An ExperimentaI Study on the Performance of Miniature Heat Sinks for

Forced Convection Air CooIing

V. Egan, J. Stafford, P. Walsh, E. Walsh, R. Grimes,
Stokes Research Institute,
Department of Mechanical & Aeronautical Engineering
University of Limerick
Limerick, Ireland
Phone: (+353) 61 213487
Fax: (+353) 61 212039
Email: vanessa.egan@ul.ie



ABSTRACT
In recent years the design of portable electronic devices must
incorporate thermal analyses to ensure the device can be
adequately cooled to acceptable temperatures. Consumer
demand for smaller, more powerful devices has lead to an
increase in the heat required to be dissipated and a reduction in
the surface area both of which result in an increased heat flux.
In this paper, an experimental study is performed on one of the
smallest commercially available miniature fans, suitable for
cooling portable electronic devices, used in conjunction with
both finned and finless heat sinks. Previous analysis has
shown that due to fan exit angle, flow does not enter the heat
sinks parallel to the fins or bounding walls. This results in a
non uniform flow rate within the channels of the finned and
finless heat sink along with impingement of the flow at the
entrance giving rise to large entrance pressure losses. In this
paper straightening diffusers were attached at the exit of the
fan which resulted in aligning the flow entering the heat sinks
with the fins and channel walls. In designing the finned heat
sink current optimization criterion for finned heat exchangers
has been applied to ensure maximum heat transfer rates; the
finless heat sink was designed to the same specifications. The
maximum overall footprint area of the cooling solution is
534mm
2
with a profile height of 5mm. The thermal
performance of each cooling solution was investigated by
quantifying its thermal resistance over a range of fan speeds
and comparing the results to cases without diffusers. In order
to investigate the flow field, detailed velocity measurements
were obtained using Particle Image Velocimetry, which
provided a further insight into the physics of the flow in such
miniature geometries and in designing the straightening
diffusers. The thermal analysis results indicate that the cooling
power of the solution is increased by up to 20% through the
introduction of a diffuser. Hence, demonstrating the need for
integrated fan and heat sink design of low profile applications.

KEY WORDS: low profile, forced convection cooling,
miniature fan

NOMENCLATURE
A
conv
convective surface area, m
2
b fin spacing, mm
FOV Field of view, m
h convective heat transfer coefficient, W/m
2
K
h
FC
forced convective heat transfer coefficient,
W/m
2
K
h
f
total Convective heat transfer coefficient,
W/m
2
K
k
heat sink
thermal conductivity of heat sink material
W/m.K
L Length of heat sink, mm
N number of pixels in interrogation region
AP pressure drop, Pa
Q volumetric flow rate, m
3
/s
Q
FC
power dissipated by forced convection, W
Q
input
input power, W
Q
Losses
power dissipated by secondary cooling
mechanisms, W
RES resolution of camera, Pixels
Rth
T
total thermal resistance,
o
C/W
Rth
FC
thermal resistance of Forced Convection
Cooling,
o
C/W
Rth
Losses
thermal resistance of Natural Convection
Cooling,
o
C/W
RPM revolutions per minute, RPM
SP static pressure rise, Pa
t
fin
fin thickness, mm
T

ambient air temperature,

o
C
T
s
surface temperature,
o
C
U
max
Maximum velocity, m/s
u
x,y
Velocity at local coordinates (x,y), m/s
W Total width of heat sink, mm
978-1-4244-1701-8/08/$25.00 2008 IEEE
497
X Half of channel width, mm
Y Half of channel height, mm
Greek Symbols
o thermal diffusivity, m
2
/s
viscosity , N.s/m
2
Subscripts
conv convective
FC forced Convection
IR interrogation Region
Max maximum
s Surface
T Total
Ambient
INTRODUCTION
The design and integration of miniature cooling solutions into
todays portable electronic devices is moving to the forefront
of electronics cooling research. The drive behind this research
is consumer demand for smaller more powerful portable
devices. As the surface area available for power dissipation
reduces, along with technology advancements increasing the
devices power consumption, the resulting heat fluxes are
considerably elevated. At present most manufacturers rely on
natural convection in conjunction with heat spreading
technologies to maintain temperatures at acceptable consumer
levels. However, a major flaw with this method of heat
dissipation is the resulting elevation in device case
temperatures. If the envisaged trend in increasing heat fluxes
continues, then natural convection and conduction
technologies will no longer be able to maintain devices at
acceptable temperatures. Hence, there is a need for research
into cooling technologies that can offer higher heat transfer
rates compared to current natural convection and heat
spreading techniques.

Forced convection cooling is the preferred new cooling
technology among designers of portable electronic devices.
This is due to large number of factors including, cost,
reliability, power consumption, footprint area, and profile
height. Liquid cooling [1] and heat pipe [2] technologies
require secondary coolers but for forced convection air is the
working medium which is in abundant supply. Unlike phase
change materials [3], forced convection is capable of
efficiently dissipating heat over extended time periods. An
example of this is the unconventional design proposed by
Walsh et al [4], which gives a thermal resistance value of
7.5C/W in a volume of 4.56mm
2
with a profile height of 4mm
including the fan. As air cooling is currently in widespread use
in industry the implementation costs are relatively low.
Finally, with a sufficiently small footprint and profile, heat can
be dissipated directly at the source via a path of low thermal
resistance, resulting in a reduction in the temperature of all
other components and surfaces of the device.

Over the past few years a number of groups have initiated
research to address the developing market of miniature forced
convective air cooling. The first series of papers on such
research examined the validity of the conventional scaling
laws [5-9] as the fan scale is reduced to miniature dimensions.
The authors found that at low Reynolds numbers boundary
layer phenomena result in reducing the flow coefficient and
elevating the power coefficient. Day et al [10] also reported a
scale effect which resulted in the flow rate delivered being
significantly less than expected when operating at low
Reynolds numbers.

