WP15

WP15 LUMILEDS | Alldatasheet

Document overview

• Manufacturer or author: Provided By ALLDATASHEET.COM(FREE DATASHEET DOWNLOAD SITE)
• PDF pages: 16

Technical content

WP15 Evaluating the Lifetime Behavior of LED Systems White Paper 20161201 ©2016 Lumileds Holding B.V. All rights reserved. White PaPer evaluating the Lifetime Behavior of LeD Systems The path to a sustainable luminaire business model One of the strongest propositions of power Light Emitting Diodes (LEDs) is their long lumen maintenance—their ability to continue producing light output for many years of use, in contrast to most conventional light sources, which force users to go through repeated and frequent failure-and-replacement cycles. The market perception of an LED’s reliability is reinforced by the widespread practice among lighting manufacturers of offering long warranties on their LED luminaires. Product testing regimes specified by industry standards make it possible for LED manufacturers to analyze lumen maintenance of a single LED, with great confidence, under virtually any operating condition. The most notable of these is IES LM-80 (LM- 80), which is an “approved method for measuring lumen depreciation of solid-state (LED) light sources, arrays and modules,” according to the U.S. Department of Energy of luminaires” and that it “Does not provide methods for estimation of life.” However, luminaire manufacturers have not had access to any additional information about LED behavior that would allow them to better understand and predict the “lifetime” behavior of the LEDs in their solutions. As a result, for some, lumen maintenance data has become a proxy for luminaire lifetime, which was clearly not the intent of LM-80 and is not an accurate assessment of luminaire lifetime. Using lumen maintenance data that describes how a single LED behaves can create unplanned business risks and potentially affect end-customer experiences with LED solutions. What is now known and understood is that an array of LEDs behaves differently than a single LED. Additionally, it is possible to account for the extremely slim chance that a quality power LED will fail completely. Luminaire manufacturers are also learning how to better account for the lifetime behavior of the many other components that are used including drivers, optics, mechanical fixings and housings. Each of these is also a factor in determining the lifetime of a luminaire (see Figure 1). Fortunately, responsible LED luminaire manufacturers are beginning to incorporate more detailed analysis and utilize the methods and tools that are covered in this paper so that they can more accurately project and report lifetime behavior. For Consideration To specify an LED array for 50,000 hours of operation while another system component is rated for just 25,000 hours raises engineering and manufacturing costs and does not maximize commercial opportunity.

the manufacturer’s bill of materials unnecessarily. environmental benefit through reduced power consumption. As an LED manufacturer, Lumileds cannot address the reliability of all the components in an LED system—no LED manufacturer can. domain, provide forecasts of lumen maintenance behavior, and separately of the catastrophic failure rates of our LEDs. information, luminaire manufacturers can design for reliability and align market promises with solution performance. reliability data as the basis for their luminaire warranties. combination of drive current and temperature from their test cells. to arrays of LEDs, so that manufacturers can know the probability of lumen maintenance failure of any given array. Figure 1. LED system reliability.

WP15 Evaluating the Lifetime Behavior of LED Systems White Paper 20161201 ©2016 Lumileds Holding B.V. All rights reserved. 3 the ‘failure’ modes of power LeDs, and their impacts on luminaire reliability The concept of ‘lumen maintenance’ is well understood in the LED lighting industry: the light output from power LEDs is highest when new, and declines gradually over time. A common specification for power LEDs is for 70% lumen maintenance (that is, output at 70% of its peak) after 50,000 hours of operation. All reputable power LED manufacturers conduct long-term performance testing of each variant of their products, and publish lumen maintenance data separately for each of these variants. A study of different lumen maintenance data sets from different manufacturers will reveal differences in their products’ performance. This is not surprising as there are stark differences between LEDs in terms of:

• the precise chemical make-up of the semiconductor and optical system [encapsulant and primary lens]
• structure of the LED die
• the chemical makeup and implementation of phosphor conversion
• the mechanical structure of the device
• the materials used and device’s thermal performance
• the materials used and consistency and quality of the LED manufacturing process In combination, these factors result in significant differences in LED performance both when new and over time. Power LEDs are not generic devices and will vary in all performance aspects from manufacturer to manufacturer. Indeed, lumen maintenance is a competitive battleground for LED manufacturers, as customers demand LEDs that sustain lumen output over longer periods, and under more stressful conditions (that is, higher temperature and higher drive current). An LED can be said to have ‘failed’ when its light output falls below a threshold expressed as a percentage of peak output. In this lumen maintenance ‘failure’ mode, of course, an LED is still producing light, but not at the specified level. But lumen maintenance is not the only failure mode of power LEDs: they can also fail catastrophically, just like a conventional light bulb, and just like every other electronic or semiconductor product. In the case of a product such as the LUXEON Rebel from Lumileds, the rate of this so-called ‘catastrophic failure’ is extremely low, so low that the myth that LEDs never fail is still widely believed. Nonetheless, should a catastrophic failure occur, it could be of material significance depending on system design and the nature of the application. In fact, there are a number of reasons why luminaire manufacturers should take account of catastrophic LED failure rates when designing a fixture. Perhaps the most obvious is that a catastrophic LED failure might leave a dark spot in an array. This could lead the user to the conclusion that the fixture is malfunctioning, even if a photometric examination showed that light output is still at or above its specified level. Second, the concept of lumen maintenance has already been rehearsed above; any catastrophic failure, which eliminates the light output of an LED, will add to the decline over time in a lighting system’s output. Third, and less obviously, a catastrophic failure in an individual LED can cause instant failure in a complete luminaire or section of a luminaire. In part, this depends on whether an LED fails electrically ‘open’ or fails electrically ‘short’. If it fails open, the power supply is cut off from every LED in the failed LED’s string (in other words, to every other LED connected in series with the failed device), and the whole string goes dark. When an LED fails ‘short’, on the other hand, current continues to flow through the string allowing the luminaire to continue functioning. LEDs like InGaN LUXEON Rebel LEDs can only fail short as there are no bond wires. When a device fails short, metal ions can still pass directly from anode to cathode, maintaining the integrity of the LED string’s electrical circuit. It should be noted at this point that the catastrophic failure rate of LUXEON Rebel LEDs is extremely low. A common “open” failure is due to a broken wire bond. Wire bonds are a feature commonly found in other power LEDs which therefore have two possible failure modes, “open” and “short.” Luminaire designers should also be aware of the risks inherent in connecting LEDs in parallel with each other, or in parallel strings, rather than in series. In a parallel topology, an electrical short of a single LED will cause the forward current to increase

an accelerating series of failures.

• Use a current source to drive the LEDs
• Avoid the implementation of parallel connections between LEDs or between LED strings. All LEDs should be connected in series or in smaller strings of series connected LEDs, each with its own current source. Provided these design for reliability practices are followed, the remaining LEDs in a string will continue to emit light as specified, even after one or more LEDs in the string have failed short. Long-term LeD performance testing: the foundation of reliability data The long-term performance of LEDs, then, is affected by the rate of lumen maintenance and by the incidence of catastrophic failures; in combination, these two effects lead to a reduction in light output over time across a population of LEDs. So the industry’s critical need is for a trusted process that allows for more accurate predictions of system performance and for higher confidence in the engineering and business decisions associated with a luminaire. But the user’s confidence in such a model, and in the data that lie behind it, is a factor of the thoroughness of product testing carried out by the LED manufacturer. Therefore we start with a description of:
• how Lumileds reliability models are derived from its tests
• what the raw test results tell users about LED behavior
• how lighting system designers can use this knowledge how reliability models are derived Each luminaire designer needs to know the predicted long-term performance of their chosen LED under the specific conditions existing in their design. There is an infinite number of such conditions, so LED manufacturers cannot test for all possible conditions. All LED reliability models are therefore the result of extrapolation from a base set of data. The extrapolation occurs in two dimensions: operating conditions (drive current, and board/junction temperature); and time. For Consideration Minimizing the number of failure modes that must be accounted for in the engineering process can save time, reduce costs and simplify the design for reliability process. FORWARD CURRENT (IF) TBOARD 0.35A 0.7A 1A 1.5A 150°C 120°C 85°C 55°C 25°C -40°C = Maximum ratings boundary, TJ ≤ 150°C, IF ≤ 1A = Operating Limit for 50,000 hour L70 expectation, TJ ≤ 135°C, IF ≤ 0.7A

Figure 2. Cells needed for LM-80-08 specification.

