Microelectronics Journal 43 (2012) 280–287
Contents lists available at SciVerse ScienceDirect
Microelectronics Journal journal homepage: ww w.elsevier.com/locate/me w.elsevier.com/locate/mejo jo
A study on the heat dissipation of high power multi-chip COB LEDs Hsueh-Han Wu a, , Kuan-Hong Lin b, Shun-Tian Lin a n
a b
Department of Mechanical Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan, ROC Department of Mechanical Engineering, Tungnan University, New Taipei City 222, Taiwan, ROC
a r t i c l e
i n f o
Article history: Received 28 June 2011 Received Received in revised form 30 December 2011 Accepted 13 January 2012 Available online 10 February 2012 Keywords: Heat dissipation Multi-chip Junction temperature Thermal resistance Computational fluid dynamics (CFD)
a b s t r a c t
In this study, the heat dissipation efficiencies of high power multi-chip COB (Chip-on-Board) LEDs with five differ different ent chip chip gaps gaps were compar compared ed by assessi assessing ng their their junctio junction n temper temperatur ature e (T j) and thermal thermal resistance resistance (Rth). Junction temperatures were measured using an IR camera and were also simulated by computational fluid dynamics (CFD) software. The effects of heat sinks with different surface areas, heat slugs made of different materials and different injection currents (different wattages) on high power LED junction temperatures are discussed. In addition, the optical characteristics of the LED, such as its lumens and luminous efficiency are evaluated. The experimental results show that a chip with a smaller gap has a higher junction temperature and more thermal resistance, and the junction temperature difference between the LEDs with the smallest and largest chip gaps is 3.12 1 C. Optical Optical performance performance analyses show that the LED with a larger chip gap has higher lumens and higher luminous efficiency. Thus, higher junction temperatures reduce the optical performance of high power LEDs. & 2012 Elsevier Ltd. All rights reserved.
1. Introduc Introduction tion
A light emitting diode (LED) is a semiconductor component composed of group II–VI or III–V elements. Semiconductors with different energy gaps emit visible light of different colors. LEDs have multiple advantages, including lower power consumption, highly directional light emission, fast response time, long lifetime and environmental protection. LEDs have come to be widely used as the LCD back light source, automotive automotive and general lighting [1– [1 –3]. Although these products are small, they require very high lumens, and the currently available single-chip LED module is not suitable for these types of applications [4] [4].. Relati Relative ve to multimulti-ch chip ip LED module modules, s, single single-c -chip hip LED LED module moduless have have more more inter interfac face e therma thermall resist resistanc ance, e, and and they they must must be placed placed on a metal structure for heat dissipation. The resulting LED products are thicker, more difficult to assemble and have a more complex circuit design. Currently, multi-chip LED modules [5] [5] and and high power COB modules [6] are are the the indu indust stry ry tren trends ds.. Thes These e two two type typess of modu module less have have much higher power than many single-chip LED modules, and their advantages advantages include simple circuit circuit design, a simple simple heat dissipation dissipation structure and greatly reduced thickness of the package (ultra-thin). However, However, multi-ch multi-chip ip LED modules and high power COB modules provide more lumens and wider light emitting areas while simultaneously producing more heat. The generation of extra heat results in many problems for LEDs, affecting their their lifetime, stability, wavelength
n
Corresponding author. Tel.: þ 886 2 27376464; fax: þ 886 2 27376460. 27376460. E-mail address:
[email protected] (H.-H. Wu). Wu) .
