts of transformer
Ptot 4.12 W Material
Parameter
Copper Ferrite Core Air
The obtained parameters for this transformer are listed Density, ρ [kg/m3] 8900 8900 1.09
in Table 1. The core and winding losses calculated for Specific heat, c [J/kgºC] 387 750 1006
transformer are equal to 230 mW and 3.89 W respectively. Thermal conductivity,
385 0.004 0.027
These losses can be used as heat generating source in λ [W/(m·K)]
modeling and simulating the transformer in ANSYS.
4. Transformer Modeling in ANSYS
In this part, transformer with two secondary windings is
modeled in ANSYS software which is shown in Figure 3.
The distance between each winding is 0.5 mm. the pri-
mary winding is 20 mm long and 2 mm wide which this
cross-sectional area includes 22 turns primary winding.
This area for secondary winding is 20 mm2 (1 mm × 20
mm) which is for one turn. Also, another secondary
winding is 20mm long and 0.5mm wide which consists of
three turns.
The electrical and thermal characteristics for different
parts of transformer are listed in Table 2.
5. Simulation Results
In this simulation, it is assumed that the heat transfer be- Figure 4. Thermal distribution in transformer.
tween winding and core of the transformer has been
naturally done (natural convection). So, the film coeffi- through the transformer is shown. In this figure, the hottest
cient varies from 10 W/m2·ºC to 25 W/m2·ºC when there point is primary winding and middle leg of ferrite core
is natural convection. The film coefficient is considered which is about 61.962ºC. By paying attention to this fig-
equal to 10 W/m2·ºC. The total cross-section of area is 70 ure, it is clear that the core corners have lower tempera-
mm2 and the power dissipation in this area is 3.89W. ture in comparison with other parts. It should be noticed
Therefore, the value of watt per square meter of area is that in this case, the secondary windings temperature are
equal to 55570 W/m2 which is used for heat generation lower than that in primary one. The heat distribution in
source in ANSYS. In Figure 4, temperature distribution ferrite core is shown in Figure 5.

Copyright © 2010 SciRes. JEMAA

630 Analysis of Thermal Behavior of High Frequency Transformers Using Finite Element Method

Figure 7. Thermal distribution in transformer (Decreasing

flux density).
Figure 5. Thermal distribution in core.
this case, one of the secondary windings temperatures
(neighbor of primary winding) is approximately equal to
Thermal flux in this case is shown in Figure 6 and it is
primary winding temperature (the highest temperature).
maximized in primary winding corners which are about
428.69 W/m2. Regarding this matter and according to 5.2. Variation of the Ambient Temperature
Figure 4, it can be pointed out that the corners tempera-
In this case, it is supposed that the transformer is used in
tures are lower than that in other parts.
40ºC temperature. So, it is observed that the maximum
5.1. Variation of Flux Density temperature inside the transformer has been slightly in-
creased and has reached to 61.971ºC. In this case, the
In this part, the effect of flux density variation on tem-
maximum temperature is in primary winding and core
perature distribution in transformer has been investigated.
middle leg as well.
For this purpose, the flux density is decreased from 80 mT
By comparing Figure 4 with Figure 8, we can make
to 120 mT. According to Equation (10), the power loss is
the conclusion that regarding the proximity of core sides
in inverse proportion with the square of flux density and
to heat source and warmer environment, the sides’ tem-
by this change, the power loss is decreased 2.25 times.
perature are higher than that in up and down of core.
The heat distribution for this condition is shown in
Figure 7 in which the hottest point is about 55.309ºC. In 5.3. Using a Fan to Cool the Winding
The film coefficient varies from 50 W/m2·ºC to 120
W/m2·ºC when there is forced convection. For this case
the film coefficient has been changed from 10 W/m2·ºC
to 50 W/m2·ºC. The temperature distribution is shown in
Figure 9.
By using the fan, almost all windings have the same
temperature of 52.508ºC. By comparing this figure with
Figure 4, it can be understood that using the fan leads to
winding temperature reduction from 61.962ºC to
52.508ºC. By this analysis, the appropriate fan can be
chosen.
5.4. Using a Fan to Cool the Core
In this case, it is supposed that core is cooled by a fan
(the film coefficient is 50 W/m2.ºC). The heat distribu-
tion in this condition is plotted in Figure 10. According
Figure 6. Thermal flux in transformer. to this figure, the core ambient temperature is almost

Copyright © 2010 SciRes. JEMAA

 Analysis of Thermal Behavior of High Frequency Transformers Using Finite Element Method 631

Figure 8. Thermal distribution in transformer (Increasing Figure 10. Thermal distribution in transformer (Using fan to
the ambient temperature). cool the core).