With regard to the integration of miniature fans with heat
sinks it is noted that impingement cooling emerges as one of
the best methods to dissipate large heat loads over small areas,
with thermal resistances as low as 1.4
o
C/W being reported
[11,12] for pin-fin heat sinks. However, to achieve such
performances, footprints areas of order 2000 mm
2
and profile
heights of order 100 mm were reported. Whilst this footprint is
relatively small, it is noted that such large profile heights
would most likely deem this technology unacceptable for use
in most portable electronic devices. Also, when considering
miniature scale fans, the flow rates and pressure rises
attainable are considerably lower than those attained in the
aforementioned studies for impingement cooling. Hence in
order to satisfy the constraints of low profile forced
convection cooling a fan and heat sink in parallel is the most
practical solution. A recent analysis of fan and heat sink
arrangement in term of performance optimisation has been
carried out by Walsh et al [13, 14]. It was shown that best
thermal performance was achieved when the fan exit angle is
matched to the heat sink fin angle, i.e. flow enters the heat
sink parallel to the fins. A recent study by Egan et al [15, 16]
investigated the thermal performance of a miniature fan used
in conjunction with a finned and finless heat sink. The overall
footprint area of the solution was 456mm
2
with a profile of
5mm. Velocity measurements of the flow in the heat sinks
showed that flow exited the fan at an angle to the heat sink
walls and the fins. In the case of the finned heat sink it was
seen that the flow strongly impinged on the heat sink fins at
the entrance to the channels. The current study is focused on
developing the previous analysis of Egan et al [15, 16] to
reflect the finding of Walsh et al [13, 14] in designing a
suitable cooling solution.

The primary objective is to determine the effect of
straightening the flow prior to entering the heat sinks through
the use of a diffuser. A number of different diffusers were
manufactured to obtain the optimum design giving parallel
flow at entry to the heat sink. The optimum diffuser design
was determined from flow velocity measurements obtained
using particle image velocimetry. Upon finding such a design
the fan, diffuser and heat sink solutions were characterised by
repeating the heat transfer measurements carried out in Egan
at al [15, 16]. Such results would conclusively show if the
diffuser resulted in an overall enhancement of the heat transfer
rate of the finned and finless solutions. Particle image
velocimetry measurements of the flow in the heat sink at
speeds ranging from 3000RPM to 8000RPM were also
obtained and provide a further insight into the flow field
within the heat sinks. The results of this study are valuable to
designers of such cooling solutions as enhanced heat transfer
498
rates may be obtained by simply adding a diffuser between the
fan and heat sink.

HEAT SINK AND DIFFUSER DESIGN
The fan tested for the current work is the Micronel U16LM-9
having a footprint of 100mm
2
and profile of 5mm (See Fig. 4).
This is currently one of the smallest commercially available
fans for forced convection cooling of electronics. The
pressure versus flow rate characteristics supplied by the
manufacturer for a nominal speed of 6000RPM are shown in
Fig. 1. In determining this curve it is possible that only a small
number of experimental data points were recorded by the
supplier, hence resulting in a sharp change in flow rate once a
certain static pressure is reached. Also shown on this graph are
the curves for 3000RPM, 5500RPM and 8000RPM. These
curves were obtained using the fan scaling laws detailed in
Bleier [17] and defined here in equations 1 and 2.

Q RPM (1)

2
SP RPM (2)

0
1
2
3
4
5
6
7
8
9
10
11
0 0.00005 0.0001 0.00015 0.0002 0.00025 0.0003
Volume Flow Rate [m
3
/s]
S
t
a
t
i
c

P
r
e
s
s
u
r
e


[
P
a
]
Supplier Curve Scaled to 8000RPM Supplier Curve Scaled to 5500RPM
Supplier Curve Scaled to 3000RPM Pressure V's Flow curve at 6000RPM

Fig. 1. Pressure versus flow rate characteristics for the
Micronel U16LM-9 low profile fan as supplied by Micronel
for 6000 RPM and scaled for speeds from 3000PRM to
8000RPM using the fan scaling laws.


Experiments were carried out in both finned and finless heat
sinks. In designing the finned heat sink the pressure versus
flow rate characteristics at a fan speed of 8000RPM were
used. Fig. 2 (a) shows a schematic of this heat sink and
highlights the variables which were optimized; these include
the fin spacing, (b), and fin thickness, (t
fin
). The former of
these was optimized using Eq. (3) which was developed by
Bejan [18] for flow between parallel plates using the
intersection of asymptotes between a fully developed flow
limit and a flat boundary layer flow limit.
1
4
2
2.73
o | |
=

A
\ .
optimum
b
L P
|
L
(3)

Fig. 2. Schematic of finned and finless heat sinks outlining
the overall dimensions and the optimised parameters, b and t
fin


In determining the optimum fin spacing, the pressure drop
(AP) across the bank of channels, was set equal to 6.5Pa so
that the fin spacing of the heat sink is optimised for a fan
speed of 8000RPM and should correspond to a total
volumetric flow rate of approximately 2.5E-04m
3
/s passing
through the heat exchanger. In order to achieve this Eq. (3)
requires that the fin spacing be approximately 1.1mm. The
second parameter examined is the thickness of the fins, t
fin
.
This was optimized by ensuring that the fin efficiency is
greater than 99% and Ellison [19] shows this to be the case
when Eq. (4) is satisfied.


