Color-consistent LED modules for general lighting

Transcription

1 Invited Paper Color-consistent LED modules for general lighting Christoph Hoelen* a, Peter van der Burgt a, Paul Jungwirth b, Matthijs Keuper c, Kwong Man b, Claudia Mutter a, and Jan-Willem ter Weeme a a Philips Lighting, Advanced Innovation Group, Mathildelaan 1, 5611 BD Eindhoven, The Netherlands; b Philips Lighting, LED Modules and Retrofits, 7700 Riverfront Gate, Burnaby, BC, Canada; c Philips Lighting, LED Modules and Retrofits, Mathildelaan 1, 5611 BD Eindhoven, The Netherlands; ABSTRACT During the last few decades the efficacy and luminous flux of LEDs have developed fast. Also the color quality of white LEDs and LED illumination systems has improved considerably. Thanks to the performance improvements and the continuously declining cost per lumen, it is now possible to create LED lighting systems with high luminous fluxes that can be applied in downlights for general lighting and in spot lights for accent lighting. One of the important requirements on lighting systems in indoor lighting applications is the color consistency. For all systems the chromaticity, or color point, of the light should be the same, i.e. within well-defined small tolerance areas. For down lighting, LED modules with high optical efficiency have been developed based on the concept of mixing light from multiple LEDs and luminescent materials, and emitting the mixed light through a translucent window. This concept is ideal for down lighting and other general illumination applications since it enables the design of luminaires with high optical efficiencies and low glare. In addition, it enables high color uniformity and excellent color consistency between modules. The module concept enables forward compatibility by well-defined interfaces and optical properties that are decoupled from the actual performance and number of LEDs. In this paper the properties with respect to color consistency of the various concepts will be discussed. By applying a phosphor remote from the blue LEDs, we have developed mediumbrightness ( kcd/m 2 ) LED-modules with high system efficacy. This is the basis of the Philips Fortimo downlight system. Based on mixing of multiple colors, the color tunable Lexel downlight module has been developed. The systems comprising multiple LED colors have feedback loops to comply with color consistency requirements. In all systems a color consistency within 5 SDCM is achieved. Keywords: LED, module, SSL, color consistency, tunable color, illumination, Fortimo, Lexel, downlight, Luxeon 1. INTRODUCTION The recent advancements in LED performance have enabled the development of various new solid state lamp architectures. Because LEDs can emit radiation in the visible spectral range, and because luminescent materials have become available that can be excited by blue or near UV LEDs, over time LED illumination systems are expected to outperform most other lighting technologies with respect to efficacy. The Stokes shift losses and (near) infrared or (near) UV emission losses that are associated with conventional lighting technologies can be omitted, or at least significantly reduced, while the wall plug efficiency of LEDs has shown a continuous increase over the years, and is still improving. On the other hand, the luminous flux emitted per LED is still relative small compared to most conventional lamps. Therefore to create lighting systems for general illumination purposes in most cases multiple LEDs have to be combined into a system. Although the luminous flux per emitter will increase significantly over time, it is to be expected that for many applications also in the future multiple emitters are needed. This offers both opportunities and challenges. Advantages are the increased design freedom, the relative ease of realizing modular building blocks, and the ease of scaling the flux output of lamps or luminaires. Challenges on the other hand are found in the thermal management requirements, cost reductions, and in the performance spread of the LEDs. The latter leads to a huge spread in luminous flux and color point of LEDs and LED modules. In this paper we will address the color consistency of various LED modules, taking into account variations in the characteristics of the input components and in operating conditions. *christoph.hoelen@philips.com; phone ; fax Light-Emitting Diodes: Materials, Devices, and Applications for Solid State Lighting XIII, edited by Klaus P. Streubel, Heonsu Jeon, Li-Wei Tu, Proc. of SPIE Vol. 7231, 72310A 2009 SPIE CCC code: X/09/$18 doi: / Proc. of SPIE Vol A-1

2 1.1 Light generation mechanisms of LED lighting systems The relative narrow-banded nature of LED light emission and the availability of colors throughout almost the whole visible spectral range naturally enables additive color mixing systems comprising different LEDs. Systems based on mixing of three or four primary colors have been presented before in theoretical exercises covering a broad range of emission wavelengths [1] as well as in various practical configurations [2],[3]. In [1], the relations between emission wavelength, color rendering properties and efficacy have been described for theoretical narrow emission spectra, while in e.g. [2] and [3] the practical performance and influence of operating conditions has been presented. One of the key items in these systems turns out to be the accuracy and stability of the color point of the light. In a second basic configuration, LED lighting systems are based on blue-pumped phosphors. The well-known cool-white LED, comprising a yellow phosphor on or close to the blue emitting LED die is the well known example of this white light generating mechanism. Mixtures of phosphors can be applied