Influence of Fan Setting Height On The Cooling Performance of A Plate-Fin-Type Heat Sink With Microprocessor

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Heat Transfer—Asian Research, 30 (6), 2001

Influence of Fan Setting Height on the Cooling Performance of a


Plate-Fin-Type Heat Sink with Microprocessor

Keizou Kawamura, Tadakatsu Nakajima, and Hitoshi Matsushima


Mechanical Engineering Research Laboratory, Hitachi, Ltd., Japan

The cooling performance of a plate-fin-type heat sink equipped with a cooling


fan was investigated experimentally. The heat sink was 80 mm long, 43 mm wide, and
24 mm in height (including the 4-mm-thick base). The cooling fan was 40 u 40 u 15
mm and was set to direct the air flow vertically in the downstream half of the heat sink.
We focused on the influence of the height (which varied from 5 to 20 mm) that the fan
was set at, on the heat transfer coefficient of the heat sink. The maximum value of the
heat transfer coefficient was achieved at a setting height of 5 mm. At this height, the
volumetric heat transfer coefficient was 1.8 times as high as that in a parallel flow
under the same fan power. This result indicates that the cooling performance of heat
sinks with a cooling fan can be improved by using this kind of compact structure.
© 2001 Scripta Technica, Heat Trans Asian Res, 30(6): 512–520, 2001

Key words: Forced-air cooling, compact heat sink, microfan, microproces-


sor

1. Introduction

The strong demand for high operation speed of recent computers is driving increases in
packing density of chips and multichip modules. For example, the heat density of the chip will increase
from the present level of 5 to 10 u 104 to 30 to 40 u 104 W/m2. In spite of this trend, from the cost
and maintenance points of view, forced air cooling is still the best means of cooling electronic
equipment. In computers, it is necessary to keep the spacing of CPU boards mounted on the
motherboard as low as possible in order to minimize the delays in exchange of information. It is
therefore also necessary to cool the CPU within a limited space. For example, the height of the heat
sink on the CPU is often required to be less than 25 mm because the spacing of the CPU boards is
often only 2.5 mm. The cooling efficiency must therefore be increased so as to reduce the electric
power consumption and packing volume. Until now, research on increasing the heat sink performance
has been very active [1–3]. Kitajo [4] developed a cooling method using a microfan attached directly
to a heat sink (hereafter termed a “heat sink with fan”). He changed the fan location and measured
the thermal resistance of the heat sink with fan. Also, he showed that putting the fan above the heat
sink achieves a lower thermal resistance (by 10%) than putting it beside the heat sink. However, when
the fan is above the heat sink, it is necessary that the space over the fan be more than 20 mm because
the thermal resistance increases rapidly when it is lower than 20 mm. And, under this space limitation,
cooling performance per unit volume of the side-fan structure is twice that of the above-fan structure.
© 2001 Scripta Technica
512
The above-fan cooling structure for cooling CPUs on stacked boards is therefore strongly restricted
by space limitations.

Accordingly, we have improved the cooling performance of the plate-fin-type heat sink with
fan installed on the heat sink side. Also, the relation between the position of the fan and the cooling;
performance per unit volume was experimentally determined. In addition, from the viewpoint of
saving electric power, the relation between cooling performance per unit volume and fan power
(electric power consumption) was also determined.

A mini-tower-personal computer type was used in the present study (Fig. 1). In this computer,
a plate-fin-type heat sink whose base area is less than that of the CPU board (processor card) is attached
to the CPU package by thermal paste. A main fan is installed inside the body for cooling the internal
parts. The main fan is located near the air outlet and power supply, which are both located in the upper
part of the PC body. An air inlet is located in the lower part of the body. The main fan sucks cooling
air from the air inlet in the lower part of the body, and this air flows across the whole area inside of
the body. After cooling the inside parts, the cooling air enters the power supply part. Then the main
fan discharges it outside of the body.

Users are often required to exchange the CPU board themselves in order to improve the CPU
performance of their personal computer. To deal with the resulting increase in heat dissipation, a heat
sink with a maximum allowable size considering the limited space is used in the present study. Also,
a new cooling structure—using a high-pressure-type microfan in the upper part of one side of the heat
sink—is introduced. In this structure, a combination of the main fan and the microfan forms a kind
of “push–pull” cooling, so they help each other. It is therefore easy to obtain the air flow rate necessary
for cooling the inside parts. However, since we want to focus on the basic performance of our new
type of heat sink with a microfan, the present study does not include the main fan in the PC body.