0026-2692/$0026-2692/$- see front matter & 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2012.01.007 doi:10.1016/j.mejo.2012.01.007
shift and luminous efficiency [7 [7– –10 10]. ]. Thus, heat management is one of the most important topics in high power LED research currently. LED heat management research includes three major areas: the package level, the board level and the system level. At the package level, Kim [11] Kim [11] used used Au/Sn eutectic bonding, Ag paste and solder paste to connect an LED chip to a heat slug, and the heat dissipation effect of the type of die attach was evaluated using a thermal transient tester (T3ster). The results demonstrated that when Au/Sn eutectic bonding is used as a die attach, it effectively reduces reduces thermal resistance resistance between the LED chip and the die attach (R j die attach ¼ 3.5 K/W). With solder paste, the resistance is 4.4–4.6 K/W, and with Ag paste, paste, the resistance resistance is R j die attach attach ¼ 4.4–4.6 R j die attach ¼ 11.5–14.2 K/W. Wierer [12] Wierer [12] used used a flip-chip method without a molding process, resulting in an LED package that has no wire bonding to block light emission, which improves lumens. Because the shortened current flow path reduces resistance, heat generation is also reduced. Packages prepared using this method dissipate dissipate heat efficient efficiently. ly. Alternati Alternatively, vely, other other research researchers ers have proposed the laser lift-off method and the wafer-level package (WLP) method for LED package preparation [13 [13– –15 15]. ]. At the board level, Yin [16] used the finite finite element element method method (FEM) and the electrical test method to evaluate the thermal performance of high power LED modules with three different heat slugs (Al, AlN and Al2O3), and the results show that the AlN heat slug is superior to the the othe otherr two becaus because e the the AlN substr substrat ate e has has bette betterr heat heat cond conduc uctiv tivity ity without a dielectric layer. In addition to the commonly used MCPCB, ceramic substrate [17], [17], there have also been studies on the heat dissipat dissipation ion of a direct direct bonding bonding copper (DBC) [18]. [18]. At the system level, Kim [19] Kim [19] used used a thermal transient tester (T3ster) to measure -
-
-
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
H.-H. Wu et al. / Microelectronics Journal 43 (2012) 280–287
chip chip temperat temperature ure.. The high power power LED module module under under study study was composed of six single chips with 2.5 cm gaps between each single LED, and a heat pipe was placed under the high power LED module for heat dissipation. The results showed that the LED chip temperature was about 100 C. If a heat pipe was not placed under the LED module, the chip temperature reached 150 C. In addition to a heat pipe, there have also been studies on heat dissipation using heat sink, sink, loop heat heat pipe and thermo thermoelec electric tric cooling cooling [20 [20– –22 22]. ]. Among Among these these,, the the packa package ge level level is closes closestt to the the heat heat sourc source; e; thus thus,, impr improvi oving ng heat dissipation problems at the package level is more efficient than that that at the the board board or system system level. level. Howe However ver,, resea researc rch h on heat heat managem management ent method methodss for high power LEDs LEDs with multi-chi multi-chip p COB modules is currently quite rare. This study used high power LEDs with five different chip gaps and a new COB package structure. The effect of different chip gaps on the heat heat dissip dissipati ation on of high high power power multi multi-ch -chip ip COB COB LEDs LEDs is discus discussed sed.. Heat Heat dissipation performance was evaluated using the junction temperature and thermal thermal resistance, resistance, and the optical optical characte characteristic risticss of the LEDs, such as lumens and luminous efficiency, were analyzed. 1
1
2. Experiment Experimental al procedures procedures
A high high power power multi multi-c -chip hip COB LED structu structure re is shown shown in Fig. 1(a). 1(a). This LED uses a 33 mm ( L) 33 mm (W ) 2.1 mm (H ) aluminum heat slug. Above the aluminum heat slug, there is a
281
dielectric layer and a solder mask structure. At the center of the aluminu aluminum m heat slug, there there is a circular circular concave concave reflector reflector.. The reflector was processed using a CNC milling machine and was machine glazed; the GaN-based LED chips was placed inside the reflector reflector (45 45 mil2, l p ¼ 455–460 455–460 nm; EPILED EPILED Co. Ltd. comcommercial chip). The chips were arranged as a 3 3 array (9 LED chips connected connected in series), and Ag paste was used as the die attach between the LED chips and the aluminum heat slug. Gold wires were used to bind the chip electrode and the aluminum heat slug electrod electrode. e. The supplied supplied electric electrical al power power of the LED is 9.92 W (350 mA). An exploded view of the high power LED locked in the aluminum heat sink ( Ø 80 mm 30 mm (H)) is shown in Fig. 1(b). 1(b). Thermal grease was applied evenly between the LED and the heat sink to reduce air thermal resistance. Fig. 2(a)–(f) 2(a)–(f) shows diagrams of five five diff differ eren entt chip chip gaps gaps of 0.5, 0.5, 1.0, 1.0, 1.5, 1.5, 2.0 2.0 and and 2.5 2.5 mm, mm, respectively, and they are referred to as A-, B-, C-, D- and E-type. Junction temperature temperature measurements measurements were recorded using an infrared infrared thermal thermal image camera camera (TVS-500EX (TVS-500EX,, NEC-SanNEC-San-ei ei Co., Japan), and computational computational fluid dynamics software (CFdesign, version 10.0) was used for simulation analysis. The chip numbering starts starts with with the the first first chip chip at the upper upper left corner, corner, and it continues from left to right and then down for nine chips in total (nos. 1–9). Among these nine chips, the #5 chip is located at the center. The optical characteristics measured include high power LED lumens and luminous efficiency; the lumens and luminous efficienc efficiency y measurem measurements ents were recorded recorded with an integrati integrating ng
Fig. 1. A high power multi-chip multi-chip COB LED (a) compound compound structure, structure, and (b) exploded exploded view diagram diagram of an LED locked in an aluminum aluminum heat sink.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
282
H.-H. Wu et al. / Microelectronics Journal 43 (2012) 280–287
sphere (T68X, d ¼ 30 cm, CYT Co., Ltd.). All of the high power LEDs with five different chip gaps used a DC power supply (SPD-360, TECPEL Co., Ltd.) and were driven under constant current mode.
Table 1 The emissivities for the materials.
3. IR camera camera measurem measurement ent
The junction temperatures ( T j) of chips were recorded using an IR camera camera (TVS-5 (TVS-500E 00EX, X, NEC-S NEC-San an-ei -ei Co., Co., Japan) Japan).. The The therm thermal al sensit sensitivi ivity ty of the the IR camera camera was higher higher than 0.05 0.05 C and and had had a spatia spatiall resolu resolutio tion n of 1.07 1.07 mrad; mrad; the IR spectr spectral al range range was 8–14 mm, which was beyond the sapphire IR transmission transmission window. window. Fig. 3 shows 3 shows the temperature measurement experimental setup. In this this study, study, the whole whole radiant radiant energy energy receiv received ed by the the IR camera included the reflected radiation of atmosphere ( ea) and the emitted radiation radiation of device device ( eo). The transmit transmitted ted radiation radiation was negligible because the sapphire substrate absorbed the longwavelength IR band. The emissivity of the LED chip was determined by the following steps:
Materials
Emissivity (e)
GaN ( LED chip epitaxia epitaxiall layer ) Au ( LED LED chip chip elec electr trod ode e) Sapphire Sapphire ( LED chip chip substrate substrate ) Al ( Heat slug )
0.50 0.02 .02 0.89 0.05
1
Step 1: The unpowered high power multi-chip COB LED was placed onto a heater with a temperature control machine, and the module was heated until it reached a set temperature. Step 2: A thermocouple was attached to the Al heat slug and record recorded ed the temper temperatu ature re value, value, which which was called called ‘‘true ‘‘true temperature.’’ In this step, we cannot directly attach a thermocouple to an LED chip surface because it affects the temperature of the LED at the point of contact. Step 3: The radiant energy of the LED chip was measured by the IR camera, and the reflected radiation of atmospheric ( ea) was isolated and nullified. Step 4: The emissivity of every pixel was computed, and we used this emissivity matrix to calculate the true temperature at every pixel. The emissivities for the materials used in the experiment are listed in Table in Table 1. 1.
4. Computatio Computational nal fluid dynamics dynamics (CFD) method method
A high high power power multi-ch multi-chip ip COB COB LED model model was built built using using CFdesign CFdesign 10.0 software. software. This software software was used for computer computer calcu calculat lation ions, s, image image genera generatio tion n and for analys analyses es of physi physica call phenomena such as fluid flow and heat dissipation. The analytical proced procedure ure follow followed ed the the Navier Navier–St –Stoke okess equati equation on;; the mass, mass, momentum and heat conservation equations [23] [23] are are as follows:
r @
@t
þ nUr r ¼ rr Un
ð1Þ
Table 2 Physical property of high power LEDs package materials.