to realize lower color temperature and/or higher color rendering properties [4]. However, the external quantum efficiency (EQE) of a blue LED is generally significantly better than that of a conventional white LED made from dies with identical properties. When increasing the phosphor load in a packaged LED, the EQE further reduces. The system performance can be significantly enhanced by reducing the absorption losses in the LED, e.g. by reducing the back scattering of blue light and the emission of converted light back into the die. To take full advantage of this, medium-brightness LED-modules (with a brightness on the order of 150 kcd/m 2 ) have been presented for general lighting applications [5],[6]. In these modules the phosphor is applied remote from the blue LEDs, enabling a high overall system efficacy (i.e., luminous flux per unit of electrical input power). Although the application of blue light results in small Stokes shift losses, the unconverted blue light is an essential part of the overall spectrum. Therefore the resulting color point is a function of the blue emission and the effective layer thickness of the luminescent converter material. Basically this is the cause of the large spread in color points of white phosphor conversion LEDs (based on blue pump dies) that the LED manufacturers are facing. A third basic configuration is based on near-uv LEDs and a mixture of phosphors [7]. The advantage in this case is that the color point is determined by the emission from the luminescent materials, which can be very stable. Of course, still a dependence on the emission spectrum of the LEDs is present due to the shape of the excitation spectra of these phosphors, but this dependence can be relatively small. The major disadvantage is the reduced efficacy of these systems due to the large Stoke shift losses and the additional conversion losses by a luminescent converter that emits in the blue spectral range. The addition of a blue emitting phosphor further increases the total phosphor load, leading to additional scattering losses of light, and thus to a reduced overall system efficacy. Finally, multiple systems based on various combinations of the light generation technologies mentioned above can be designed. As an example, blue LEDs may be provided with a yellow-green phosphor and combined with red LEDs. In this way low color temperatures can be achieved that would otherwise not be possible when using the very popular, stable and efficient aluminum garnet phosphors. In other configurations, red phosphor conversion LEDs (where a red emitting phosphor is pumped by e.g. an intrinsically blue emitting LED) are combined with blue and green emitting LEDs, thus resulting in a lighting system comprising only LEDs with epitaxial layers based on the InGaN material system [8]. This has intrinsically more stable emission properties than systems comprising multiple LED material systems, in particular because of the large difference in thermal dependencies of InGaN LEDs emitting in the blue-green region versus AlInGaP LEDs emitting in the amber-red region [9]. In addition, the color rendering properties can be significantly enhanced. However, as the conversion efficiency of red phosphor conversion LEDs is still relative low, the overall system efficacy is negatively impacted in this case. 1.2 The need for color control loops. In systems where at least two different color contributions can be controlled independently the overall color point can be controlled by adjusting the ratio of the contributions. It has been shown before that for LED spot lamps the application of red, green and blue (RGB), amber, green and blue (AGB), and the 4-color combination red, amber, green and blue (RAGB) LEDs, leads to unacceptable color point deviations due to temperature variations in absence of some form of color control loop [2],[10],[8]. Typically, in spot lamps the temperature may vary over 50 C due to luminaire characteristics and operating conditions (dimming, color point selection, ambient temperature). When driving such lamps with a neutral white output of ca. 4000K in an open loop configuration, a temperature variation of 25 C does already induce a color point change of about 0.01 in the CIE u v chromaticity space [10]. This corresponds, depending on the direction in the chromaticity diagram, with 7 to 10 times the standard deviation of color matching (SDCM, or MacAdam step). As the just noticeable color difference is about three times the SDCM [11], and because temperature variation is Proc. of SPIE Vol A-2

3 not the only cause of color point shifts, for such lamps color differences are clearly noticeable when driven in open loop configuration. For light sources used in general lighting applications color point variations are commonly accepted if these do not exceed the area of an ellipse of 5 SDCM around a target color point. Therefore, also for LED modules for general illumination we aim for color consistency within a 5 SDCM ellipse around a target color point under all operating conditions. It is clear that for RGB color mixing systems a control loop is inevitable. 