Analysis of the evaluation of LED light output intensity decline

Abstract: Scientific evaluation of LED light output degradation and color change to ensure the reliability of LED use requires not only the establishment of predictive models, but also the need to guarantee a certain test time. This paper mainly introduces several prediction models of output light intensity degradation caused by LED current accelerated aging experiments, and analyzes its limitations. At the same time, the results of predicting the lumen maintenance of power LEDs by Weibull distribution function are discussed. It is hoped that through these studies, the industry will be able to propose more predictive models to more objectively describe the changes in light output and color versus time of the lighting industry for the lighting industry.

1 Introduction

With the continuous rapid development of LED technology, especially the power and efficiency of white light, semiconductor lighting devices represented by high-power LEDs have been rapidly developed in recent years, and in the past few years in different lighting systems (for example, export Signal lights, traffic lights, channel letter symbols and other display forms are increasingly used. Compared with traditional lighting sources, LEDs have many advantages such as long life, low driving voltage and fast response time. LED efficacy is generally 80 to 90 lm/W [1, 2], which is significantly higher than the efficiency of 100W incandescent lamps (17 lm/W), although there are reports that the efficiency of incandescent lamps can reach 30 lm/W by 2010. [3], still far lower than the current LED efficacy. Compared with fluorescent light (85 ~ 105 lm / W), the best white LED (90 lm / W 100W) is still slightly lower than the efficacy of fluorescent lamps. However, with the rapid improvement of the efficiency of solid-state light sources, especially the improvement of material growth technology and the improvement of design level, it is inevitable to produce more efficient illumination sources. For example, 131lm/W LED (8W) has been reported here [ 4] and 150 lm/W (9 lm) [2] efficiency, it is expected that by 2020, LED light source will penetrate into the general lighting market, the efficacy of white LED is expected to exceed 200 lm / W [5].

LED In addition to the many advantages mentioned above, there are also many factors that affect the quality of its efficacy, such as the optical initial variables, temperature and electrical properties in the same batch of LEDs, and the inconsistencies in their associated variables over time. Behavior, these factors need to be considered in the structural design, production and maintenance of LED lighting products. For example, for a new batch of LEDs, the initial optical properties may vary due to defects in the growing material and manufacturing processes. Changes in junction temperature or external environment will affect the LED's light output characteristics and electrical parameters, and the brightness and color of the LED will also decay as the ignition time increases. As a result, different LEDs from the same batch may have different light output intensity degradation rates and color changes. If multiple LEDs are integrated into one lighting system, this change will inevitably result in LED array space color and luminous intensity. Non-uniformity, which can be a big problem for large color displays, especially for various imaging systems.

This paper summarizes several predictive models of current output light intensity degradation based on current accelerated aging reported. The limitations were analyzed and the results of the LM-80 lumen maintenance of the power LEDs predicted by the Weibull distribution function were analyzed. It is hoped that through these analyses, the industry will be able to propose more predictive models to more objectively describe the changes in light output and color versus flash time for the lighting industry.

2 Current aging test for light intensity decay assessment

The degradation of the LED spectrum can significantly affect the performance of LED products. The degradation of the LED spectrum, especially the variability of the decay process of the same batch of LEDs, such as differences in package thermal resistance, current load and working substrate temperature. Since the factors mentioned above may affect the quality of LED products, it is necessary to find an objective method for objectively evaluating the quality and performance of LED products. There are many reports of causes of spectral degradation, such as an increase in the internal junction temperature of the LED due to high applied current stress, and an increase in the non-radiative recombination center of the illuminating region, causing spectral degradation [6-8]. Some scholars believe that the excessive conversion of heat from the PN junction results in a decrease in the efficiency of phosphor conversion, which leads to the degradation of the LED spectrum [9]. In addition, the degradation of encapsulated epoxy resin is also considered to be the cause of LED spectral intensity degradation [10,11], because the use of epoxy packaging materials to limit the LED operating temperature can not exceed 120 degrees. Because of the variety of white LEDs on the market, whether from the growth of the front chip or the packaging of the back channel, the heat transfer capability from the chip to the surrounding environment is different. Therefore, it is reasonable to estimate that different products have different decay rates for junction temperatures.

