Académique Documents
Professionnel Documents
Culture Documents
IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGY, VOL. 33, NO. 1, MARCH 2010
Abstract This paper explores the potential of micro-pinfin heat sinks as an effective alternative to microchannel heat
sinks for dissipating high heat fluxes from small areas. The
overall goal is to compare microchannel and micro-pin-fin heat
sinks based on three metrics: thermal performance, hydraulic
performance, and cost of manufacturing. The channels and pins
of the microchannel and micro-pin-fin heat sinks, respectively,
have a width of 200 m and a height of 670 m. A comparison of
the thermal-hydraulic performance shows that the micro-pin-fin
heat sink has a lower convection thermal resistance at liquid flow
rates above approximately 60 g/min, though this is accompanied
by a higher pressure drop. Methods that could feasibly fabricate
the two heat sinks are reviewed, with references outlining current
capabilities and limitations. A case study on micro-end-milling
of the heat sinks is included. This paper includes equations that
separate the fabrication cost into the independent variables that
contribute to material cost, machining cost, and machining time.
It is concluded that, with micro-end-milling, the machining time
is the primary factor in determining cost, and, due to the
additional machining time required, the micro-pin-fin heat sinks
are roughly three times as expensive to make. It is also noted
that improvements in the fabrication process, including spindle
speed and tool coatings, will decrease the difference in cost.
Index Terms Micro heat sink, micro-manufacturing, micromachining, pin-fin heat sink.
N OMENCLATURE
nf
tchange
tcleaning
tmachining
tsetup
ttoolchange
dstraight
dstgpin
Manuscript received November 25, 2008; revised February 19, 2009. First
version published October 13, 2009; current version published March 10,
2010. This work was supported by the National Science Foundation, Grant
CBET-0729693 at the University of Wisconsin-Madison, and Grant CBET0730315 at the University of Hawaii at Manoa. Recommended for publication
by Associate Editor A. Bhattacharya upon evaluation of the reviewers
comments.
Y. Jeon was with the Department of Mechanical Engineering, University of
Wisconsin-Madison, Madison, WI 53706 USA. He is now with the Hyundai
Motors, Seoul, South Korea (e-mail: yjeon@hyundai-motor.com).
K. T. Turner, B. A. Jasperson, and F. E. Pfefferkorn are with the Department of Mechanical Engineering, University of Wisconsin-Madison, Madison, WI 53706 USA (e-mail: kturner@engr.wisc.edu; bajasperson@wisc.edu;
pfefferk@engr.wisc.edu).
W. Qu is with the Department of Mechanical Engineering, University of
Hawaii at Manoa, Honolulu, HI 96822 USA (e-mail: qu@hawaii.edu).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TCAPT.2009.2023980
Chipload [mm].
Operator rate [$/hr].
Cost per tool [$].
Width of heat sink for case study [mm].
Length of heat sink for case study [mm].
Spindle speed [rpm].
Number of tools required for fabrication of one
heat sink.
Number of flutes.
Time to change one tool [min].
Time to clean up after machining [min].
Machining time [min].
Setup time before machining [min].
Total time to change tools [min].
Tool path to machine straight channel heat sink
[mm].
Tool path to machine staggered pin fin heat sink
[mm].
Manufacturing Variables
Ctotal
CT
CM
fr
t
tc
R
CT
w
l
Ns
Ntools
h
Hfin
L hs
P
Pdh
Pfh
qeff
Rconv
Tf
Tw
Whs
Wch
Wfin
Wt
Greek Symbols
149
SL = 400 m
Hfin = 670 m
Hfin = 670 m
Lfin = 200 m
Wfin = 200 m
Wch = 200 m
Wfin = 200 m
Flow
Flow
(a)
Fig. 1.
Average.
Liquid (water).
Inlet.
Microchannel heat sink.
Micro-pin-fin heat sink.
Outlet.
Wall.