2. LED MODULE CONFIGURATIONS FOR DOWNLIGHTING The principle of using an optical mixing cavity with a translucent window to emit the light of multiple LEDs is particularly advantageous for applications that do not require a high source brightness (as is required in general for narrow beam spot lamps), as this is a configuration that is future proof for improvements in LEDs. For typical downlight applications, one to several klm of luminous flux is requested with a form factor of the luminaire that allows for a diameter of the (apparent) light source up to about 6 cm. This corresponds with a source brightness of 100 to several 100 knit. With LED efficacies of lm/w under operation (hot) conditions and a power dissipation per LED in the range of 1-4 W, this can easily be achieved. Including optical losses up to about 20%, this would require an LED count in the range of 4 to 60. With the current LED dimensions, e.g. of the Luxeon Rebel emitter, this number of LEDs can well be accommodated in a cavity with such a maximum diameter. In this paper we compare the following module configurations that in principle are suitable for downlighting applications: 1. a diffuse reflective box with a translucent emission window comprising blue LEDs and a remote phosphor layer. The phosphor layer is applied on the emission window material and comprises an aluminum garnet phosphor emitting in the yellow-green spectral range, and a nitride phosphor emitting in the red-orange spectral range. More details on red nitride phosphors can be found elsewhere in literature [7],[13]-[16]. The LEDs are Luxeon Rebel emitters [17], which are about equidistantly positioned to optimize the uniformity in the output window. This concept is applied in the Philips Fortimo Downlight Module. 2. a diffuse reflective box with a translucent emission window comprising red, green, blue and white (blue LED die covered with a luminescent layer) LEDs. This system enables full color tunability. In addition, high CRI values can be achieved. Thanks to the individual addressability of the four groups of emitter types, color point setting can be realized under optimization of either the luminous output or the CRI. The Luxeon Rebel LEDs are positioned on a circle to maximize the color mixing and, therefore, the color uniformity in the output window. This is the color concept of the color tunable Philips Lexel Downlight Module. 3. a diffuse reflective box with a translucent emission window comprising blue and red LEDs and a remote phosphor layer. Again the phosphor layer is applied on the emission window material, and comprises an aluminum garnet phosphor emitting in the yellow spectral range. This configuration enables high CRI values thanks to the peaked red LED emission and a relative high overall efficacy, although the latter is strongly impacted by the actual case temperature. Some color point tuning is possible, although this is limited to the line interconnecting the color points of the red LEDs and the resulting color point of the blue LEDs and the yellow phosphor. Therefore this option does not seem to offer added value in the application. The LEDs are about equidistantly positioned to optimize the uniformity in the output window. For the color tunable RGBW system a closed loop feedback system has been developed to control the color point. Thanks to the presence of four independent color channels, this also enables optimization of an additional parameter like efficacy, color rendering properties or luminous flux. The control loop is based on a combination of a temperature feed forward and a flux feedback system, where the various color strings are fed with pulse width modulated drive currents. The system is factory-calibrated to enable accurate compensation for thermal effects and spread in the properties of the LEDs. The temperature feedback is used to predict changes in the color points of the individual channels as a function of temperature. This enables, together with the flux information, an accurate setting of the color point. 3. SENSITIVITIES OF VARIOUS COLOR CONCEPTS FOR INPUT VARIABLES The light properties of LED lighting systems such as luminous flux, color of the light and color rendering depend on many parameters. These parameters are e.g. the type of LEDs and luminescent materials used, the LED drive current, LED wavelength, LED efficacy, the efficiency and optical thickness of luminescent materials, the temperature of the system, and the optical efficiency of the system. But also the properties of color control loops (when present) can play an Proc. of SPIE Vol A-3

4 important role. Which of these are dominant depends on the concept and on the extent of the spread in the input parameter (settings, component and control characteristics, ambient conditions). Although the luminous flux or the efficacy of a system is important, the human eye is not very sensitive to variations in the luminous flux level. Also moderate variations in color rendering are in general not easily observed. Very sensitive, however, is the human eye for color variations; already small changes in the spectrum often lead to changes in the color point that exceed the discrimination threshold. In general lighting schemes it is very important that the differences in the color of the light between the individual luminaires are sufficiently small since big differences will lead to user dissatisfaction. Therefore, in this section we will focus on the color point variations of the various LED module configurations as presented in the previous section. In practice, the most important parameters affecting the color point are the LED wavelength and wall plug efficiency, the drive current, the temperature, and the thickness of luminescent conversion materials. For the configurations mentioned before we will discuss the influence of practical spread in the most relevant input parameters on the chromaticity of the system. 