Component material
Thermal conductivity (W/mK)
Mass density (kg/m3)
Specific heat (J/kg K)
Al heat slug Dielectric layer Solder mask Ag paste GaN-base chip Silicone ring Thermal grease Al heat sink
204 0.35 0.25 8 130 0.17 3.6 204
2707 19 9 0 432 2300 61 5 0 9 80 1180 2707
896 92 0 8 37 6 71 4 17 1173 1044 8 96
n
r
@
@t
rc p
þ nUr n ¼ r p þ r g
@T @t
2 r ðmr nÞ þ 2r ðmS Þ 3
U
2 3
ð2Þ
U
T ¼ r UðK r T ÞÞ r T þ nUr T mðr UnÞ2 þ 2ðmS : S Þ þ bT U
D p Dt
ð3Þ
where r is the density, t the time, v the velocity vector, p the pressure, g the gravity, m the viscosity, S the strain rate tensor, k the thermal conductivity, conductivity, T the the absolute temperature, temperature, and b is the expansiveness. In this study, a high power multi-chip COB LED model was built upon a real high power LED consisting of Al heat slug, Ag paste, paste, GaN-based GaN-based LED chips, chips, silicone silicone ring, thermal thermal grease grease and alumin aluminum um heat heat sink. sink. A solder solder mask and and a dielec dielectri tricc layer layer are located above the aluminum heat slug. Table 2 Physical 2 Physical property of high power LEDs package materials. The high-power multi-chip COB LED model assumes perfect bonding of every component, and the GaN-based LED chip epitaxial structure is simplified. In additi addition on,, the effect effect of the the gold gold wire wire on the simulat simulation ion is not not considered, and 80% of the input power is converted to heat that is distri distribu buted ted evenly evenly among among the nine nine LED chips chips.. The initia initiall temperature of the environment and the heat sink base are set at 25 C, and only heat dissipation under natural convection is considered. 1
5. Results Results and discus discussion sion
The junction temperatures ( T j) of high power LEDs with A-, Band C-type C-type package package structur structures es operatin operating g in constant constant current mode mode at 350 mA were were measur measured ed using using an IR camera camera,, and and the results are shown in Fig. in Fig. 4(a)–(c). 4(a)–(c). The IR camera revealed that the T j of the A-type structure (with a smaller chip gap) was higher
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
283
H.-H. Wu et al. / Microelectronics Journal 43 (2012) 280–287
Fig. 4. Junction temperature ( T j) measurements for high power LEDs with different package structures and different chip gaps operating in constant current mode at 350 mA obtained using an IR camera: (a) A-type (0.5 mm), (b) B-type (1.0 mm) and (c) C-type (1.5 mm).
Fig. 5. Junction temperature ( T j) measurements for chips #1–9 in a high power LED with five different package structures at 350 mA in constant current mode.