I. I NTRODUCTION
(b)
Structure and dimension of (a) microchannel heat sink and (b) micro-pin-fin heat sink.
Subscripts
ave
f
in
mc
mpf
out
W
Wch = 200 m
150
IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGY, VOL. 33, NO. 1, MARCH 2010
TABLE I
P OTENTIAL M ANUFACTURING M ETHODS FOR M ICRO H EAT S INKS
Mfg Method
Can Mfg
Mass Production
Suitability*
Prototyping
Suitability*
Cost Comparison
of Designs
EDM
Wire
Plunge
Channel
Both
Poor
Poor
Average
Average
NA
Pin Channel
Etching
LIGA/Electroforming
Both
Good
Good
Pin Channel
Casting
Both
Very Good
Poor
Pin Channel
Extrusion
Channel
Very Good
Poor
NA
Machining
End Mill
Slot/Form Mill
Both
Channel
Average
Good
Very Good
Very Good
Both
Very Good
Poor
Pin Channel
Sintering
151
Wire Direction
Workpiece
Direction
(a)
(b)
(c)
(d)
(e)
(f)
(a)
(b)
Fig. 3. Schematic of LIGA process: (a) deposit a conductive seed layer, (b)
spin on a thick layer of photoresist, (c) expose photoresist to high-energy
X-rays through a mask, (d) develop photoresist removing X-ray exposed
material, (e) deposit metal into photoresist mold, and (f) dissolve photoresist
mold.
(c)
Fig. 2. Illustrations of three electrodischarge machining techniques: (a) wire
EDM, (b) die sinking, and (c) EDM milling.
152
IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGY, VOL. 33, NO. 1, MARCH 2010
Heating/Cooling Passages
(a)
(b)
(c)
Fig. 4. Schematic of die casting: (a) metal molds with runner and cooling passages, (b) molds pressed together with molten metal being inserted, and
(c) separation of molds and removal of part.
153
Billet
Die
(a)
Fig. 5.
Fig. 7.
(a)
(b)
(c)
(d)
Fig. 6. Schematic of micro powder injection molding, (a) inject metalbinder mix into mold, (b) heat to remove binder, (c) sinter metal powder, and
(d) remove part.
(b)
6) Machining: a) Slot/form milling: Slot milling is a potential method of manufacturing the straight channel heat sinks.
A single circular cutter with teeth on the outer portion of the
bit or a cluster of cutters [Fig. 7(a)] can be used. Slotting saws
as thin as 150 m are commercially available, sufficient for the
feature sizes on the micro heat sinks being compared in this
paper.
b) End milling: Micro-end-milling [Fig. 7(b)] refers to an
end-milling process that uses cutting tools between 5 and
1000 m in diameter to create microscale features on micro-,
meso-, and macroscale parts [51]. It is a direct method of
creating true 3-D shapes in myriad materials, frequently in
a single process step. The fact that the geometry of interest
is created by a part program that controls the movement
of the end mill makes this method flexible. Therefore, it is
clearly suited for prototyping metal heat sinks and low-volume
production.
To maintain the same cutting speed as the diameter of
an end mill decreases, the spindle speed must be increased
proportionally. For example, to achieve the recommended
cutting speed for wrought aluminum alloys being end-milled
with a tungsten carbide tool (3.15 m/s [52]) a 200 m-diameter
end mill requires a spindle speed of 300 000 rpm. Currently,
there are 200 000 rpm spindles commercially available, and
ongoing research aims to develop spindles that can achieve
more than 1 million rpm [53]. However, most micro-endmilling is done with spindles between 50 000 and 100 000 rpm,
because it is not yet known if the cutting speeds that decades
of empirical data have shown to work well at the macroscale
are optimal for micromachining [54], [55]. Micro-end-mills
remove small amounts of material with each rotation, thus
high-speed spindles do not need to be powerful with costs
ranging from approximately $5,000 for a 50 000 rpm air-drive
spindle (fixed rpm) to $25 000 for a 200 000 rpm electric-drive
spindle with variable rpm.
