Microsoft Word - 2-faris-53.doc
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
17
EXPERIMENTAL AND NUMERICAL ANALYSIS OF
ELECTRONICS HEAT SINK
AHMAD F. ISMAIL, MIRGHANI I. AHMED, AND YOUSIF A. ABAKR
Department of Mechanical Engineering, Faculty of Engineering, IIUM, 53100, Kuala Lumpur,
Malaysia.
e-mail: faris@iiu.edu.my
Abstract: Cooling of electronic components continues to attract many research and
development activities towards achieving an effective way of cooling. Computational
fluid dynamics (CFD) tools may be considered as a cheap substitute for expensive
experimental testing methods. In this work the cooling of a simulated electronic board
was modeled using FLUENTTM CFD software, and experimental procedures were
followed to validate the estimated results, and to understand the factors that would
affect the software capability to predict the actual measured values. Results showed
good agreement between the simulation and experimental results. The software was
found to be capable to predict the exact values at the locations where the temperature
values were similar to the board mean temperature. The maximum percentage error
was found to be limited to 4.7%, and the capability of the software to estimate the
exact measured values was found to be affected by the function of thermal wake
generation.
Keywords: CFD, Electronic cooling, Heat sink, Simulation.
1. INTRODUCTION
The progressive decrease of electronics device sizes and the higher processing rates
resulted in an increasing rate of heat generation per unit surface area, which means
increased cooling requirements. The possibility of electronic equipment failure
increases with the increase of temperature. High thermal stresses resulting from
temperature variations in the solder joints of electronic components mounted on circuit
boards are major causes of failure. Figure 1 shows the increasing rate of failure as the
junction temperature increases. Therefore, cooling has become increasingly important in
the design and operation of electronic equipment. Air-cooled forced convection cooling
is the most used technique. Fan speed control offers numerous benefits including
reduced noise, low energy consumption and extended operating life of the fan.
The need of cooling of electronic components was known since Thomas Edison
discovered the vacuum diode in 1883. The new era of electronics started in 1958 by the
invention of the silicon integrated circuit (IC). Integrated circuits contain several
Author for correspondence
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
18
components such as diodes, transistors, resistors, and capacitors in a single chip. The
number of components per chip has been increasing steadily. Electronic component
suppliers estimate that for every 10 C rise of junction temperature, the device failure
rate doubles [1]. When a device exceeds the maximum temperature, the semiconductor
performance, life and reliability are tremendously reduced. Therefore, enhancing the
heat transfer rate from electronic devices became a very important research issue.
Fig. 1: The relation between the junction temperature and the failure rate.[1]
Modify solution parameters or
grid
No Yes
No
Set the solution parameters
Initialize the solution
Enable the solution monitors of interest
Calculate a solution
Check for convergence
Check for accuracy
Stop
Yes
Fig. 2: Flow chart for solution procedure.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
19
Studies of the cooling of electronic devices are based on the fluid flow around obstacles
mounted on a wall base, on which the phenomenon of mixed convection is observed
around the localized heat sources. Many numerical and experimental studies were
conducted previously [2].
Heat transfer and flow characteristics of four heat source locations were studied by
Kennedy and Zebib [3]. Studies by Kang et al. [4] found that heat transfer can be
improved significantly by buoyancy-driven secondary flow. But the effect of thermal
wake generation was not fully understood because of use of a single heat source.
Numerical study of the unsteady and transitional characteristics of mixed convection of
airflow in a channel was conducted by Huang and Lin [5]. Kerki et al. [6] showed the
significant effect of the thermal boundary flow pattern on the secondary flow pattern
and on the heat flow distribution over the surface. Effects of the varying duct
dimensions and obstruction dimensions on the convection from the surface of the heat
generating obstructions to the airflow were studied experimentally by Leung et al. [7].
Choi and Kim [8] proposed a “modified 5% deviation rule” to define the natural, mixed
and forced convection regimes. Chin et al. [9] investigated experimentally the optimum
spacing problem in mixed convection. Jubran and Al-Salaymeh [10] investigated
experimentally the effects of using ribs to enhance the convective heat transfer.
In most of the previous work the thermal wake generation, and the actual effect of
buoyancy driven secondary flow has not been studied thoroughly. The impact of the
transitional characteristics of mixed convection on the accuracy of the numerically
predicted results has not been investigated.
This paper will focus on the sensitivity of the CFD obtained results to the local
conditions and try to define the parameters that will result in accurate estimations that
approach the experimental results very closely.
