Measurement of colour coordinates of LEDs used in the automotive exterior lighting

Article deals with dichromatic white light-emitting diode (LED’s) color coordinates used in automotive exterior lighting. This article also describes basic white automotive LED functionality and basic physical processes that create white light of these LEDs. It focuses on measuring color coordinates of white automotive LEDs with different temperature of LED and how the LED’s color depends on LED’s temperature. The article is comparing very important datasheet information of LED producers and values measured in the laboratory at university. The article contains statistical results of measurements and graphical representation of measured values and declared color bins which are very important for producers of headlamps for automotive companies.


INTRODUCTION
Color coordinates of LED are important when adjusting color temperature of various light sources in car head or rear lamps. New trends in automotive lighting combine various light sources as halogen bulbs, high-intensity discharge (HID, also known as xenon lights), and LEDs. Each of these sources has different color and each customer has different requirements on color matching of headlamp functions. Usually, color of the white light is expressed as color temperature. The temperature of an ideal black body radiator that radiates light is meant by the color temperature of a light source. For illustration, the color temperature of halogen bulbs is appx. 3500 K, the color temperature of HID is usually 4500 K and the color temperature of LED is changing from 3000 K to 6000 K, depending on producers and a LED type. Generally, bluish LEDs (from 4500 K to 6000 K) are usually chosen for car white lighting functions [1]. When more types of light sources are used in the car's headlamp, e.g. halogen bulb for high beam, HID for low beam and LED for position and day time running lights (DRL) functions, the color of light sources is becoming a very important design feature. It is also possible to measure the dominant wavelength, peak wavelength, correlated color temperature (CTT) or the color rendering index (CRI) [2,3], as well as spectral characteristics of the LED sources which were measured with the help of semi-empirical model based on the solid-state theory depending on the temperature. In this article, the composite author also dealt with an idea of temperature influences on colorimetric changes (chromaticity) [4]. Generally, the white LEDs for indoor illumination and similar applications are described by the parameters CCT or CRI [2]. White LEDs are usually described by chromaticity coordinates [5] which is only the way how Ì ISSN: 2088-8708 to characterize different white light sources. Measuring and unification of white light emitted by the source is not only the task for the automotive industry. In general, it is also the task for illumination applications [6].
In the past, the temperature efficiency practical examination of the white LED Luxeon REBEL under different temperature conditions were carried out. In the past, the temperature efficiency practical examination of the white LED Luxeon REBEL under different temperature conditions were carried out [7]. Measuring LEDs characteristics using the integration sphere has been carried out by many researchers, which is one of the most reliable ways how to measure LEDs [1,8,9]. Various researches about measuring LED junction temperature were also made. One of the ways how to control LED temperature is by measuring the forward voltage of LED which was in the oil bath of specific temperature [8]. It has been proved that output luminous flux substantially changes with the temperature by implementation of the LED into the integration sphere and brings in fatal errors in determining the LED properties [10]. Measuring the temperature of the LED transition with the help of various experiments was reached by contact and contactless measurement [11]. Measurement of optical and electrical properties and calculation of the transition temperature was attained in the publication [12]. Another approach is mathematical-LED junction temperature can be calculated from its forward voltage [9]. This solution is very accurate, but it is limited by accuracy of measurement equipment. Modelling of temperature changes of LED diodes by computational fluid dynamics (CFD) analysis was also discussed in the article [13][14][15]. The Fan et. al. focused on degradation of LED aging and changes in spectral power distribution and efficiency. The LED chip parameters and its aging over time, as well as temperature, phosphorus and polymer used for the LED were measured. The process of the shift of LED elements, which is not linear, was monitored, and non-linear state-space model was developed. It predicts the time of color shift failure (CSF) [16]. The physical properties of the LEDs in time were also tested. It was monitored how their parameters would change, e. g. illumination, current through the LED on operation time, and efficiency coefficient on operation time [17]. A team of authors Qu focused on the LED temperature measurement and aging of installations while using pulse width modulation (PWM) for the red, green, blue (RGB) LED. It has been found out that the PWM modulation technique can effectively control color/luminosity without the necessity of feedback [18]. Similar results were achieved by another team of Umar and others [19] who designed a DC-DC converter with PWM lighting control or Dyble [20] and Gu [21]. Other way how to measure LED temperature is by using a sensor which contacts with printed circuit boards (PCB) near LED solder point [22,23]. The measurements in this article aimed to find a process of measuring the color coordinates of white LEDs in the real applications in the company Varroc lighting system without using any special laboratory equipment. This process should be very simple easy enough for repeating even by a person that has only basic optical and electrical knowledge. For this reason, measuring of the color coordinate by integrating sphere and spectrometer and measuring the temperature of LED by the thermal camera was chosen and recommended to the company Varroc lighting systems. In the Table 1 we can see list of measured LEDs. Based on the measured data, conclusions and recommendations on how to work with different types of LED batches from different suppliers are made to minimize changes in the light emitted and negative impact on production.

AUTOMOTIVE WHITE LED
In the automotive industry, is usually used only one type of white LED-blue LED with luminophore that is converting blue light into white. In that case a luminophore, yttrium-aluminum garnet activated by cerium is usually used. This material is activated by blue light of the InGaN diode. In automotive industry, usually only one type of white LED is used-blue LED with luminophore. The luminophore is yttrium-aluminium garnet (YAG) doped by cerium in most cases. The YAG is activated by blue light from InGaN diode and it converts a part of blue light in yellow light. This yellow light is mixed together with the rest of blue light.

Int J Elec & Comp Eng
ISSN: 2088-8708 Ì 2713 Correct proportions of lights (blue and yellow) create white light. LED producers divide each type of LED into various groups according to the color coordinates in CIE x, y chromaticity diagram. These groups are called color bins. Differences in LED color and dividing LED into various bins is caused by technology of LED production. For LED production, a simple substrate is used. On the surface of the substrate thin crystalloid layer is grown. This process is called epitaxy. As a result, semiconductor forming die of blue LED is created. This semiconductor is cut to small pieces afterwards. The electrical connections are inserted and layer or suspension of luminophore is added into its surrounding. This setup is encapsulated and creates a white LED. The luminophore layer is thin in order 10 −2 mm, epitaxy processes create natural differences in LED properties [8].
Even the manufacturers are trying to minimize these differences; it is not possible to create highly consistent and strictly controlled LED production with the same characteristics. For this reason, LEDs are divided into groups with similar characteristics. These groups are called "bins" and they can be organized into three main groups: by luminous flux, color coordinates and voltage.

Sorting automotive LEDs according to colour coordinates
Manufacturers sort LEDs to color bins according to their x and y coordinates in chromaticity diagram defined in CIE 1931. These bins are defined usually by 4 points that create convex rectangle in chromaticity diagram. Pentagons or hexagons defining color bin are not unusual, too. Size of the bins depends on manufacturing quality-generally, the more accurate production, the smaller color bins can be produced.

EXPERIMENT REQUIREMENTS AND DESCRIPTION
The task of the measurement was to measure color coordinates of white automotive LEDs in the range of different temperature of LED and to evaluate the difference between color coordinates of LEDs in different temperatures. All measurements were aimed at automotive industry needs, so measuring of color coordinates was chosen for this purpose. A dark room was used to prevent direct light to affect the results of experiments. Measurements were made in a special dark room situated in VSB-Technical University of Ostrava. Sizes of the dark room were 2.6 × 3 × 2.5 m. Amount of photons in this room was only 7 photons per second (Declared dark counts of photodetector 1 cps). Setup scheme consists of LED, Peltier cooler, CPU cooler and fan as it is shown in Figure 1. The real final setup can be seen in Figure 2. All main components are described in the paragraph above. All experimental measurements were made 30× for each LED and each temperature. The types of LEDs used for measurements can be found in Table 2. All LEDs are certified for using in the automotive industry and are proven as light sources for the automotive industry. Seoul LEDs are a simple and cheap example of automotive LED; they are widely used owing to their low price and a good luminous flux. Luxeon Rebel LEDs are known because of their great thermal management and thermal stability. Osram OSTAR LED used in this experiment is powerful 4-chip LED with a very high luminous output. Aim of the measurement was to verify the hypothesis that the colour coordinates change with the growing temperature of LED. As the P-N junction is situated inside of the LED structure, temperature of the LED's surface T s was measured.  -Connecting electronic components (Peltier cooler, fan, LED) to power supplies.
-Setting the nominal current of LED according to producers datasheet.
-Turning on power supplies for Peltier cooler and fan.
-Monitoring LED's surface using thermocamera, regulation of temperature using Peltier cooler and LED's forward current until the temperature on LED's surface and forward current reach the values defined in vendor's datasheet.
-Measuring color coordinates using Avantes spectrometer and integration sphere.
-Changing LED's surface temperature by adjusting Peltier cooler current.
-Adjusting LED's forward current to nominal.
All results were placed into the graph with proper bin definitions and bin's tolerances. Osram OSTAR LE UW D1W4 01 nominal temperature was not measured according to the datasheet, because the vendor provides information about color shift according to temperature changes and the LED was already assembled on its cooler. This does not allow fitting it on the cooling system described in Figure 2. Results for each LED were compared for nominal temperature and heated LED. All results were evaluated using exploratory statistic methods of data analysis and the statistical data were calculated: Mean (µ), Median ( x), Mode ( x), 75% . Hypotheses were tested for 95 % confidence interval separately for x and y coordinates. The following chapters gradually describe the results of experimental measurements and then the processing by means of the Statgraphics Centurion XV. The different types of LED boards with LEDs were always labelled according to the bin to which the LED type fell. As can be seen in Figure 2, a zero series resistor was used for all the LED samples, and therefore, it was not necessary to perform exhaustive calculations to determine the electric current flowing through the tested LEDs.

