Thermal analysis of COB array soldered on heat sink

Thermal analysis of COB array soldered on heat sink

International Communications in Heat and Mass Transfer 59 (2014) 55–60 Contents lists available at ScienceDirect International Communications in Hea...

2MB Sizes 0 Downloads 4 Views

International Communications in Heat and Mass Transfer 59 (2014) 55–60

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer journal homepage: www.elsevier.com/locate/ichmt

Thermal analysis of COB array soldered on heat sink☆ Fan He, Qinghua Chen ⁎, Juanfang Liu, Jiao Liu Key Laboratory of Low-Grade Energy Utilization Technologies and Systems of Ministry of Education, Chongqing University, Chongqing 400030, China College of Power Engineering, Chongqing University, Chongqing 400030, China

a r t i c l e

i n f o

Available online 23 October 2014 Keywords: High-power LED Cold-spray COB Heat dissipation FEM

a b s t r a c t In this paper, the finite element method (FEM) and experiment were adopted to evaluate the thermal performance of a multi-chip COB LED lamp based on the cold spray technology. The results show that the junction temperature can be limited under 110 °C. Compared to an aluminum-substrate LED module pressed on the heat sink, the soldering structure of a copper-substrate LED can reduce the junction temperature. The junction temperature of the LED module soldered on the heat sink is only 97.3 °C, while one of the pressing structures is 103.5 °C. It is further found that the chip gaps and the thickness of the copper-circuit layer have significant effect on the heat spread. Increasing the chip gaps and the thickness of the copper-circuit layer can effectively reduce the junction temperature and improve the LED lamp's thermal performance. © 2014 Elsevier Ltd. All rights reserved.

1. Introduction LED (Light Emitting Diode) is one kind of solid light which is made from semiconductor materials. It is widely used for back light, signal light and general light because of its long lifetime, energy saving and environmental protection advantages [1–3]. However, as to the high power LED, only 20 to 30% of input power is transferred to light while the rest are all converted to heat. The heat flux is up to 100 W/cm2 in a small chip area [4]. Without effective heat dissipation measurement, it will lead the LED junction temperature to rise up. This will make light extraction efficiency lower down, life time get shortened or ever the LED completely useless [5]. Because temperature has a direct impact on the performance of electronic, it is important to keep the peak temperature within acceptable temperature levels. To coper with this issue, there are several methods proposed recently [6–11]. Some have recognized the opportunity to improve performance by changing the distribution of discrete heat sources, such as refs [6,7]. Generally, single-chip LED module's power is 1–4 W. In order to obtain sufficient lumens to replace traditional lightings, the array of LED is essential and becomes the industry trends. The chips of LED array are the heat sources, and the distribution of chips is analyzed to improve the thermal performance in the paper. Currently, there are two ways to build up the LED array. One is to arrange surface mount type (SMT) high power LED packages on printed circuit board (PCB) [2,12] and the other is to arrange chips directly on PCB, which is called chip-on-board (COB) array [13]. Compared to the SMT high power LED array, COB array has many advantages on cost packaging densities and thermal management. Thermal management

of LED array includes three areas: the package level, the board level and the system level [14]. At the package level, Yung [15] proposed a new placement configuration for the LED array which can lower down the LED temperature. Horng [16] and Lion [17] improve the thermal performance by changing the die bonding material. The materials of substrates [18,19] and heat sinks [20] can also be changed to improve the thermal management at the board and system levels. However, the interface material between the LED array and heat sink is rarely studied. COB packages have two types: ceramic substrate COB package and metal substrate COB package [21]. Metal substrate of COB structure, as shown in Fig. 1. It is composed of copper layer, dielectric layer and base (metal Core). Hereinto, the copper layer is also called as a copper-circuit layer, it's used for the chip circuit connection. To avoid copper circuit conduction with metal layer, insulation must be added between these two layers. In most applications [14], the metal base is attached to the heat sink with thermal grease and screws. In this way, the distribution of stresses is nonuniform which leads to crack between chip and sink [22]. Moreover, thermal grease is easy to appear the aging problem at the high temperature. The high power COB array can produce more lumens while it generates more extra heat. To overcome these problem, cold spray—a new technology which can improve the thermal conductivity of interface materials is applied to connect the heat sink and COB LED array. The heat dissipation capability of this new structure LED lamp was analyzed and optimized in the paper. 2. Simulation and experiment 2.1. The cold spray technology

☆ Communicated by W.J. Minkowycz. ⁎ Corresponding author. E-mail address: [email protected] (Q. Chen).

http://dx.doi.org/10.1016/j.icheatmasstransfer.2014.10.021 0735-1933/© 2014 Elsevier Ltd. All rights reserved.