2. NUMERICAL MODELLING
The steady state heat convection of the air moving from the electronic components is
found by considering the heat generated by the chips, the heat conducted through the
board and the heat radiated to the surroundings. Equation (1) shows the heat balance for
a steady state condition,
c t r lQ Q Q Q= - - (1)
Both the total amount of heat to be dissipated and the density of the air must be known.
According to the first Law of Thermodynamics (Conservation of Energy) for a steady
state and steady flow process, the total amount of heat dissipated in a system is
determined as follows:
tQ H K E PEº D + D + D (2)
FLUENTTM CFD software is based on the finite volume method. In this technique, the
equations of motion are treated in balance form for finite sized control volumes (CV).
Formal balance of a property transported with the fluid can be expressed by
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
20
dVSdddV
t AAV
AAV
Unsteady Convection Diffusion Generation
(3)
A 2D model was first constructed using GAMBIT software. The geometry was modeled
with an appropriate number of nodes assigned to each line. The nodes for the modules
and the board were set to 0.5 since more computational effort is needed in this region to
gain good results. Finally the whole model was meshed using triangular element type. A
total of 18983 nodes and 36467 elements were used to build up this model. The
appropriate boundary conditions were set to the model. Fig. 3 shows the 2D model
before and after meshing.
(a) Model before mesh.
(b) Model after mesh.
Fig. 3: The 2-D Modeling and mesh generation.
Due to the large number of nodes for modeling in 3D space, only sub-region of the total
test rig was modeled. The modules, the board and the zone above them were chosen for
the purpose of modeling as shown in Fig. 4. A total of 43079 nodes and 185215
tetrahedral elements were used to build up this model.
Due to low air velocities involved, this problem has been considered as an
incompressible flow. The segregated solver was selected for both cases. The maximum
Reynolds number based on the modules length is 1418, which mean laminar flow. The
option to solve the energy equation was enabled because there is heat-generation. The
property of the air was set to the default values. Properties of aluminum block and the
hard board were set according to the values used for heat transfer analysis. Using a
computer of Pentium II processor with a speed of 266 MHz, The program takes about
two hours to finish the calculations. Graphical residual plotting is shown in Fig. 5,
which shows the progress of the solution to convergence.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
21
Fig. 5: Residual plotting for convergence.
3. EXPERIMENTAL SETUP
The experimental apparatus is made of a wind tunnel of a rectangular section with a
dummy electronic components mounted on a circuit board. Electric heating elements
were imbedded inside the dummy electronic components, and sets of thermocouples
were used to read the surface temperature of the components. The setup arrangement is
shown in Fig. 6. Isometric view of the test rig is shown in Fig. 7.
The setup consists of the test rig, DC power supply for electronic modules, brushless
blower and a number of digital thermocouples. The design and the coordination
assignments for the modules were constructed as in Fig. 7. The test rig shown in Fig. 8
is divided into eight zones of airflow impedance. The zones are numbered sequentially,
following the direction of airflow.
3 Digital thermocouples
Fan D
C
H
eater D
C
Supply
Fig. 6: Experimental setup.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
22
Abrupt contraction
outlet
Flow straightener unit
Simulated circuit board
Simulated electronics modules
Y
Z
X
Gradual
contraction
inlet
Fig. 7: Isometric view of the test rig and its components.
1 3 2 4
5
6
7 8
Fig. 8: Side view of the test rig.
Temperature measurements were taken only for the modules of row one, three and five.
Temperatures of the modules in row two and four were approximated from the
corresponding column. This procedure will be followed in the heat transfer analysis
part. The heat transfer coefficient for each module was calculated from,
( )s ac ref
q
h
A T T
=
-
(4)
The temperature of the modules closer to the air exit is affected by the thermal wake
phenomenon. The thermal wake happens when heat-generating components are placed
along the flow stream direction. This function is defined as,
i ref
i
ac ref
T T
T T
q
-
=
-
(5)
The volumetric maximum flow rate is taken as 11.8 10-3 m3/s. This is the maximum
flow rate of the blower that will be used in the test rig. Each zone in the test rig will be
analyzed individually for cross-section area changes and head loss. Details of zone 5,
which contains the simulated board, are shown in Fig. 9.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
23
Fig. 9: Zone 5 (Isometric view).
4. EXPERIMENTAL PROCEDURE
Natural convection heat transfer measurements were taken while the fan power supply
is kept off. After turning on the power supply of the heater, enough time was allowed
for the system to reach steady state condition. When the steady state condition was
achieved (modules temperature does not change with time), the local temperatures of
the modules were then recorded. Similar procedure was then followed for the forced
convection heat transfer measurements. Then, the fan power supply is turned on and set
at 4 volts. At steady state condition the temperature measurements of the two modules
were recorded.
