LED Lighting: A Case Study in Thermal Management This article examines the thermal management of a light emitting diode (LED)-based lighting system. First, we discuss the environment in which the lighting system will be used. Then, we look at the system’s cooling needs and the various analyses used to confirm that the LED thermal requirements are being met. The article concludes with a comparison of the results. LED-Based Lighting System Requirements An LED-based lighting system was to be designed to replace a halogen-based downlight. A downlight is typically installed in a hollow opening in a ceiling and provides a concentrated output in the downward direction. A thermal management analysis was needed to properly design a cooling method for the LED system, which had to include a natural convection heat sink. This environment is shown in Figure 1 Ceiling air temperature Product Requirements The lifetime of an LED relates to its junction temperature and forward current. The new downlight includes three InGaNbased LUXEON cool white K2 LEDs at a forward current of 1000 mA. The maximum operational junction temperature for these cool white LEDs is 150ºC [1]. The downlight has a lifetime requirement of 60,000 hours. Figure 2 shows the lifetimes of the cool white LED for different forward currents, junction temperatures, and for the B10, L70 lifetime condition (which implies that for a specific lifetime, 10% of the LEDs are expected to fail at the specified junction temperature and forward current.) The failure criterion is when the light output of the LED has been reduced to 70% of its original light out. To achieve the 60,000 hours lifetime with a B10, L70 condition, the junction temperatures required for specific forward currents are shown in Table 1. From Table 1, with a forward current of 1000 mA, the junction temperature needs to be kept below 124ºC tolifetimes achieve a 60,000 hours lifetime. (B50, L70) lifetimes InGaN LUXEONK2 K2 (B50, L70) forforInGaN LUXEON 70,000 60,000 Ceiling Room ambient temperature Lifetime (Hours) Downlight 350mA 50,000 700mA 40,000 1A 30,000 1.5A 20,000 10,000 0 90 Figure 1. A Typical Downlight Environment. 6 100 110 120 130 140 150 160 Junction Temperature Junction temperature (C) (C) 170 180 190 200 Figure for 7. Expected (B50, L70) lifetimes forVersions InGaN LUXEONof K2 the LUXEON Figure 2. Lifetimes Different InGaN K2 LED [2]. Table 1. Required Junction Temperatures of LUXEON K2 LEDs for Specific Forward Currents to Achieve 60,000 Hours Lifetime Under the B10, L70 Lifetime Condition [2]. Forward Current [mA] Max Junction Temperature [ºC] 350 154 700 134 1000 124 1500 112 For this study, in order to achieve a 60,000 hours lifetime the LED junction temperature must be kept under 124ºC, with an average year-round temperature of 20ºC. Under maximum temperature conditions, the junction temperature must be less than 150ºC at an ambient temperature of 40ºC. Thermal Management Analysis The lifetime and maximum temperature conditions were determined previously; now, a thermal management analysis is applied to each condition. This is a confidence level analysis performed to build in safety margins for all unknowns in all engineering phases. The analysis comprises three sections: analytical, numerical (CFD) and experimental. 1. Analytical analysis a. Based on the unknowns in the analysis and shortcomings of empirical and experimental correlations, assumptions made in order to do the analysis 2. Numerical or CFD analysis a. Unknowns and assumptions made in order to do the analysis b. Shortcomings in the numerical code 3. Experimental a. Incorrect thermocouple placement b. Variations in thermocouple response c. Errors in velocity probe calibration d. Power input measurement Equation 1 is used for the confidence level analysis, where Tj is the required junction temperature and CFL is the confidence level being applied. Additionally, Tj,condiction is the specified junction temperature and Treference is the reference or ambient temperature. The temperature difference between the required junction temperature and the reference temperature, ΔTcondition, is used when comparing different conditions. Tj − Treference ∆Tcondition = ≤ CFL (1) Tj,condition − Treference Tj,condition − Treference Table 2. Confidence Factor Level, CFL, for Different Types of Analyses. Type of Analysis CFL Analytical 80% Numerical 80% to 85% Experimental 90% A confidence level of 90% is used in this study. Re-arranging Equation 1 yields Equation 2. Applying the lifetime conditions to Equation 2 determines the temperature difference for the lifetime condition. ( ) (2) ∆Tlifetime = 0.9 × (124 − 20 ) = 93.6 K (3) ∆Tcondition = CFL × Tj,condition − Treference The maximum temperature difference determined, as shown in Equation 4. can also ∆Tmaximum = 0.9 × (150 − 40 ) = 99 K be (4) From Equations 3 and 4, the lifetime condition is the most severe condition. Re-arranging Equation 1 yields Equation 5, which, applied to the lifetime condition, gives us Equation 6. ( ) Tj =× CFL Tj,condition − Treference + Treference September 2009 |Qpedia (5) 7 Tj = 0.9 × (124 − 20 ) + 20 = 113.3 ºC (6) (10) Pe= Pl + Q j Re-arranging Equation 10 yields Equation 11: Therefore, the junction temperatures to be determined by different analyses must be less than 113.3ºC at an ambient of 20ºC. Analytical Analysis As a starting point, an LED junction temperature of 108ºC is assumed, with a required forward current of 1000 mA. The usable light tool [3] gives a light efficiency of 9.4% and electrical power dissipation, Pe, of 3.53 W. The light efficiency is the ratio of the light power, Pl that the LED emits to the electrical power input, Pe. This is also given by Equation 7, which can be re-arranged in the form shown in Equation 8. ηl =Pl Pe (7) Pl = ηlPe (8) (11) Q= Pe + Pl j Substituting Equation 8 into 11 yields Equation 12. Rearranging Equation 12 gives Equation 13: = P − η P Q j e l e = Q Pe (1 − ηl ) j (13) Because all other values of Equation 13 are known, the heat dissipated by the LED can be calculated. (12) = 3.53 × (1 − 0.094= Q ) 3.2 W j (14) Pl Standard FR-4 boards can be used for LEDs with up to 0.5 W of dissipation, but metallic substrates are required for higher levels [4]. Because the LED heat dissipation is 3.2 W, a metal core board type PCB was used. Figure 4 is a sketch of the LED junction to heat sink. It shows each material that the heat from the LED must transfer through before it reaches the heat sink. Figure 4 also provides a thermal resistance diagram based on the sketch. The resistances are considered to be in series. Pe . Qj Figure 3. Control Volume Around an LED. Consider the control volume around the LED in Figure 3. The electrical power input, Pe enters the control volume while and the light power, P leave the the heat dissipated, Q j l control volume. Applying an energy balance to the control volume yields Equation 10. ∑ Ein = ∑ Eout The metal core board’s spreading resistance, Rmetalcore can be determined using the spreading resistance calculation method explained in [5]. The effective in-plane thermal conductivity can be calculated using Equation 15, as described in [6]: where t is the total thickness of the PCB, tc,i and tg,i are the Ng Nc (9) ∑ k c t c,i + ∑ k g t g,i =i 1 =i 1 k p,e = (15) t September 2009 |Qpedia 9 FUTURE COOLING P re p re g 1 5 0 µm R j-s M e ta l co re b o a rd R s-solder R m etalcore A lu m in iu m 1 .6 m m In te rfa ce m a te ria l C o pp e r 1 7 .5 -70 µm S o ld e rp aste 1 5 0 µm R interface R hs base, spreasding H e a tslu g H e a tsink b ase Figure 4. Heat Sink-to-LED Junction and Corresponding Thermal Resistance Diagram. thicknesses of the copper and glass-epoxy or prepreg/ dielectric layers, and kc and kg are the thermal conductivities of the copper and glass-epoxy, respectively. Equation 15 can be modified to accommodate the PCB’s aluminum layer, as shown in Equation 16. Additionally, the coverage percentage of each layer can be taken into account by the factor βi, Nc Ng ∑ βik c t c,i + ∑ βik g t g,i =i 1 =i 1 k p,e = + β ALk AL t AL Material Coverage [%] Conductivity [W/m∙K] Thickness [µm] Copper 50 385 70 Dielectric 100 3 150 Aluminum 100 180 1600 (16) t The other thermal resistance needs are as follows: where tAL is the aluminum thickness and kAL is the thermal conductivity of the aluminum. The PCB’s material properties are shown in Table 3. Using the spreading resistance calculation and effective in-plane thermal conductivity methods previously mentioned, along with the PCB’s material properties, the spreading resistance in the metal core board was calculated as Rmetalcore = 1 K/W. Table 3. PCB Material Properties. 10 1. Rs-solder is the thermal resistance in the solder under FUTURE COOLING the LED slug. It is 146 µm thick with a thermal conductivity of 50 W/m∙K and an area of 22.5 mm². This results in a thermal resistance of 0.13 K/W. 2. The interface resistance is assumed to be 0.2 K/W. This is comparable to the resistance of Chomerics T405-R thermal interface material. 3. The spreading in the heat sink base, Rhs base,spreading is assumed to be zero. 4. The junction-to-heat slug thermal resistance of the LED is 9 K/W [1]. Consider the thermal resistance in the heat transfer path from the junction to the heat sink base shown in Figure 4. These resistances are considered to be in series, and the junction-to-heat sink resistance is the sum of the individual resistances. Using Fourier’s law of heat conduction in a onedimensional differential form, the heat transfer rate between the junction and the heat sink can be expressed by Equation 17. Because the required lifetime junction temperature, the heat dissipated by the LED, and the thermal resistance from the junction to heat sink are known, Equation 17 can be rearranged to calculate the heat sink temperature, Equation 18. = Ths − Tamb Q hs Rhs−amb Rhs−amb = (17) R Ths= Tj − Q j j−hs (18) Ths − Tamb Q (21) hs = Rhs amb 80.34 20 = 6.28 K/W 9.6 (22) Figure 5. Thermal Resistance Diagram of LED Junction to Tj = 113.3 ºC . Qj . Qj . Qj Rj-hs Rj-hs Rj-hs Rhs-amb = Tj − Ths Q j R j−hs (20) . . Qhs = 3xQj = 9.6 W Ths = 80.34 ºC Tamb = 20 ºC Ambient. Ths= 113.3 − 3.2 × 10.3= 80.34 ºC (19) Because there are three LEDs on the