Multi-LED Package design, fabrication and thermalanalysis

Transcription

1 Multi-LED Package design, fabrication and thermalanalysis R.H. Poelma 1, S. Tarashioon 1, H.W. van Zeijl 1, S. Goldbach 2, J.L.J. Zijl 3 and G.Q. Zhang 1,2 1 Delft University of Technology, Delft, The Netherlands 2 Philips Lighting, Eindhoven, The Netherlands. 3 Fico/Besi, Duiven, The Netherlands. Abstract: An ultra-thin multi-led package is designed, manufactured and its thermal performance is characterized. The objective of this study is to develop an efficient thermal modelling approach for this system which can be used for optimization of the thermal-performance of future ultra-thin designs. A high-resolution thermal imaging camera, and thermo couples were used to measure the temperature distribution of the multi-led package and the LED-die temperature for different operating powers. Finally, we compare the thermal measurements with the Finite element simulation results. It is concluded that the modelling approach can assist in the thermal optimization of future multi-led package designs. Keywords: LED, packaging, optics molding, high-resolution thermal imaging, thermal modeling and measurements. 1 Introduction Light emitting diodes (LEDs) are currently the most energy efficient light sources available. Consequentially, high brightness LEDs become the cheapest light source over time compared to incandescent light bulbs and compact fluorescent lamps for lighting applications [1, 2]. However, the thermal management of high brightness multi-led packages is still a challenge that can be improved. Currently, about 65% to 85% of the input power of the best commercial available high brightness LEDs is converted into heat while the rest is transformed into light [3, 4]. Improving the heat dissipation from the LEDs is crucial for reducing the LED junction temperature. The temperature is the primary factor influencing the total lifetime, the power efficiency and droop of LEDs [3-5]. Modelling the thermal performance of electronic systems and optimization of the design, is becoming a useful tool which can reduce the development time and improve the quality of the final product. For this study we designed and fabricated a novel ultra-thin multi-led package which is briefly discussed in section 2. In section 3 we perform thermal measurements on the multi-led package at different operating powers. In section 4 we discuss, and validate the thermal model of the system. Finally, we give conclusions on how to thermally optimize future designs of multi-led packages. 2 Multi-LED package design and fabrication Market research has shown that lighting application areas for retail and hospitality have demanding requirements for new LED spots in terms of thickness, lumen output, colour rendering index (CRI), efficacy and lifetime [2]. The lifetime of electrical systems is strongly affected by the temperature. The temperature is driven upwards by the miniaturization and increasing power density of the multi-led packages. Therefore, long lifetime is arguably the most difficult requirement to achieve. Consequently, no luminaire system or LED-spot is available that meets these specific requirements. The aim of this paper is to design a system for this application area. Table 1 gives an overview of the stringent user requirements for the LED-spot for retail and hospitality.

2 Table 1: User requirements for LED-Spot for retail and hospitality applications. Parameters Values Lateral size mm Thickness 1-5 mm Lumen output lm Efficacy K Lifetime 50k hours (a) 90mm 8mm 90mm 35mm 35mm A specific LED type was chosen that can meet the requirements, see Table 2 for the LED properties. A grid array (4 x 4) of this LED type are placed inside a simplified package. The specific LED performance and electrical properties are given in Table 2. The power of these LEDs and their efficiency rating are used for the calculation of the amount of energy that is transformed into heat. (b) (c) Table 2: LED properties Parameters Values Max. T junc 135 o C Power 2 Watt Voltage 2.8 V Current o C Efficiency 83% (d) 2.1 Geometric parameters Figure 1a shows a schematic illustration of the design of the LED package placed on the fin heat sink. Figure 1b shows the multi-led package after the silicone molding step and Figure 1c shows the package after interconnect patterning. The substrate material is a Cu foil with thickness t Cu = 70µm. The silicone thickness is t S = 1mm. The assembly of the LED package on the passively cooled copper heat sink is shown in Figure 1d. The package is glued using a thermal adhesive. Finally the working sample is shown in Figure 1e. Figure 2 shows the fabrication steps. We start by: 1) Solder resist and solder paste deposition using stencil printing. 2) The second step is connecting the LEDs to the foil by pick and place and solder reflow. 3) The third step is Silicone encapsulation of the LEDs and Cu foil performed by the company Besi/Fico, The Netherlands. 4) The final step is the Cu foil backside patterning and interconnect fabrication. (e) Figure 1: a) Schematic illustration of the multi-led package placed on the fin-heatsink showing the locations of the thermo couples, and. b) Package after LED placement and silicone molding and before interconnect patterning. c) Package after interconnect patterning. d) Final assembly on the passively cooled copper heat sink. e) Illumination at low power.