lumen maintenance performance available in the industry. median performance for a LUXEON Rebel LED at 85°C, 0.35A. 50% of LEDs, and understates the performance of the rest. information that luminaire manufacturers need in order to predict the lumen maintenance behavior of a population of LEDs. operating conditions. An example of such a probability curve is shown in Figure 6. Ta within –5°C of Ts, in accordance with LM80. Figure 5. Long-term lumen maintenance data and L70 extrapolation.

that cells “X” are higher operating conditions than the maximum ratings. Note that there are three blue lines. The heavy blue line shows the estimated catastrophic failure rates based on the actual tests. 90% Upper Confidence Level (or best case). Lumileds recommends the use of the more conservative 90% lower confidence line. Figure 8. Catastrophic failure distribution from catastrophic failure model.

management reduces catastrophic failure rates. Lowering the drive current also helps to reduce catastrophic failure rates. can choose the drive current supplied to LEDs, and they can control junction temperature at the LED via thermal management. to be expressed in combination. maintenance threshold, or through catastrophic failure. Such a probability curve is shown in Figure 13. combination of catastrophic failures in some LEDs, and gradual decline in output—the lumen maintenance effect—in the rest. probability of failure of a complete array for any given operating conditions and Lxx threshold. holding a hand of 10 cards, which is known to contain one ace. Draw one card from the hand: the probability that it is an ace is 10%. Figure 13. Combined lumen maintenance and catastrophic failure model.

(0.001%) chance that you will draw 18 aces. distributed within groups of LEDs, just as much as to the way in which aces are dealt from hands of cards. aces from the 60 hands of cards. arrays will contain none. One in every 1,000 luminaires will contain 18 failures. failure of a complete array of LEDs is flawed. ‘Monte Carlo method’ to accurately predict the probability of system failure based on known component failure rates. for each of the LEDs in the array. So the time to system Lxx occurs when the combined light output falls below this limit. and lumen maintenance, will cross the manufacturer’s chosen Lxx threshold after a certain number of operating hours. performing LEDs (in lumen maintenance terms) could also be different. Figure 14. Monte Carlo simulation for a 32-LED system.

WP15 Evaluating the Lifetime Behavior of LED Systems White Paper 20161201 ©2016 Lumileds Holding B.V. All rights reserved. 13 In fact, every random sample of 32 LEDs will perform slightly differently, producing a spread of performance across the population of 32-LED luminaires. What the Monte Carlo method does is to take many such random samples of 32 LEDs, and plot for each one the point at which it crosses the system Lxx threshold. The curve joining these many points provides a model for predicting the probability of system Lxx failure. This curve can be displayed graphically (see Figure 14). This graph, then, predicts the probability that any given combination of 32 LEDs will fail in terms of system light output— the combination of lumen maintenance failures and catastrophic failures—at the given operating conditions. This shows a spread of outcomes, from the worst performing array to the best performing array. A luminaire manufacturer can use this information to help determine, for instance, a sustainable warranty period based on a known number of LED array failures at the warranty’s expiration. And because Lumileds models for lumen maintenance and for catastrophic failure predict LED behavior for any combination of drive current and junction temperature, the model can determine the probability of system Lxx failure for any operating condition (within the data sheet range) and for arrays of any number of LEDs. This system reliability approach:

• provides a comprehensive picture of LED performance at the system (luminaire) level. Unlike LED reliability measures commonly published today, it shows for any given set of operating conditions and size of luminaire how the worst luminaire will perform, how the best will perform, and the spread between them.
• informs investment and risk-management decisions. It enables luminaire manufacturers to accurately predict the percentage of units that will fail to survive their warranty period. Previous reliability models that model only for median performance place large, unforeseeable risks on the luminaire manufacturer, because they do not show how far short of median performance any individual luminaire is likely to fall.
• enables luminaire manufacturers to quickly and easily optimize designs for reliability and cost. The simplicity of the approach allows users to quickly evaluate many different choices for drive current, temperature, system size and light- output target. Designers can, for instance, avoid over-specifying systems that would maintain system light output for much longer than use-case assessments suggest was necessary. We started with the assertion that the current lighting industry practice of taking LED lumen maintenance ratings as a proxy for a luminaire’s lifetime rating was flawed. Figure 15 is an example that shows the potential commercial impact of this practice. At a drive condition of 0.35A and a junction temperature of 85°C, the LEDs have a median L70 of 176,000 hours. Note from the lower left graph that there is a 15% probability of L70 lumen maintenance failure at 100,000 hours. So the manufacturer might set 100,000 hours as the warranty duration. As shown by the lower right graph, the catastrophic failure model shows that the LEDs when driven at 0.35A and at a junction temperature of 85°C have a probability of catastrophic failure of about 2% at around 100,000 hours. And in this example, the system failure probability curve produced by the Monte Carlo method shows that luminaires will perform better than individual LED lumen maintenance suggests—there will in fact be no system failures at 100,000 hours. But the same approach when applied to a luminaire in which the LED junction temperature is 135°C is commercially disastrous. At a drive condition of 0.35A and a junction temperature of 135°C, the LEDs have a median L70 of 150,000 hours. Note from the lower left graph that there is a 20% probability of L70 lumen maintenance failures at 100,000 hours. So the manufacturer might set 100,000 hours as the warranty duration. But Figure 16 shows that at the system level—rather than at the level of a single LED—that 30% of the arrays will have failed at 100,000 hours. As shown above in the lower right graph, the catastrophic failure model shows that the LEDs when driven at 0.35A and at a junction temperature of 135°C have a probability of catastrophic failure of about 8% at around 100,000 hours. This example shows that high operating temperatures are a strong driver for catastrophic failures, and these are the main reason that system light output rapidly drops below L70. The manufacturer now faces the commercially disastrous prospect of replacing a number of luminaires.