As will be shown in the case study, the cost of machining
a heat sink is inversely proportional to the time it takes to
machine a part (productivity), which is mainly a function of the
feed rate (mm/min) at which a micro-end-mill can be moved
through the material. The feedrate fr is the product of the
chipload tc number of flutes (cutting edges) n f , and spindle
speed Ns
f r = tc n f N s .
(1)
Hence, doubling the spindle speed or number of flutes
will double the feedrate and cut the time to machine a
154
IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGY, VOL. 33, NO. 1, MARCH 2010
Chip Load
[ft/tooth]
Cutter
Rotation
Material
Feed
Fig. 8.
(3)
the microchannel heat sink is determined using the heat transfer and pressure drop models summarized in Table II [56],
[57].
The performance of the two micro heat sink geometries
are compared assuming identical heat sink substrate material,
single-phase liquid coolant, overall dimensions, microscale
structure dimensions, and operating conditions. Fig. 1(a) illustrates the structure and key characteristic dimensions of the
microchannel heat sink along side those of the micro-pin-fin
heat sink. In particular, the (Wfin , Wch , Hfin ) combination for
the microchannel heat sink is chosen to be (200 m, 200 m,
670 m), which is the same as that for the micro-pin-fin heat
sink. Average convection thermal resistance for the microchannel heat sink is similarly evaluated from (2) using the heat
transfer models provided in Table II. Pressure drop across the
microchannel heat sink P is the sum of the pressure drop
across the upstream hydrodynamically developing entrance
region Pdh and the pressure drop across the downstream
fully developed region Pfh . Analytical models for evaluating
the two pressure drop components are provided in Table II.
Fig. 10(a) and (b) compare the average convection thermal
resistance for the micro-pin-fin heat sink and microchannel
heat sink for Tin = 30 and 60 C, respectively. The solid
line and dashed line in these figures are power-law curves
to best-fit the micro-pin-fin heat sink and microchannel heat
sink data, respectively, and are used to indicate the overall
data trend. It can be seen from Fig. 10(a) and (b) that Rconv
for the microchannel heat sink is fairly constant throughout the
total flow rate Wt range, while Rconv for the micro-pin-fin heat
sink is more sensitive to Wt and decreases significantly with
increasing Wt . In the low Wt range, Rconv for the micro-pin-fin
heat sink is higher than that for the microchannel heat sink,
but becomes lower at a higher Wt . The comparison indicates
a better micro-pin-fin heat sink thermal performance at an
elevated cooling water flow rate.
Fig. 11(a) and (b) compare the pressure drop across the
micro-pin-fin heat sink and microchannel heat sink for Tin =
30 and 60 C, respectively. It can be seen from Fig. 11(a)
and (b) that P in the micro-pin-fin heat sink is significantly
higher than that in the microchannel heat sink at all flow rates
tested.
IV. C ASE S TUDY: M ICRO -E ND -M ILLING
The authors have the most experience with micro-endmilling [Fig. 7(b)] and have used it to manufacture copper
micro heat sinks (Fig. 9). In this section, the cost of micro-endmilling two different heat sink geometries, pin-fin and straight
channel (Fig. 1), is compared. Only relative differences will
be highlighted since any productivity improvements that would
be applied to machining of one design would also be applied
to the other. The goal of this case study is to determine which
heat sink is more expensive to manufacture by micro-endmilling and why.
In order to compare the geometries directly, the same base
material (110 copper), and hence material cost and overall
heat sink geometry (width = 1.0 cm, length = 3.38 cm) are
assumed. Likewise, the pin width and gap (200 m) is consistent between heat sinks, and hence the same tool diameter
155
(a)
Fig. 9.
(b)
Photographs of (a) copper heat sink and (b) pin fin geometry created by micro-end-milling.
