Automated Scanning Acoustic Microscopy for Package‑Level Defect Screening of QFN Devices
Automated Scanning Acoustic Microscopy for Package‑Level Defect Screening of QFN Devices
Wednesday, October 7, 2026: 6:40 PM
Summary:
Semiconductor packages used in high‑reliability aerospace, defense, automotive, and medical applications are increasingly sourced as commercial off‑the‑shelf (COTS) devices, requiring additional up‑screening beyond standard manufacturing controls. For Quad Flat No‑Lead (QFN) packages, Scanning Acoustic Microscopy (SAM) is widely employed to detect internal defects such as die‑attach voiding and mold or lead‑frame delamination; however, inspection outcomes are often dependent on manual image interpretation, introducing subjectivity, variability, and significant labor burden. This paper presents an automated, pattern‑recognition‑based SAM screening approach applied at full JEDEC‑tray scale for QFN packages. Using commercially available SAM software, repeatable package features are identified and aligned despite part‑to‑part positional and rotational variation, enabling consistent defect quantification within defined regions of interest. Known defect modes are reliably detected and mapped at the tray level, allowing rapid identification and removal of non‑conforming devices. Compared to manual inspection, the automated workflow significantly reduced inspection time and operator dependence, with estimated labor savings of approximately $500 per 1,000 devices at ~95% yield. An attribute agreement analysis framework is proposed to further assess inspection consistency and agreement between human interpretation and automated screening, supporting the role of automation as a scalable, objective front‑end filter that complements engineering judgment.
Semiconductor packages used in high‑reliability aerospace, defense, automotive, and medical applications are increasingly sourced as commercial off‑the‑shelf (COTS) devices, requiring additional up‑screening beyond standard manufacturing controls. For Quad Flat No‑Lead (QFN) packages, Scanning Acoustic Microscopy (SAM) is widely employed to detect internal defects such as die‑attach voiding and mold or lead‑frame delamination; however, inspection outcomes are often dependent on manual image interpretation, introducing subjectivity, variability, and significant labor burden. This paper presents an automated, pattern‑recognition‑based SAM screening approach applied at full JEDEC‑tray scale for QFN packages. Using commercially available SAM software, repeatable package features are identified and aligned despite part‑to‑part positional and rotational variation, enabling consistent defect quantification within defined regions of interest. Known defect modes are reliably detected and mapped at the tray level, allowing rapid identification and removal of non‑conforming devices. Compared to manual inspection, the automated workflow significantly reduced inspection time and operator dependence, with estimated labor savings of approximately $500 per 1,000 devices at ~95% yield. An attribute agreement analysis framework is proposed to further assess inspection consistency and agreement between human interpretation and automated screening, supporting the role of automation as a scalable, objective front‑end filter that complements engineering judgment.