Failure Analysis of Semiconductor Devices – LEDs, Photocells, and Beyond
Oftentimes, discussion of failure analysis services for semiconductor devices tends to focus on the most complex of devices – microprocessors with millions of transistors, intricately designed printed circuit boards, or the fantastically precise silicon sensors called MEMS (micro-electro-mechanical systems). The reality of the industry, however, is that the vast majority of electronic components are far more simple, running the gamut from passive discretes like resistors and capacitors to simple active components like light-emitting diodes (LEDs) and power transistors. Almost every television remote control system, for example, uses a combination of LEDs and photodetectors to allow a user to channel surf. Despite their relative ubiquity, these types of components are just as susceptible to failure as any other – and, therefore, failure analysis can be just as useful in their improvement.
Failure analysis of a semiconductor LED can actually be quite complex, as there’s really only one failure mode – insufficient or completely absent light output – that can be seen. For obvious reasons, photoemission microscopy is tailor-made for finding defects on LEDs. The photoemission system will often reveal that what the naked eye perceives as a lower-than-specified light output is in fact uneven illumination (in other words, only part of the LED is lighting up, not the whole device). This uneven lighting can often be attributed to a damaged junction; often, a cross-section of the failing device will show a junction that has been spiked with metal causing a localized short, an ESD flash point creating a leaky junction, or even a processing defect.
Similarly, failure analysis of semiconductor photodetectors often depends on leveraging the unique characteristics of these types of devices. For example, failure analysis of a photovoltaic die (usually an array of diodes that generates a voltage when exposed to light) may involve shining a pulsed light source onto the surface of the die and monitoring the output of the device with an oscilloscope. By characterizing the photovoltaic’s output, an analyst may be able to narrow down the type of failure afflicting the array. An output that is stuck low, for example, may imply something completely different than an output that reaches its nominal voltage, but decays back to nothing extremely quickly. At this point, an analyst must combine experimental data with experience and knowledge of semiconductor physics in order to successfully proceed with the failure analysis.
Though tearing apart cutting-edge technology is certainly the most (relatively) glamorous part of semiconductor failure analysis, analysis of more commonplace devices like LEDs is just as important. Indeed, finding the failure on an LED or photocell can be at least as complex, if not more so, than hunting down a failure on a microprocessor.
Derek Snider is a failure analyst at Insight Analytical Labs, where he has worked since 2004. He is currently an undergraduate student at the University of Colorado, Colorado Springs, where he is pursuing a Bachelors of Science degree in Electrical Engineering.