2
fin heat sink
40 > t hL k (4)

Forced convection heat transfer coefficients for gases are
typically in the range 25-250 W/m
2
.K [20]. In the current
study the expected maximum velocities, based on the fan
blade top speed, were approximately 0.5m/s, hence a value of
60 W/m
2
K was chosen for hin Eq. (4). Using copper as the
heat sink material, Eq. (4) reveals that the fin thickness should
be at least 0.06mm. However the technique used to
manufacture the heat sink required this dimension to be of
order 0. 3mm, hence ensuring a more than adequate fin
thickness. The result is that a finned heat sink with a
maximum of 6 channels could be manufactured whilst
maintaining the heat sink footprint constraints that were
applied. The finless heat sink was manufactured to have the
identical specifications of profile, footprint area and external
wall thickness and is shown in Fig. 2 (b).

As previously mentioned in order to achieve the optimum
thermal performance of a fan and heat sink in parallel, flow
must enter the heat sink aligned with the heat sink walls and
fins. Fig. 3 is a schematic of a velocity vector diagram for a
backward curved fan with a rotational speed, e, and radial
499
velocity, V
r
. The angle of the flow exiting the fan casing can
be measured using PIV and is dependant on system resistance.
A high system resistance such as that created by a finned heat
sink will result in a lower flow rate through the fan. This in
turn will result in reducing only the radial flow component and
hence the angle at which fluid leaves the fan. It can be seen
that high system resistances will reduce , whilst low system
resistances will result in increasing this angle.

In order to align the flow exiting the fan with the heat sink
walls and fins a number of diffusers with angles varying
between 25
o
and 50
o
with both curved and straight walls were
manufactured from polycarbonate. As discussed above, it was
found that the optimum angle for flow alignment varied
between the finned and finless case. Each diffuser was tested
using PIV to measure the velocity of the flow in each of the
heat sinks and it was determined from the measurements that a
25
o
curved diffuser and a 50
o
straight diffuser resulted in flow
alignment within the finless and finned heat sinks respectively.


Fig. 3. Velocity triangle showing fan exit angle.

EXPERIMENTATION
This section is structured as follows. Firstly the cooling
solution, encompassing the fan, diffuser and heat sink, is
described followed by a description of the experimental
technique employed in determining the thermal resistance.
Particle Image Velocimetry (PIV) was also used to visualize
the change in flow structures encountered by incorporating a
diffuser into the design; hence this technique is described in
the latter part of this section.

Experimental Configuration

The experimental configuration is shown in Fig. 4 consisting
of a finned and finless set-up. The results presented in this
paper are detailed in table 1. The current work presents results
from all four cases however, case 1 and 2 are detailed in a in
Egan et al [15,16] and encompass tests performed without a
diffuser.



Fig. 4. Photograph of the experimental configuration
employing the straightening diffusers with the finned and
finless heat sinks.



TABLE 1. Experimental Test Description
Thermal Resistance Measurements

The experimental configuration consisted of a Micronel fan, a
flow diffuser and the optimized heat sink that was designed in
the preceding section. During experimentation the top cover of
the heat sink, as shown in Fig. 4, was in position so as to
create a fully closed channel. This cooling solution has a
footprint of 256mm
2
for the fan, either 231mm
2
or 278mm
2
for
the heat sink and diffuser and a profile of only 5mm. The fan
was powered using a DC power supply and rotational speeds
ranging from 1000RPM to 8500RPM were measured using an
optical tachometer. The heat source used for experimentation
was a 66mm thin-film heater, available from Minco Inc., and
was attached to the base of the heat sinks. Three K-type
thermocouples attached using a heat resistant tape to the centre
of the top, base and left wall of the heat sink. The ambient air
temperature was also recorded. The four thermocouples were
connected to a National Instruments 9211 USB High Speed
Carrier and temperatures were recorded and plotted using
LabVIEW 8. During experimentation, the entire heat sink was
found to be almost isothermal, as a maximum temperature
difference between the hottest (base) and coldest (top) surface
of approximately 5% was measured. Also of significance, was
that the external surfaces of the heat sink were covered using
insulating tape to reduce the heat loss due to secondary
cooling mechanisms such as natural convection and radiation.
The purpose of this is to attempt to simulate the operational
environment for such a system in a typical portable electronic
device where the total set-up is likely to be confined within a
very small cavity.
Test Number Description
1 Finless Heat, No diffuser
2 Finned Heat, No diffuser
3 Finless Heat, 25
o
curved diffuser
4 Finned Heat, 50
o
straight diffuser
Diffusers
Top cover of
heat sink
500
The thermal performance of any cooling solution can be
characterized by its thermal resistance which is defined by
equation (5).

( )
s
T
Input
th
T T
R
Q

= (5)
Q
Input
is the total power supplied to the thin-film heater, T
s
is
the steady state temperature of the heat sink surface and T

is
the temperature of the ambient air. However, it should be
noted that this power was not only dissipated by forced
convection, but also by the secondary cooling mechanisms
already mentioned so that that total power supplied to the heat
sink is dissipated by:

Input FC Losses
Q Q Q = + (6)

Hence, the total thermal resistance of the cooling solution can
also be defined by:
( )(
( )
)
=
+
FC Losses
T
FC Losses
th th
th
th th
R R
R
R R
(7)

and the forced convection thermal resistance by

( )( )
( )
T Losses
FC
T Losses
th th
th
th th
R R
R
R R
=

(8)

It should be noted, however, that the power dissipated by these
secondary mechanisms is highly dependant on factors that
vary from application to application. Such factors include:
available air for free convection currents; temperature of
surrounding surfaces and any contact of conducting materials
with the heat sink surface. In order to remove this ambiguous
power dissipation from the characteristics reported herein, all
of the experiments were conducted whilst maintaining the heat
sink surface temperature constant at 70
o
C and ambient at
25
o
C. An experiment was then conducted with the fan
switched off (i.e. Q
FC
= 0) and resulted in directly measuring
Q
Losses
. Since the same temperatures were maintained
throughout all experiments, the heat lost through secondary
mechanisms should remain constant throughout. This allowed
for the calculation of the heat dissipated by just forced
convection giving a more representative result of the forced
convection cooling.