heat sink, the sink must be able to transfer 3 x 3.2 W = 9.6 W from a heat sink temperature of 80.34ºC to an ambient of 20ºC. Using the thermal resistance diagram shown in Figure 5, the thermal resistance from the heat sink to ambient can be calculated using Equation 21. From Equation 22, the heat sink thermal resistance must be less then 6.28 K/W or the heat sink must be able to dissipated 9.6 W at a temperature difference of 60.34 K. 12 For the analytical simulation, two methods available to determine the heat sink thermal resistance. The first is to refer to the heat sink’s data sheet, which, in this study, shows that 9.6 W can be dissipated at a 56.3K temperature difference (see Figure 6.) This is less than the required 63.4 K temperature difference. The second method is to use an analytical model of the heat sink (whose part number is ATSEU-077B-C4-R0.) The results of the analytical analyses are shown in Table 4. and tested. This was done in order to verify the results of the analytical and numerical analyses. The LEDs were calibrated using the forward voltage method, also referred to as the electrical method. In the forward voltage method, the LED is calibrated at a sense current. Thereafter, the LED is tested at the required forward current of 1000 mA. When steady-state has been reached, the junction voltage at the sense current is measured and the junction temperature can be calculated from the calibration curve. A detailed example of the forward voltage/electrical test method is given in [8]. ce G raph 9.00 10.00 11.00 12.00 ation [W ] ed DThs-am bient [K ] T herm al P erform ance G raph 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 11.00 12.00 P ow er D issipation [W ] Vertic al m ounted Figure 6. Experimental Results using the ATSEU-077B-C4-R0 Heat Sink. Numerical Results Based on the analytical results, a model of the downlight was created. It was simulated in a free air environment. The boundary conditions for a free air environment are discussed in [7]. The results of the numerical analysis are shown in Table 4. A B Qj Tam b 2 0 °C Tj, m ax = 1 1 3 °C He a t sin k Ba se In te rfa ce material MCPCB S id e o f h o u sin g not shown C o m pa ct model LED Figure 7. Numerical Results of the Downlight Analysis. Experimental Results An experimental model of the downlight was manufactured C (c) Figure 8. Images of the Experimental Analysis showing the LEDs (a), the Experimental Set Up (b) and an Infrared Picture of the Lighting System (c). Comparing the Analytical, Numerical and Experimental Results (c) Table 4 summarizes the analytical, numerical and experimental results for the LED lighting system. The table shows that the results obtained using the different methods are within 10% of each other and have a high confidence level. The maximum temperature difference calculated for the CFD results is 93 K. Further, the experimental results September 2009 |Qpedia 13 FUTURE COOLING Table 4. Comparison of the Analytical, Numerical and Experimental Data, Normalized to an Ambient Temperature of 20ºC. Units Analytical, with experimental hs-data Analytical, only CFD Experimental Tambient °C 20 20 20 20 Iforward mA 1000 1000 1000 1000 Light efficiency % 9% 9% 9% 9% Pdissipated,total W 9.6 9.6 9.6 9.6 Theatsink base °C 68 76 75 71 Tboard °C 73 81 84 78 Tj, led °C 102 110 113 107 Comparison of methods % 95% 103% 105% 100% ΔTj-amb K 82 90 93 87 TRUE TRUE TRUE TRUE Parameter Less than the required temperature difference of 93.6K have a temperature of 87 K. Both of these results are below the required 93.6 K for the lifetime condition. Therefore, the analyses have shown that the LED-based downlight system satisfies the lifetime temperature condition. The LED-based downlight lighting system end product is shown in Figure 9. Figure 9. LED-based Downlight Whose Cooling Solution was Developed by ATS Europe B.V. Summary This article explains the development of an LED-based downlight system. The LED lighting system uses 3 LUXEON K2 LEDs at a forward current of 1000 mA. The article discusses analytical, numerical and experimental analysis methods A comparison of the different analysis results are 14 given. For reliability, it is recommended that at least two independent results be obtained, and that these not differ by more than 20%. References: 1. Luxeon K2 Technical data sheet, DS51, http://www. philipslumileds.com/pdfs/DS51.pdf, September 2009. 2. Luxeon K2 Reliability Datasheet RD06, http://www. philipslumileds.com/pdfs/RD06.pdf, September 2009. 3. Future Electronics, Usable Light Tool, www. futurelightingsolutions.com. 4. Petroski, J., Thermal Challenges in LED Cooling, Electronics Cooling Magazine, November 2006. 5. Spreading Thermal Resistance: Its Definition and Control, Qpedia eMagazine, September 2007. 6. Shabany, Y, Component Size and Effective Thermal Conductivity of Printed Circuit Boards, ITHERM, 2002. 7. Boundary Conditions for Natural Convection CFD Simulations, Qpedia eMagazine, June 2009. 8. Hulett, J. and Kelly, C., Measuring LED Junction Temperature, http://www.photonics.com/Content/ ReadArticle.aspx?ArticleID=34316, September 2009/