3 Copper foil Stencil print Cure Stencil print Pick & Place Solder reflow Molding Photo-etch Test Solder resist Solder paste LED package Silicone Optics Interconnects Figure 2: Process flowchart of the multi-led package. Due to the low cost and high reliability, fin heat sinks in natural convection and radiation are still widely used for cooling in various applications [6]. Furthermore, the modelling of passively cooled heat sinks is well-defined and helps in characterizing the thermal performance of the package assembly. Therefore, we connect the LED package to a 9 cm by 9 cm Copper polished heat sink using a thermal dielectric adhesive as is shown in Figure 1. The heat sink has a base thickness of t b = 2 mm, 55 plate fins with a length L fin of 21 mm, a fin thickness of t fin = 300 µm and a fin pitch of 1.6 mm. This approach has the possibility of reducing the effective thermal resistance between the LED-die and the heat sink and improving the thermal performance. Between the heatsink and the LED package we placed a 550 µm thick Silicon wafer, coated with a 500 nm thick electrically insulating Nitride film. 3 Thermal measurements We use two measurement techniques to characterize the temperature distribution of the LED package for different operating power in a temperature controlled environment (T Room = 297 K ). One method consists of measuring the temperature using thermocouples placed on the locations:, and on the heat sink, see Figure 1a. The other approach is using a thermal imaging camera to characterize the heat distribution over the surface and estimate the LED junction temperature. The orientation of passively cooled heat sinks strongly affects their efficiency, we have considered the worst-case horizontal scenario (Figure 3) and have compared the results to the thermal model. The points of interest are the maximum temperature and the location of hotspots. The LED package mounted on the Cu heat sink is placed horizontally in a temperature controlled environment and is allowed to be cooled by natural convection. The thermal imaging camera is placed directly above the package to get a top view of the system. Figure 3a shows the measurement setup and Figure 3b shows the thermal image of the system. The thermal imager used was the Fluke Ti10, which has a spatial resolution of 2.5mRad and an accuracy of ±5 O C on non-reflective surfaces. (a) (b) Figure 3: a) Measurement setup. b) Thermal image of the multi-led package at an input power of respectively 35 Watt. The color indicates the surface temperature. Figure 3b shows that the top row of LEDs have a slightly higher temperature. This is probably caused by the soldered wires that prevented a good

4 thermal contact. The measurement results are discussed in more detail and compared to the simulation results in Section 4. 4 Thermal simulation 4.1 Model description The differential form of Fourier s law, gives the relationship between q the heat flux, k the thermal conductivity of a solid material and T the corresponding temperature gradient [7, 8]: T+ kq= 0 (4.1) The temperature gradient and heat flux are defined by the prescribed boundary conditions and the material properties. Finite element simulation is used to study the thermal performance of the LED package containing a square 4 x 4 LED array. We study how the heat generation in the LEDs defined by Q LED, affects the temperature T of the system and we compare the results to the measurements. Contact resistance between the LED package and heat sink is defined by the thermal conductivity and thickness of the thermal adhesive, respectively: k TIM = 0.1 W/mK and t TIM = 140 µm. The material thermal conductivity of the silicone, copper and silicon are respectively: 0.26, 400 and 130 W/mK. The initial temperature and that of the surroundings is 297 Kelvin. Figure 4, shows the different film thickness of the materials. The thermal interface material is between the heat sink and the silicon and between the copper foil and the silicon. tsi Q LED h c,0 Figure 4: Side-view of the LED package and the corresponding geometric parameters. 4.2 Boundary conditions Symmetry boundary conditions can be applied that reduce the size of the model to one fourth of its original size. On all the exposed surfaces of the heat sink and package a heat flux is prescribed that allows for convective heat transfer to the surroundings. The heat transfer coefficient h, depends on ts = 1mm h c t = 70µ tb = 2mm 500µ m = 21mm Cu Lfin the surface orientation and temperature. The heat transfer coefficient [7] of a surface in ambient air is estimated by, Txy (, ) n T(, ) n s xy T 0 hci, = C = C L fin L fin (4.2) where T s (x,y) is the surface temperature on location (x,y) and T 0 is the ambient temperature. Table 3 shows the constants and exponents for the calculation of the heat transfer coefficient for different surface orientations. The length of the fin is defined by L fin, which is 21 mm when the heat sink is horizontal and 90 mm when vertical. Table 3: The values of constant C and exponent n for different surface orientations. Plate orientation C n Vertical Horizontal (Top-side) Horizontal (Bottom-side) Thermal radiation can be ignored for the polished Cu heat sink since the surface emissivity coefficient is about 0.05 [-]. The heat transfer coefficient as predicted by Eq. 4.2 is plotted in Figure 5 for the three different orientations. Heat transfer coeficient h c (W/(m 2 K)) Vertical plate 2 Top-side horizontal plate Bottom-side horizontal plate Surface temperature T s (K) Figure 5: Heat transfer coefficient plotted against different temperatures of the surface and for several orientations. The ambient temperature is 300 K. Figure 5 shows that restriction of the airflow by changing the orientation strongly affects the efficiency of the heat sink. For the simulations and experiments we therefore simulate and test for the worst case scenario to define the maximum operating condition of the package.