3.1 Fortimo downlight concept: direct blue emitters with remote luminescent layer Various test units were built comprising different blue wavelength LEDs. These units were operated at various drive currents and at various case temperatures. The resulting experimental color points were used to model the influence of these variables on the color point. Color point spread of RP downlight module due to input spread 0,4 0,39 CIE y 0,38 0,37 BBL 4000 K 5 SDCM Varying PWL, If, T Spread boundary 0,36 0,355 0,365 0,375 0,385 0,395 0,405 CIE x Fig. 1 Variation of the color point under variation of the drive current ( ma), case temperature (25-65 C) and blue pump wavelength ( nm). Data points are for all combinations of the minimum, median and maximum values in these ranges. Target color point is 4000K, with the 5 SDCM ellipse around that point as a reference. In Fig. 1 the color points are plotted that result from this modeled variation in the blue wavelength range between 450 and 455 nm (i.e., the typical width of a wavelength bin) in combination with variation in the drive current between 200 and 700 ma (i.e., a wide range around the standard drive current of 350 ma), and with variation in the case temperature between 25 and 65 C (i.e., from room temperature to the maximum operating case temperature). In the figure, all combinations of the minimum, the median, and the maximum values of the ranges of these input parameters are plotted. The result is an area within the 5-step MacAdam ellipse with a boundary indicated by the dashed curve. This area is much smaller than the 5 SDCM ellipse and well centered at the target color point. For this input spread, the color point spread stays within a 3-step MacAdam ellipse around the target color point. About a 10 nm wavelength range of blue LEDs can be used without any sort of wavelength bin selection while the color point of the module stays within the requested 5-step MacAdam ellipse. Therefore, this system is robust with respect to the blue LED wavelength, the drive current and the case temperature. For larger pump wavelength ranges, longer wavelength LEDs are to be combined with shorter wavelength LEDs to fulfill the color consistency specification. Not included in this graphs is spread due to variation in the layer thickness of the luminescent material, nor due to spread in the composition of this material. Other experiments have shown that at the 4000K color point a 10% increase in the load of the luminescent material results in a Proc. of SPIE Vol A-4

5 color point shift of up to 4 SDCM, and a 10% change in the ratio of red-orange and yellow-green luminescent materials results in a color point shift of less than 2 SDCM. Because it is easy to control the phosphor load with an accuracy better than 5% and the ratio between different luminescent materials with an accuracy better than a few percent, the total color point deviation of the Fortimo Downlight concept may be expected to remain well within a 5-step MacAdam ellipse around the target. 3.2 Lexel Downlight concept: color mixing LED modules An illumination system comprising four independent color channels enables, apart from setting the requested color point, optimization of an additional parameter like efficacy, CRI or flux. For the color tunable RGBW configuration of the Lexel Downlight system a closed loop feedback system has been developed that allows for selection of flux optimization, CRI optimization, or a mix of both. The relative flux contributions from the four color channels are adjusted by pulse width modulation of the drive currents for the various LED colors. Open versus closed loop performance for varying temperature 3000K and 4100K Targets Planckian Target 3000K CIE y CIE x Target 4100K 5-step MacAdam ellipse 3000K Open Loop 3000K Closed Loop 4100K Open Loop 4100K Closed Loop Fig. 2 Chromaticity points of a Lexel Downlight module as a function of the case temperature when driven in open loop and in closed loop operation. Target settings are 3000 and 4100 K. The case temperature was varied from 25 C to 65 C. The 5-SDCM ellipse is plotted as a reference for each target color point. The feedback system not only maintains the color accuracy, but also prevents the luminous flux output to drop from the set target value when the case temperature increases. This is particularly important for the RGBW system because of the strong temperature sensitivity of the red-orange AlInGaP LEDs. It has been shown before that for RGB and RAGB color mixing systems a feedback loop is necessary to achieve color consistency during varying operating conditions or module settings. In the Lexel Downlight color concept, RGB LEDs are combined with white LEDs, resulting in other color point stability when driven in open loop than for the RGB and RAGB systems due to the addition of relative stable white LEDs. For a single test module, the color point drift induced by a case temperature increase when the RGBW system is driven in open loop configuration is plotted for two target color temperatures (3000 and 4100K) in Fig. 2. For the same module and settings, but with enabled closed loop feedback system, the color points of the light output are also plotted in the figure. For the 3000K target, the open loop configuration results in a huge color point drift caused by a case temperature increase of 40 C. This drift amounts to about 10 SDCM. At a color temperature target of about 4000K, the temperatureinduced color point drift is significantly smaller, but still exceeds the 5 SDCM limit for a temperature variation of 40 C. Around this color temperature the system is more stable thanks to the relative large contribution from the white LEDs that are much more stable than the red, and to a lesser degree, the green LEDs. The closed loop feedback system results in a very stable color tunable system, that remains well within the targeted 5-step MacAdam ellipse. We see that there is a small setpoint error for this particular module. This error is caused by calibration inaccuracies. Apart from the setpoint Proc. of SPIE Vol A-5

