«SOLID-STATE LIGHTING PRODUCT QUALITY INITIATIVE THIRD EDITION SEPTEMBER 2014 Next Generation Lighting Industry Alliance LED Systems Reliability ...»
18 At the LED level, color shift can arise from changes in the emitter, phosphors used to convert emitted blue light into white light, clear encapsulant, and the package containing the LED chip, encapsulant, and the phosphor.21, 22 Common degradation pathways occurring in LED emitters that will affect color point stability include changes in the emission flux and wavelength. Likewise, aging-related changes in the emission flux and emission spectra of common phosphors used in LEDs can also cause a color shift in light produced by the LED. The magnitude and direction of these shifts is strongly dependent on the phosphor mix, the operating temperature of the phosphor, and ambient contaminants.
Packaging materials used to protect LEDs can also cause a color shift especially through the oxidation of the plastics used as encapsulants and lenses attached to LEDs. The same effect can also occur at the luminaire level where oxidation of plastics used for optical components such as lenses, diffusers, and reflectors can cause a shift in color. Polymer oxidation tends to produce disproportionately greater light absorption at shorter wavelengths (e.g., blue and green) than at longer wavelengths (e.g., yellow and red), which will shift the color point. The relative influence of these degradation pathways will vary depending on light source type (high brightness LED, mid-power LED, chip-on-board arrays, remote phosphor) and the materials and design used by the product’s manufacturer. The relative ease of oxidation of the materials commonly used in luminaire manufacturing is generally known,23, 24, 25 and adequate design choices can be made by consulting the manufacturers of optical materials.
Most, but not all, of the degradation processes associated with color shift in LEDs and light engines are temperature activated, so control of the LED junction temperature is important in maintaining a consistent color point for installations containing SSL luminaires. In reviewing the LM-80 data of LEDs for potential color shifts over the product lifetime, it is important to examine the propensity of the LED to experience color shift under the expected operating conditions (current and temperature) as well as higher stress conditions. The higher stress conditions can provide an indication of the long-term aging characteristics of the LED for longer test times under milder conditions. However, at this time, standard methods to project LED color point do not currently exist but are under development by standards organizations such as IES.
Other environmental stresses such as moisture or chemical exposure can also accelerate color shift in luminaires unless adequate precautions are taken to protect the optical materials from ingress into the optical cavity. Color shifts in some LED packages due to corrosion of silver surfaces that are used as mirrors to enhanced light extraction have been observed in the presence of sulfur or other corrosives materials.26 Common sources of sulfur that should be avoided in SSL luminaires include industrial chemicals and vulcanized rubber (e.g., used in gaskets). In a similar manner, volatile organic carbon (VOC) ingress can produce significant discoloration of LEDs resulting in color shifts and lumen depreciation.27 Common sources of VOCs include outgassing from adhesives used in luminaire assembly and environmental contaminants. LED damage due to corrosive chemicals is 21 van Driel and Fan, Solid-State Lighting Reliability.
22 G. Meneghesso et.al. “Recent results on the degradation of white LEDs for lighting,” Journal of Physics D: Applied Physics 43, no. 35 (2010): 354007. doi:10.1088/0022-3727/43/35/354007.
23 M. Yazdan Mehr et al. “Reliability and optical properties of LED lens plates under high temperature stress,” Optical Materials, 05/2014.
24 M. Yazdan Mehr et al. “Photodegradation of bisphenol A polycarbonate under blue light radiation and its effect on optical properties,” Optical Materials 35, no. 3 (2013): 504-508. doi: 10.1016/j.optmat.2012.10.001.
25 J.L. Davis et al. “Insights into accelerated aging of SSL luminaires,” Proc. SPIE 8835, LED-based Illumination Systems, 88350L (September 30, 2013). doi: 10.1117/12.2025295.
26 Osram Application Note, “Preventing LED Failures Caused by Corrosive Materials,” June 2013.
27 Cree Application Note, “Cree® XLamp® LEDs Chemical Compatibility,” Document CLD-AP63 Rev 5, June 2014.
19 generally irreversible while that due to VOC absorption may be reversible in some cases. In addition, there is a strong blue light stress influence on both failure modes but exact relations are unknown.
The luminaire power supply can also affect color shift, especially in luminaire systems employing red-green-blue (RGB) emitters and those with hybrid LEDs (e.g., cool white LEDs with red emitters). In addition to the differences in aging characteristics of different emitters, the power supplies operating the different LED colors may also age differently over time and produce a color shift in the luminaire output.28 This effect can be corrected to some extent in luminaires that have either active color control or a feedback loop to control color point.
Figure 4 illustrates the challenge in predicting color shift over time. The chart, extracted from a publicly available LM-80 test report, shows the chromaticity change of a particular LED package model. The data spans 9000 hours and measurements were made every 1000 hours. Each curve is the average of 3 identically-operated LEDs. The u’v’ coordinates were shifted to make trends easier to see. The starting point is indicated with a circular marker. In each case, in the early hours of life, the chromaticity shifted towards the lower left. Color change would then slow near the extreme lower left of each curve. For those LEDs that were driven harder (either higher ambient temperature or higher current), the chromaticity began to shift back towards the (upper) right again. It seems unlikely from these data that a model as simple as that used for lumen maintenance will be able to describe color shift. It is also not likely that all LEDs from all manufacturers will follow the same trend.
Figure 4. Chromaticity change of a single LED package model.
The data spans 9000 hours and measurements were made every 1000 hours. Each curve is the average of three identically operated LEDs. The u’v’ coordinates for each individual LED were shifted to make trends easier to see. The u' and v' grid scales indicated on the chart show the magnitude of the changes.