Fig. 1. Experimental apparatus.

513
Nomenclature

Ab: base area of heat sink (m2)


Ac: cross-sectional area of fin passage (m2)
Af: total area of fins (m2)
As: inside area of fin passage (m2)
C: temperature incline (K/m)
Cp: specific heat (J/kgK)
de: hydraulic diameter between fins (m)
G: fin gap (m)
H: fin height (m)
L: fin length (m)
Lb:
BBB distance from the air inlet (m)
Nu: average Nusselt number (—)
P: fan power (W)
Pr: Prandtl number (—)
Qa: flow rate of microfan (m3/min)
Qh: heat load to heater (W)
Re: Reynolds number (—)
Ta: inlet air temperature (K)
Tb 1–3: fin base temperature (K)
Tf: fin base average temperature (K)
Vf: flow velocity between fins in upstream half (m/s)
yb: half of fin thickness at the base (m)
Db: heat transfer coefficient based on fin base area (W/m2K)
Dbv: heat transfer coefficient per unit volume (W/m3K)
Dt: heat transfer coefficient based on total area of fins (W/m2K)
DBmB: average heat transfer coefficient (W/m2K)
J: density (kg/m3)
'P: static pressure (Pa)
'Tb: temperature difference (K)
Oair: thermal conductivity of air (W/mK)
Ofin: thermal conductivity of fin (W/mK)
n: kinematic viscosity (m2/s)

2. Experimental Apparatus and Procedure

Figure 2 shows an outline of the heat sink structure used in the present study. The heat sink
is made of aluminum with a fin base size of 43 u 80 u 4 mm, a fin thickness of 1 mm, a fin height of
20 mm, and a fin gap of 5 mm. The size of the microfan is 40 u 40 u 15 mm with an external wing
diameter of 38 mm and a motor diameter of 25 mm. To install the fan on the heat sink, the downstream
half of the heat sink was cut as shown in Fig. 2. There is, therefore, a sudden contraction of air flow
at the center of the heat sink. Also, contracted air is discharged from the fan to the upper side. When
fin height H is set at 5 mm, the upper surface of the microfan and the top of the fins in the upstream
half are at the same level. The maximum cutting depth is equivalent to a fin height H of 2.5 mm when
the fan height is lower than the heat sink height in the upstream half. Moreover, the minimum cutting

514
Fig. 2. Schematic of heat sink.

depth is equivalent to a fin height of 20 mm. The upper side of the heat sink in the upstream half and
the edge of the downstream half are closed by aluminum tape in order to prevent recirculation of
cooling air. The heat sink is heated by a copper heater block (size 40 u 40 u 4 mm), to which a film
heater is attached. The heater block is connected to the fin base by thermal paste.

The cooling performance of the heat sink is evaluated by heat transfer coefficients determined
using the following equations:

Dt @ Qh/Af/(Tf – Ta) (1)

Db @ Qh/Ab/(Tf – Ta) (2)

Here, Dt is the heat transfer coefficient based on total area of fins, and Db is the heat transfer coefficient
based on fin base area. Tf is the average temperature at the fin base (11 places), Ta is the temperature
of the inlet air, Qh is the heat load to the heater, and Af is the total area of the fins. The temperatures
of each position shown in Fig. 2 are measured by copper-constantan thermocouples with a diameter
of 0.1 mm. Six kinds of test pieces were used: five with fan setting heights of 2.5, 5, 10, 15, and 20
mm, respectively, and one for parallel flow (for reference).

Regarding the heat sink for parallel flow, there is no cutting and the fan is set beside the heat
sink and a 40-mm-square duct (270 mm in length) is connected to the outlet of the fan.