Currently, the best LEDs will have a lifetime of up to 10,000 hours, and it is somewhat impossible to verify the spectral intensity as a function of ignition time under normal operating conditions. The use of accelerated aging quality assessment methods can quantitatively evaluate the important factors of LED quality in the shortest possible time. Junction temperature is an important reason that affects LED light attenuation. The rise of junction temperature will cause LED light to decay quickly. LEDs generate more heat under high current operating conditions, which accelerates aging. There have been many reports on temperature and LED electrical, optical (intensity and color) characteristics. Among them, the mode of decline of LED intensity with the ignition time is the most interesting and often debated focus. At present, many prediction models for output light intensity degradation based on current accelerated aging are successively proposed. Below, we will mainly discuss the following four evaluation models.

Yanagisawa [12] used The three-parameter model studies the accelerated degradation behavior of GaN-based blue LED light output under applied forward current stress conditions. The model is based on intensity decay as a function of time t and current i. Through an experimental study of a single type of low power GaN white LED, it was found that under 40, 60, 80 mA current stress, the test time of decay to 50% lumen maintenance and the estimated time of the model showed a good correlation. It is a pity that the validity of the model prediction results is not verified due to the lack of experimental data under normal current stress conditions to verify the model.

Chuang [13] proposed a theoretical kinetic model for the attenuation of light output of II-VI compound semiconductor LEDs under continuous current aging conditions. The model takes into account the effect of applied current and the increase in LED PN junction temperature on the light output. By comparing the experimental results with the low-power B-LED packaged in a 3mm resin case, the theoretically obtained results are basically consistent with the results of the experimentally attenuated light output caused by current aging. However, the model is purely based on theoretical derivation, which is derived from the linear relationship between the LED light output intensity decay mode and the initial quantum efficiency of the LED, the applied current, and the junction temperature of the LED. When using this model for theoretical evaluation, it is required to experimentally test the initial light of all LEDs used, or know the linear shape of the decay mode in advance. However, the linear shape of the decay mode is difficult to do because the decay mode of the LEDs between different types sometimes changes very much.

More importantly, the model is only to investigate the suitability of II-VI compound semiconductor LEDs, GaN-based III-V compound semiconductor LEDs to be further verified. Because the II-VI compound semiconductor LED represented by AlInGaP and the III-V compound semiconductor represented by GaN have great structural differences, early research results show that GaAs and GaAlAs substrate LEDs emit intensity in accelerated life experiments. The attenuation is proportional to the square root of the ignition time. However, such a relationship is not very consistent with GaN-based LED related data [14]. This brings difficulties to the universality of the model prediction.

Narendara [15] used current acceleration tests and changes in junction temperature [16] to estimate LED lifetime. The method is to fit a normalized intensity decay curve by an exponential model. Experimentally, the attenuation coefficient obtained by selecting different injection currents or different LED junction temperatures To determine the parameters of the model, and then predict the life of the LED under normal operating conditions.

(2) where i is the applied current, For the attenuation factor at any operating current, The attenuation factor for the junction temperature for different jobs. Although the experimental results are relatively consistent with the predicted results, there is a fatal flaw in using the tested attenuation coefficient and exponential extrapolation to obtain the prediction of LED life at any current or any junction temperature. The prediction error cannot be evaluated.

Grillot et al. [17] passed The study of the decay of luminous intensity proposes the following empirical formula for qualitatively evaluating the relationship between luminous intensity and applied current stress.

In the formula (3), J represents the current density applied to the LED chip, and D1, D2, D3, and D4 are constants irrespective of the current stress and the applied current stress time. Experimentally, experimental results of applying different current stresses to LEDs have found that as long as the applied current stress is long enough and the current density is high enough, the results obtained by prediction and measurement are also very consistent. The downside is that although this decay model considers the effect of applied current and applied time on the LED light output characteristics, it does not consider the effect of the increase in junction temperature of the LED on the test results due to the applied current. And we also know that LED junction temperature is one of the most critical factors affecting the photoelectric properties of LEDs.

