The IES is considering feedback from the LED industry on projection methods for testing and defining additional long-term behavior changes in LEDs that are not addressed in the current version of TM-21. JIANZHONG JIAO describes possible models and updates under investigation.
By providing a simple yet reliable method for making long-term luminous flux maintenance projections, IES TM-21-11 has greatly benefited the LED lighting industry (http://bit.ly/1MuDYuH). When TM-21 was established, the experts focused on the urgent need of capturing the luminous flux degradation behavior of LEDs so that producers, users, and specifiers had a widely acceptable approach to evaluating the luminous flux maintenance life for LED lighting products. TM-21 was established based on an objective engineering judgment and a statistical confidence. Now the IES is look to apply a similar practice to projection of the impact of current change to lumen maintenance, and to chromaticity and wavelength shift based on current and thermal impacts.
With the continuing growth of the LED lighting market and applications, projection methods for other long-term behavior changes in LEDs have been requested by practitioners. One of the requests is for the drive current or forward current impact on LEDs at a given case temperature. If the LEDs are tested per LM-80 at two different drive-current levels for evaluating the luminous flux change over time, but while in real-life applications the LEDs are operated at a different current level than was tested, the users would like to have a method to interpolate the test data so that these LEDs will not need to be tested again. A second request relates to wavelength or color change of LEDs over time. The methods for projecting these changes have been requested by the industry, and the IES Testing Procedures Committee (TPC) has initiated working proposals to develop the required projection methods.
Engineering models for projection
The TM-21 projection method is based on the physical picture of an LED. When light output from an LED changes over time, the physics behind the changes can be described as contributions from the LED die, phosphor, dielectric or optical materials, and the structures of the LED component. These contributors to the change in light output may play various roles in the different operational timeframes throughout the life of the LED. With the thorough understanding of each change contributor in the LED, the mathematical expressions for a single contributor or combined effects can be established. These mathematical expressions can be verified or proved by the experimental data. When writing TM-21, the experts developed eight mathematical models and adopted one empirical model, which are all engineering-based analysis models. These engineering models sufficiently describe the light output change over time, or luminous flux maintenance.
When asked which engineering model(s) can best describe the lumen decay behavior of all LEDs, it was clear that the answers are not simple due to the variations in the LEDs themselves. LEDs may all be constructed with dies, phosphor, materials, and structures, but each of these elements can be different. The challenge was determining which model could represent the most LEDs.
The team of experts first took an objective approach by running each model against the data being collected. At the time, there were more than 40 sets of LM-80 equivalent data contributed by the major LED manufacturers. A statistician in the team of experts made more than 500 runs for each set of data using each model. The calculated values from running the data with different models were compared with real data to judge the model's statistical confidence, or accuracy.
Once the team exhausted all the data and model runs, it was clear that no one model was superior to the others in terms of confidence level and accuracy. Based on this conclusion, the experts made an engineering judgment call to choose the simplest model. Such a model may present the worst LED lumen-decay case, yet it may have the least punishment to the most-luminous-flux-stable LEDs. The experts also made recommendations on the conditions of data usage: what test duration of data should be used in the model, what the sample size should be, and associated projected length of the test. With these conditions, the experts re-ran all data sets against the selected model to ensure the method was the most appropriate approach for all LEDs, ultimately resulting in the TM-21 method.
Potential updates for projection methods
All the engineering models and data studies were focused on the light output or luminous flux change as a function of the LED case temperature in developing TM-21. Due to the limitation of the data collection, the experts did not propose approaches or models for describing current dependency, color changes, and wavelength shifts. After four years into the practice of using TM-21, these new topics are on the table for discussion. IES is looking into the possibilities of the following technical topics for updating the projection methods.
The first technical topic is regarding the LED current dependency and interpolation. In early discussion on LED decay behavior, the priority was temperature dependency even though the experts knew that current also matters. Some experts believed current and temperature can be two independent variables. Both higher case or junction temperature and higher forward current should lead to faster lumen decay over time. However, the LEDs' individual dependency, either to the temperature or to the current, can be described separately. For the lumen decay contributors within an LED package, unlike temperature, the current impact may not affect all the contributing elements, or at least not in the same way. In general, current has more direct impact to the die and its impact to other elements may be indirect or none. Therefore, a new mathematical model may be needed to describe the current effects. Once model(s) are established, similar to the previous TM-21 development work, data collection needs to take place. The working group must obtain sufficient current-dependence data in order to validate and prove the model(s). If the conclusions are made and the IES is able to make current interpolation and projection recommendations, the information may be integrated into TM-21.
The second technical topic is the LEDs' color change over time. Recent information provided by LED manufacturers demonstrates the possibility of severe color changes or chromaticity shifts in LEDs being tested for more than 10,000 hours. Based on the changing behavior of the die, phosphors, optical materials, and structures, white LEDs may experience color shift toward green, yellow, or blue. Although the contributors to the color changes are from the same elements causing the luminous flux decay, the changing mechanism is different. Additionally, the chromaticity measures are using two-dimensional metrics - namely x,y or u',v' values. Therefore, new mathematical model(s) are needed to describe these changing behaviors. These models will need additional verification or proof with real test data with extensive testing hours. The Lighting Research Center of Rensselaer Polytechnic Institute conducted research and published information for color change projection on the LED luminaire level a few years ago. This information may be valuable for the IES Working Group in developing recommended methods in the new document for color change projection.
The third topic concerns the recent expansion of LED lighting in remote phosphor and horticultural applications (see the recent LED Standards feature "Stakeholders make progress on LED horticultural lighting standard"). More single-colored or monochromatic LEDs used for these applications are now tested per LM-80, just as for the white LEDs. These monochromatic LEDs are tested for their radiant flux or photon flux changes instead of luminous flux change. They are also tested for the peak wavelength or centroid wavelength shifts instead of chromaticity value changes. In general, the wavelength shifts over time mainly relate to the LED die and optical materials used in the LED package. These shifts may be described as one-dimensional changes, compared to x,y or u',v' value changes in a two-dimensional color coordinate. Over time, the wavelength of monochromatic LEDs may likely become shorter for red LEDs or longer for blue LEDs. In some observations, such a wavelength shift in comparison to the initial wavelength is often a positive value (wavelength increases) or a negative value (wavelength decreases), and it does not necessarily fluctuate like white LEDs. From the limited data obtained in testing monochromatic LEDs, the peak wavelength shifts within the test durations are often minimal, especially for blue LEDs. Whether these wavelength shifts should be negligible, how much change is considered to be negligible, and what method should be used to project or determine long-term wavelength shifts or stability are all topics of discussions in developing the new document.
The activities for developing projection methods at IES will use the same standard development principle - reflecting the industry's best practice. The goal is to be as objective as possible without speculation or subjective judgment. If no conclusions are reached for a viable model(s) that can describe the current impact for projection or interpolation, color changes or wavelength shifts, the IES may provide rationale or explanation as to why.
DR. JIANZHONG JIAO, director of Regulations and Emerging Technologies at OSRAM Opto Semiconductors, Inc., is an internationally recognized lighting expert. He has been actively involved in LED and SSL standard development activities. He serves as the past chairman of the SAE Lighting Committee, past chairman of NGLIA, past Chairman of the NEMA SSL Technical Committee, active member of IESNA Testing Procedure Committee, Roadway Lighting Committee, and Computer Committee, ANSI SSL Working Groups, Standard Technical Panel of UL8750, standard committees in IEEE, CIE USA, SEMI, JEDEC and other organizations. He can be reached at firstname.lastname@example.org.