515
Fig. 3. Performance characteristics of a microfan.

Moreover, the air velocity between fins (Vf) is varied by changing the input power of the fan.
Figure 3 shows the flow rate Qa and static pressure 'P of the microfan. The velocity between fins in
the upstream half, Vf, is calculated from the velocity which is measured by using a velocity probe
just in front of the heat sink. The ratio of cross-sectional areas between fins and the frontal area of the
heat sink is 1/1.2. The velocity is measured at 21 measurement points as shown in Fig. 2. They are
located along the center lines of the cross-sectional area between the fins (2 mm upstream from the
heat sink inlet). We used the average velocity at these 21 points for calculating Vf. Measurements
were performed after confirmation of reaching a fully steady-state condition of the flows and
temperatures at each location. The measurement error of the velocity is estimated to be within r3%,
that of temperature is estimated to be within r2%, and the calculated error of the heat transfer
coefficient is estimated to be within r5%.

3. Experimental Results and Discussion

The relation between Vf and the heat transfer coefficient based on total fin surface area Dt, is
shown in Fig. 4. The heat transfer coefficient increases in proportion to the velocity Vf. Dt increases
from 39.6 to 49.0 W/m2K when the velocity increases from 1.5 to 2.5 m/s for H = 2.5 mm. It increases
from 29.7 to 40.9 W/m2K under the same increase in Vf for H = 5 mm. Therefore, Dt increases by
23.7 to 37.7% when Vf increases by 67%. Figure 4 also compares the calculated Dt (shown by solid
line) with that measured in parallel flow (A): the agreement within 2% to 5% of these two lines
confirms the validity of the measurement in the present experiment. The calculated Dt for parallel
flow is obtained as [5]

BBB
1 (3)
Dt = 1  e\ ˜ Dm ˜ I
\

where

516
Fig. 4. Relation between inlet velocity and heat transfer coefficient based on total area of fins.

BBB BBB
­ ½
°\
°
Dm ˜ I ˜ As
J ˜ Cp ˜ Vf ˜ Ac
, I
tanh Ub
Ub
, Ub H
Dm
CCC
—
Ofin˜ yb
, °
°
° de °
° BBB BBB Oair BBB 0.104 Re ˜ Pr
L ° (4)
® Dm Nu , Nu 3.65  ¾
° de de °
1  0.016 Re ˜ Pr 0.8
° L °
° H˜G, Vf ˜de °
° de 2 Re °
¯ HG Q ¿

Figure 5 shows the relation between fin height H (setting height of the microfan) and the heat
transfer coefficient based on total fin surface area Dt. Increase in H from 2.5 to 20 mm reduces Dt by
about 50% (at each air velocity Vf). This reduction is due to the decrease in air velocity in the
downstream half when H increases. Figure 6 shows the relationship between H and the heat transfer
coefficient based on the fin base area Db. Db has a minimum value between H = 10 and 15 mm for all
Vf. As H increases, Db decreases rapidly, and then increases again at H over 15 mm. When H is low,
the heat transfer area in the downstream half becomes small. However, the dramatic increase in surface
heat transfer coefficient due to the sudden contraction there sufficiently compensates for the reduction
in surface area. On the other hand, the increase in surface area of the heat sink is evident at higher H
and is more dominant than the decrease in surface heat transfer. Moreover, the fin height producing
the minimum Db decreases from 15 to 10 mm as the flow velocity Vf increases. This change in fin
height at the minimum Db shows the possibility of a different flow structure of turbulence and wake

517
Fig. 5. Downstream fin height and heat transfer Fig. 6. Downstream fin height and heat transfer
coefficient at total area of fins. coefficient based on fin base area.

especially in the downstream half of the heat sink when Vf varies. This change in flow structure affects
the heat balance of the heat sink in both the upstream and downstream half.

Figure 7 shows the relation between temperature difference 'Tb and distance from the air
inlet Lb at an air velocity Vf of 1.5 m/s. Here, 'Tb is the temperature difference between fin base
(Tb1, Tb2, Tb3) and inlet air (Ta). 'Tb increases almost linearly with Lb in the range from 20 to 60
mm for all fin heights. The approximate value of the increase of 'Tb is about 1.1 to 2.0 K. Figure 8
shows the temperature inclines (C) obtained from the slope of these lines. It is seen that C has a
maximum at a fin height of about 10 mm. This implies a change of the conduction heat flux through
the fin base from downstream to upstream of the heat sink when the fin height varies. Figure 8 also
shows that the thermal resistance in the downstream half at an air velocity (Vf) of 1.5 m/s has a
maximum value at H = 10 mm, because C is directly related to the thermal resistance in the
downstream half.

The relation between the heat transfer coefficient per unit heat-sink volume Dbv and fan power
P (electric power consumption) is shown in Fig. 9. Here, the volume is calculated from the fin base
area multiplied by the maximum height of either the heat sink or the microfan. However, the volume
of the duct is excluded from the calculation in parallel flow. The fan power is obtained directly from
the measurement of the input voltage and current to the fan. Dbv increases with increasing P. Figure
10 shows the relationship between Dbv and H at P = 0.5 W.

Dbv increases with decreasing H from 20 to 5 mm, and it has a maximum point at H = 5 mm.
However, the heat transfer coefficient decreases rapidly when H is below 2.5 mm. This is a kind of

518
Fig. 7. Temperature distribution of heat sink Fig. 8. Downstream fin height and temperature
base. incline.

trade-off relationship. A decrease in H causes a reduction of surface area in the downstream half and
an increase in heat transfer coefficient due to a velocity increase. However, a reduction of H causes
a rapid increase of pressure loss. When the fin height is lower than 5 mm, the pressure loss is significant
and it reduces the volume flow rate dramatically (under constant fan power). Figure 10 shows that
the fan should be completely buried in the heat sink. That is, when the microfan height and the heat
sink height are the same, the cooling efficiency is best (1.8 times that of parallel flow).

Fig. 9. Fan power and volumetric heat transfer Fig. 10. Downstream fin height and volumetric
coefficient. heat transfer coefficient (P = 0.5 W).

519
4. Summary

A new cooling structure, which uses a parallel-plate-fin-type heat sink and an axial-type microfan
in the downstream half of the heat sink, was developed. This structure significantly improves the heat sink
under a limited cooling volume. The influence of the setting height of the microfan on the cooling
performance of the heat sink was investigated and the following conclusions were obtained.

x The new structure provides a heat transfer coefficient per unit volume 1.8 times higher than that
of a typical parallel-flow structure using a similar heat sink and fan (at a fan power of 0.5 W).
The best cooling performance is achieved by setting the level of the upper surface of the fan at
the same height as that of the fin top of the heat sink in the upstream half.

x The heat transfer coefficient based on total surface area increases with increasing air velocity
between the fins in the upstream half of the heat sink. For example, the heat transfer coefficient
increases to 49.0 (from 39.6 W/m2K) when the fan setting height is 2.5 mm, and the air velocity
increases by 67% from 1.5 to 2.5 m/s. The heat transfer coefficient increases from 29.7 to 40.9
W/m2K when the setting height is 5 mm. The heat transfer coefficient therefore increases 23.7
to 37.7%.

x The heat transfer coefficient based on total surface area decreases by about 50% when the setting
height of the fan increases from 2.5 to 20 mm.

x The heat transfer coefficient based on the fin base area has a minimum value at a setting height
of 10 to 15 mm. And it becomes very large at low setting heights.

x The heat transfer coefficient per unit volume at a fan power of 0.5 W has a maximum at the
setting height of 5 mm. It increases with decreasing setting height from 20 to 5 mm. However,
it decreases rapidly at a setting height lower than 5 mm.

Literature Cited

1. Shinohara K et al. Cooling performance of pin-fin-type heat sink by an axial fan. Trans JSME
Proceedings of Thermal Engineering Conference 1997, p. 158. (in Japanese)
2. Kondou Y, Matsushima H. Experimental study of impingement cooling of heat sinks for LSI
packages with pin-fin arrays. Trans JSME 1995;61:339. (in Japanese)
3. Komatsu T, Matsushima H. Effect of cooling-air direction on cooling characteristics of heat
sinks. Trans JSME Thermal Engineering Symp 1994, p 23. (in Japanese)
4. Kitajo S. Estimation for heat sink cooling performance with cooling fans. Trans JSME Thermal
Engineering Symp 1993, p 97. (in Japanese)
5. Kays WM. Numerical solutions for laminar-flow heat transfer in circular tubes. Trans ASME
1955;77:1265–1274.

"F F F"

Originally published in Trans JSME B, 65, 1999, 1377–1382.


Translated by Keizou Kawamura, Mechanical Engineering Research Laboratory, Hitachi, Ltd., 502
Kandatsu, Tsuchiura, Ibaraki, 300 Japan.

520

You might also like