• driver module, which can either fail catastrophically or can cause the drive current to change over time
• electrical connections and solder joints, which tend to fail open, thus causing a complete string of LEDs to go dark
• secondary optics, which can degrade over time, reducing light output. (The rate of degradation is a function of the materials choices made by the luminaire manufacturer.) Light path changes also affect system light output, and weather and other environmental factors can affect the light path.
• inconsistencies of manufacturing (missing screws, etc.) can impair a thermal interface (thus increasing the system’s thermal resistance) or a light path (thus reducing system light output). Any increase in thermal resistance will affect junction temperature, which is a crucial factor in LED component reliability. Analysis of the reliability of each of these components in a system, and the time-to-failure of each across a population of luminaires, should be carried out in the same way as for LEDs. Such analysis should then be fed back into the LED selection and system design process, to ensure that the LED sub-system is not over-specified. For instance, for a given design, assume that the best-performing 95% of LED arrays will reach the Lxx threshold after 150,000 hours or longer. Assume again that analysis of the driver module shows that 5% of the modules will have failed after 50,000 hours, and as a result the manufacturer decides to warranty the luminaire for no more than 50,000 hours of operation.

Figure 15. Top graph shows Monte Carlo simulation of 32-LED system. Bottom graphs show lumen maintenance and catastrophic failure models for LED component.

longer than the product’s warranty requires. help to reduce the manufacturer’s bill-of-materials cost and produce a luminaire that is more fit for purpose. source, LED, is probably not the weakest link but in fact just one of a number of components whose lifetime must be understood. manufacturers can maximized design, minimize costs and appropriately warranty their systems. Representative or Lumileds Technical Support Manager to tailor reliability predictions to your specific application. Figure 16. Top graph shows Monte Carlo simulation of 32-LED system. Bottom graphs show lumen maintenance and catastrophic failure models for LED component.

©2016 Lumileds Holding B.V. All rights reserved. LUXEON is a registered trademark of the Lumileds Holding B.V. in the United States and other countries. lumileds.com Neither Lumileds Holding B.V. nor its affiliates shall be liable for any kind of loss of data or any other damages, direct, indirect or consequential, resulting from the use of the provided information and data. Although Lumileds Holding B.V. and/or its affiliates have attempted to provide the most accurate information and data, the materials and services information and data are provided “as is,” and neither Lumileds Holding B.V. nor its affiliates warrants or guarantees the contents and correctness of the provided information and data. Lumileds Holding B.V. and its affiliates reserve the right to make changes without notice. You as user agree to this disclaimer and user agreement with the download or use of the provided materials, information and data. WP15 Evaluating the Lifetime Behavior of LED Systems White Paper 20161201 About Lumileds Lumileds is the global leader in light engine technology. The company develops, manufactures and distributes groundbreaking LEDs and automotive lighting products that shatter the status quo and help customers gain and maintain a competitive edge. With a rich history of industry “firsts,” Lumileds is uniquely positioned to deliver lighting advancements well into the future by maintaining an unwavering focus on quality, innovation and reliability. To learn more about our portfolio of light engines, visit lumileds.com.