«SOLID-STATE LIGHTING PRODUCT QUALITY INITIATIVE THIRD EDITION SEPTEMBER 2014 Next Generation Lighting Industry Alliance LED Systems Reliability ...»
• Test-to-pass demonstration testing, or zero failure acceptance testing, is an approach in which the component or product must undergo a certain number of test cycles without the occurrence of failures.
Test-to-pass only provides pass/fail results, which do not provide any information with respect to reliability as a function of time (or cycles). These limitations are addressed by test to failure.
29 For more information on LED product and test standards development, see http://apps1.eere.energy.gov/buildings/publications/pdfs/ssl/richman_standards_lightfair2014.pdf.
For the principal components in a LED product, it is feasible to design a test-to-fail approach, possibly using meaningful accelerated tests. For LED products as a system, however, a test-to-pass approach may be the most practical method, as it is very difficult to accelerate failures in a fully assembled complex system without
introducing new failure modes or other complexities. Some generic rules for these tests are:
Each principal component in a system exhibits its own failure behavior that needs to be captured by:
o Experiments by using at least 3 accelerated testing conditions o Numerical/analytical models that describe the reliability physics or physics of failure Interactions between the components are captured by:
o Testing the total system o Slightly accelerating environmental user conditions in a physically correct manner In most industries, standard tests are used in order to quantify the reliability performance of the components and systems. Examples are the MIL standards for military and the JEDEC standards for electronics. Those methods would also apply to the electronic components of an LED system.
Accelerated life tests will tend to emphasize failure modes that are activated by the environmental stress values chosen for the test and may engage other failure modes that are unaffected by the test parameters. For example, temperature shock tests will tend to emphasize failure modes associated with mechanical processes of expansion and contraction of materials in the DUT. In contrast, a high temperature bake of the same DUT will emphasize the chemical reactivity and thermal stability of the materials but will not provide any information on compliance to temperature excursions. In this example, both tests can provide meaningful information if the proper interpretation is applied to the data.
Accelerated life tests will typically be performed on the DUTs until all of the devices fail. Because of the desire to achieve failure of the DUTs without creating new failure modes, accelerated life tests can be long and expensive to perform and interpret. In some cases, the accelerated life test may be terminated before all DUTs have failed in order to save time and money. However, sufficient failure rates must occur in the accelerated test to obtain a meaningful failure distribution that can be extrapolated to normal operational conditions. In accelerated testing, it is important to ensure that failure modes are determined for each DUT in order to build correlations to expected operational lifetime.
Screening tests (also known as “robustness” tests, which could be passed without failure) can also performed on devices in order to ensure that they meet some minimal performance threshold that may have an empirical correlation to expected lifetime. Screening tests are often less expensive and simpler to perform than full accelerated life tests because there is no requirement for any failures to occur. A typical screening test would involve exposing the DUT to one or more environmental stresses for a pre-determined period of time and simply counting the number of operational and failed parts at the end of the test. A common example of a screening test in the LED industry is the use of temperature shock tests to study solder joint integrity, and a typical screening test may last for up to 1,000 cycles between high and low temperatures without any failures.
255.1.1 ACCELERATION FACTORS
When accelerated tests are designed properly, the ratio of the expected DUT lifetime under normal conditions and what is observed under accelerated testing provide the acceleration factor for the test. Accurate determination of acceleration factors requires extensive experimentation. However, the investment in time and money to determine proper acceleration factors is often cost effective in the long run since it can reduce test time.
The calculation of acceleration factors will depend upon the type of accelerated test being performed. An overview of common acceleration factor calculations for SSL devices can be found elsewhere30 and more
detailed discussions can be found in several textbooks, such as Wayne B. Nelson’s Accelerated Testing:
Statistical Models, Test Plans, and Data Analysis.31 Assuming consistent test method standards have been established for the critical parameters, a valid objective is then to reduce the test burden for an SSL device with integrated electronic components by accelerating the test.
For example, the Arrhenius model, which is suitable to characterize many failures that depend on chemical reactions or diffusion, assumes that the time to failure is exponentially dependent on the ratio of the activation energy and the product of Boltzmann’s constant and absolute temperature, allowing a simple estimate of an acceleration factor. However, materials selected for the LED system may not able to withstand extremely high temperatures, forcing the use of lower temperatures (assuming the component cannot be removed) that would limit the amount of acceleration possible.
There are models for other accelerants such as humidity, voltage, vibration, and so forth. Each has limits depending on the materials in the system. The common relationship for any acceleration model, empirical or not, is related to the system design and materials used. Understanding the chemical and physical properties of the materials used in an LED system can provide the necessary guide to which tests can be performed to best accelerate aging and which models are best to interpret the results. This approach can lead to the determination of acceleration models/factors which do not exceed the limits of the established standard test method for the critical test parameters. Different parts of the luminaire may respond disproportionately to environmental stresses necessitating the need to test some system components separately. For example, LEDs and some other electrical components can be tested at a higher temperature than many plastics. In such instances it may be beneficial to test the system parts individually rather than test the assembled luminaire.
Color shift behavior is quite dependent on drive current and temperature, as noted earlier, and different LEDs behave in different ways. After 9,000 hours of testing, some LEDs are still shifting in different directions from others. In order to create a model for this behavior, data extending over a longer time span is needed.
We have already noted that the complexity of color shift data available suggests that predicting the changes will be difficult. While it may be possible to find suitable accelerants, there are a number of issues that need to be
better understood before this work will be able to advance much more:
Possible variations in the direction of shift over the product life.
Possible different shift mechanisms depending on operating conditions.
There is a variation in shift from LED model to LED model Accelerated tests would be of great assistance to chromaticity studies, but they may need to be quite product specific. The placement and processing of the phosphor can vary widely among suppliers, which alone will likely affect the behavior. In fact, because of the proliferation of designs and materials in LED luminaires, any highly accelerated test is likely to be quite specific to a product or group of products and is best applied to materials and subsystems rather than entire products, so a general standard protocol may be impractical.
5.1.2 TESTING INTEGRATED LUMINAIRES
For integrated luminaires, disassembly of the LEDs and drivers may be difficult. However, often the LEDs and drivers can be tested as a unit while still attached to the housing. While this does place a limit on the types of tests that can be performed, it can also be a convenient approach to investigate both an LED and a driver design.
In these cases it may be beneficial to remove any plastic parts (e.g., lenses) especially if temperatures above the limits of the component specification(s) will be used in the testing.
Historically electronic components have been stress tested for qualification in varied applications such as automotive, consumer electronics, and commercial usage. Military standard testing methods have been established for electronic components used in space and military applications. For general lighting in the U.S., the American National Standards Institute (ANSI) and the National Electronic Manufacturers Association (NEMA) have established certain standards for performance metrics of incumbent lighting technology such as metal halide lamps and new compatibility requirements for solid-state lighting.
ANSI C78.43-2013, American National Standard for Electric Lamps: Single-Ended Metal Halide Lamps.32 Sets forth the physical and electrical requirements for single-ended metal halide lamps operated on 60 Hz ballasts to ensure interchangeability and safety.
NEMA SSL 7A-2013, Phase Cut Dimming for Solid State Lighting—Basic Compatibility.33 Provides compatibility requirements when a forward phase cut dimmer is combined with one or more dimmable LED light engines (LLEs).
The SSL industry could, in principle, create new stress test standards for integrated electronics, but it may also be possible to leverage the existing electronic stress test standards for incumbent lighting electronics. There are clear advantages to the latter, but appropriate data are not always available. The type of reporting that would
be helpful has been described above, but the reasoning is as follows:
Electronic component manufacturers typically determine lifetime expectancy differently compared to, for example, the SSL industry TM-21 standard reference. Most electronic component manufactures quote MTTF/MTBF statistics for their customers. Alternatively, electronic component manufacturers may quote failure rates in billions of component hours of operation (i.e., FITs).34 The details of these statistics are very important.
Just like TM-21, statistical calculations require a sample size and test time for each different test condition, so electronic component manufacturers should provide the same testing information which established the MTBF or MTTF data. As an extreme example, if an electronic component manufacturer decides to test one million components for one hour and all the units survive, the MTBF is one million hours. This does not necessarily mean the same component will survive 50,000 hours in an integrated SSL luminaire. To gain more confidence in the MTBF value, a longer test time of the component should be provided. The TM-21 document is clear: 6,000 hours is a minimum test time for a minimum sample size of 20 units to establish a 6X acceleration factor for lifetime at the specific test condition for drive current and case temperature per LM-80. Establishing a similar test condition standard with respect to ambient temperature (or if possible the operating electronic component temperature would be best) and drive current conditions for the integrated electronic components in SSL luminaires would be helpful.
18.104.22.168 HANDLING LUMINAIRES BUILT FROM DISCRETE ASSEMBLIES Luminaires built from discrete assemblies that can be broken down into individual components provide the greatest flexibility in testing. Each component can be stressed independently allowing the maximum flexibility in available testing parameters.
Plastic parts are commonly used in lenses, reflectors, and other optical components in SSL devices. Knowledge of the polymer chemistry of these plastic parts is important in designing accelerated tests, especially if temperatures exceed the limits of the component specification(s). Some common polymeric materials such as polyolefins will warp and distort at elevated temperatures, which will impact the performance of the luminaire in testing. A handbook of material properties and recommended use temperatures can be consulted before testing to determine if the plastic parts in the SSL device can withstand the planned temperature excursions. If the plastic parts will be adversely affected by the accelerated life test, it is recommended that they be removed and, if possible, be reinserted prior to light measurements on the SSL device. When parts are added back to the luminaire, it is essential that no foreign materials are introduced to the device. The removed parts should be evaluated using other known methods.
Acquiring failure rate information about electronics products, either through field returns or experimental testing, is invaluable in determining the reliability of SSL devices. In the absence of such information there are a
As mentioned in the previous section regarding accelerated testing, the proliferation of LED luminaire products precludes the possibility of a universal model to predict reliability of any design. Nevertheless, it is feasible to develop limited models that adequately describe a particular product or group of products, and many manufacturers have done so to aid in their design-for-reliability process and to give them confidence in their warranty claims. This section is meant to be a brief overview of the considerations that go into a reliability model. Specific manufacturers may be willing to provide additional information on the analysis they use to give their buyers some assurance that their representations of reliability are well-supported.
Many textbooks are available that describe reliability, ranging from its history, (accelerated) testing, system reliability, reliability predictions, and reliability standards. Refer to the References section for a brief list. In this section, only the basic principles and those reliability theories that are important for LED products are discussed.