6 error, the additional color point drift of a Lexel Downlight module in (standard) closed loop operation is less than 1 SDCM. Therefore we can conclude that the closed loop color control algorithm of the Lexel Downlight systems shows an excellent stability with respect to temperature variations, and a color point accuracy that is well within the required specs. 3.3 Remote yellow phosphor with direct blue and red emitters By adjusting the load of luminescent material that is excited by light from blue LEDs, any color point on an almost straight line between the color point of the blue LEDs and of the luminescent material can be obtained. 1 0, Spectral output vs LED PWL variation; ca. 3500K 0,8 Blue LED + Y phosphor ,7 0,6 0,5 0,4 Blue+RP_Y+ Red: 2700K Blue+RP_Y+ Red: 4000K Spectral density (1/nm) ,3 0,2 0, ,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 Fig. 3 CIE x,y chromaticity diagram with typical color points of blue LED, yellow phosphor, and red LED. The phosphor load determines the position on the interconnect between the blue LED and the phosphor. For two phosphor loads the trend lines are shown when flux from the red LED is added. For the two combinations, the intersection with the BBL is at 2700K and 4000K respectively Wavelength (nm) Fig. 4 Spread in the spectral output distribution of an LED module due to spread in the wavelength of the red (627 ± 5 nm) and blue (450 ± 5 nm) LEDs. The wall plug efficiency of the LEDs, the yellow luminescent conversion layer, the housing temperature and the LED drive currents are not subject to spread in this example. The spectra in the graph are based on combinations of experimental spectral data, and show all combinations of the minimum, median, and maximum values of the wavelengths of the red and blue LEDs. Target color point at the nominal values is 3500K. For a yellow-green phosphor, the BBL is crossed at a relative high color temperature. By further increasing the phosphor load, a color point above the BBL results. In combination with light from red LEDs, optimization of the phosphor layer thickness and the relative drive currents of the red and blue LEDs enables color point targeting on the BBL for any color temperature below that where the blue LED to phosphor line intersects the BBL. In Fig. 3 this is visualized for two combinations that result in a color point on the BBL with color temperatures of 2700K and 4000K respectively. In this way the color rendering properties of the light can be improved without significantly reducing the system efficacy thanks to the narrow spectral band of the red LEDs. Spread in the red and blue wavelength results in significant changes in the overall spectrum. Fig. 4 shows, for a target color point of 3500K, how the spectrum changes upon variation of the blue and red peak wavelengths within a 10 nm range centered around the nominal values. These spectra were derived from experimental spectra obtained with various blue and red wavelengths. This significantly affects the color point and color rendering properties. However, there are other parameters that play an even more important role with respect to the color point. Due to the strong temperature dependence of AlInGaP LEDs with respect to luminous flux and peak wavelength, temperature variation of the LEDs significantly impacts the overall color point. In addition, relative small variations in the drive currents lead to significant drift of the color point. Proc. of SPIE Vol A-6

7 Blue and red-orange LEDs in a module with a remote yellow phosphor; variable currents and housing temp BBL 0.41 Color point spread due to variation in T, I_b, and I_r (blue, yellow and red peak ref. condition fixed) K 0.40 BBL CIE y B C Increasing drive current of red LEDs ~> A) 100,130,160 ma; (136 BBL) B) 130,160,200 ma; (167 BBL) C) 160,220,280 ma; (227 BBL) D) 160,287,650 ma; (400 BBL) CIE x D A 350mA blue; 25ºC 350mA blue; 65ºC 700mA blue; 25ºC 700mA blue; 65ºC CIE y CIE x 3500 K 5 SDCM ellipse CP due to input variation Fig. 5 Resulting color points of light from an LED module when driven under different temperature conditions and drive settings. The module comprises blue LEDs, red LEDs, and a translucent yellow luminescent layer. The target color point is 3500K on the BBL. Parameters that were varied are the housing temperature, the blue LED drive current, and the red LED drive current. The blue and red LED peak wavelengths are 455 nm and 627 nm, respectively. Fig. 6 Color point spread of LED module with direct red emitters and (remote) yellow luminescent layer pumped by blue LEDs, as resulting from a model fit to the experimental data. LED wavelength variation and flux variation is not included. The housing temperature is varied between T 0-10 C and T C, and the red and blue LED drive currents are varied between 90% and 110% of I R,0 and I B,0 respectively, where T 0, I R,0 and I B,0 are the nominal values to reach the target (3500K). As a reference the 5 SDCM ellipse is shown for the target color point. In Fig. 5 some experimental color points around a target color point of 3500K are plotted. These color points result from a fixed set of blue emitters, a fixed set of red emitters, and a fixed remote phosphor plate. As the luminescent material, the well known aluminum garnet YAG:Ce was used. This configuration enables high color rendering properties. The only variables in this experiment are the blue and red LED drive currents and the case temperature. A model was fitted to these experimental data to analyze the transfer functions between the input variation and the resulting variation in the color point. Fig. 6 shows the resulting color point drift when the housing temperature is varied in a rather limited range of only 20 C, and the drive currents are varied from 90% to 110% of their nominal values. These nominal values were chosen such that at a case temperature of 45 C (i.e., the selected nominal temperature), with the nominal drive currents the target color point was achieved. The figure shows that even with these limited input variations the color point deviation is already far more than the maximum 5 SDCM limit. For an inaccuracy of the drive currents of ±5%, to remain within a 5-step MacAdam ellipse the total variation in the case temperature should be less than about 10 C in absence of any other source of color point spread. It is clear that this system configuration is not useful for general illumination when driven in an open loop configuration. Implementation of a control loop as well as accurate calibration of each individual system would be necessary to fulfill the color consistency requirements. 4. COLOR CONSISTENCY OF FORTIMO AND LEXEL DOWNLIGHT SYSTEMS The principles of remote phosphor application, described briefly in the previous sections and elsewhere [5] in more detail, have been applied to develop a downlight module that provides under operating conditions (i.e., thermal interface of the module at 65 C) 1100 or 2000 lm at 4000K with a CRI of 80. The downlight module has been developed together with a driver, resulting in the Philips Fortimo downlight system [6] that can easily be implemented by luminaire manufacturers in their products without having to deal with all the details and variations of LEDs. The Fortimo LED downlight system is depicted in Fig. 7. This system has become commercially available by fall Fortimo LED downlight modules have been specifically designed for use as a functional light source in luminaires designed for general lighting. The dimensions of the light module are equal for both versions delivering 1100 lm and 2000 lm, respectively. The use of different remote phosphor components allows for other CCT options in the near future. Proc. of SPIE Vol A-7

8 Fig. 7 The Philips Fortimo downlight system, comprising a light emitting module and a driver. The downlight module delivers 1100 or 2000 lm at 4000K with a CRI of 80. The heat is transferred via a thermal interface to an additional external heat sink. The Philips Lexel downlight system, depicted in Fig. 8, is the tunable version of the downlight module architecture of the Fortimo downlight system. The Lexel Downlight system offers full color tunability, including color temperature selection in the range from 2700 to 6500K. The typical CRI is larger than 80 for all color temperatures of 3000K and higher. The Lexel downlight system typically delivers 1000 lm at 3000K and 1100 lm at 4100K under standard operating conditions, i.e., with a case temperature of 65 C. The LED driver is placed in the same housing as the optical mixing cavity to secure EMC compatibility. The system has a dimming range of 100% to 10% of the initial output while maintaining color consistency. As will be shown below, the chromaticity accuracy of the Lexel modules is better than 5 MacAdam steps under all operating conditions. Fig. 8 The Philips Lexel downlight system, comprising a light emitting module comprising the multi channel driver, and a power supply unit comprising also the control interface. The downlight module is tunable in color, covering the color temperature range from 2700 to 6500K, and delivers 1100 lm at 4000K with a CRI of at least 80. The heat is transferred to a heat sink via a well defined thermal interface. Both downlight systems are future proof, meaning that the latest advances in LED efficacy can be easily implemented, without changing the luminous flux, the driver, and the dimensions of the LED module. Improvements in LED efficacy will be translated into reduced power consumption of the module. This represents stability for the luminaire manufacturers, and reassures to the lighting designers and specifiers that the Fortimo and the Lexel LED downlight systems will remain available for their projects in years to come. This is in sharp contrast with luminaires that use LED packages as the building block (i.e., as discrete components), as the LED development shows a strong evolution in performance which leads to the need for frequent redesign of the luminaires for given luminous flux and color rendering specifications. In the following two sections the color consistency of these two downlight systems is discussed in more detail. Because these products have already gone through their complete development phase (and introduced recently in the market), performance data on multiple units are now available including production spread and color point drift over time. Proc. of SPIE Vol A-8

9 4.1 Color consistency of the Fortimo Downlight system under varying operating conditions To determine the performance of production Fortimo Downlight systems, 4000 units were tested under reference conditions. The color points of these units were determined at their standard operating condition of 65 C, and plotted in Fig. 9. The target color point is 4000K, and the 5-step MacAdam ellipse around this point is also plotted to show the limits of acceptable color point deviations. We see that all modules have a color point that is well within the 5 SDCM ellipse (in fact, the spread is limited to about a 4-step MacAdam ellipse). This spread includes the influence of LED wavelength spread, spread in the composition and layer thickness of the luminescent layer, and other production spread. In addition, from the measurements the color points were determined at a case temperature of 25 C and 85 C. These results are plotted in Fig. 10. We see that the color point stays within 5 MacAdam steps from the target color point for all modules in the complete case temperature range of (at least) 25 C to 85 C. In practical operation the variation will even be smaller, as the system is provided with an automatic