In summary, all of the prediction methods have been proposed for one purpose, which is to obtain objective LED life predictions in the shortest possible time. Since many factors can contribute to the attenuation of LED light output, the output light intensity decay model based on current accelerated aging should incorporate as many reliable evaluation methods as possible to identify LED degradation parameters to identify different decay modes of LEDs for more scientific classification. The current problem is that manufacturers often classify LEDs according to their initial strength, color, and forward voltage. However, with the pursuit of LED lighting vision and the increasing use of LED lighting products, this classification method clearly does not meet the needs of application development, because the variables given on the trademark deviate from the lighting requirements after classification. Very big. In addition, LEDs typically experience rapid intensity changes at the beginning of the work phase and then enter a more regularly degraded mode. Although this change is not useful for most products, for higher-demand LED lighting products, the purchased LEDs can be aged under normal operating conditions and then classified.

3 Weibull distribution function predicts the lifetime of power LEDs

There is no term for lamp life in the IESNA Lighting Manual, but traditionally the time to 50% test lamp failure is usually defined as the lamp life. Unlike traditional lighting sources, LED light sources have very few catastrophic failures, but their light output slowly decreases with the ignition time. Based on the previous study of the human eye's response to reduced light output levels [18], and considering the LED luminous intensity and color decay as a function of junction temperature under rated load current conditions, ASSIST gives two LEDs for general illumination. The definition of component and system life defines the lifetime of the general lighting facility as 70% lumen maintenance, but this definition is not intended for all lighting sources. For example, for decorative lighting sources, the lifetime is defined as 50%. The lumens lasted for a while.

Recently, Philip took the lead in releasing the LM-80 test report for the maintenance of lumens online according to LES LM-80-08 "Methods for Measuring the Approval of LED Light Sources" [19]. The report gives detailed conditions for uniform stress conditions and injection current conditions ( For example, in the DB03:LM-80 test report on page 5, the stress conditions during the test are: 85 ° C, 0.35 A, susceptor temperature Ts = 85 ° C, ambient temperature Ta = 84 ° C) 80 LEDs at 1000 h, 1500 h, The light output data of 2000h, 3000h, 4000h, 5000h and 6000h is collected, and the data is normalized by the 24h output luminous flux, and the service life of the power LED is predicted by the Weibull distribution function. However, from the results of the index extrapolation, it is unacceptable that there is a negative L70 life. Also listed on the published DR-03 with Value. among them, For the epitaxial regression coefficient, For the correlation coefficient, Represents the regression square, if The closer to 1, the better the degree of fit. DR-03 single LED Value (for example, stress conditions at the test: 55 ° C, 0.35 A, susceptor temperature Ts = 69 ° C, ambient temperature Ta = 73 ° C), in the 80 test samples Less than half of the value of 0.5 indicates that the degree of fit is not good, which implies that at least for the sample batch being tested, the use of such an extrapolation model does not predict the life very well. Moreover, the decline of LEDs is not static. One mode continues. Some LED index declines may occur after several thousand hours of ignition, and Cree also reported that the light-passed mode of decline after 5,000 hours There are significant differences compared to the early days [20].

Due to differences in materials and manufacturing processes, the lumen maintenance performance of LEDs from different manufacturers may vary greatly. LEDs are tested after 1000 hours of ignition, and then the lumen maintenance of LED L70 or L50 is estimated by exponential extension to estimate the LED lifetime. It is not universal, at least from the results already reported. Perhaps for some manufacturers' LED products, 2000 hours, 3000 hours or more is the best time to start the test. Therefore, it is necessary to study in depth the factors affecting the luminous flux maintenance rate of different types of power LEDs, and hope to use these studies to propose a more universal prediction model to more accurately describe the LED lumen maintenance characteristics for the lighting industry.

