A Solder Joint Reliability Model for the Philips Lumileds LUXEON Rebel LED Carrier Using Physics of Failure Methodology
International Symposium on Microelectronics
Light Emitting Diodes (LEDs) are quickly evolving as the dominant lighting solution for a wide variety of applications. With the elimination of incandescent light bulbs and the toxic limitations of fluorescent bulbs, there has been a dramatic increase in the interest in high-brightness light emitting diodes (HB-LEDs). Getting the light out of the die, with reliable color, while maintaining appropriate thermal control over a long service life is a challenge. These issues must be understood and
... hieved to meet the needs of unique applications, such as solidstate-lighting, automotive, signage, and medical applications. These applications have requirements for 15-25 years of operation making their reliability of critical importance. The LUXEON Rebel has been accepted as an industry leading LED product, widely used in Mean-Time-Between-Failure (MTBF) sensitive applications. Customers use various mounting platforms, such as FR4 Printed Circuit Board (PCB), FR4 PCB with thermal via's, Aluminum & Copper Metal Core printed Circuit Boards (MCPCB), Super MCPCB, etc. As in other LEDs, when mounting to a platform where a large Coefficient of Thermal Expansion (CTE) exists between the LED & the PCB, Solder fatigue could become an issue that may affect system level lifetime. In this paper we have examined extreme cases and how a solder joint can impact system level reliability. We have modeled the conditions and formed a means to predict system level reliability. We have compared the prediction modeling with empirical tests for validation of the models. It is vital to understand system level reliability factors to build lighting solutions that match the application and customer expectations. It is impractical to test LEDs and other components for 50k hours ~5 years since the device evolution is much faster than thaton average one LED generation every 12-18 month. Hence we need models and prediction methods .....