Automated Scanning Acoustic Microscopy for Package-Level Defect Screening of QFN Devices

Abstract

Semiconductor packages are widely used across aerospace, automotive, military/government, and consumer electronics applications. In high-reliability environments that demand long-term performance under harsh operating conditions, standard commercial manufacturing controls are often insufficient, necessitating additional screening of commercial-off-the-shelf (COTS) devices. While inspection methods such as X-ray and Scanning Acoustic Microscopy (SAM) are commonly employed, defect identification is still largely dependent on manual image interpretation. This reliance on manual screening increases the risk of data misinterpretation, missed defects, and elevated labor cost— particularly in high-mix, high-volume manufacturing. This paper presents a comparison of manual interpretation of SAM image of a Quad Flat No-Lead (QFN) package and automated, pattern recognition-based screening applied at full JEDEC-tray scale for QFN packages. Automated analysis reliably identified package features and defect modes, including die-attach voiding and perimeter delamination, while accommodating part-to-part positional and rotational variation. The approach significantly reduced inspection burden and operator subjectivity, with estimated labor savings of approximately $500 per 1,000 devices inspected at ~95% yield based on internal time studies.