CIS FBM Classification Using Machine Learning Algorithm and Specific Defect Prediction Analysis

• Published in: ASMC proceedings pp. 1 - 4
• Main Authors: Lee, Saerom, Lim, Jihye, Kim, Jinseok, Im, Sooseok, Son, Seungjun, Won, Seokjun
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2023
• Subjects: CMOS image sensor; Failure analysis; FBM Classification; Machine learning
• ISSN: 2376-6697
• DOI: 10.1109/ASMC57536.2023.10121108

As CMOS image sensor (CIS) manufacturing has been enlarged its quantities and scaled to a smaller pixel, not only the amount of failure test results has been increased, but it has also diversified. It is essential to interpret delicate test results in consideration of CIS characteristics, which are analog semiconductor chips. The importance of classification analysis and target analysis for defect failures is increasingly emphasized. Defective failure is able to be expressed as data called CIS fail bit map (FBM) through the several stages of image processing. In this paper, we introduce the automated CIS FBM classification system created by in a machine learning method. Also, focusing on the practical needs, this study aims to develop analysis methods based on fab data, utilizing a CIS FBM classification application. Robust performance is shown based on experiments using real physical failure analysis (PFA) data.