The starting point is indicated with a circular marker and shifts then follow the line. The legend shows the temperature and operating current for each group. Source: Philips Lighting
The previous edition of these recommendations contained an example failure-distribution chart that showed the frequency of various field failure modes that had been documented for a family of outdoor SSL luminaires from a manufacturer’s installed base. The overall failure rate of this product family was low, representing a 5% cumulative failure rate in the field across 7+ years of production deliveries. The chart showed the relative incidence of failure modes, valuable in trying to understand reliability.
Figure 5 below, an updated version of that chart, cannot be generalized across all types of products. Specifically, in this case, there is a fairly good-sized segment entitled “housing integrity” which generally would not apply to indoor or less-exposed products. Failure of housing integrity can possibly lead to LED package failure or to driver failure, artificially inflating those factors. A different type of product would show different relative failure mode incidences. Regardless of those caveats, field data is a valuable tool for the manufacturer to understand and monitor a given product’s performance in actual use. Such data is not generally published but may be available to the buyer from the manufacturer to support claims of reliability and life.
Figure 5. SSL luminaire failure modes, across 212 million field hours.
Source: Appalachian Lighting Systems, Inc.
The failure categories in this chart are defined as follows:
Driver (Power Supply) includes traditional power supplies and contains all failures related to the power supply or its inability to perform as specified by the luminaire manufacturer.
Driver (Control Circuit) includes control board(s) or other control devices, if they are separate and unique from the power supply. These devices are often used to split and/or electronically condition the output of the power supply, and in some cases may also include wireless, wired or other types of controls that monitor and/or manage the luminaire’s operational state.
As indicated in the first edition of these recommendations, the role of warranty is important and should be considered every bit as carefully as the performance criteria. Absent standard accelerated life testing, which is likely to be the case for most products, the user depends on a reputable, qualified manufacturer to provide a useful and effective warranty. Especially for smaller installations, this may be the only protection a buyer has.
During the warranty period, the manufacturer takes on the risk of the cost to replace a failed LED-based luminaire. After the warranty period is over, that risk is passed to the consumer. So, the provider and the buyer truly share in the risk of failure.
Since most failure modes directly relate to the numerous design decisions known only by the manufacturer, the manufacturer may be in the best position to estimate the time to catastrophic failure and the time for the light output to degrade to a set level. The time covered by the warranty, therefore, will relate to how conservative or aggressive the design is and issues such as serviceability and replacement cost. Since our first edition, most reputable manufacturers have had the time to gather the information needed to assess expected performance over time and offer an appropriate warranty. Warranties of up to 10 years are now available. Accordingly, when comparing two different LED-based luminaires, the warranty period should be an important part of the selection process.
At the same time, since design choices made by each manufacturer do strongly impact expected luminaire and replacement lamp life, it is incumbent on the buyer to carefully review and understand the warranty. Usually the warranty will provide a replacement unit but will not pay for the labor cost of uninstalling the defective unit and replacing it. On a more basic level, one must understand what the manufacturer considers to be a failure and how the user can determine if the unit has failed other than catastrophically. Lumen depreciation and other gradual failures mostly depend on operating hours, not time from purchase. In the case of a warranty that is stated as a fixed time, one should know how many operating hours per day the manufacturer has assumed and if that assumption is consistent with the contemplated usage. It is possible that a certain number of operating hours, not “10 years,” is a better description of what is warranted. If the unit is serviceable, the service and parts costs and expected frequency of service will also enter into the overall economics of the system.
235 RELIABILITY TESTING AND STANDARDS
Finding accelerated-testing methods that accurately forecast LED lumen depreciation and color shift has proven to be difficult. The previous section reviews efforts along those lines since publication of the last report. As far as the LED light source (package, array, or module) is concerned, we continue to recommend LM-80 data and the IES-approved TM-21 method to project lumen depreciation. In situ temperature measurement tests, available as a part of the LM-79 testing protocol, may be useful to buyers to assure that operating temperature is wellcontrolled and to compare products, although it does not directly predict lifetime. IES LM-84-14 is the approved method for measuring luminous flux and color maintenance of LED lamps, light engines, and luminaires, but color maintenance should be considered an “add-on” requirement, not necessary for many, if not most, applications.29 Accelerated methods do show promise for reducing test times of principal components, and many manufacturers have developed proprietary protocols for internal reliability verification. Temperature has been the most common accelerator, but there are limits, especially when considering that integrated products as failures not relevant for normal operation may be introduced. Individual manufacturers may use other stressors in their protocols, but in each case the acceleration factors must be determined and verified. Some existing standard methods may be available for characterizing driver electronics but may need modification to deal with the longer lifetimes expected for LED luminaires. The LSRC recommends users qualify their suppliers and ask directly for evidence of reliability for the products they purchase, but there is no new recommendation as to accelerated testing.
5.1 ACCELERATED TESTING
To describe system reliability one would need to test the reliability performance of both the components and the total system. If the total system is intended for long lifetimes, which is usually the case for LED products, a common way of tackling this requirement is to expose the device to sufficient overstress to bring the time to failure to an acceptable, practical level and then to “extrapolate” the information obtained under overstress to normal use conditions. Accelerated life testing (ALT) conditions (stressors) may involve a higher level of temperature, pressure, voltage, load, vibration, and so on, than the corresponding levels occurring in normal use conditions. One must use care in choosing the stressors and interpreting the results: meaningful failure distribution should not include root causes of failures beyond the limiting factors of the materials or design of the component or system itself. For example, if accelerated tests are performed at temperatures beyond the limit of the LED components, this should not be considered in the failure rate model.
There are basically two different reliability test approaches: