TECH NOTES: Illuminating evidence - a breakthrough in power LED lifetime data helps manufacturers build and deploy reliable lighting solutions
A graphical representation of lifetime in terms of both drive current and junction temperature provides the optimum data for designers, says Steve Landau of Philips Lumileds.
For all the obvious benefits, however, there are also hurdles for the user to overcome before completing a successful design with power LEDs. One of the most significant of these is determining the relationship between drive current, thermal management and effective lifetime of a luminaire or installation that uses power LEDs.
It is this challenge that Philips Lumileds has sought to address with a new category of lifetime and reliability data. This new data set improves on the industry-norm information on lumen maintenance that power LED manufacturers have presented to date. And designers who have used the new data give strong backing for claims that it is easier to work with, and provides a higher level of confidence in operational lifetime forecasts.
B and L lifetime data
Every lighting designer is familiar with the 'mean time between failure (MTBF)' data commonly provided by conventional lamp manufacturers. Lighting designers use MTBF as a guideline to determine when re-lamping must occur. Such a simple rating is appropriate for conventional lamps, which tend to fail catastrophically after a relatively short period of time.
Power LEDs, however, behave differently; they rarely fail completely, but instead their light output declines gradually over a long period as a function of drive current and temperature. Thus the simple MTBF figure applied to conventional lamps is inapplicable to power LEDs.
Indeed, the complex relationship between drive current, temperature and light output makes it much more difficult for lighting companies to accurately model the behaviour of power LED systems than it is to model the behaviour of conventional lighting systems. To solve this problem, Philips Lumileds devised a new tool based in part on research by the Alliance for Solid State Illumination Systems and Technologies (ASSIST).
This new 'graphical reliability data' model consists of two parts, a statement of failure (B) and a statement of lumen maintenance (L). Failure in the case of a power LED is defined as lumen maintenance below a specified level. (It should be noted that, unlike the conventional lamp model, even when an LED is considered a ‘failure’ it is likely to still be providing useful light output.)
L is the minimum acceptable lumen maintenance figure as required by the application. So, if L = 70 (70% lumen maintenance) then an LED will be considered a failure if its lumen maintenance is 69% or lower.
Using these definitions, we can begin to describe the lifetime behaviour of power LEDs. For instance, if the LEDs are rated for B50/L70 at 50,000 hours, then we would expect that half of the LEDs in an array would have lumen maintenance below 70% at 50,000 hours.
While this B50/L70 measure has the benefit of being clear, on its own it is not sufficient to enable the designer to optimise an LED lighting design. This is because both the operational lifetime of power LEDs and their light output are dramatically affected by two factors: one is drive current – how much power you supply to the device; and the other is temperature, both ambient and internal to the LED.
Lighting engineers have to weigh up their system’s costs and performance requirements, and find a balance between the number of LEDs, drive current and temperature that provides the best commercial outcome.
In other words, for each luminaire design, a specified amount of light is required: this can be achieved by using fewer LEDs driven at a higher current, or more LEDs driven at a lower current.
But how much does a decrease in drive current extend the LEDs’ lifetime? How much heat needs to be extracted from the light source in order to hit the designer’s target lifetime?
The plot of B/L data can be described for virtually any combination. Certain applications – such as wall washing – cannot tolerate as much as 30% degradation in light output, and can require 80% or 90% lumen maintenance typically. Others – such as decorative lighting – can sustain a degradation of as much as 50% of peak light output, so different representations are made for different combinations of B and L.
One long-time power LED user that has used these new data sets is IST Ltd, a professional lighting company specialising in innovative architectural and entertainment lighting solutions. IST is on the verge of strengthening its commercial and domestic product offering with the launch of a range of downlighters which use LUXEON Rebel power LEDs.
Guaranteeing the level of light output for the lifetime of IST’s products is essential because its downlighter is designed to be ‘Part L’-compliant. This latest version of the regulatory document for new buildings in the UK came into effect in April 2006. This regulation includes provisions for the efficiency of light fittings, and their absolute light output.
Matt Fitzpatrick, director of IST, explains: "For commercial applications you need to reach 45 lumens/Watt, for domestic you need 40lm/W. The downlighter market is a high-volume, attractive market to play in, but to get adoption you need it as aggressively priced as possible while also meeting the requirements for Part L."
This presents a difficult design challenge: by driving fewer LEDs harder, IST could reduce the cost of the product, but this could put at risk Part L compliance if it meant lumen maintenance was compromised and total light output too quickly dropped below that specified by the regulations.
By the same token, if the company played safe – using many more LEDs at a greatly reduced drive current in order to greatly prolong peak lumen output – the cost of the product would go up beyond what the market could accept. It was essential to get the balance exactly right.
IST’s Fitzpatrick says that the data from Philips Lumileds allowed product designers to do a comprehensive cost/performance comparison on designs using different numbers of LEDs at different drive currents, using temperature data that they were able to get from prototypes.
"Based on the information from Philips Lumileds, I know what the light output is going to be and how many LEDs I need to put in to the product to get the light output I want, over the operational lifetime that I am guaranteeing to customers. It is exactly the kind of information that I believe the industry has been looking for," he says.
The ready availability of such data has given IST an important competitive edge, allowing it to get to market faster after a shorter research and development process. As Fitzpatrick explains, "without this information we would have to do more development work and more practical experimentation with prototypes to understand exactly how the LED performed over time in our application."
The impact of comprehensive lifetime data is also imperative in applications where premature failure could be commercially damaging. Chromatica, an architectural lighting manufacturer, has been working with power LEDs for more than three years. It was recently responsible for what, at the time, was the largest installation utilising Rebel LEDs in the world: an architectural lighting installation for an office complex comprising in excess of 35,000 LEDs.
Kevin Clark, director of Chromatica, explains: "It is critical to have factual information to predict the “end of life”, which for us means performance after 50,000 hours of operation. We guarantee our products for this duration, so we in turn have to be able to rely on our LED supplier’s data."
Clark says that this is a key reason Chromatica primarily uses LUXEON LEDs. "In our experience, Philips Lumileds supplies the most comprehensive data sets for power LEDs in the industry," he says.
The Philips Lumileds reliability tools are distinguished by their ability to predict operational lifetimes in any combination of the two key variables: drive current and temperature. If Chromatica were to use different data sets that did not expose the effect of current and temperature differences, the company would risk making false assumptions about the operational lifetime of its light sources. This in turn could lead to costly in-warranty failures – a commercially disastrous consequence of poor rating data.
The alternative would be to withhold or limit the lifetime warranty. As Clark says, such installations are "not the type of application we want to target, as the price pressures in them are much tighter."
Indeed, cases such as Chromatica’s highlight the importance for LED users of understanding the validity of the rating data that vendors supply.
Using the Weibull distribution function – a universally respected statistical technique – to extrapolate product performance, the data from Philips Lumileds shows an extremely low divergence between forecasted and actual performance. Based on the quantity of data it now holds, and using extremely prudent statistical principles, Philips Lumileds is able to accurately predict lumen maintenance for 60,000 hours with a 90% confidence rate. It is worth noting that data from fewer parts on test for fewer hours would support performance predictions that extended less far into the future.
The experiences of IST and Chromatica show that power LED customers should be aware of the parameters that affect device lifetimes, and demand that their suppliers provide comprehensive data showing the interactions of average lifetime, lumen maintenance, drive current and temperature.
Only by understanding and applying this crucial data can manufacturers and their customers continue to develop reliable lighting solutions that can be guaranteed for the whole of their intended lifetime.