package structure has a small or large chip gap. This result is obtained because the heat generated by the surrounding chips during LED operation affects the central chip, resulting in apparent heat accumulation at the central chip and higher temperature. In this study, the T j j of the high power LED chip described below is that of chip 5, located at the center of the package structure. The experimental results show that the T j of an A-type chip at the center of the structure is 84.32 C and that the T j j of an E-type chip at the center of the structure is 81.20 C; thus, the difference in T j j between between the central central chips of the package structures structures with the smallest and largest chip gaps is 3.12 C. The difference between the highest and lowest T j j values measured in the A-type structure is 0.69 0.69 C, where whereas as this this differ differen ence ce in the the E-typ E-type e struc structu ture re is 0.39 C, indicating that a larger chip gap structure reduces heat 1
1
1
1
1
The high-powered LEDs operated at constant current mode of 350 mA, the measured measured voltage was 28.342 28.342 V, and the supplied supplied electrical power of the LED was 9.92 W. The high-powered LEDs were supplied supplied with a constant constant power; power; a portion portion of the power transformed into light, and the remaining portion triggered heat. There Therefor fore, e, the heati heating ng power power must must be confi confirme rmed d before before CFD simula simulatio tion. n. In this this study, study, a test test point point temper temperatu ature re was first first obtained, and the same condition model was subsequently constructed by numerical simulation. Several heating powers were set up, up, and and the corre correspo spondi nding ng temper temperatu ature re was obtain obtained ed by nume numeric rical al calcu calculat lation ion.. Finall Finally, y, the heatin heating g watt watt of the highhighpowered powered LEDs was confirme confirmed d by comparin comparing g the temperatu temperature re acquired using two distinct methods. Fig. 6(a) 6(a) shows the position of a J-type thermal couple at heat slug slug compo componen nentt of LEDs LEDs for temper temperatu ature re measur measureme ement nt.. The The posi positi tion on of the the test test poin pointt was was the the same same as that that in the the CFD CFD simulation. simulation. Fig. 6(b) 6(b) shows the comparison comparison of the temperature temperature of test points when various electrical energies transform the heat energ energy y ratio ratio by measur measureme ement nt and and simula simulatio tion n metho methods. ds. The The results indicated that, when the electrical energy transforms the heat energy ratio by 80%, these two methods are consistent, with a difference of only 1.32 C. Therefore, the transformation of the electr electrica icall energ energy y into into heat heat energy energy was appro approxim ximate ately ly 80% at heating power of 7.94 W. To compare the effects of different chip gaps on heat dissipation by high power LEDs, this study also used CFD software for thermal simulation. Temperature distribution simulation results for the heat generated by nine LED chips and dissipated via an aluminum slug are shown in Fig. 7(a)–(d) 7(a)–(d) for structures A, B, D and E, respectively. The simulation results show that the T j j at the center of the A-type structure is 84.09 C, whereas the T j at the 1
1
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
284
H.-H. Wu et al. / Microelectronics Journal 43 (2012) 280–287
Fig. 6. (a) The position of the test point by thermal couple measurement and CFD simulation, (b) a comparison of test point temperatures when various electrical energies transform the heat energy ratio by measurement and simulation methods.
Fig. 7. Temperature distributions simulated by CFD software for high power LEDs with package structures of (a) A-type (0.5 mm), (b) B-type (1.0 mm), (c) D-type (2.0 mm) and (d) E-type (2.5 mm) at a total heating power of 7.94 W.