TABLE II
A NALYTICAL M ODELS FOR M ICROCHANNEL H EAT S INK [16], [17]
Heat transfer coefficient
For L 0.2 (thermally fully-developed flow),
k
w 0.14
h = N u3 d f
ST Wfin
Hfin
L = Red z Pr
h
f
For L 0.2 (thermally developing flow),
N u 3 k f w 0.14
h = N u 4 + 8.68(103 L )0.506 exp (9.9776 ln () 26.379) L
Nu
d
4
Pdh
2 f app,dh f u 2f L dh
Pdh =
; L dh = (0.06 + 0.07 0.04 2 )Rein dh
dh
+ 0.5
Re3.44
L
+
f
0.5 K () 4L +
fh
dh
dh
w 0.58
1 3.44 L +
f app,dh = Re
+
2
b
dh
+
1+C L dh
Pfh =
2 f fh w b
dh
f u f L fh
; L fh = L L dh
(4)
156
IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGY, VOL. 33, NO. 1, MARCH 2010
0.16
0.16
Micro-pin-fin heat sink
Microchannel heat sink
0.14
0.14
0.12
Rconv[C/W]
Rconv[C/W]
0.12
0.10
0.08
0.02
30
0.10
0.08
0.06
0.06
0.04
0.04
Water
Tin = 30 C
40
50
60
wt[g/min]
70
80
0.02
30
90
Water
Tin = 60 C
40
50
(a)
Fig. 10.
60
wt[g/min]
70
90
(b)
Comparison of micro-pin-fin heat sink and microchannel heat sink average convection thermal resistance for (a) Tin = 30 C and (b) Tin = 60 C.
0.12
0.12
0.10
0.10
0.08
0.08
P [bar]
P [bar]
0.06
0.06
0.04
0.04
0.02
0.02
Water
Tin = 60 C
Water
Tin = 30 C
0.00
30
40
50
60
wt[g/min]
70
80
90
(a)
Fig. 11.
80
0.00
30
40
50
60
wt[g/min]
70
80
90
(b)
Comparison of micro-pin-fin heat sink and microchannel heat sink pressure drop for (a) Tin = 30 C and (b) Tin = 60 C.
(5)
Since the cost per tool does not change between heat sinks, the
total tool cost only varies with the number of tools. The tool
life and tool geometry are assumed constant between heat
sinks, meaning that the number of tools varies solely with
the final part geometry of the heat sinks.
The material costs C M are dependent on the volume of
material required and the per unit cost. The heat sinks
have the same overall dimensions and are made out of the
same material, so C M does not vary between the two heat
sinks.
The final term in (4) is the cost of the processing time.
The cost rate R includes the capital cost of machinery, any
overhead and utilities required for operation, additional training required for machining/setup of process, labor, etc. These
factors do not change, regardless of fabrication geometry, and
as such can be excluded from the current comparison.
The total time t required to manufacture the heat sink is a
function of the machining time, the amount of time required
for tool changes, setup time, and cleaning time (6). The
amount of time spent machining is a function of the feedrate
and geometry of the final design. More complex designs
require more passes with the tool, resulting in longer machining time (7). Feedrate is a complex function of multiple parameters, including material properties, tool strength/geometry,
157
wfin
wfin
2(wfin)
2(wfin)
(a)
Fig. 12.
(b)
(8)
(9)
(c)
(d)
(e)
(f)
Fig. 13. Illustration of tool path for milling staggered pin heat sink: (a) first
pass, illustrating the effect of tool radius on the corners of the pins, (b) second
pass, which finishes the first column of pins and makes the first cut on the
second column, and (c) through (f) repeating the process to make multiple
columns of pins.
158
IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGY, VOL. 33, NO. 1, MARCH 2010
140
120
Staggered Pins
Straigth Pins
100
80
60
40
20
0
100
200
300
400
500
600
Pin/Wall Width, wfin [microns]
Fig. 14. Total machining distance (tool path as a function of pin/wall width
for a 1 cm 3.38 cm area).
159
160
IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGY, VOL. 33, NO. 1, MARCH 2010