The total and forced convective heat transfer coefficients were
calculated using Eq. (9) and Eq. (10) respectively.
1
T
T
th conv
h
R A
= (9)

1
=
FC
FC
th conv
h
R A
(10)
Where, h, is the heat transfer coefficient in W/m
2
K and A
conv

is the convective surface area of the finned heat sink in m
2
.
The convective surface area of the finned and finless heat
sinks was calculated to be 0.000984m
2
and 0.00044m
2

respectively with an accuracy of 1E-06m
2
. The heat transfer
coefficient is inversely proportional to the convective surface
area of the heat sink as shown in Eq. (9) and (10), this means
that the finless heat sink with a lower A
conv
will result in a
higher heat transfer coefficient.

Uncertainty analysis
An uncertainty analysis was performed, [21], giving the
thermal resistance an error of less than 5%. The uncertainty
was greatest for natural convection and decreased with
increasing fan speed.

The optical tachometer used to measure the rotational speed of
the fan has a resolution of 1RPM. However, during
experimentation it was observed that the speed varied up to
50RPM.
Velocity Measurements

The second set of experiments involved detailed velocity
measurements of the flow within the channels of the heat sink.
Particle Image Velocimetry (PIV) was employed to obtain
these velocity measurements. However, in order to obtain the
optical access required for this technique, the copper top cover
of the heat sink was replaced with a glass slide. Hence all PIV
results were obtained without heat transfer. The following
paragraphs detail the principle of the technique and the set-up
used during experimentation.

PIV works on the principle of introducing seeding or tracer
particles to the flow and tracking their motion over known
time intervals. Ideally the particles should be small and their
density matching the working fluid density. Gravitational
forces will produce an error depending on the density
mismatch of the particles and if the particles are large they
will be unable to follow the exact motion of the fluid. A water-
based oil consisting of monopropalene glycols or CO
2
fog was
used for the seeding which was produced using a JEM Stage
Hazer 2000. The size of seeding particles produced were of
approximately 10m diameter with relative density of
1.05kg/m
3
at 293K, [22]. In order to contain the seeding
particles the experimental apparatus was placed in a glass
walled enclosure of dimensions 400mm(L) 250mm(H)
250mm(W). Once the particles were released into the
enclosure a laser illuminated a plane in the heat sink and the
light subsequently emitted by the particles was recorded on a
camera positioned perpendicular to the illuminated plane.
Velocity measurements were taken for a plane at mid-depth in
the heat sink channels. A schematic of the experimental set-up
is shown in Fig. 5

501


Fig. 5. PIV Experimental Set-up showing camera and laser
positions with respect to fan and heat sink. The diffuser is
inserted between the fan and heat sink.


Raffel et al. [23] discuss the science and technical aspects of a
PIV system in full. Almost all PIV systems use a laser system
as lasers emit intense and monochromatic light, which after
being passed through a series of lenses is easily shaped into a
thin light sheet for illuminating the seeding particles. A
Nd:YAG, 532nm, semiconductor laser was used giving a
1.5mm thick light sheet. To measure the velocity of the
particles the laser was pulsed twice with a time delay chosen
so that the distance travelled by the seeding particles between
images is discernable and that particles do not leave the
illuminated plane within the period. The exposure time of the
camera was long enough to capture the position of the
particles and short enough to avoid blurring of the image. To
ensure this, TSI Inc [24] suggest that the time between pulses
be calculated from:

s in IR
max
*
4U
pixel
FOV
N
RES
(11)

where FOV is the field of view, RES is the resolution of the
camera in the direction of flow, N
pixels
in IR is the number of
pixels in an interrogation region and V
max
is the expected
maximum in-plane velocity. Prior to experimentation, in order
to calibrate the subsequent measurements, a scaling device
must be inserted in the test section to obtain the correct scaling
factor for the velocity measurements. This was achieved by
placing a calibrated ruler in the measurement plane.

A PowerViewTM Plus 2MP CCD camera of resolution 1600
1200 with a 60mm focal length lens was used to record the
scattered light from the illuminated particles and their images
appear as bright spots on a dark background. The images are
recorded on separate frames from the double pulsing of the
laser. This technique is called multi-frame/single exposure
PIV. A statistical PIV evaluation using Insight 3G software
supplied by TSI Incorporated was applied to the recorded data
in order to extract the required displacement information. The
recordings were divided into regions known as interrogation
areas and a correlation technique cross-correlation, was
applied to determine the displacement of the particles within
the interrogation area. The size of the interrogation area was
set according to the time between pulses so that the seeding
particles would not travel more than 25% of the interrogation
area length. If seeding particles travel much more than this
distance then too many seeding particles will have left the
interrogation area when the second image is taken. These
particles will not be available for velocity measurement and
will be a source of noise. This problem can be overcome by
overlapping interrogation areas but again care must be taken to
ensure that seeding particle displacements between pulses are
not too small. Interrogation area overlap was set to 50% of the
interrogation area height and width for the current results and
the interrogation area size was set at 32 32 pixels. The
velocities were calculated from the displacements and the
duration of the illumination pulse. A series of post-processing
steps were applied to the PIV data.
RESULTS AND DISCUSSION
The objectives of this work are to provide optimized thermal
management solutions based on a previous analyses, [13-16],
using one of the smallest commercially available fans
combined with finned and finless heat sink geometries.
Optimization of such solutions has been considered using
velocity field and thermal resistance measurements presented
in this section. Velocity measurements are considered first in
order to develop a suitable diffuser for each case based on
design criterions previously discussed. Thermal resistance
measurements, heat transfer and flow rate results proceed from
the flow field analysis, quantifying the influence of the
introduction of diffusers on both heat sinks over a range of fan
rotational speeds.

