Optimal Design of Thermal Management System in A LED Package

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

23rd International Symposium on Transport Phenomena

Auckland, New Zealand


1922 November 2012

Optimal Design of Thermal Management System in a LED Package

M.W. Jeong1, Y. Jang2, J. Heo2, H. Kang2 and Y. Kim2


1
Graduate school of Mechanical Engineering
Korea University, Seoul 136-713, Korea
2
Department of Mechanical Engineering
Korea University, Seoul 136-713, Korea

Abstract various fields as a light source. Heat dissipation is being more


and more important in LEDs because of increasing light output
LEDs (Light emitting diodes) are a powerful light source because and integration. Generally, it is known that only about 15 to 20%
of high energy efficiency, long life, and eco-friendliness. Most of of input power is converted into light, and the remaining 80 to
the electricity in a LED becomes heat rather than light. If this 85% is converted into heat. Life time of a LED decreases from
heat is not removed, the LEDs run at a high temperature, which 40,000 to 20,000 hours when junction temperature increases from
not only lowers their efficiency, but also makes the LED less 40 to 50oC [3]. Also, heat generated from LEDs can cause a
reliable. Therefore, the thermal characteristic of the LED package critical failure due to difference of thermal expansion coefficient
should be carefully considered in the design process because the in the materials and degradation of light output by changing
heat generated from junction degrades reliability and efficiency colors in package and lens [5]. Park et al. (2004) studied about
of the LEDs. In this study, the effect of heat spreading was the effect of junction temperature on the performance of a LED.
studied with the variation of geometrical design parameter using They reported that the LED efficiency decreased by 70% at the
a commercial CFD (computational fluid dynamics) program. An junction temperature of 85oC [2]. Therefore, thermal
effect of the thermal modeling of the LED model was evaluated characteristics of LEDs should be carefully considered in the
with the variation of the thickness of a board. In addition, thermal design process for efficiency and reliability. In addition, thermal
characteristics were studied for the composition of a material spreading is an important factor to be considered in thermal
layer of a board. In thermal design, the variation of thermal design [1]. Thermal resistance is composed of Rz (z-direction
performance with the variation of the thickness had different resistance; vertical direction) and Rs (spreading resistance;
tendencies between in thermal modeling with TIM and without horizontal direction). The was the lowest thermal resistance at the
TIM. The reasonable tendency was observed in the thermal thickness of 0.6 mm due to the effect of heat spreading, when the
model with TIM because the effect of heat spreading was thickness of Cu substrate is varied in a multi-chip LED [4].
included. The effect of heat spreading was varied by the In this study, the heat spreading characteristics of LEDs were
composition of layers in the board. studied with the variation of geometrical design parameters. The
effect of the thermal modeling in LEDs was evaluated with the
Nomenclature variation of the thickness of a board. In addition, thermal
k thermal conductivity [W/mK] characteristics were studied for the composition of layers in a
Q heat transfer rate [W] board. The thermal performance of a LED package model was
R thermal resistance [K/W] analyzed using a commercial CFD program.
T temperature [oC]
t thickness [-] Thermal Modeling
angle of heat spreading through material [o] Figure 1 shows the model structure in analysis. The working fluid
is air. The ambient air temperature is 25oC and the air is assumed
Subscripts to be incompressible ideal gas. LED chips are on the board in air
a air surrounding. The size of air surrounding is 232310 mm3 to
avg average consider the experimental condition. The bottom of surrounding
b board is at isothermal condition whose temperature is 25oC.
D.A. die attach Gravitational acceleration is -9.81 m/s2 in the z-direction
H high considering natural convection.
j junction
L low
s spreading
z z-direction