Seoul semiconductors Z5
First of the tested LED was Seoul semiconductors (SSC) Z5 LED shown in Figure 3. In Figure 4 shows distribution of heat on PCB and areas on PCB used for the thermal management measurements by thermocamera for 25 • C, or for 93.9 • C on the LED sample SSC Z5 LED. The thermocamera was aimed at the LED. As can be seen on the left part of the picture, these areas have higher temperature as they are leading a heat away. On the right part of the picture there is the cooler that can not handle all the heat.  Table 2. After, we found the basic statistical functions and outputs it was proceeded to more thorough analysis based on hypotheses through t-test.
Hypothesis test for x-coordinates: Hypothesis test shows that temperature change has influenced color coordinates x and y for LED SSC Z5 U2C3H. Comparison of measured results for T s with data declared by the vendor is shown in Figure 5.

U2E7I
Statistical results for LED SSC Z5 binned as U2E7I are in Table 3, where we can observe that the changes range at very low levels.
Hypothesis test for x-coordinates: P − V alue = 0 0.05 −→ rejecting H 0 on confidence interval 95%. Hypothesis test for y-coordinates: P − V alue = 0 0.05 −→ rejecting H 0 on confidence interval 95%. Hypothesis test shows that temperature change has influenced color coordinates x and y for LED SSC Z5 U2E7I. Comparison of measured results of T s with data declared by vendor is shown in the Figure 6.  Figure 6. Comparison of measured values to company's information for SSC Z5 LED binned as U2E7I

U3C8I
The basic statistical analyses were again used according to the exploratory analysis through the Statgraphics Centurion XV software environment for the LED SSC Z5 binned as U3C8I. Statistical results for LED are shown in Table 4. Afterwards, the analysis through the t-test based on the hypotheses was proceeded to find out whether the mean values differ at a significant level.
Hypothesis test for x-coordinates:  Hypothesis test for y-coordinates:

U1D5H
Statistical results for LED SSC Z5 binned as U1D5H are shown in Table 5.
Ì ISSN: 2088-8708 P − V alue = 2.82712 · 10 −11 0.05 −→ rejecting H 0 on confidence interval 95%. Hypothesis test for y-coordinates: P − V alue = 0 0.05 −→ rejecting H 0 on confidence interval 95%. Hypothesis test shows that temperature change has influenced color coordinates x and y for LED SSC Z5 U1D5H. Comparison of measured results of T s with data declared by vendor is shown in the Figure 8.

Seoul semiconductors P8
For SSC P8 LED measurement, the same procedure was used as for SSC Z5. As this LED was only for demonstrational purposes, it was not binned in any color bin. Nominal current for this LED is 350 mA, lower temperature limit 25 • C, the highest measured temperature was 100 • C. Photo of SSC P8 can be seen in Figure 9, heat distribution on PCB is shown in Figure 10. Figure 9. LED seoul semiconductors P8 assembled on PCB PCB for this LED has little bit worse ability for heat leading. This is the reason why the holes for heat leading not clearly visible on thermocamera pictures. Statistical results for LED SSC P8 are shown in Table 6.
Hypothesis test for x-coordinates: P − V alue = 0 0.05 −→ rejecting H 0 on confidence interval 95%. Hypothesis test shows that temperature change has influenced color coordinates x and y for LED SSC P8. Comparison of measured results of different T s is shown in the Figure 11.   Figure 11. Comparison of colour coordinates between different temperatures of LED SSC P8