Fig. 6 and Fig. 7 depict the flow at the entrance regions to the
heat sink for finned and finless geometries without the use of a
diffuser [15,16]. While Fig. 8 and Fig. 9 present the velocity
magnitude data for flow in the finless and finned heat sinks
with the inclusion of diffusers at the entrance to the heat sink.
A range of diffusers were examined for both heat sink
geometries producing various heat sink fan orientations from
25
o
to 60
o
. Therefore due to the quantity of PIV results
analyzed, the data presented for this work is only considered
for the diffusers which would optimize flow alignment and
distribution in both heat sink cases. It was found that a 25
o

curved diffuser for the finless heat sink, and a 50
o
straight
diffuser for the finned heat sink were beneficial.

The velocity results presented in all figures are velocity
magnitudes with vector length and color varying accordingly;
longer vectors represent high velocity regions. The
relationship between color and velocity magnitude is indicated
in the legend provided with each figure. The data presented for
the current work was obtained by averaging 70 instantaneous
results, each taken at a frequency of 14.5Hz.

502
The PIV results for the finless heat sink and 25
o
curved
diffuser are shown in Fig. 8 for fan rotational speeds between
3000RPM and 8000RPM. Each velocity magnitude plot
shown in Fig. 8 presents the fluid flow through the diffuser
and heat sink, with the dashed line indicating the entrance to
the heat sink from the diffuser. Considering the data presented
in Fig. 8 and Fig. 6 for a fan speed of 3000RPM, it is possible
to distinguish noticeable differences in the flow distribution
due to the introduction of a diffuser. In Fig. 8 the flow is
distributed relatively evenly in the channel of the finless heat
sink for all three speeds tested. This implies that the effects of
forced convection should also be relatively even across the
channel section, reducing the possibility of raised
temperatures in localized areas of the heat sink. However, in
Fig. 6, a large region of low velocity exists near the right wall
at the entrance. Hence forced convective cooling in this region
will have little effect on the right wall temperature. By
introducing the diffuser at the entrance to the heat sink, flow
now enters the finless channel parallel to the heat sink walls as
seen from the velocity magnitude plot of Fig. 8.

The results from PIV analysis on the finned heat sink and 50
o

straight diffuser are presented in Fig. 9. The diffuser and heat
sink are labeled in Fig. 9(a) and also color contrasted to
distinguish the fins (black) from their shadows in the diffuser
(grey). These were unavoidably created by the laser beam
used to illuminate the seeding particles within the heat sink.
Consequently particle tracking was not possible in these
regions of the diffuser. However, this does not affect the
validity of the current work as flow structures are still visible
in the diffuser, and concentration on flow characteristics
within the heat sink is of primary interest for optimizing this
thermal management solution.



Fig. 6. PIV velocity magnitude plot, in m/s, from Egan et al
[16] of the flow in the entrance region of the finless heat sink
with no diffuser attached for 3000RPM. All velocities are
normalised with respect to the maximum velocity of 4m/s.

For each case considered in Fig. 9, the flow entering the heat
sink shows minimal impingement on the fins as opposed to
Fig. 7, where no diffuser is used at 8000RPM. Under fully
developed conditions the maximum and minimum velocity
and hence flow rate is recorded in channel 3 and channel 6
respectively for all three fan speeds (see equations 12 and 13).
The percentage difference in flow rates between channel 3 and
6 for fan speeds of 3000RPM, 5500RPM and 8000RPM is
42%, 48% and 43% respectively. However the standard
deviation of the flow rates and velocities in for all six channels
is higher for a fan speed of 8000RPM suggesting that the flow
distribution between channels is not as uniform at higher
speeds. It can be seen from Fig. 9(c) that the angle of the flow
entering the diffuser changes at 8000RPM and that a longer
diffuser but of the same angle may be required.


Fig. 7. PIV velocity magnitude plot, in m/s, from Egan et al.
[15] showing the flow field in the entrance region to the finned
heat sink at 8000RPM with no diffuser. All velocities are
normalised with respect to the maximum velocity of 2.7m/s.

Following the selection of a diffuser for the finned and finless
heat sinks using the flow field analyses discussed, the
experimentation was arranged to obtain a measure of the total
thermal resistance for a range of fan speeds from 0RPM to
8000RPM. The data recorded are plotted in Fig. 10. This
figure, along with all subsequent figures discussed, also
compares data recorded for the finned and finless cases where
no diffuser was used [15,16]. From Fig. 10 it can be seen that
a reduction in thermal resistance can be obtained through the
use of a diffuser. For the finned heat sink at low velocities, the
measured thermal resistances are of similar magnitude for
both set-ups. However, as rotational speed of the fan
increases, the heat sink and diffuser combination clearly out
performs the solitary heat sink case. For a finless heat sink
there is a constant reduction in total thermal resistance
achieved when using a diffuser between approximately
4000RPM and 8000RPM. Below this range of speeds, large
reductions in thermal resistance are seen by using a diffuser to
align the flow exiting the fan. In order to achieve a thermal
resistance of lower than 40
o
C/W using a finned heat sink and
diffuser combination, the fan must operate at a rotational
speed greater than 5000RPM. Alternatively, the use of a
finless heat sink and diffuser combination produces similar
thermal resistances from a fan speed of just 3000RPM. A
benefit of this is that lower operating speeds are sufficient for
similar cooling effects when using the optimized finless
arrangement. In terms of fan reliability, noise and pumping
power this proves highly advantageous.
503


