R. Pahwa, S. Gopalakrishnan, Huang Su, Ong Ee Ping, Haiwen Dai, D. Wee, Ren Qin, V. S. Rao
{"title":"自动空洞检测tsv从2D x射线扫描使用监督学习与3D x射线扫描","authors":"R. Pahwa, S. Gopalakrishnan, Huang Su, Ong Ee Ping, Haiwen Dai, D. Wee, Ren Qin, V. S. Rao","doi":"10.1109/ECTC32696.2021.00143","DOIUrl":null,"url":null,"abstract":"Yield improvement is a critical component of semiconductor manufacturing. It is done by collecting, analyzing, identifying the causes of defects, and then coming up with a practical solution to resolve the root causes. Semiconductor components such as Through Silicon Vias (TSVs) and other package interconnects are getting smaller and smaller with the ongoing miniaturization progress in the industry. Detecting defects in these buried interconnects is becoming both more difficult and more important. We collect both 2D and 3D X-Ray scans of defective TSVs containing defects such as voids. We label the data in 3D and perform registration between 2D and 3D scans. We use this registration information to locate the TSVs and void defects in these 2D X-ray scans which would be difficult to label manually as these voids look very fuzzy in 2D scans. Thereafter we use a state-of-the-art deep-learning segmentation network to train models to identify foreground (TSV, void defects) from the background. We show that our model can accurately identify the TSVs and their voids in images where it is impossible to locate the defects manually. We report a dice score of 0.87 for TSV segmentation and a dice score of 0.67 for void detection. The dice score for voids demonstrates the capability of our models to detect these difficult buried defects in 2D directly.","PeriodicalId":351817,"journal":{"name":"2021 IEEE 71st Electronic Components and Technology Conference (ECTC)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2021-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":"{\"title\":\"Automated Void Detection in TSVs from 2D X-Ray Scans using Supervised Learning with 3D X-Ray Scans\",\"authors\":\"R. Pahwa, S. Gopalakrishnan, Huang Su, Ong Ee Ping, Haiwen Dai, D. Wee, Ren Qin, V. S. Rao\",\"doi\":\"10.1109/ECTC32696.2021.00143\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Yield improvement is a critical component of semiconductor manufacturing. It is done by collecting, analyzing, identifying the causes of defects, and then coming up with a practical solution to resolve the root causes. Semiconductor components such as Through Silicon Vias (TSVs) and other package interconnects are getting smaller and smaller with the ongoing miniaturization progress in the industry. Detecting defects in these buried interconnects is becoming both more difficult and more important. We collect both 2D and 3D X-Ray scans of defective TSVs containing defects such as voids. We label the data in 3D and perform registration between 2D and 3D scans. We use this registration information to locate the TSVs and void defects in these 2D X-ray scans which would be difficult to label manually as these voids look very fuzzy in 2D scans. Thereafter we use a state-of-the-art deep-learning segmentation network to train models to identify foreground (TSV, void defects) from the background. We show that our model can accurately identify the TSVs and their voids in images where it is impossible to locate the defects manually. We report a dice score of 0.87 for TSV segmentation and a dice score of 0.67 for void detection. The dice score for voids demonstrates the capability of our models to detect these difficult buried defects in 2D directly.\",\"PeriodicalId\":351817,\"journal\":{\"name\":\"2021 IEEE 71st Electronic Components and Technology Conference (ECTC)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-06-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"3\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2021 IEEE 71st Electronic Components and Technology Conference (ECTC)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ECTC32696.2021.00143\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 IEEE 71st Electronic Components and Technology Conference (ECTC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ECTC32696.2021.00143","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Automated Void Detection in TSVs from 2D X-Ray Scans using Supervised Learning with 3D X-Ray Scans
Yield improvement is a critical component of semiconductor manufacturing. It is done by collecting, analyzing, identifying the causes of defects, and then coming up with a practical solution to resolve the root causes. Semiconductor components such as Through Silicon Vias (TSVs) and other package interconnects are getting smaller and smaller with the ongoing miniaturization progress in the industry. Detecting defects in these buried interconnects is becoming both more difficult and more important. We collect both 2D and 3D X-Ray scans of defective TSVs containing defects such as voids. We label the data in 3D and perform registration between 2D and 3D scans. We use this registration information to locate the TSVs and void defects in these 2D X-ray scans which would be difficult to label manually as these voids look very fuzzy in 2D scans. Thereafter we use a state-of-the-art deep-learning segmentation network to train models to identify foreground (TSV, void defects) from the background. We show that our model can accurately identify the TSVs and their voids in images where it is impossible to locate the defects manually. We report a dice score of 0.87 for TSV segmentation and a dice score of 0.67 for void detection. The dice score for voids demonstrates the capability of our models to detect these difficult buried defects in 2D directly.