cent center er of the the LED. LED. Nota Notabl bly, y, when when the the GaNGaN-ba base sed d chip chip gap gap increa increases ses from from 0.5 mm to 2.0 mm (4 gap differenc difference), e), chipchipgenerated heat is dissipated effectively. In addition, a comparison of the the meas measur ured ed resu result ltss with with thos those e of the the heat heat simu simula lati tion on demonstrates that these two methods lead to T j values for high power LEDs that are very similar. The difference of measure value and and simu simula lati tion on resu result lt was was less less than than 2%. 2%. Thus Thus,, if a ther therma mall
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
285
H.-H. Wu et al. / Microelectronics Journal 43 (2012) 280–287
LEDs. LEDs. The thermal resistance resistance of a semicondu semiconducto ctorr componen componentt is defined by the JEDEC (Joint Electron Devices Engineering Council) 51-1 standard as follows [25 follows [25– –26 26]: ]: Rth j ,
x
-
¼
ð4Þ
1
-
1
where Rth,j x is the thermal resistance between device junction and the specific specific environm environment, ent, T j the junction junction temperatu temperature re of device in a steady state condition, T x the reference temperature for the specific environment, and P H the heating power (voltage current electrical energy transforms heat energy ratio). Substitution of the previously measured T j into Eq. (4) yields the thermal resistances (Rth,j a) of the high power LEDs with five different package structures at the temperature of 25 C. These resistances are 7.47 C/W (A-type), 7.29 C/W (B-type), 7.17 C/W (C-type), 7.10 C/W (D-type) and 7.08 C/W (E-type). Thus, high power LEDs with D- and E-type package structures have lower thermal resistance. This This study study employ employed ed CFD softwa software re to evalua evaluate te the the heat heat dissipation capabilities of heat sink of four different thicknesses used to dissipate heat from high power LED modules. The surface area areass of the the heat heat sink sink are are 27,7 27,766 66 mm2 (thickne (thickness ss 15 mm), 2 48,660 mm (thickness 30 mm), 67,660 mm2 (thickness 45 mm) and 87,367 mm2 (thickness 60 mm). Fig. mm). Fig. 9(a) 9(a) and (b) shows that as the surface area of the heat sink increases, the T j of the high power power LED decrea decreases ses linear linearly, ly, and and the the same same trend trend applie appliess to thermal resistance. The T j j of an A-type sample with a 15 mm heat sink is 87.94 C. When the heat sink thickness increases to 60 mm, T j decrea decreases ses to 75.09 75.09 C. The The differ differenc ence e betwee between n these these two measurements is 12.85 C, corresponding to a thermal resistance (Rth,j a) difference of 20.38%. By comparison, the thermal resistance (Rth,j a) difference between the 15 mm and 45 mm heat sink is 14.49%. CFD software simulation-derived simulation-derived temperature distributions for high power LEDs with an A-type package structure locked in heat sink of different thicknesses (different surface areas) are shown in Fig. 10(a)–(d) 10(a)–(d) for the total heating power of 7.94 W. The results demonstr demonstrate ate that increasing increasing the heat sink thicknes thicknesss effectivel effectively y reduces LED T j. Currently, LED heat management at the system level includes the commonly used heat sink as well as heat pipes, loop heat pipes and thermoelectric cooling [19 [19,,21 21,,22 22]. ]. Fig. 11(a) 11(a) shows the effects of using four different materials for the heat slug as simulated by CFD software on the heat dissipation of high power LEDs with five different package structures. The simulation results indicate that the T j of a high power LED with an Al2O3 heat slug package structure is higher than that of LED LED packag package e struc structu tures res using using the other other three three materi materials als.. The The copper heat slug package structure has the lowest T j, followed -
-
1
1
1
1
1
1
1
1
-
1
1
T j T x P H
-
by the aluminum heat slug package structure. The T j j of an A-type sample with an Al 2O3 heat slug is 91.45 C, and the T j j of an E-type samp sample le with with an Al2O3 heat heat slug slug is 85.1 85.13 3 C. The The differ differenc ence e betwee between n these these two struc structur tures es is 6.32 6.32 C, corre correspo spond nding ing to a thermal resistance (Rth,j a) differenc difference e of 9.36%. 