(a) Finless Heat Sink at 3000RPM



(b) Finless Heat Sink at 5500RPM



(c) Finless Heat Sink at 8000RPM


Fig. 8. PIV velocity magnitude plots depicting the flow within the diffuser and finless heat sink for fan speeds of (a)
3000RPM (b) 5500RPM and (c) 8000RPM respectively. The respective velocity scale bars in m/s are shown above each image.
The dashes lines represent the entrance to the heat sink from the diffuser.
504
(a) Finned Heat Sink 3000RPM


1
2
3
4
5
6


Diffuser Heat Sink


(b) Finned Heat Sink 5500RPM





(c) Finned Heat Sink 8000RPM



Fig. 9. PIV velocity magnitude plots of flow in finned heat sink and 50
o
straight diffuser at 3000RPM, 5500RPM and
8000RPM. Respective scale bars for velocities in m/s are given above each result. Flow is from left to right. Regions in gray are
shadows from the heat sink fins which result from the laser light entering from the right.















505
A plot of the forced convection thermal resistance comparing
all cases with and without diffusers is presented in Fig. 11.
This plots only the forced convection cooling achieved from
the fan and not the heat dissipated by secondary cooling
mechanisms. For the current work these secondary cooling
mechanisms were found to give a thermal resistance, R
th-Losses
,
of 63
o
C/W. Hence, by using Eq. (9) and the results for R
th-T

presented in Fig. 10, a corresponding forced convection plot of
the forced convection thermal resistance, R
th-FC
was obtained
and is presented in Fig. 11 . For a fan speed of 8000RPM,
using the diffuser designed for the finned heat sink, a decrease
of approximately 23% in forced convection thermal resistance
is obtained. At the same operating speed, the finless heat sink
and combined diffuser provide a decrease of 15.3% over a
finless heat sink without diffuser. It is therefore apparent that
in the case of a fan and heat sink in parallel aligning the flow
with the heat sink fins produces lower thermal resistances than
impinging flows [13].

At 8000RPM the finned diffuser case results in a lower
thermal resistance compared to the finless diffuser solution.
This is an expected result as the design of the finned heat sink
was optimized for 8000RPM, based on boundary layer
analyses discussed in the preceding section. As previously
mentioned, a difference in velocity magnitudes for both
geometries can be seen in Fig. 8 and Fig. 9. This can be
accredited to a higher system resistance due to the finned heat
sink, which implies that higher mass flow rates are evident in
the finless heat sink for each rotational speed recorded. Higher
mass flow rates commonly result in increased heat transfer
rates; however Fig. 10 and Fig. 11 enforce the influence of
boundary layer analysis used in the finned heat sink design,
i.e. the optimum spacing of fins in a heat sink for a given
pressure drop is determined based on the boundary layers
merging just at the exit of the heat sink. Thermal resistance
measurements for the finned case drop below that of the
finless case at approximately 7500RPM. It is also apparent
that the boundary layer optimization criterion is proven for
both situations with and without the use of a diffuser for low
profile solutions. However, the thermal resistance values for
the finless-diffuser case are lower compared to finned case.
(test 1). Even at 8000RPM where the forced convective
thermal resistance appears to be reaching a constant value with
increased rotor speed, there is a reduction of 11.3% over the
forced convective thermal resistance of the finned heat sink
without a diffuser. This is significant considering the 25
o

diffuser used for the finless heat sink only increases the
overall footprint area of the thermal solution by approximately
6%. It was also noted that the 50
o
diffuser used to optimize the
finned heat sink increases the overall footprint area of the
thermal solution by approximately 16%. This result
demonstrates the need to design integrated and optimized fan
and heat sink solutions for the low profile market.

0
10
20
30
40
50
60
70
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Rotor Speed (RPM)
R
t
h

(
o
C
/
W
)
Finned + diffuser
Finless + diffuser
Finned
Finless

Fig. 10. Plot showing total thermal resistance for fan speed
ranges between 0RPM and 8300RPM.

0
25
50
75
100
125
150
175
200
225
250
275
300
3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000
Rotor Speed (RPM)
R
t
h
F
C

(
o
C
/
W
)
Finned + Diffuser
Finless + Diffuser
Finned
Finless
Losses

Fig. 11. Plot of forced convection thermal resistance with
varying RPM. Also indicated are the effects due to secondary
cooling mechanisms (Losses).

Similarly to a combined fan and heat sink cooling solution,
sizing issues may also be associated with the power supply
necessary for operation. Considering the power requirements
of a cooling solution, it is reasonable to measure shaft power
supplied to the rotor when characterizing fan performance.
This provides designers with power requirements for the fan
alone which, may deviate considerably from the motor power
required especially when considering small scale fans. The
current work is not examined in this manner, as there is an
enclosed motor and fan assembly designed for combined use.
Power requirements are therefore presented for the motor and
fan collectively. The manufacturer provides a nominal
operating speed of 6000RPM. To achieve this rotational
speed, a power input of 0.163W is required. This value
506
increases to 0.320W to reach an operating speed of 8000RPM.
The scale law for power consumption predicts a 1.77 factor
[(8000/6000)
3
] increase however, the power is seen to almost
double. This is due to the efficiency of the driving motor
dropping off at higher speeds as bearing losses become larger.
Power is therefore conserved to a greater extent when
operating the cooling solution below the nominal operating
speed of 6000RPM as opposed to above this speed where
power requirements rise sharply for small increases in
rotational speed. Hence, depending on the level of restrictions
evident with power supply choice and the thermal
performance requirements, it may be beneficial to integrate a
finless heat sink and diffuser design as it outperforms the
finned heat sink cases below 6000RPM, shown in Fig. 10 and
Fig. 11. It is also of interest to note that this level of power
consumption could be supplied by a high end mobile phone
battery (Nokia BP-4L Battery) for a period of 34hrs of
continuous use. In reality however, the fan would only be in
use while the device is in high processing mode so the actual
lifetime is likely to be limited by phone functionality rather
than by cooling solution.

In Fig. 12 and Fig. 13, the heat transfer coefficient is presented
for combined cooling and forced convection cooling. Similar
to the thermal resistance measurements, this data is based on
the average heat transfer coefficient of the heat sink. The data
presented in Fig. 12 is that of the total heat transfer coefficient
and it can be seen that the finless cases achieve more than
twice the average heat transfer coefficient compared to the
finned cases over the rotor speeds tested. This is highly
dependant on the convective surface area of both heat sinks as
given in Eq. (9). With reference to Fig. 10 and Fig. 11, which
show thermal resistance plotted against fan speed, it is
possible to conclude that in general over the range of speeds
examined, the finless design will dissipate a similar, and in
some cases favorable magnitude of heat per unit of
temperature compared to the finned design. The heat transfer
coefficient quantifies this relative to the convective surface
area of the heat sink. The finless heat sink considered here has
a convective surface area 44.7% less than the convective
surface area of the finned heat sink. It is therefore expected
that the heat transfer coefficient values will be greater for the
finless design as it dissipates a similar magnitude of heat using
almost half the convective surface area of the finned design.
This also applies to Fig. 13 for the forced convection heat
transfer coefficient data. Increasing rotational speed results in
an increasing difference in heat transfer coefficient between
heat sinks using diffusers and heat sinks without diffusers. At
8000RPM, the finned heat sink achieves an increase of
approximately 27% through the introduction of the selected
diffuser. For the same rotor speed, an increase of
approximately 15% is achieved for the finless heat sink using
the diffuser to align the flow. The effect of convective surface
area results in a heat transfer coefficient for the finned case
with diffuser to be 51% of the heat transfer coefficient for the
finless case with diffuser, considered for a rotor speed of
8000RPM.

0
10
20
30
40
50
60
70
80
90
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Rotor Speed (RPM)
h

(
W
/
m
2

o
C
)
Finned + diffuser
Finless + diffuser
Finned
Finless

Fig. 12. The total heat transfer coefficient for both finned and
finless heat sinks.

0
5
10
15
20
25
30
35
40
45
50
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Rotor Speed (RPM)
h
F
C

(
W
/
m
2

o
C
)
Finned + diffuser
Finless + diffuser
Finned
Finless

Fig. 13. The forced convection heat transfer coefficient for
both finned and finless heat sinks.
As seen in Fig. 9 the flow within the finned heat sink channels
is seen to fully develop which makes it possible to describe the
velocity profile across the entire channel cross section using
the solution provided by Purday [25]. This solution describes
the velocity distribution in rectangular channels of varying
aspect ratios and is adapted in Eq. (12) to describe the flow in
channels with an aspect ratio of , such as those in the heat
sink analysed herein. In this equation X and Y correspond
to half the channel width and half the profile height
respectively. While x and y correspond to the local
positions measured from the mid-width and mid-depth of the
channel respectively.

507
2 3
( , ) max
1 1
x y
x y
u U
X Y
| ||
| | | |
= |
| |
|
\ . \ .
\ .\
.7
|
|
|
.
(12)

Hence, the volumetric flow rate in each channel can be
calculated from Eq. (13).

( , ) max
2 4.7
4
3 3.7
Y X
x y
Y X
Q u dydx XY U

| || |
= =
| |
\ .\ .
) )
(13)

The maximum velocity measured using PIV within each of the
heat sink channels was substituted into Eq. (13) which yielded
the corresponding flow rates in each channel. These were
summed for all 6 channels to obtain the total volumetric flow
rate passing through the system. The results are presented in
Fig. 14 and compared to volumetric flow rates for the tests
carried out without a diffuser [15]. It is seen that the flow rate
in the finned heat sink with the diffuser attached at the
entrance is on average 15% higher that just the fan and heat
sink arrangement. However the increase in volumetric flow
rate at 8000RPM is lower compared to 3000RPM and
5500RPM. A possible explanation for this is a higher variation
in flow rate between the heat sink channels at 8000RPM due
to higher speeds affecting the flow distribution. The accuracy
of the velocity measurements are due to a number of
experimental factors and are estimated at 5%.
0.E+00
1.E-05
2.E-05
3.E-05
4.E-05
5.E-05
6.E-05
7.E-05
8.E-05
9.E-05
0 1500 3000 4500 6000 7500 9000
Rot at ional Speed (RPM)
T
o
t
a
l

V
o
l
u
m
e
t
r
i
c

F
l
o
w

R
a
t
e

(
m
3
/
s
)
Heat Sink +Diffuser
Heat Sink
Fig. 14. Graph depicting the volumetric flow rates through
heat sink measured using Eq. (12) along with the PIV data
presented in Fig. 9 and [13] over the three fan speed tested
CONCLUSIONS
A miniature low profile cooling solution suitable for
integration into a range of portable electronics devices has
been characterized through thermal resistance and velocity
field measurements.
This paper presents an analysis on fan exit angle with results
showing that a reduction in thermal resistances are obtained
for flow entering the heat sink aligned with the fins. Flow
alignment has been achieved through the introduction of a
diffuser at the entrance to the heat sink.
By accounting for fan exit angles, the thermal
resistance has been reduced by up to 23% for a
finned heat sink and 15% for a finless heat sink. In
order to achieve these reductions the footprint area
has been increased by 16% and 6% respectively.
A finless heat sink and diffuser outperforms the
conventional finned heat sink solution over a range
of fan speed tested.
The finned heat sink was designed using an
optimization criterion based on the fin spacing which
results in thermal boundary layers merging at the exit of
the heat sink. This was applied for a fan speed of
8000RPM.
Results presented show that this optimization
criterion was confirmed for miniature scale
solutions as the finned heat sink outperforms the
finless case at 8000RPM
Finally the results also provide a basis for designers of heats
to analyze fan exit angles and a move towards integrated fan
and heat sink solutions.
REFERENCES