Similar procedures were followed to repeat the test for the different fan speeds.
Thermocouples were placed at different modules and test procedures were repeated
again. A vane anemometer was placed at the exit of the blower to measure the air
velocity for every set of applied fan speed. The system usually takes about 1 hour and
30 minutes to reach the steady state.
5. RESULTS
From the measured data it was noticed that the modules in the middle column have a
higher temperature relative to the other two columns. Also the hottest module was the
module at the middle of the board. This is mainly because of heat accumulated at the
centre of the circuit board.
In Fig. 10 the average module surface temperatures were plotted at different positions
along the board longitudinal axis. From this curve we can notice that the surface
temperature is maximum at the middle of the board for the free convection condition.
The position of the maximum surface temperature tends to change towards the direction
of the exit as the air velocity increases.
Figure 11 shows the results of the variation of the modules surface temperature as the
air speed is increased gradually. The curve is plotted for three different axial locations.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
24
At the free convection condition the modules temperature reaches values more than
80˚C, and when he air is forced to move at high velocities the temperature drops to less
than 60˚C for the first row.
Free Convection
Re = 495
Re = 1006
Re = 762
Re = 1121
Re = 1418
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
0 10 20 30 40 50 60
Position along test rig axis (Cm)
S
ur
fa
ce
T
em
pe
ra
tu
re
(o
C
)
Fig. 10: Average modules surface temperature.
Fifth Row
Third Row
First Row
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
80.00
90.00
0 1 2 3 4 5 6 7 8
Air Velocity( m/s)
B
ot
to
m
S
ur
fa
ce
T
em
pe
ra
tu
re
(
oC
)
Fig. 11: Variation of the modules surface temperatures with air velocity.
Figure 12 shows the variation of local rate of heat removal by convection at different air
velocities. From this figure we can see that as the air velocity exceeds a value of 4 m/s
the rate of heat removed becomes almost uniform all through the circuit board, while for
low velocities it increases from the inlet position towards the exit.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
25
First Row
Second Row
Third Row
Fourth Row
Fifth Row
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
0 1 2 3 4 5 6 7 8
Air Velocity (m/s)
Lo
ca
l r
at
e
of
H
ea
r
co
nv
ec
tio
n
(W
at
t)
Fig. 12: Variation of local rate of heat convection with air velocity.
The percentage of the amount of heat removed by convection to the total amount of heat
generated at the chip is shown on Fig. 13. It can be noticed from this figure that at
different axial locations the maximum value is around 90% while the minimum is
slightly below 60%, which occurs at natural convection conditions.
Re = 1418
Re = 1121
Re = 1006
Re = 762
Re = 495
Free Convection
1.00
1.10
1.20
1.30
1.40
1.50
1.60
1.70
1.80
1.90
0 10 20 30 40 50 60
Position (Cm)
R
at
e
of
H
ea
t
C
on
ve
ct
ed
(
W
at
t)
Fig. 13: Percentage of heat convected at different axial position.
The overall thermal wake generation correlation for different Reynolds number (ReL) is
shown in Fig. 14. The thermal wake generation was found to be increasing at high
constant rate up to the middle of the board, then the rate decreases and remains constant
up to the end of the board regardless of the value of the Reynolds number.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
26
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
1 2 3 4 5
No. of row
(
di
m
en
si
on
le
ss
)
Re =495
Re =762
Re =1006
Re = 1121
Re =1418
Fig. 14: Thermal wake generation function.
0
10
20
30
40
50
60
450 650 850 1050 1250 1450
Re
L
N
u L
row 1
row 2
row 3
row 4
row 5
Fig. 15: Forced convection correlation (local).
It can generally be stated that the thermal wake generation function increases at the
downstream region. Comparing the thermal wake generation correlation for each case,
we can notice that, the thermal wake effect is high for high values of Reynolds number
and vice versa.
According to the dimensionless correlated data shown in Fig. 15, the rate of cooling is
highest near the air inlet position and decreases as it approaches the outlet. Modules in
the first row experiences better cooling and as the row number increases the cooling
efficiency reduces. The reduction of the cooling rate is caused partially by the thermal
wake generation.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
27
6. DISCUSSION
The analysis and evaluation of the FLUENTTM CFD simulation results revealed a very
good agreement with the obtained experimental results. Comparison between the
estimated and the actual modules surface temperatures is shown in Fig. 16 and Fig. 17
on which the estimated and the measured values are found to be very close.
The error in the estimated values was found to be related to the values of the Reynolds
numbers and the axial position on the board. Lower values of Reynolds number were
found to coincide with better estimates at locations near to the inlet. Larger values of
Reynolds number were accompanied with better estimates of surface temperature at
downstream regions. Figure 18 shows the variation of the percentage error values as
Reynolds number increases.
313.00
323.00
333.00
343.00
353.00
363.00
373.00
10 15 20 25 30 35 40 45 50
AXIAL DISTANCE (Cm)
TE
M
P
E
R
A
TU
R
E
(
K
)
EXPERIMENTAL VALUES
Fig. 16: Comparison between the CFD and the experimental results at Re = 1006.
313.00
323.00
333.00
343.00
353.00
363.00
373.00
10 15 20 25 30 35 40 45 50
AXIAL DISTANCE(Cm)
T
E
M
P
E
R
A
T
U
R
E
(K
)
EXPERIMENTAL VALUES
Fig. 17: Comparison between the CFD and the experimental results at Re = 1211.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
28
ROW 5
ROW 4
ROW 3
ROW 2
ROW 1
-5.00
-4.00
-3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
400 600 800 1000 1200 1400 1600
REYNOLDS NUMBER
E
R
R
O
R
%
Fig. 18: Variation of the percentage error with Reynolds number
One of the interesting observations is the mode of change of the position at which the
software estimated exactly the measured value of surface temperature. This position was
noticed to change axially towards the downstream direction as Reynolds numbers
increases. This phenomenon is illustrated in Fig. 19.
In addition to this fact, the modules surface temperatures that were estimated exactly by
the software as the measured experimental values were found to be approximately equal
to the average board temperature. Figure 20 shows the values of temperature at different
positions when the CFD software succeeded to estimate exactly the measured values.
0
200
400
600
800
1000
1200
1400
1600
1800
0 10 20 30 40 50 60
Axial Position (Cm)
R
ey
no
ld
s
nu
m
be
r
Fig. 19: Reynolds number at axial positions when the CFD software estimated the exact
values of temperature.
When the comparison was made taking Reynolds number variation at fixed positions,
the standard deviation was found to be increasing as Reynolds number increases to a
maximum value of 800, then it start to decrease further. This is illustrated in Fig. 22.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
29
313
318
323
328
333
338
343
348
353
358
10 15 20 25 30 35 40 45 50
Axial position(Cm)
Te
m
pe
ra
tu
re
(
K
)
Fig. 20: Values of temperature at axial position at which CFD and experimental results
were equal.
1.55
1.6
1.65
1.7
1.75
1.8
1.85
10 15 20 25 30 35 40 45 50
AXIAL DISTANCE (Cm)
E
R
R
O
R
S
TA
N
D
A
R
D
D
E
V
IA
TI
O
N
Fig. 21: Standard deviation of the error at different axial locations.
0.00
0.50
1.00
1.50
2.00
2.50
3.00
400 600 800 1000 1200 1400 1600
Reynolds Number
E
rr
or
S
ta
nd
ar
d
D
ev
ia
tio
n
Fig. 22: Standard deviation of error at different Reynolds number values.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
30
7. CONCLUSION
This work shows a combination of CFD and experimental investigation of a heat-
generating matrix that simulates the behavior of an electronic circuit board. Initially the
board was simulated using FLUENTTM CFD software, then the simulation results were
validated experimentally.
The estimated values of the surface temperature were found to be in good agreement
with the actual measured values. The maximum percentage error was found to be 4.9 %,
with a maximum standard deviation of 2.5 when considering different Reynolds
number, for the entire plate, and 1.7 when considering the local error variations at same
locations.
Values of Reynolds number, at which the software estimated exactly the measured
values, were found to be increasing following the trend of thermal wake generation.
This phenomenon may be related partially to the instability of the thermal boundary
layer when the thermal wake exceeds a certain value with respect to different Reynolds
numbers.
The local surface temperatures, at which the software was able to predict exactly the
measured values, were found to be always the same values. And they were generally
noticed to be very close to the mean temperature of the board.
ACKNOWLEDGEMENT
The authors would like to thank IIUM for funding this project. The authors would like
also to thank M. Farid M. Ismail and Hilmi F. Bachok for their assistance in fabricating
the experimental setup and running the CFD simulation.
REFERENCES
[1] R. Remsburg, “Advanced Thermal Design of Electronic Equipment”, International
Thomson Publishing, 1998.
[2] F. P. Incropera, “Convection heat transfer in electronic equipment cooling”, Journal of
Heat Transfer, Vol. 110, pp. 1097-1111, 1988.