9.36%. Converse Conversely, ly, the T j of an A-type sample with a copper heat slug is 82.44 C, and the T j of an E-type sample with a copper heat slug is 80.27 C. The difference between these two structures is 2.17 C, corresponding corresponding to a thermal resistance ( Rth,j a) difference of 3.79%. On the basis of the package material material thermal thermal conductivit conductivity y coefficient, i.e., 46 W/mK for Al 2O3, 59 W/mK for Fe, 204 W/mK for for Al and and 380 380 W/mK W/mK for for Cu, Cu, the the copp copper er heat heat slug slug pack packag age e structure facilitates the rapid dissipation of the heat generated by high power LEDs and effectively effectively reduces the Joule Joule heating heating effe effect ct.. The The majo majori rity ty of the the heat heat gene genera rate ted d by an LED LED chip chip dissipates through the back of the chip by heat conduction, and the heat then dissipates into the air via the slug or heat sink through convection and radiation. Thus, the back of an LED chip should should have high heat dissipation dissipation capability, capability, or the chip will overheat easily. LED chips can tolerate temperatures up to about 130–140 C; therefore, a proper heat dissipation design is necessary to reduce LED T j j. The above analyses demonstrate that LED T j j is closely related to both package material and heat dissipation path, path, and and the heat heat dissip dissipati ation on capaci capacity ty is relate related d to both both the the geometrical shape and the thermal conductivity coefficient of the material. A higher thermal conductivity coefficient and a shorter heat dissipation path result in better heat dissipation. As shown in Table Table 3, the T j of high high power power LEDs with five different package structures always rises linearly with increasing heat heatin ing g powe power. r. At the the lowe lowerr heat heatin ing g powe powerr of 3.14 3.14 W, the the difference in T j j between A- and E-type samples is 1.19 C. When heating power increases to 7.94 W, the difference in T j between the A- and E-type samples increases to 3.18 C. The lumens and luminous efficiency of high power LEDs with five different package structures driven by150 400 mA in constant current mode are shown in Fig. 12(a) 12(a) and (b). This figure shows shows that that the the lumen lumenss rise rise linear linearly ly as the the drivi driving ng curre current nt increases. When the high power LED is driven by a fixed current of 150 mA, the lumens of the E-type structure are 6.16% higher than those of the A-type structure. When the high power LED is driven by a fixed current of 350 mA, the lumens of the E-type structure are 12.38% higher than those of the A-type structure. However, under higher operational currents, the high power LED emits more lumens, but the luminous efficiency is reduced as the driving current increases, as shown in Fig. in Fig. 12(b). 12(b). This effect arises beca becaus use e the the high high powe powerr LED LED has has a high higher er T j under under higher higher operation operational al currents, currents, and a higher higher T j results results in non-rad non-radiativ iative e recombination [27] [27],, which reduces the luminous efficiency.
1
1
1
-
1
1
1
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
286
H.-H. Wu et al. / Microelectronics Journal 43 (2012) 280–287
Fig. 10. Temperature distributions simulated by CFD software for high power LEDs with an A-type package structure locked in heat sink with four different thicknesses of (a) 15 mm, (b) 30 mm, (c) 45 mm and (d) 60 mm, at a total heating power of 7.94 W.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
The world’s largest digital library
Try Scribd FREE for 30 days to access over 125 million titles without ads or interruptions! Start Free Trial Cancel Anytime.
H.-H. Wu et al. / Microelectronics Journal 43 (2012) 280–287
287
Fig. 12. The (a) relative relative lumen and (b) relative luminous luminous efficiency efficiency of five different different high power LED modules driven by constant current current ranging from 150 to 400 mA.
Acknowledgment Acknowledgment