[1] Vogel, M.R., Liquid Cooling Performance for 3-D
Multichip Module and Miniature Heat Sink, IEEE
Transactions on components, packaging and
manufacturing technology, Part A, Vol. 18 (1) (1995),
68-73.
[2] Launay, A., 2004, Experimental Study on Silicon
Micro-Heat Pipe Arrays, J. Applied Thermal
Engineering, Vol. 24 (2-3), 233-243.
[3] Tan, F.L., 2004, Cooling of mobile Electronic Devices
using Phase Change Materials, J. Applied Thermal
Engineering, Vol. 24 (2-3), 159-169.
[4] Walsh, E.J, Grimes, R. and Punch, J. 2007, "A cooling
device", WO2007057871
[5] Grimes, R., Walsh, E.J., Quin, D., and Davies, M.R.D.,
2005, Effect of Geometric Scaling on Aerodynamic
Performance, Journal of AIAA, Vol. 43 (11), 2293-
2298.
[6] R. Grimes, Walsh, P., Walsh, E., and Egan, V., 2007,
The Effects of Diameter and Rotational Speed on the
Aerodynamic Performance of Low Profile Miniature
Radial Flow Fans, Puebla, Mexico, ICNMM2007-
30185.
[7] K. Hanly, Grimes, R., Walsh, E., Rodgers, B., and
Punch, J., 2003, The Effect of Reynolds Number on the
Aerodynamic Performance of Micro Radial Flow Fans,
Proceedings of ASME Heat Transfer Conference, San
Francisco, California, USA, HT2005-72514.
[8] P. Walsh, V. Egan, R. Grimes, E.J. Walsh, 2007,
Scaling of Flow Characteristics and Power
Consumption with Profile Height for Miniature
508
Centrifugal Fans, Proceedings of the 5th International
Conference on Nanochannels, Microchannels and
Minichannels, ICNMM2007, Puebla, Mexico, 18th
20th June, ICNMM2007-30136.
[9] Grimes, R., Walsh, E.J., Walsh, P. and V. Egan, 2007,
The effects of diameter and rotational speed on the
aerodynamic performance of low profile radial
miniature flow fans, Proceedings of the 5th
International Conference on Nanochannels,
Microchannels and Minichannels, ICNMM2007,
Puebla, Mexico, 18th 20th June.
[10] S.W. Day, Lemire, P.P., Flack, R.D., and McDaniel,
J.C., 2003, Effect of Reynolds Number on Performance
of a small Centrifugal Pump, Proceedings of the 4th
ASME_JSME Joint Fluids Engineering Conference,
Honolulu, Hawaii, USA, FEDSM2003-45686.
[11] Chien, H., 2001, The Study of Micro-Fin Heat Sinks
for Electronic Cooling Applications, 17
th
IEEE SEMI-
THERM Symposium.
[12] J.G. Maveety and Jung, H.H., 2002 Heat Transfer
from Square Pin-Fin Heat Sinks using Air
Impingement Cooling, IEEE Transactions on
components, packaging and manufacturing technology
Vol. 25 (3), pp. 459-469.
[13] Walsh, E., Walsh, P., Grimes, R., Egan, V., 2007,
Thermal Management of Low Profile Applications,
Proceedings of the 5
th
International Conference on
Nanochannels, Microchannels and Minichannels, 18
th

20
th
June, Peubla, Mexico.
[14] Walsh, E.J., 2006, Low profile thermal management,
EPEC Seminar on High Temperature Electronics and
Thermal Management, 9
th
10
th
November, sud-west
park conference center, Nuremburg, Germany.
[15] Egan, V., Walsh, P., Walsh, E., Grimes, R., 2007, On
the Characterisation of Finned and Finless Heat Sinks
for Portable Electronics, 5
th
International Conference
on Nanochannels, Microchannels and Minichannels,
18
th
20
th
June, Peubla, Mexico.
[16] Egan V., Walsh P., Walsh E.J., Grimes R., 2007,
Characterisation of a Miniature Low Profile Cooling
Solution for Portable Electronics, Sixth International
Conference on Enhanced, Compact and Ultra-Compact
Heat Exchangers: Science, Engineering and
Technology, Potsdam, Germany, 16th 21st
September.
[17] F.P. Bleier, 1997, Fan Handbook: Selection,
Application and Design, McGraw-Hill.
[18] Bejan, A., Convective Heat Transfer, John Wiley &
Sons, Inc. (2004).
[19] G.N. Ellison, 1989, Thermal Computations for
Electronic Equipment, Krieger Publishing Company
[20] Incropera F.P., DeWitt D.P., Fundamentals of Heat and
Mass Transfer, John Wiley & Sond, Inc (1996)
[21] Moffat J.R., 1999. Uncertainty analysis. Electronics
Cooling. 5(2).
[22] JEM, Jem "Safety Data Sheet" Pro Haze Fluid. SDS
No. 112., (2005).
[23] M. Raffel, 1998, Particle Image Velocimetry: A
practical guide, Springer.
[24] TSI Inc., TSI Insight 3G Software Manual, (2006).
[25] Purday, H. F. P. 1949, An Introduction to the
Mechanics of Viscous Flow (Streamline Flow) Dover
Publications Inc., New York.

509