5 4.3 Simulation and measurement results Figure 6: Temperature distribution of a 4x4 LED array (only a quarter shown) at 40 Watt operating power. The package is horizontally orientated, worst case. Temperature (K) LED temperature Contact resistance Simulation Measurement Arc-length (m) Figure 7: Temperature profile measured by the IR camera compared to the numerical simulation. The profile is measured along the cross-section - for the horizontal orientation of the heat sink at operating power of 35 Watt. Figure 6 shows the numerical simulation result of the LED package placed on the heat sink operating at 40 Watts. The high temperature gap of about 15 Kelvin is caused by the thermal contact resistance. Figure 7 shows the comparison between the simulation and the measurements of the IR camera. There is a good agreement between the model and the measurement. The differences can be explained by Joule heating of the interconnects that were not accounted for in the model. Furthermore, the IR temperature measurement is influenced by the different emissivity coefficients of the materials of the package and heat sink material. The IR camera doesn t distinguish between the different emissivity coefficients and therefore a small measurement error can be introduced of +- 2 Kelvin. Figure 8 compares the temperature measurements of the thermocouples with the numerical simulation for increasing operating power. The temperature reading of and overlap although they are positioned in different locations. Temperature (K) Thermocouple, and measurements Simulation Operating Power (W) Figure 8: Temperature at the measurement locations - for increasing operating power compared to simulation. Figure 8 shows that the measurements with the thermocouples are in good agreement with the simulation. The package appears to heat up at a higher rate than the model predicts. We assume this can be explained by the Joule heating of the wire interconnects that were not accounted for. 5 Conclusions An ultra-thin multi-led package was designed, manufactured and its thermal performance was studied. There is a good agreement between the thermal simulation and experiment when the model accounts for a high thermal resistance between the LED package and the heat sink. The simulation and experimental results show that the best way to optimize the thermal performance is by reducing the contact thermal resistance. The results of this study will help in the design of larger and more complex packages leading to high lumen spots up to 5000 lumens using high brightness LEDs. Acknowledgements The authors would like to thank CATRENE L Project CA502 SEEL Solutions for Energy-Efficient Lighting - for financial support and the people from DEMO and DIMES TU Delft for their help and technical expertise.

6 References 1. Lee, S.-J., et al., High-Brightness GaN-Based Light-Emitting Diodes on Si Using Wafer Bonding Technology. Applied Physics Express, (6). 2. Agrawal, V., et al., Solid State Lighting LED Manufacturing Roundtable Summary, 2011, U.S. Department of Energy. p Cheng, T., et al., Thermal analysis and optimization of multiple LED packaging based on a general analytical solution. International Journal of Thermal Sciences, (1): p Efremov, A., et al., Effect of the joule heating on the quantum efficiency and choice of thermal conditions for high-power blue InGaN/GaN LEDs. Semiconductors, (5): p Wang, J., et al., High-efficiency diode lasers at high output power. Applied Physics Letters, (11): p Sahray, D., et al., Study and Optimization of Horizontal-Base Pin-Fin Heat Sinks in Natural Convection and Radiation. Journal of Heat Transfer, (1): p Mills, A.F., Basic Heat and Mass Transfer1999: Prentice Hall. 8. Öchsner, A., M. Tane, and H. Nakajima, Prediction of the thermal properties of lotus-type and quasi-isotropic porous metals: Numerical and analytical methods. Materials Letters, (21-22): p