One of the cornerstones of non-destructive failure analysis of packaged integrated circuits, allowing an analyst a relatively simple way of examining the structural integrity of a device, is Scanning Acoustic Microscopy (SAM). By using tightly focused pulses of ultrasonic waves and analyzing the sound reflected by and transmitted through a sample, it is possible to create a detailed, accurate image of a packaged semiconductor device, showing any pockets of air or delamination that may contribute to early-life failure. SAM has been an invaluable tool in performing analysis on the types of parts that have traditionally been the most prevalent in the industry – plastic encapsulated, wire-bonded ICs. Though the industry may be shifting away from these types of devices in favor of packaging technologies like flip-chip ball grid arrays (FCBGAs) due to the more efficient use of bonding space and potential for increased thermal compensation, the SAM is not obsolete; indeed, with a few changes, SAM can provide invaluable data on these cutting-edge technologies.
One of the challenges posed by FCBGAs for scanning acoustic microscopy is the much smaller scale of features that must be inspected. Finding a defect that can be as tiny as a broken die bump requires much higher precision than SAM is traditionally known for. This increased spatial resolution is achieved through the use of ultra-high frequency transducers, which can provide sound at frequencies between 110 and 250MHz (depending on the design of the transducer). The narrower wavelengths of these transducers allow them to resolve even the smallest of defects that might be lurking on the device; however, these high-resolution images require additional interpretation that may not always be necessary on other devices.
Generally, the scanning acoustic microscope will have built-in algorithms for automatically identifying defects – it may look at the phase of the echo waveform, for example, or set amplitude thresholds that flag any areas where the returned sound pulse is too “loud”. While these algorithms are often sufficient for analysis of traditional devices, they may fail when applied to FCBGAs – the difference between good and bad devices is often small enough as to go undetected by any sort of automated inspection, and the different material composition of the FCBGA is a confounding factor that can prevent direct analysis of a waveform’s phase. Fortunately, the well-trained, inquisitive failure analyst has yet to be replaced by a mindless automaton (perhaps to the dismay of accountants and science-fiction writers everywhere); though a machine may not be able to detect the subtle signs of a dewetted die bump, a trained eye can pick it out from a lineup of properly formed connections with ease.
Though scanning acoustic microscopy has most often been associated with the failure analysis of more traditional semiconductor devices, it is more than capable of producing data for more modern processes. Indeed, these types of processes often benefit the most from analysis with SAM, since they are certainly much less mature than the traditional methods of packaging!