Introduction
LEDs draw strong attention as a next generation light source.
LEDs have a tenth energy consumption and 50 times life time
comparing with an incandescent lamp, and 10 times life time
comparing with a fluorescent lamp [3, 5]. Therefore, LEDs have
a powerful light source because of high energy efficiency, long
life, and eco-friendliness. Recently, LEDs have been used in Figure 1. Schematic of thermal modeling.
Thermal conductivity
Material
(W/m K)
Chip 46
Die bonding (AuSn) 57
Dielectric layer 2
Board (Aluminum) 140
TIM
0.6
(Thermal interface material)
Table 1. The thermal conductivity of materials in LED package
Figure 3. Schematic of the layer structure in the board: (a) Base
The temperature of the LED model was solved by a CFD tool model of the layer structure, (b) Two types in composition of the
with the FVM (finite volume method). The used CFD tool was layer.
the commercial CFD program, Ansys Icepak 13.0. Mesh quality
was satisfied to be over 0.5. The number of elements was about Heat spreading is not a trivial issue on thermal problem of a LED
1,945,000. The time variation was steady state. The convergence including micro scale heat transfer [1]. In this study, we
during solving was investigated to consider the residuals and to attempted to investigate the effects of the modeling method when
monitor the maximum chip temperature. analyzing thermal characteristics of the LED packages with
various geometries.
Model to Change the Board Thickness
Two cases of the LED model compared: one is the thermal model
Figure 2 shows a schematic of the LED model. The LED package without TIM and the other is the thermal model with TIM. The
is a COB (chip on board) type. There are 4 chips on the board. uniform temperature of 25oC was fixed on the bottom of the
The size of the chip is 1.01.00.1 mm3. The heat generation is board in the thermal model without TIM and on the bottom of
1.68 W per a chip. Chips are attached on the board by die TIM in the thermal model with TIM. Figure 4 compares the
attaching (AuSn). The thickness of the die attaching is 0.02 mm. junction temperatures in the thermal model without TIM with
The board consists of a dielectric layer and aluminum base plate. that in the thermal model with TIM according to the increase in
The dielectric layer is anodizing oxide layer. The size of the the thickness of the board from 0.21 to 2.10 mm. In the thermal
dielectric layer is 16.0017.000.01 mm3. The thickness of TIM model without TIM, the junction temperature increased from
(thermal interface material) is 0.23 mm. Table 1 shows the 36.5 to 41.8oC with the increase in the board thickness. In the
thermal conductivities of materials used in the LED package. thermal model with TIM, the junction temperature decreased
Model to Change the Structure of the Board from 87.8 to 59.0oC with the increase in the board thickness. The
junction temperature in the thermal model with TIM was 51.3 to
Figure 3 shows a schematic of the layer structure in the board. As 17.2oC higher than that in the thermal model without TIM at the
shown in Fig. 3(b), two types are considered in this study. The thickness of the board from 0.21 to 2.10 mm. The junction
first case is named as kH-kL, which consists of chip-kH and -kL temperature increased in the thermal model without TIM and
layers in order. The second case is named as kL-kH, which is decreased in the thermal model with TIM according to the
composed of chip-kL and -kH layers in order. Thermal thickness of the board. This is due to the effect of heat spreading.
conductivities of kL and kH layers are 20.24 W/mK and 202.40
W/mK, respectively. The heat generation from the chip is 3 W. Figures 5 and 6 show temperature distributions of the thermal
The thicknesses of kL and kH layers are 0.2 mm and 1 mm, model without TIM and with TIM, respectively. As shown in
respectively. Figure 6, the high temperature region was expanded from the
center to side on the surface of the board with the increase in the
TIM thickness because of the decrease in the junction
Result and Discussion
temperature. However, as shown in Figure 5 for the thermal
Effects of Modeling Method and Thermal Boundary Condition model without TIM, the spreading of the high temperature region
restricted due to the increase in the thermal resistance with the
The boundary condition and modeling method are important for thickness of the board. The effect of heat spreading did not
the numerical analysis using CFD. appear in uniform temperature condition on the bottom of the
board when varying the thickness of the board. The high
temperature region became larger on the board surface with the

90 Junction temperature
Without TIM
80 WithTIM
Temperature (oC)

70

60

50 Tambient : 25oC
40

30
0.0 0.5 1.0 1.5 2.0 2.5
Base(Al plate) thickness (mm)

Figure 2. Schematic of the LEDs: (a) Exterior of the LED Figure 4. Comparison of junction temperature according to the
package (b) Structure of the LED package. thickness of the base.
8
Rjb 90
7 85

Thermal resistance (oC/W)


Tj
6 80

Temperature (oC)
Tambient : 25oC 75
5 70
4 65
60
3
55
2 50
0 1 2 3 4 5 6
Figure 5. Variation of temperature distribution with the variation Base plate thickness (mm)
of the thickness of the base in the thermal model without TIM.
Figure 7. Variation of thermal performance with the increase in
the thickness of the base in the board.

Rb Rz Rs
= + (3)
dt dt dt

Where Rb is the thermal resistance of the board, Rz is the vertical


thermal resistance, Rs is the horizontal thermal resistance
(spreading resistance). Rz is l/kA and A is vertical area to
direction of heat transfer. Rs decreased, but Rz increased with the
increase in the thickness from 0.21 to 2.10 mm. In addition, Rb
decreased with the increase in the thickness from 0.21 to 2.10
Figure 6. Variation of temperature distribution with the variation mm. After 2.10 mm, the increase in Rz was consistent with the
of the thickness of the base in the thermal model with TIM. decreasing in Rs.
was increase in the board thickness from 0.21 to 2.10 mm in the Effect of Layer Composition in Board
thermal model with TIM. It was consistent with the result that
heat spreading increased with the increase in the thickness of a Figure 8 compares junction temperatures by combination of the
substrate in a specific thickness range [1,4]. Therefore, it is layers. The board was composed of two layers The junction
required to consider the thermal model with TIM when temperatures were 74.5oC in the kH-kL and 101.7oC in the kL-kH.
investigating the thermal performance of LEDs with the variation The kH-kL showed 27.3oC higher junction temperature than the
of the board thickness. kL-kH. The area of sink substrate was larger than that of heat
source. Therefore, heat spreading occurred when heat moved
Thermal Characteristics with a Thickness of Al Plate from the heat source to the heat sink substrate. The effect of heat
Figure 7 shows the variation of thermal characteristics of a LED spreading can investigate from the spreading angle. The
package with the variation of base (Al plate) thickness. The spreading angle was calculated by equation (4) [6].
thermal resistance (Rjb) is calculated by equation (1).
k Layer 1
T j Tb ,avg = tan 1 (4)
k Layer
R jb = (1) 2
Q
The angles of heat spreading were 84.29o in kH-kL and 5.71o in
kL-kH. The wider angle of heat spreading indicates the better heat
The junction temperature was from 89.8 to 58.3oC at the
spreading.
thickness of Al base plate from 0.21 to 5.25 mm. The minimum
junction temperature was 58.2oC at the thickness of 4.20 mm The
thermal resistance from junction to board was from 7.32 to 2.71
K/W and the minimum value was 2.70 K/W at the thickness of 150
4.20 mm. However, the difference values of thermal performance kH-kL Tambient : 25oC
were very small after 2.10 mm. Therefore, the thermal kL-kH
performance became almost constant after 2.10 mm within the
Ttemperature (oC)

thickness from 0.21 mm to 5.25 mm. The thermal resistance at 100


the thickness of 2.10 mm decreased by 62.2% comparing with
that at 0.21 mm. The thermal resistance from the junction to
board can be expressed by equation (2).
50
R jb = Rchip + RD. A. + Rd + Rb (2)

In addition, the variation of the thermal resistance of the board 0


can be expressed as equation (3) by Yang et al. [4].
Figure 8. Comparison of junction temperature according to layer
composition in the board.
L2 Conclusions
In thermal design, the variation of thermal performance with the
L1
variation of the thickness had the different tendency between in
the thermal modeling with TIM and without TIM. This is
chip important in thermal design of LEDs. The thermal model with
Layer1 TIM showed more reasonable tendency because the effect of heat
Layer2 spreading was considered. In the thickness of Al base from 0.21
to 5.25 mm, the thermal performance decreased before the
thickness of 2.10 mm and remained almost constant after the
thickness of 2.10 mm. The effect of heat spreading was different
by combination of the layers in the board. The angle of spreading
Figure 9. Cross-section of layout for thermal spreading in a board. and average cross-sectional areas of layer1 became larger when
the layer of higher thermal conductivity was close to the heat
Figure 9 shows concepts of length of heat spreading, average source.
length of heat spreading, average cross-sectional area for the
effect of heat spreading. The length of heat spreading can be Acknowledgments
calculated by equation (5). The average length of heat spreading
can be calculated by equation (6). The average cross-sectional This research was supported by the Korean Evaluation Institute
area by heat spreading effect can be calculated by equation (7) of Industrial Technology (Grant No. 10035530).
[6].
References
L2 = 2t Layer1 tan ( ) + L1 (5)
[1] Lasance, C.J.M., How to Estimate Heat Spreading Effects in
L L Practice, Journal of Electronic Packaging, 132, 2010,
LLayer1,avg = 2 1 + L1 (6) 0310041.
2 [2] Park, J., Shin, M.H., Lee, C.C., Measurement of
Temperature Profiles on Visible Light-Emitting Diodes by
ALayer1 , avg = L2Layer1 , avg (7) Use of a Nematic Liquid Crystal and an Infrared Laser,
Optics letters, 29(22), 2004, 26562658.
The average lengths of heat spreading were 11.00 mm in kH-kL [3] Park, J.W., Rapid Thermal Diffusion Devices Development
and 1.02 mm in kL-kH. The average cross-sectional areas of layer1 of the High Power LED Lighting system, Proceedings of
by heat spreading effect were 121.02 mm2 in kH-kL and 1.04 mm2 Korea Society of Precision Engineering Autumn Conference,
in kL-kH. The kH-kL would have the larger average cross-sectional 2010, 593-594.
areas than the kL-kH because kH has higher thermal conductivity [4] Yang, C., Liu, W., Liu, C., Measurement of Thermal
than kL so that better heat spreading occurs in the kH. Therefore, Resistance of First-level Cu Substrate used in High-power
cross-sectional area of layer1 is dominant for heat transfer, Multi-chips LED Package, Microelectronics Reliability, 52,
because the cross-sectional area increases continuously through 2012, 855860.
all layers. Under the limitation of design, to set the higher [5] Ok, C.W., A Study on Optimization of the Heat Sink for
thermal conductivity material close to heat source in PCB yielded Reliability Improvement of High Power Light Emitting
better thermal design. Diode, Chonbuk National University, Thesis for the Degree
of Master, 2006.
[6] Ulrich, R.K. Brown, W.D., Advanced Electronic Packaging,
IEEE Press, 2006.

You might also like