This work was supported by the National Science Council of Taiwan Taiwan under under contrac contractt no. NSC 98-2622-E 98-2622-E-011-011-002-C 002-CC2. C2. The autho authors rs greatl greatly y apprec appreciat iate e the the financ financial ial suppor supportt of Nation National al Science Council in this study. References [1] Ivan Ching-Cherng Sun, Shih-Hsun Moreno, Wei-Ting Chung, Chung, Chih-To Chien, Hsieh, Hsieh, Tsung-Hsu Tsung-Hsun n Yang, Brightness Brightness management in a direct direct LED backlight backlight for LCD TVs, J.Soc.Inf. Disp. 16 (2008) 519–526. [2] [2] Yan Yan Lai, Lai, Nico Nicola la´ s Cord Corder ero, o, Fran Frank k Bart Barthe hel, l, Fran Frank k Tebbe, Tebbe, J org ¨ Kuhn, Kuhn, ¨ Robert Apfelbeck, Dagmar Wurtenberger, Liquid cooling of bright LEDs for automotive applications, Appl. Therm. Eng. 29 (2009) 1239–1244. [3] Zhang Lei, Guo Xia, Liang Ting, Gu Xiaoling, Lin Qiao Ming, Shen Guangdi, Guangdi, Color rendering rendering and luminous luminous efficacy efficacy of trichromat trichromatic ic and tetrachromatic tetrachromatic LED-based LED-based white LEDs, Microelectro Microelectron. n. J. 38 (2007) 1–6. [4] Chun-Jen Chun-Jen Weng, Advanced thermal enhancement enhancement and management management of LED packages, Int. Commun.Heat Mass Transfer 36 (2009) 245–248. [5] Lan Kim, Kim, Woong Woong Joon Joon Hwang, Hwang, and Moo Whan Whan Shin, Shin, Therma Thermall resist resistanc ance e analysis of high power LEDs with multi-chip package, Proceedings – Electronic Components and Technology Conference 2006 (2006) pp. 1076–1081. [6] M.Y. Tsai, C.H. Chen, and C.S. Kang, Thermal measuremen measurements ts and analyses analyses of low-cost-p low-cost-power ower LED packages packages and their modules, Microelectron. Microelectron. Reliab., Reliab., (2011) doi:10.1016/j.microrel.2011.04.008, (2011) doi:10.1016/j.microrel.2011.04.008, in in press. [7] Tetsushi Tamura, Tatsumi Setomoto, Tsunemasa Tsunemasa Taguchi, Illumination characteristics of lighting array using 10 cd-class white LEDs under AC 100 V operation, J. Lumin. 87-89 (2000) 1180–1182. [8] N. Narendran, Y. Gu, J.P. Freyssinier, H. Yu, L. Deng, Solid-state lighting:failure analysis analysis of white LEDs, J. Cryst. Cryst. Growth 268 (2004) 449–456. [9] M. Arik, C. Becker, S. Weaver, Weaver, J. Petroski, Petroski, Thermal management management of LEDs:
[13] Cheng-Chung Chiang, Dong-Sing Wuu, Ray-Hua Horng, An ultrathin ( 100 thick) flexible flexible light plate fabricated fabricated using self-alignm self-alignment ent and lift-off lift-off mm thick) techniques, Appl. Phys. Lett. 91 (2007) 181108-1–181108-3. [14] Yu-Sheng Huang Chingfu Tsou, Silicon-based packaging platform for lightemitting emitting diode, IEEE Trans. Trans. Adv. Packag. 29 (2006) 607–614. [15] F.K. Yam, Z. Hassan, Innovative advances in LED technology, Microelectron. J. 36 (2005) 129–137. [16] Lianqiao Luqiao Yin, Weiqiao Yang, Yansheng Yang, Kejun Guo, Shuzhi Ma, Jianhua Zhang Li, Thermal design and analysis of multi-chip LED module with ceramic substrate, Solid-State Electron. 54 (2010) 1520–1524. [17] Sunho Lianqiao Yang, Woongjoon Jang, Moowhan Shin Hwang, Hwang, Thermal Thermal analysis of high power GaN-based LEDs with ceramic package, Thermochim. Acta 455 (2007) 95–99. [18] Renli Hong He, Deliu Fu, Xiufeng Wang, Min Jing Song, A new method for preparation preparation of direct bonding bonding copper substrate substrate on Al2O3, Mater. Mater. Lett. Lett. 61 (2007) 4131–4133. [19] Lan Kim, Jong Hwa Choi, Sun Ho Jang, Moo Whan Shin, Thermal analysis of LED array system with heat pipe, Thermochim. Thermochim. Acta 455 (2007) 21–25. [20] Xiaobing Luo, Wei Xiong, Ting Cheng, and Sheng Liu, Design and optimization of horizontally-lo horizontally-located cated plate fin heat sink for high power LED street street lamps, lamps, Proceedings – Electronic Components and Technology Conference (2009) pp. 854–859. [21] Xiang-You Lu, Tse-Chao Hua, Mei-Jing Mei-Jing Liu, Yuan-Xia Cheng, Thermal Thermal analysis of loop heat pipe used for high-power LED, Thermochim. Acta 493 (2009) 25–29. [22] Nan Wang, Chang-Hong Wang, Jun-Xi Jun-Xi Lei, Dong-Sheng Zhu, Numerical study study on thermal thermal management management of LED packaging packaging by using using thermoelectr thermoelectric ic cooling, cooling, Internation International al Conference Conference on Electronic Electronic Packaging Technology Technology and High Density Packaging (2009) 433–437. [23] Jamil Wakil, Thermal performance impacts of heat spreading lids on flip chip packages: packages: with and without heat sinks, sinks, Microelectro Microelectron. n. Reliab. 46 (2006) 380–385. [24] Adam Christensen, Samuel Graham, Graham, Thermal effects in packaging high power light emitting emitting diode arrays, arrays, Appl. Therm. Eng. 29 (2009) 364–371.