May_EDFA_Digital

edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 2 6 Because anLED is adiodedevice, when current applied to anode (A) and cathode (K), before the voltage reaches the barrier potential (for this LEDmodel, it is about 2.4 V), the current will go through the leakage path. As the leakage is more like Ohm-contact rather than Schottky contact, the voltage on R leakage will increase as the current increases. However, when this voltage reaches the LED’s barrier potential ( V fw ), the LEDwill start to conduct and clamp the voltage between anode and cathode, so the current I leakage will stop to increase as the current I LED dominates, thus, ρ will start to increase and (1 — ρ ) decrease. As the driving current keeps increasing, the ratio between I leakage and I Device will become less and less. This means the efficiencyof the LEDs is increasingas thedriving current increases. This investigation revealed PWM dimming is used in the LED and LCM/TDM functional tester designed by up- stream suppliers, whichmeans the brightness function is tested by adjusting the duty cycle of the driving current. When an LED is on, it is always driven by constant rated current (13 mA in this case). Consider an LED with 50 µA leakage current; the energy efficiency drop is about 50 µA/13mA = 0.38%, which iswithin the LED spec and out of humandetection limit. [4] Therefore, in the LED supplier’s reliability test, output light intensity under rated current is used to judge the degradation and is not able to catch such a failure. The investigation also revealed that in the portable computer device product design, the LED strings are driven with DC constant current. Thus, analog dimming (DC dimming) is used. The LED driving circuitry on moth- erboards adjusts the DC current going through the LED for brightness adjustment. When changing the brightness to low level, the current going through the LEDwill decrease. Consider an LED string with 1 mA driving current, a 50 µA leakage current will cause 50 µA/1 mA = 5% energy efficiency drop, which is far exceeding the threshold and can be seen with the naked eye. Though this model could explain the test coverage gap between upsteam suppliers and final assembly test, it is not perfect, as the leakage will bring more negative impacts to further reduce efficiency. The local high tem- perature at leakage path will cause unbalanced current distribution, and impact the ignition location of an LED, for example. Analog dimming will bring uniformity and white balance issues, but it could remove the flickering impact of PWM dimming. In many new designs, hybrid driving is used: when LCM is working in high brightness mode, analog dimming is used; when LCM is working in low brightness mode, PWM dimming is used. Hybrid driving also reduces the risk of showing hotspots for LEDs with slight leakage. FAILURE ANALYSIS FINDINGS IN T5 SUPPLIER Along the long supply chain from finished goods to LED component (Fig. 1), failing LED samples are passed down tier by tier and initial failure analysis was done by the T5 LED supplier. Much of the previous research regarding leakage cur- rent caused energy efficiency drop are focused on elec- trostatic discharge (ESD) damage on the GaN die, as ESD damage is the most common failure mode for GaN LEDs. Failing WLEDs are de-capped (removing silicone glue and phosphor with chemicals) to examine the burn mark caused by an ESD event. By optical inspection, a tiny crack near the P-electrode end tipwas identified as a burnmark. Then focused ion beam (FIB) is performed at the crack location that confirmed the crack. Based on the findings, the T5 LED supplier concluded the failure as ESD damage (Fig. 6). As such, suggestions of corrective actions included ESD control and increasing silicone glue thickness. Fig. 6 T5 LED supplier FA findings: tiny crack near the P-electrode end tip.