Philips Luxeon Rebel PWN1 0100 NTPD
Vendor Philips has other criteria for color coordinates measuring temperature. They recommend measuring the temperature of the LED using pad situated on the PCB near LED solder point known as T pad . For this Ì ISSN: 2088-8708 measurement, the same technology of measuring LED temperature is used for all other LEDs. Generally, the equation T pad = T s − 17.64 can be used for conversion between T pad and T s . Vendor determines inaccuracy in bin definitions 0.005 for both coordinates. PCB has a system of holes used for great thermal management, as can be seen in the Figure 12. As can be seen in Figure 13, has much better thermal management than previous LEDs. Heat is distributed evenly all over PCB. Statistical results for LED Luxeon Rebel NTPD are shown in Table 7.
P − V alue = 0 0.05 −→ rejecting H 0 on confidence interval 95%. Hypothesis test for y-coordinates: P − V alue = 0 0.05 −→ rejecting H 0 on confidence interval 95%. Hypothesis test shows that temperature change has influenced color coordinates x and y for LED Luxeon Rebel NTPD. Comparison of measured results of T s with data declared by vendor is shown in the Figure 14.  However, the temperature on the surface of LED has reached 120 • C, as shown in Figure 16. Once the LED reaches this value, the temperature of LED is stabilized and no significant swings of temperature can be observed. As this LED was assembled on the heatsink, it was not possible to add the Peltier cooler. Vendor provides information about color coordinates shift in LED's datasheet. Value for color shift for 120 • C and nominal current 700 mA is 0.012 for x-coordinate and 0.014 for y-coordinate. Vendor guarantees tolerance of color coordinates binning equal to 0.005. For verifying LED's binning, color bin definitions were shifted according to information about color shifting in datasheet. Comparison of measured values and shifted vendor's definitions can be seen in Figure 17.

RESULTS OF EXPERIMENT AND DISCUSSION
All experimental measurements proved that color coordinates of LEDs are changing according to the LED temperature. Average shift in coordinates is 0.011 for x-coordinate and 0.018 for y-coordinate. In these numbers, results of measuring two prototype LEDs have been counted-unfortunately, concrete results of these two prototype LEDs cannot be published as they are still in process of development. According to the hypotheses test that rejects equality of means of measured color coordinates we can claim that there is a shift of color coordinates caused by the change of LED's temperature. This shift is caused by change of the intensity of LED die. As mentioned before, the spectrum of white LED used for automotive industry is created by blue part of LED die and yellow/orange luminophore layer providing color conversion. LED die's luminous output is getting lower with rising temperature, but luminophore layer's optical properties do not change dramatically with temperature change. This leads to a fact that less amount of blue light is converted by the same amount of luminophore layer. This means the color coordinates shift to the yellow part of chromaticity diagram. Results of this experiment can be used in automotive lighting industry for matching colors of various types of light sources as the temperature of LED changes during car operation time. Head lamps of any cars with LEDs will be very important in the future in case of security issues of vehicle to vehicle (V2V) communications via visible light communication (VLC) [24][25][26].

CONCLUSION
The measurements of different characteristics and their dependence on the temperature of the P-N transition for the most commonly used types of LED samples in the automotive industry, or more precisely for manufactures of lighting components, were made within the experimental measurements. The first hypothesis through the t-test was that the change of the temperature of the P-N transition had an effect on the shift of color coordinates. This hypothesis was confirmed after processing the measured data with statistical methods. At the same time, it has been proved that the manufacturer Seoul designates incorrectly the temperature, at which the color bins are defined, as the ambient temperature. It has been shown that the temperature control of the P-N transition is more important than the ambient temperature control in terms of achieving values given in the data sheet. The verification also determined the average value of the color coordinate shift at the temperature change of the P-N transition for the selected LED samples. The color coordinates of the specified LEDs correspond to the bin definitions given by the manufacturers and into which they were classified. In the future it will be necessary to make further measurements of other LED bathes, as the latest trends in the development and use of lighting or automotive components are driven by the demand for high-quality radiation sources that enable to reduce energy consumption and, therefore, to achieve greener operation combined with further possible use of these radiation sources, e.g. for communication via VLC.