Commercial claims for LED-based products in terms of lumen maintenance are fully based on TM-21 extrapolations using LM-80 data. This paper indicates that there may be a risk in doing this as TM-21 only relies on the behavior of the average LED degradation, instead of taking into account the degradation of all individual LEDs. Therefore, we propose a more profound statistical analysis in order to make the appropriate step from TM-21 extrapolation to lumen maintenance on a product level. This is needed as some commercial claims are based on 10 years of warranty and some service bids provide periods of 20 to 25 years of operation. This paper reviews the different approaches currently available to perform lumen maintenance extrapolations. We propose a new method to analyze and extrapolate LM-80 data using a more profound statistical approach. Highlights The main highlights of the presented research are: • A new statistical method to extrapolate LM-80 data • The method outperforms the currently available ones as it is statistically founded • Five cases were executed and benchmarked with the TM-21 method • A full statistical acceleration model is now available for lifetime assessment of LEDs

No abstract

No abstract

Reliability is an essential scientific and technological domain intrinsically linked with system integration. Nowadays, semiconductor industries are confronted with ever-increasing design complexity, dramatically decreasing design margins, increasing chances for and consequences of failures, shortening of product development and qualification time, and increasing difficulties to meet quality, robustness, and reliability requirements. The scientific successes of many micro/nano-related technology developments cannot lead to business success with-out innovation and breakthroughs in the way that we address reliability through the whole value chain. The aim of reliability is to predict, optimize and design upfront the reliability of micro/nanoelectronics and systems, an area denoted as 'Design for Reliability (DfR)'. While virtual schemes based on numerical simulation are widely used for functional design, they lack a systematic approach when used for reliability assessments. Besides this, lifetime predictions are still based on old standards assuming a constant failure rate behavior. In this paper, we will present the reliability and failures found in solid-state lighting systems. It includes both degradation and catastrophic failure modes from observation towards a full description of its mechanism obtained by extensive use of acceleration tests using knowledge-based qualification methods.