temperature limiting control loop. This limits the case temperature to a maximum value of 75 C by reducing the input power when needed. Colour distribution at standard operating conditions. Colour distribution at extremes of temperature operating range CIE-Y 0.38 CIE-Y CIE-X CIE-X 0.4 Fig. 9 Total color point spread of Fortimo Downlight modules at standard operating conditions with a heat sink temperature of 65 C. The data include production spread of 4000 modules. The 5 SDCM ellipse around the target color point at 4000K is shown as a reference. Fig. 10 Color point spread of 4000 Fortimo Downlight modules at the extremes of practical operating housing temperatures. Blue markers are for the spread at 25 C, red markers at 85 C heat sink temperature. The 5 SDCM ellipse around the target color point at 4000K is shown as a reference. Finally, for a set of 40 test units also the color point drift over time under standard operating conditions with a case temperature of 65 C was measured. For the sake of clearness, the initial color points of all these 40 units were translated to the target color point to show the drift over time independent of the initial production spread. The resulting color points shifts after 100 hrs, 300 hrs and 3000 hrs are plotted in Fig. 11. The initial spread at t=0 hrs is also shown as a reference. We see that the largest color point shift occurs during the first 300 hrs of operation, after which the color points stabilize. The total additionally induced spread over time is limited to about 2 SDCM. This means that at 25 C case temperature some modules are expected to show a color point deviation from the target of just somewhat more than 5 SDCM, but at all practical operating conditions and over time (and even up to significantly higher case temperatures than will occur in practice) the color point stays within a 5-step MacAdam ellipse around the target color point. Proc. of SPIE Vol A-9

10 Colour shift over time 0.39 CieY T=0 T=100 hr T=300 hr T=3000 hr CieX 0.4 Fig. 11 Color point shift of 40 test modules over time under standard operating conditions (65 C case temperature) after 100 hrs, 300 hrs and 3000 hrs. All shifts presented relative to the target color point at 4000K. The initial spread of 4000 modules and the 5-step McAdam ellipse at the target color point are plotted as a reference. 4.2 Color consistency of the Lexel Downlight system under varying operating conditions To evaluate the color consistency of the Lexel Downlight modules over time when driven at standard operating conditions, for 11 modules the color point was measured at 0 hrs, 168 hrs, 500 hrs and 1000h hrs. At these evaluation points in time, the actual color point was determined when the target color point of the module was set to 2700K, 3000K, 3500K, 4100K, 5000K and 6000K with a case temperature of 25 C, 45 C, and 65 C. The spread in color point due to production spread between these modules and due to the case temperature variations are plotted in Fig. 12 for the various target color temperature settings. The initial color points are plotted together with the color points after 168 hrs and after 1000 hrs. The production spread includes wavelength spread and luminous flux spread of the LEDs as well as sensor sensitivity spread. The data show that the initial calibration of the modules results in very accurate color targeting, with a deviation from the target color points within about 1 SDCM step. Within the first 168 hrs, in particular at the lower color temperatures the modules show some color point deviation, but this stays well within the 5 SDCM ellipses. After that, the color points do hardly change any more; the data at 1000 hrs overlap almost completely with the data at 168 hrs. The same holds for the measurement data in between that were collected at 500 hrs of operation (not plotted here). The system turns out to be extremely robust at color temperature settings of about 4000K, with a color point accuracy within 2 SDCM steps. At higher color temperatures the color accuracy is better than about 3 SDCM steps, and at the lowest color temperatures the accuracy is well within 5 SDCM steps. But although this is the maximum deviation from the target, the graph shows that the actual change in the color points is much smaller; this stays within a 3-step MacAdam ellipse at 2700K, within a 2- step MacAcam ellipse at 6000K, and within a 1-step MacAdam ellipse at 4100K. And after a burn-in period of less than 167 hrs the total color change in time and spread between modules including a 40 C temperature range of operation stays within a 2-step MacAdam ellipse at all color temperature settings (at least as far as has been determined with the measurements up to 1000 hrs of operation). Proc. of SPIE Vol A-10

11 0.43 Long Term Chromaticity Performance Pilot Units, Case temp = 25-65C Planckian 0.41 CIE-y SDCM ellipses at target color points 0 hr hr hr CIE-x Fig. 12 Color consistency of Lexel Downlight units, plotted as the actual color points of 11 units in response to setting the units at target color temperatures of 2700K, 3000K, 3500K, 4100K, 5000K and 6000K with case temperatures of 25 C, 45 C, and 65 C. Data sets were collected after 0 hrs, 168 hrs and 1000 hrs of operation at standard conditions. The 5-step SDCM ellipses at the target color points are shown as a reference. The flux over all the CCT points was measured for the 11 test units at a stabilized 65C case temperature. For color temperatures of 3000K and higher, the average initial output was ca lm with a CRI well above 80. After 1000 hrs of operation, the luminous flux had reduced to ca. 95% of its initial value. At 2700K, the luminous output is somewhat lower (average output of the 11 test units ca. 870 lm), but the flux depreciation was less than 1%. Therefore, the Lexel Downlight system is an accurate system with a very good color consistency that stays well within 5-step SDCM ellipses at all color temperature settings. 