Statistical modeling and analysis of wafer test fail counts

• Published in: 13th Annual IEEE/SEMI Advanced Semiconductor Manufacturing Conference. Advancing the Science and Technology of Semiconductor Manufacturing. ASMC 2002 (Cat. No.02CH37259) pp. 266 - 271
• Main Author: Melzner, H.
• Format: Conference Proceeding
• Language: English
• Published: Piscataway NJ IEEE 2002
• Subjects: Applied sciences; Circuit testing; Data mining; Design. Technologies. Operation analysis. Testing; Electronics; Exact sciences and technology; Fabrication; Failure analysis; Integrated circuit technology; Integrated circuit yield; Integrated circuits; Life testing; Linear regression; Microelectronic fabrication (materials and surfaces technology); Principal component analysis; Semiconductor device modeling; Semiconductor electronics. Microelectronics. Optoelectronics. Solid state devices; Testing, measurement, noise and reliability; Statistical method; Microelectronic fabrication; Patterning; Test; Linear regression; Integrated circuit; Problem solving; Modeling; Wafer; Data reduction; Principal component analysis
• ISBN: 9780780371583; 0780371585
• ISSN: 1078-8743; 2376-6697
• DOI: 10.1109/ASMC.2002.1001616

This paper presents a yield analysis technique based on test fail counts, as these are the most comprehensive and fundamental yield data available. Obviously, this requires the analysis of large volumes of data. Using powerful statistical techniques, such as Principal Component Analysis (PCA) and Multiple Linear Regression (MLR), efficient data reduction is achieved. A basic concept for the modeling of both defect related and parametric fails is presented. Based on a real life examples, means, variances, and covariances of test fail counts are analyzed. As covariance turns out to play a significant role, it is further analyzed using PCA to work out major independent sources of variation. MLR is then applied to partition total yield loss, resulting in the complete representation of actual yield data by just a few relevant patterns. Identification of physical root causes is consequently greatly simplified and accelerated, leading to fast problem solving and yield improvement.