Reliability Study of LED driver

• Main Author: Lan, Song
• Other Authors: Jong Ching Chuen
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/65371
• Institution: Nanyang Technological University

Solid-state lighting (SSL) is expected to replace the traditional incandescent and fluorescent light sources. In contrast to the conventional light sources, LED chip promises better efficiency and longer lifetime. Unfortunately, LED chip is not the life-determined component and LED driver circuit is reported to be the weak point of SSL system. However, LED driver reliability is seldom reported. There are two types of the LED drivers, namely linear-mode driver and switched-mode driver. The reliability of switched-mode LED driver is studied through the aging of the external electrolytic capacitor at output node without considering the internal circuit degradation. Thus, this method is not suitable for predicting the reliability of linear-mode driver as the electrolytic capacitor is not used for this type of driver, and understanding the degradation of the internal circuit of the LED driver becomes important. Generally, IC reliability is studied via examining the reliability of interconnects, dielectrics and transistors, and their degradation impact on circuit performance is simulated. However, this approach may not be possible for a complete circuit because circuit designers will design the circuit to be fault tolerance, and also multiple failure mechanisms can occur simultaneously. Thus, the conventional approach can lead to inaccurate conclusion. In this work, we employ a "top-bottom" approach, where the circuit is studied first and circuit health index is defined. Based on the health index degradation, possible failure mechanism can be identified. In this dissertation, a comprehensive review on LED drivers’ reliability, IC reliability and conventional reliability study approach are presented first. The objective of the literature review is to provide basic understanding of the possible failure mechanisms for different components in IC and LED drivers. Besides, the conventional acceleration testing method and extrapolation model using statistical approach are also presented. The literature provides important reference for researchers who wish to study the IC reliability. To study the reliability of a commercial linear-mode LED driver, a black box testing method is proposed, in which the detailed circuit information is not required. Black box test is "top-bottom" method that the circuit degradation index is studied first. From the degradation mechanisms and stress conditions, the cause of the degradation and possible failure mechanism can be found. Black box testing method is important to IC reliability study. First of all, it simplifies the circuit analysis, since the equivalent circuit is constructed using basic block diagrams and the circuit information is typically confidential to reliability engineer. Besides, without circuit information, the conventional IC reliability study, which uses "bottom-top" approach, is no longer practical. To accelerate the LED driver aging process, the test units are stressed with high voltage/high temperature. Under high stress condition, the degradation of the LED driver I-V knee point voltage is observed, where the knee point voltage increases with time. This knee point voltage is the minimum required output voltage to properly bias the LED driver. The knee point degradation is due to the increase in output impedance caused by either electromigration in the metal interconnect or hot carrier injection of output transistor. Although, these two failure mechanism have similar failure mode, it can be differentiated according to the shapes of their different degradation paths. For hot carrier effect, the degradation follows power law relationship and the degradation path shape is concave and the degradation rate decreases with stress time. For electromigration, the degradation path shape is convex, and the impedance increase is normally observed at later stage, which is close to break-off point. In this work, it is found that under high voltage low temperature, the degradation is caused by HCI only. Under high voltage and high temperature, both EM and HCI are involved in the degradation. The failure mechanism are verified by using circuit simulation, TCAD device model and failure analysis. Based on the failure mechanism, an analytical degradation model is derived for high voltage/room temperature stress condition, where HCI is the only failure mechanism. The degradation model is verified through the fitting of the degradation data with the proposed model. As the degradation model is accurately estimated, it can predict the knee point voltage degradation in time-span. To improve the estimation accuracy, particle filter is implemented with non-linear mixed effect estimation (NLME). By using particle filter, the degradation is characterized by a stochastic dynamic system. The degradation state is represented by a set of weighted particles, where the particles with high weight are close to the true state. Thus, the measurement noise/residual deviation is reduced. On the other hand, NLME is a regression tool used to estimate the degradation model parameters by maximizing the marginal likelihood function of the random parameters. Thus, the bias from single test unit is minimized. With the proposed method, the saturation of the model parameters’ values is observed. The saturation of the model parameters’ values is consistent to the saturation mechanism of HCI and it can indicate the minimum required test time. The estimated degradation model after the saturation point is verified to be accurate in this work. The knee point voltage is an important health index to the driver, since the driver cannot maintain its output current as constant once the knee point voltage exceeds applied voltage. Therefore, the lifetime of the LED driver can be estimated through knee point degradation. In this dissertation, a new physical-experimental lifetime extrapolation model is proposed based on circuit degradation mechanism, where the correlation between the degradation model parameters and applied stress are deduced. Thus, the lifetime under different applied stress can be predicted through this correlation.