Figure 9. Thermal distribution in transformer (Using fan to

cool the winding). Figure 11. Thermal distribution in transformer (Increasing
thermal conductivity).
equal. This figure shows that changing the film coeffi-
cient causes suitable increase in core temperature former, the generation heat is computed then it is totally
(61.95ºC). simulated in ANSYS software. In this simulation, the
thermal analysis and heat distribution were studied. Then,
5.5. Variation of Thermal Conductivity
the effects of parameters such as flux density, winding
On of the other studied parameters is thermal conductivity loss value, the film coefficient have been investigated.
that it is changed from 0.004 W/(m·K) to 0.008 W/(m.K). According to obtained results, in case of using a fan to
Thermal distribution for this mode is shown in Figure cool the winding, the maximum temperature of trans-
11. former has been decreased. Furthermore, it was shown
This figure shows that the thermal conductivity has a that by decreasing the flux density, the temperature of
little effect (61.95ºC) on the temperature but the ther- the hottest point in transformer has been decreased which
maldistribution around the transformer is uniform as a is due to the power loss reduction in windings. Moreover,
circle from centre of transformer. it is explained in the paper that the core temperature can
be controlled by increasing the film coefficient (by using
6. Conclusions
a fan). As a result, with the findings of this analysis, an
In this paper, the thermal behavior of different cores has appropriate fan can be chosen in order to cool the whole
been studied. For this purpose, first for a typical trans- transformer. In the paper, it is also elaborated that the

Copyright © 2010 SciRes. JEMAA

632 Analysis of Thermal Behavior of High Frequency Transformers Using Finite Element Method

core corners have the lower temperature in comparison [8] J. P. Vandelac and P. D. Ziogas, “A Novel Approach for
with other parts. This part has the maximum thermal flux Minimizing High Frequency Transformer Copper
Losses,” IEEE Transaction on Power Electronics, Vol. 3,
as well.
No. 2, 1988, pp. 266-277.
[9] B. Carsten. “High Frequency Conductor Losses in Switched
REFERENCES Mode Magnetic,” in Proc. HFPC Con$ Ventura, CA:
[1] S. N. Talukdar and J. R. Bailey, “Hysteresis Models for Intertec Communications Inc., May 1986.
System Studies,” IEEE Transactions on Power Apparatus [10] Power Conversion and Line Filter Applications.
and Systems, Vol. 95, No. 4, 1976, pp. 1429-1434.
[11] Design of low profile high frequency transformers-A new
[2] C. C. Wong, “A Dynamic Hysteresis Model,” IEEE tool in SMPS design-. Philips Magnetic Products. Appli-
Transactions on Magnetics, Vol. 24, No. 2, 1988, pp. cation note. May.1990.
1966-1968.
[12] Kool Mu Powder Cores Data Book. Magnetics.
[3] D. R. Bennion, H. D. Crane and D. Nitzan, “Digital
[13] L. M. Escribano, R. Prieto, J. A. Cobos and J. Uceda,
Magnetic Logic,” McGraw-Hill, New York, 1969.
Thermal Modeling of Magnetic Components: A Survey,”
[4] F. Presiach, “On the Magnetic Aftereffect,” Zeitschrift für Proceeding of 28th Annual Industrial Electronics Society
Physik, Vol. 94, 1935, pp. 227-302. Conference of the IEEE, Sevilla, 5-8 November 2002, pp.
[5] R. D. Vecchio, “An Efficient Procedure for Modeling 1336-1341.
Complex Hysteresis Processes in Ferromagnetic Materi- [14] L. M. Escribano, R. Prieto, J. A. Oliver, J. A. Cobos and J.
als,” IEEE Transactions on Magnetics, Vol. 16, No. 5, Uceda, “Analytical Thermal Model for Magnetic Com-
1980, pp. 809-811. ponents,” Proceedings of 34th Annual Power Electronics
[6] D. L. Atherton, B. Szpunar and J. A. Szpunar, “A New Specialist Conference of the IEEE, 2003, pp. 861-866.
Approach to Presiach Diagrams,” IEEE Transaction on [15] J. C. S. Fagundes, A. J. Batista and P. Viarouge, “Thermal
Magnetics, vol. 23, No. 3, May 1987, pp. 1856-1865. Modeling of Pot Core Magnetic Components Used in
[7] B. Szpunar, D. Atherton and M. Schonbachler, “An Ex- High Frequency Static Converters,” IEEE Transactions on
tended Presiach Model for Hysteresis Processes,” IEEE Magnetic, Vol. 33, No. 2, March 1997, pp. 1710-1713.
Transactions on Magnetics, Vol. 23, No. 5, 1987, pp. [16] R. W. Ericksonand D. Maksimo, “Fundamentals of Power
3199-3201. Electronics,” 2nd Edition, Springer, Berlin, 2002.

Nomenclature
ρ Wire effective resistivity (Ω-cm)
Itot Total rms winding current, ref to primary (A)
n2/n1, n3/n1, etc Desired turns ratios (V-sec)
λ1 Applied primary volt-sec
Ptot Allowed total power dissipation (W)
Ku Winding fill factor
β Core loss exponent
Kfe Core loss coefficient (W/cm3Tβ)
Ac Core cross-sectional area (cm2)
WA Core window area (cm2)
MLT Mean length per turn (cm)
le Magnetic path length (cm)
Aw1, … Wire areas (cm2)
ΔB Peak ac flux density (T)

Copyright © 2010 SciRes. JEMAA