The principle of cold spray is shown in Fig. 2. In the cold spray process, a gas is accelerated to the supersonic speed in a Laval-type nozzle.

56

F. He et al. / International Communications in Heat and Mass Transfer 59 (2014) 55–60

Copper Dielectric

Circuit Layer

Layer Base (Metal Core) Fig. 1. Schematic of a metal substrate.

The coating material is injected into the gas stream in powder form at the inlet of the nozzle, accelerated by the gas in the nozzle, and propelled toward the substrate to be coated. The metal base of a COB LED module can be either aluminum or copper. Most of heat sinks for the present LED are aluminum alloy. Oxide layer with a high melting point is easily formed on the surface of the aluminum or aluminum alloy at the high temperature, so it is hard to solder. However, if there is a copper coating on the heat sink, the copper base LED module can be easily soldered on the heat sink. With the help of the cold spray technology, the soldering to connect the heat sink and the LED module can be achieved. Fig. 3. The soldering structure of COB arrays.

2.2. The model of a soldering structure LED lamp Heat dissipation on a COB LED array by using the cold spray technology is studied by a three-dimensional finite element simulation with ANSYS 14.5. A LED lamp model is set up, as shown in Fig. 3. It consists of heat sink, copper coat, solder layers, and copper base COB array from bottom to top. This experiment applies a simple heat sink (as illustrated above). The material of the heat sink material is aluminum alloy, thickness: 1.5 mm, inner diameter: 90 mm, height: 30 mm. In order to improve the convective heat transfer, 30 small holes around the heat sink are processed and their diameter is 3 mm. To realize the soldering between the copper base and heat sink, a spray copper coat with 0.5 mm thickness is achieved on the surface of the heat sink by cold spray technology. The thickness between the dielectric layer and the copper-circuit layer is 0.1 mm, and the diameter is 10 mm. There are 3 ∗ 3 array gallium nitride semiconductors with 1 mm ∗ 1 mm ∗ 0.2 mm on the copper substrate. The gap of the COB array is 0.3 mm. The model is conducted to simulate the temperature field of the LED lamp. Most heat generated by chips dissipates from the substrate and the heat sink to the environment, a steady-state thermal conductivity equation is defined as follows:       ∂ ∂T ∂ ∂T ∂ ∂T þ þ þQ ¼0 kx ky kz ∂x ∂x ∂y ∂y ∂z ∂z

ð1Þ

where T and Q represent temperature and heat flow from the chips and kx, ky, and kz refer to thermal conductivity of material on X, Y, and Z directions, the material's thermal conductivity is isotropic in this article, the discrete values are listed in Table 1. Heat from the heat sink diffuses to the environment by natural convection. Ignoring heat passed through lens and sealing material from chips [9], the heat flow between heat sink and air can be calculated by: q ¼ h  ðts –ta Þ:

ð2Þ

In this formula, ts and ta refer to surface temperature and environment temperature, respectively. In the simulation, natural convection is the only considered heat transfer around the heat sink and has a uniform convection heat transfer coefficient, h = 10 W/(m2 · K),at the ambient temperature of 20 °C. The total input power to chips is 10 W. Without fluorescent layer, luminous efficacy is 25%. Although the accuracy of the solution can be improved with a fine mesh, using a large number of elements is limited by the computer memory and simulation time. After an adaptive meshing, a local mesh refinement is adopted. After these works, the simulation result can be got as shown in Fig. 4, in which the hot spot is in the center of the chips, which is 97.298 °C. This temperature is lower than the junction temperature's limitation 110 °C.

heat sink

Cu powder

gas

Fig. 2. Schematic of the cold spray technology.

F. He et al. / International Communications in Heat and Mass Transfer 59 (2014) 55–60

57

Table 1 Thermal conductivity of the materials. Material

Aluminum Spray Cu Solder Cu

Thermal conductivity 150 (W · m−1 · K−1)

160

40

Dielectric Glue GaN

398 1.1

3

160

Fig. 6. Testing points of T-type thermocouple.