5. CONCLUSIONS Thanks to the robustness of our remote phosphor concept and our color mixing concept with closed loop control, the various downlight systems do not need any bin structure at system level as is the case for current LED packages. Multiple LED bins can be applied without causing the systems to get out of spec. Color consistency is in particular of importance for LED illumination systems. Many of those systems show unacceptable color point drifts when driven in open loop configuration. As for traditional lamps, the fixed color Fortimo Downlight system as well as the tunable color Lexel Downlight system have been designed to provide a color consistency within a 5-step MacAdam ellipse around their target color point. We have demonstrated that the Fortimo Downlight system, based on blue emitters and a luminescent layer mounted remote from the LEDs, is an inherently stable system. The color points of these modules, including spread in LED properties, luminescent layer properties, operating settings and operating conditions, remain well within a 5-step MacAdam ellipse at the target color point. The Lexel Downlight system, based on four independent channels with blue, green, red and white emitters is not stable in open loop configuration. However, a closed loop system, based on a combination of a temperature feed forward and a flux feedback control loop enables a very stable system. The color point has been shown to remain well within 5 SDCM from any selected color point on the Planckian locus under all operating conditions. Moreover, the control loop has been designed to allow for performance optimization with respect to the luminous flux, the CRI, or a combination of both. The Lexel Downlight system will by default optimize for CRI performance with the luminous flux as a boundary condition. Proc. of SPIE Vol A-11

12 Systems based on the alternative color concept of combining blue and red direct emitters with a luminescent layer mounted remote from the LEDs show intrinsically significant color point drift and set point errors. To enable applications for general lighting, a color control loop is necessary. While the control loop cannot compensate for deviations in the luminescent layer, this layer is produced using the same technology as for the Fortimo Downlight system. Therefore also these systems are expected to be stable when provided with an appropriate control loop. ACKNOWLEDGEMENTS The authors would like to thank Johan Ansems, Huub Borel, Paul Deeben, Wido van Duijneveldt, Jan de Graaf, René Hendriks, René Hochstenbach, Bart-Hendrik Huisman, Boike Kropman, Regina Mueller-Mach, Dennis van Oers, Jamshid Ourmazdi, Mart Peeters, Nico Peeters, Menno Schakel, Peter Schmidt, Wouter Schrama, Lars Waumans, René Wegh and Alex Wong for their valuable contributions in the realization of this work. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] A. Zukauskas, M. Shur and R. Caska, [Introduction to solid-state lighting], J. Wiley & Sons, New York, (2002). C. Hoelen, J. Ansems, P. Deurenberg, T. Treurniet, E. van Lier, O. Chao, V. Mercier, G. Calon, K. van Os, G. Lijten, and J. Sondag-Huethorst, Multi-chip color variable LED spot modules, Proc. SPIE 5941 (2005). L. Wang, M. Riemeijer, G. Calon, P. Deurenberg, T. Treurniet, E. van Lier, J. Ansems, O. Chao, V. Mercier, K. van Os, and G. Lijten, Multi-chip color variable LED linear modules, Proc. SPIE 6183 (2006). R. Mueller-Mach, G. O. Mueller, M. R. Krames, and T. Trottier, High-Power Phosphor-Converted Light-Emitting Diodes Based on III-Nitrides, IEEE J. Selected Topics Quantum Electron., vol. 8, No. 2, (2002). C. Hoelen, H. Borel, J. de Graaf, M. Keuper, M. Lankhorst, C. Mutter, L. Waumans, and R. Wegh, Remote phosphor LED modules for general illumination towards 200 lm/w general lighting LED light sources, Proc. SPIE 7058 (2008). W. M. Yen, S. Shionoya, and H. Yamamoto, eds., [Practical Applications of Phosphors], CRC Press, Boca Raton, (2007). C. Hoelen, J. Ansems, P. Deurenberg, W. van Duijneveldt, M. Peeters, G. Steenbruggen, T. Treurniet, A. Valster, and J.W. ter Weeme, Color tunable LED spot lighting, Proc. SPIE 6337 (2006). E. F. Schubert, [Light-Emitting Diodes], Cambridge University Press, Cambridge, 97-98, (2003). P. Deurenberg, C. Hoelen, J. van Meurs, and J. Ansems, Achieving color point stability in RGB multi-chip modules using various color correction methods, Proc. SPIE 5914 (2005). G. Wyszecki and W.S. Stiles, [Color Science Concepts and Methods, Quantitative Data and Formulae], 2 nd edition, J. Wiley & Sons, New York, (1982). C. Hoelen, Development of high-power, color tunable LED modules for high brightness applications, Proc. LED and Lighting Seminar, Int. Seminar on LEDs, Display, and Lighting (ISLDL) 2007, (2007). R. Mueller-Mach, G. Mueller, M. R. Krames, H. A. Höppe, F. Stadler, W. Schnick, T. Juestel, and P. Schmidt, Highly efficient all-nitride phosphor-converted white light emitting diode, Phys. Stat. Sol. (A), 1-6 (2005). P. Schmidt, A. Tuecks, J. Meyer, H. Bechtel, D. Wiechert, R. Mueller-Mach, G. Mueller, and W. Schnick, Materials design and properties of nitride phosphors for LEDs, Proc. SPIE 6669 (2007). K. Uheda, N. Hirosaki, and H. Yamamoto, Host lattice materials in the system Ca 3 N 2 -AlN-Si 3 N 4 for white light emitting diode, Phys. Stat. Sol. (A) 203(11), (2006). H. Watanabe, H. Wada, K. Seki, M. Itou, and N. Kijima, Synthetic Method and Luminescence Properties of SrxCa1 xalsin3:eu2+ Mixed Nitride Phosphors, J. Electrochem. Soc., 155(3), F31-F36 (2008). Proc. of SPIE Vol A-12