4 Conclusion

The different thermal resistance of the LED external packaging material will inevitably lead to different junction temperatures under the same current drive, which will inevitably result in different spectral decay rates of the same batch of LEDs. In addition, the temperature gradient on the current loop of the end product design is The difference will also result in a difference in junction temperature between the LEDs in the module, causing different LED spectral degradation rates. Therefore, it is highly desirable to evaluate the spectral intensity and color degradation of LEDs over time in the same batch, especially for LED products that require higher spectral stability, such as LED color displays and The illumination of the instrument's visual system, etc. As stated earlier, the long life of LED light sources is a major advantage. Not only a single LED, but the final integrated system will also work longer due to the use of long-life LEDs. But one thing is certain, the long life of a single LED does not mean that the entire lighting system has a long life after the entire integrated system. How to ensure the stability of the LED system while also having a longer life, there is still a lot of work to be studied here. Among them, the selection of quality parameters affecting optical performance and the quality reliability evaluation model involving single LED and multiple integrated LED universality are still to be further studied. In short, in order to truly enter the field of large lighting, in addition to high standard initial characteristics and good cost performance, LED must also be supported by convincing reliability. In the near future, LED as a major high-efficiency illumination source may become the mainstream of lighting applications, so understand and evaluate the factors that affect LED quality, especially the stability of the intensity during use and the consistency of color for the future of LED Development is becoming more and more important. As long as we thoroughly understand the various factors affecting the quality of LEDs, we design a reasonable structure for the mechanism of light output attenuation and color degradation, select appropriate high-reliability raw materials, and strictly follow the process conditions and operating standards. It is certain that LEDs with high inherent reliability can be manufactured to meet the growing market demand.

references

[1] Nichia Corporation URL http:// (30 March 2008)
[2] Seoul Semiconductor URL http:// (30 March 2008)
[3]DOE publishes updated R&D plan for solid-state lighting
Http:// (05 Apr 2010)
[4] General Electric Company New high-efficiency lamps targeted for market by 2010 URL
Http:// (30 March 2008)
[5] Cree Inc. URL http:// (30 March 2008)
[6] Uddin A, Wei AC and Andersson TG , Thin Solid Films 483 378-81 2005
[7] Cao XA, Sandvik PM, LeBoeuf SF and Arthur SD, Microelectron. Reliab. 43 1987-91 2003
[8] Rossi F et al, J. Appl. Phys. 99 053104 2006
7
[9]Meneghesso G, Levada S, Zanoni E, Scamarcio G, Mura G, Podda S, Vanzi M, Du S and
Eliashevich I, Microelectron. Reliab. 43 1737-42 2003
[10] Narendran N, Gu Y, Freyssinier JP, Yu H and Deng L, J. Cryst. Growth 268 449-56 2004
[11] Chen ZZ et al, Phys. Status Solidi b 241 2664-7 2004
[12] Yanagisawa T, Microelectron. Reliab. 37 1239-41 1997
[13] Chuang SL, Ishibashi A, Kijima S, Nakayama N, Ukita M and Taniguchi S, IEEE J.
Quantum Electron. 33 970-9 1997
[14]Yiting ZHU and Nadarajah NAREENRAN, J.Light & Vos.Env.Vol.32, No.2, 2008
[15] Narendran N, Bullough JD, Maliyagoda N and Bierman A, J. Illum. Eng. Soc. 30 57-67 2001
[16] Narendran N, Gu Y, Freyssinier JP, Yu H and Deng L, J. Cryst. Growth 268 449-56 2004
[17]Grillot PN, Krames MR and Zhao HM, IEEE Trans. Device Mater. Reliab. 6 564-74 2006
[18] Akashi Y and Neches J, J. Illum. Eng. Soc. 33 3-13 2004
[19]Philips, “DR-03: LM-80 Test Report,” http://, 2004
[20]Cree, “XLamp Long-Term Lumens Maintenance,” Cree, Inc., 2009