Fig. 4. Temperature distribution of a Cu substrate COB array with soldering structure.

by repeating test and taking average. Comparison of the experimental results and simulated data is listed in Table 2, in which the point 1 is the junction temperature. Table 2 indicated that the relative errors between the test and simulation (the finite element method) results are in 5%. Therefore 10 W copper substrate COB array connected with a simple heat sink by the solder technology can control the chip junction temperature below 110 °C, and the LED lamp can satisfy the requirement of heat dissipation. By making comparison with the test and simulation results, the simulation results can be considered reliable, and the finite element method can be further applied to optimize heat dissipation of the LED chips.

2.3. Experiment on the LED lamp 3. Result analysis and heat dissipation optimization To verify the actual heat dissipation capability and the simulation accuracy of this LED lamp, the heat dissipation experiment is carried out, as shown in Fig. 5a. The environment temperature is set up to be at 20 °C. We use direct current to supply power, T type thermocouple to test temperature, and data acquisition instrument to collect the temperature signal. Fig. 5b is the heat sink of the LED lamp. From Fig. 2, we know Cu powers can be accelerated to the supersonic speed and impact onto the aluminum-alloy heat sink to form the spray copper coat, as shown in the yellow part of Fig. 5b. Then soldering the copper substrate COB array on the heat sink can form the LED lamp (see Fig. 5a) The measuring points of the temperature are located at different positions and illustrated in Fig. 6. In order to get temperature at the steady state, the test time interval is 2 h. We minimize the measurement errors

3.1. Different connection ways Fig. 7a shows the aluminum-substrate pressing structure. Therein, an aluminum substrate COB array is pressed on the aluminum alloy heat sink by thermal glue, whose thermal coefficient is 3 W/(m · k), and the screw part is ignored in the thermal analysis. By adopting the same boundary conditions and the meshing method as the copper substrate LED lamp, simulation result is shown in Fig. 7b. On the same condition, the junction temperature of the coppersubstrate LED lamp with a soldering structure is only 97.3 °C, while one of the aluminum-substrate LED lamp is 103.5 °C. Because the thermal conductivity of copper and solder in the copper-substrate LED lamp is significantly larger than the thermal conductivity of aluminum and

Fig. 5. a. Schematic of the experimental layout. b. Heat sink with a thin copper coat.

58

F. He et al. / International Communications in Heat and Mass Transfer 59 (2014) 55–60

3.3. Different the copper-coat thickness

Table 2 Comparison of experimental data and simulation results. Texting point

1

2

3

4

5

6

Simulation result (°C) Experimental result (°C) Relative errors

97.3 95.5 1.9%

61.5 60.5 1.6%

52.9 52.0 1.7%

51.3 50.2 2.2%

43.9 42.6 3.1%

42.4 41.0 3.4%

glue in the aluminum-substrate LED Lamp. The conduction resistance from chips to the environment reduces, thereby improving the heatdissipation capability. However, the junction temperature falls down to 97 °C is in situation that environment temperature is only 20 °C. So in order to enlarge the range of lamp application, we should do further optimization on heat dissipation.

3.2. Different the chip gaps In multi-chip COB LEDs, changing the chip gap will impact evidently on the junction temperature. The chips are arranged in 3 ∗ 3 array with 0.3 mm pitch. When the pitch taking the following value: 0.1, 0.2, 0.4, 0.5, 1.0, and 1.5 mm, the simulation results are plotted in Fig. 8 and listed in Table 3. It is manifested from Table 3 that the hot spots are all in the center of the chip. The bigger chip gap it gets, the more uniform circuit layer's temperature it is and the lower chip's junction temperature it will be. This conclusion is as same as Christensen's heat dissipation research [7], which is that increasing the LED's pitches can bring the temperature down while other conditions are constant. In addition, we can analyze the equivalent model of thermal resistance according to Fig. 9. Singlechip COB LED module thermal resistance consists of chip resistance, glue resistance, substrate resistance, TIM (Thermal Interface Materials) resistance and spreading resistance in series connection. The whole thermal resistance of a COB array is composed of 9 pieces of single-chip LED thermal resistance in parallel connection, and then connected with heat dissipation in series. Compared to the soldering structure of a copper substrate, the aluminum substrate pressed on heat sink increases the substrate resistance and TIM resistance, and the spread resistance in single LED thermal resistance will change when the chip gaps change. According to Table 3, when the gap increased from 0.1 mm to 0.5 mm, 1.0 mm, and 1.5 mm, the chip junction temperature dropped 5.7 °C, 4.2 °C and 2.6 °C respectively. Enlarging the chip gap can diffuse heat effectively; it can also reduce the spread resistance and reduce the junction temperature. In addition, this simulation is limited by the substrate size so that the chip pitch could only maximize to 1.5 mm.