[3] K. J. Kennedy, A. Zebib, “combined free and forced convection between horizontal
parallel plates: some case studies”, International Journal of Heat and Mass Transfer, Vol.
26, pp. 471-474, 1990.
[4] B. H. Kang, Y. Jaluria, S. S. Tewari, “Mixed convection air cooling of an isolated
rectangular heat source module on a horizontal plate”, ASME Proceedings of Nat, Heat
Transfer Conference, pp. 59-66,.
[5] C. C. Huang, T. F. Lin, “Buoyancy induced flow transition mixed convective flow of air
through a bottom heated horizontal rectangular duct”. International Journal of Heat and
Mass Transfer, Vol. 37, pp. 1235-1255, 1994.
[6] K. C. Kerki, P. S. Sathyamurthy, S. V. Patankar, “Laminar mixed convection in a
horizontal semicircular duct with axial nonuniform thermal boundary condition on the flat
wall”, Numerical Heat Transfer: Part A, Vol. 25, pp.171-189, 1994.
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
31
[7] C. W. Leung, H. J. Kang, “Convective heat from simulated air-cooled printed-circuit board
assembly on horizontal or vertical orientation”. International Comm. Heat and Mass
Transfer, Vol. 25, pp. 67-80, 1998.
[8] C. Y. Choi, S. J. Kim, “Conjugate mixed convection in a channel: modified five- percent
rule”, International Journal of Heat and Mass Transfer, Vol. 39, pp. 1223-1234, 1996.
[9] S. Chin, Y. Liu, S. F. Chan, Leung CW, T. L. Chan, “Experimental study of optimum
spacing problem in the cooling of simulated electronic package”, Journal of Heat and Mass
Transfer, Vol. 37, pp. 251-257, 2001.
[10] B. A. Jubran, A. S. Al-Salaymeh, “Heat transfer enhancement in electronic modules using
ribs and (film-cooling-like) techniques”, Journal of Heat and Fluid Flow, Vol. 17, pp. 148-
154, 1996.
NOMENCLATURE
A Area (m2)
H Coefficient of heat convection (kW/m2K)
S Heat generation term (kJ)
Tac Modules actual temperature (K)
Ti Modules initial temperature (K)
Tref Reference temperature (K)
q Convective heat flow rate (kW)
Qc Amount of heat convected by air (kJ)
Ql Amount of heat loss by conduction (kJ)
Qr Amount of heat loss by radiation (kJ)
Qt Total heat generated by the chip (kJ)
V Volume (m3)
H Change of total enthalpy (kJ)
K.E Change of kinetic energy (kJ)
P.E Change of potential energy (kJ)
Species concentration (kg/kg)
i Thermal wake function
General field parameter
Density (kg/m3)
IIUM Engineering Journal, Vol. 3, No. 2, 2002 A. F. Ismail et al.
32
BIOGRAPHIES
Ahmad F. Ismail was born in Kelantan, Malaysia in 1966. He received his B. Sc.
degree in Chemical Engineering from the University of Houston, Texas in 1989, and a
PhD degree from Rice University, Texas in 1993. He is currently an Associate
Professor at the Department of Mechanical Engineering at the International Islamic
University Malaysia (IIUM). Dr. Ismail's primary research interests are energy and
environmental systems, computational fluid dynamics and heat transfer, combustion and
pyrolysis, modeling and simulation, and digital image processing.
Mirghani I. Ahmed was born in Singa, Sudan in 1956. He received his M Sc. and PhD
degree in Mechanical Engineering from the Budapest Technical University Hungary in
1987. From 1987 to 1989 he worked with the Sudan University of Science and
Technology as Assistant Professor He later joined IBM – Canada as a Research
Associate from 1989 to 1996. He is currently an Associate Professor at the Department
of Mechanical Engineering at the International Islamic University Malaysia (IIUM). Dr.
Ahmed’s primary research interests are renewable energy, thermal comfort problems,
electronic component cooling, modeling and simulation, computational fluid dynamics
applications and heat transfer.
Yousif A. Abakr was born in Khartoum, Sudan in 1962. He received his B. Sc. degree
in Mechanical Engineering and M.Sc. degree from the University of Khartoum in 1987
and 1995, respectively. He joined the Energy Research Council, Sudan as assistant
researcher in 1987. He joined the Sudan University of Science and Technology as a
lecturer in 1991. Currently, he is at the final stages of his PhD study at the International
Islamic University, Malaysia. His primary research interest is the renewable energy
applications, combustion, evaporation and condensation, and computational fluid
dynamics applications.