In previous simulation, the thicknesses of the copper coat and copper circuit layer are 0.5 mm and 0.1 mm, respectively. The junction temperatures for different chip pitches are stated as Fig. 10 when we keep the copper-circuit-layer thickness constant and change the copper-coat thickness. The total thermal resistance can be calculated as follows: R¼

ΔT T j −T a ¼ P P

ð3Þ

where Tj represents the chip junction temperature, and P is the input heat power. When the environment temperature and input power are constant, the junction temperature is related to the whole thermal resistance. It is manifested in Fig. 10 when we make changes on copper-coat thickness, the junction temperatures for different chip pitches are almost fixed. It predicts that the copper-coat thickness cannot change the whole thermal resistance within a small range. So, when the cold spray is applied in the soldering LED lamp, the copper coat can be as thin as possible to save the cost. There is no need to make a big area coating to improve the heat spread even though the thermal conductivity of the coat is bigger than that of aluminum. 3.4. Different thicknesses of the copper circuit layer Fig. 11 gives the junction temperature for different thicknesses of the copper-circuit layer. If other thermal resistances are constant and the thickness of the copper-circuit layer increases, the junction temperature should evaluate from the perspective of one dimensional resistance model. However, from Fig. 11 we can see that the junction temperature drops, it is especially obvious in this situation where the chip's pitch get smaller. The main reason is that heat can spread well in front of the dielectric layer and the thermal resistance got smaller when we increase the copper-circuit-layer thickness. The thermal conductivity of the original copper-circuit layer is up to 398 W/(m · K) while the dielectric layer is only 1.1 W/(m · K). When a big thermal resistances exist in the heat conduction process, which leads to a higher partial thermal flux and an elevated junction temperature. The figure also indicates that the influence of the pitch on the junction temperature is gradually reduced when the thickness is increased from 0.2 mm to 0.5 mm. From the previous analysis we can state that the chip gap can also affect the spread resistance. With the copper-circuit-layer thickness increasing, the spread resistance decreases and the influence of the gap on the junction temperature also reduces. So we can improve the thermal performance by increasing the chip gap and the thickness of the

Fig. 7. a. Pressing connection of a aluminum substrate. b. Temperature distribution of a LED in pressing way.

F. He et al. / International Communications in Heat and Mass Transfer 59 (2014) 55–60

Fig. 8. Temperature contours with pitch increasing.

Table 3 Junction temperature of different pitch.

59

60

F. He et al. / International Communications in Heat and Mass Transfer 59 (2014) 55–60

To sum up, the cost of a copper-substrate COB array is higher than one of an aluminum substrate, but the copper-substrate chip can be combined better with the cold spray solder technology for saving heat dissipation cost.

Fig. 9. Model of the single-chip LED thermal resistance.

102

Thickness

References

0.5mm 0.2mm 0.1mm

100 98

7M ഒ

96 94 92 90 88 86 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Pitch (mm) Fig. 10. Change of the junction temperature with the gap and copper coating.

102

Thickness

0.5mm 0.2mm 0.1mm

99

7M ഒ

96

93

90

87 0.0

0.2

0.4

0.6

0.8

1.0

Pitch (mm)

1.2

1.4

1.6

Fig. 11. Change of the junction temperature with the pitch and copper circuit layer.

copper-circuit layer. When the chip gap is 1.5 mm and the coppercircuit-layer thickness is 0.5 mm, the junction temperature becomes 85.7 °C and is lower 11.6 °C than the original model. 4. Conclusions By using FEM and making experiment on thermal performance of COB array in different metal base and different connection ways, we can make conclusions as follows:

[1] Ivan Ching-Cherng Sun, Shih-Hsun Moreno, Wei-Ting Chung, Chih-To Chen, Hsieh, Tsung-Hsun Yang, Brightness management in a direct LED backlight for LCD TVs, J. Soc. Inf. Disp. 16 (2008) 519–526. [2] Zhang Lei, Guo Xia, Liang Ting, Gu Xiaoling, Lin Qiao Ming, Shen Guangdi, Color rendering and luminous efficacy of trichromatic and tetrachromatic LED-based white LEDs, Microelectron. J. 38 (2007) 1–6. [3] Hui Huang Cheng, De-Shau Huang, Ming-Tzer Lin, Heat dissipation design and anasysis of high power LED array using the finite element method, Microelectron. Reliab. 52 (2012) 905–911. [4] S. Haque, D. Steigerwald, S. Rudaz, B. Steward, Packaging Challenges of High Power LEDs for Solid State Lighting, IMAPS, 2003. [5] M. Arik, J. Petroski, S. Weaver, Thermal challenges in the future generation solid state lighting applications: light emitting diodes, 8th Inter Society Conference on Thermal Phenomena, 2002, pp. 113–120. [6] A.K. da Silva, S. Lorente, A. Bejan, Optimal distribution of discrete heat sources on a plate with laminar forced convection, Int. J. Heat Mass Transfer 47 (2004) 2139–2148. [7] C.Y. Wang, Optimum placement of heat-source in forced convection, J. Heat Transfer 114 (1994) 508–510. [8] S. Chen, Y. Liu, S.F. Chan, C.W. Leung, T.L. Chan, Experimental study of optimum spacing problem in cooling of simulated electronic package, Int. J. Heat Mass Transfer 37 (2011) 251–257. [9] M.R. Hajmohammadi, S. Poozesh, M. Rahmani, A. Campo, Heat transfer improvement due to the imposition of non-uniform wall heating for in-tube laminar forced convection, Appl. Therm. Eng. 61 (2013) 268–277. [10] M.R. Hajmohamadi, M. Moulod, O. Joneydi sharlatzadeh, A. Campo, Effect of a thick plate on the excess temperature of iso-heat flux heat sources cooled by laminar forced convection flow; conjugate analysis, Numer. Heat Transfer, Part A 66 (2014) 205–216. [11] M.R. Hajmohamadi, M. Rahmani, A. Campo, Optimal design of unequal heat flux elements for optimized heat transfer inside a rectangular duct, Energy 68 (2014) 609–616. [12] Adam Christensen, Samuel Graham, Thermal effects in packaging high power-light emitting diode array[J], Appl. Therm. Eng. 29 (2009) 364–371. [13] Ji Li, Feng Lin, Daming Wang, et al., A loop-heat-pipe heat sink with parallel condensers for high-power integrated LED chips[J], Appl. Therm. Eng. 56 (2013) 18–26. [14] H.H. Wu, K.H. Lin, S.T. Lin, A study on the heat dissipation of high power multi-chip COB LEDs[J], J. Microelectron. 49 (2012) 280–287. [15] K.C. Yang, H. Liem, H.S. Choy, Heat transfer analysis of a high-brightness LED array on PCB under different placement configurations[J], Int. Commun. Heat Mass Transfer 53 (2014) 79–86. [16] R.H. Horng, J.S. Hong, Y.L. Tsai, D.S. Wu, C.M. Chen, C.J. Chen, Optimized thermal management from a chip to heat sink for high-power GaN-based light-emitting diodes, IEEE Trans. Electron Devices 57 (2010) 2203–2207. [17] B.H. Liou, C.M. Chen, R.H. Horng, Y.C. Chiang, D.S. Wu, Improvement of thermal management of high-power GaN-based light-emitting diodes, Microelectron. Reliab. 52 (2012) 861–865. [18] Sunho Lianqiao Yang, Woongjoon Jang, Moowhan Shin Hwang, Thermal analysis of high power GaN-based LEDs with ceramic package, Thermochim. Acta 455 (2007) 95–99. [19] L. Yin, L. Yang, W. Yang, Y. Guo, K. Ma, S. Li, J. Zhang, Thermal design and analysis of multi-chip LED module with ceramic substrate, Solid State Electron. 54 (2010) 1520–1524. [20] Kai-shing Yang, Chi-Huang Chuang, Ming-Tsang Lee, An experimental study on the heat dissipation of LED lighting module using metal/carbon foam, Int. Commun. Heat Mass Transfer 48 (2013) 73–79. [21] M. Ha, S. Graham, Development of a thermal resistance model for chip-on-board packaging of high power LED array[J], Microelectron. Reliab. 52 (5) (2012) 836–844. [22] Avoidance of ceramic-substrate-based LED chip cracking induced by PCB bending or flexing, http://www.cree.com.