利用實驗設計法改善CMOS晶粒清洗過程微塵附著問題-以T晶圓封裝廠為例

Abstract

2013年智慧型手機出貨量可望上看9億支，平板電腦出貨量上看2億台，此情況帶來龐大的商品需求，也同時帶動影像感測器的強勁需求。因此現今普及的行動電話大多具有照相功能，這事實也反應了影像感測器無論在資訊產品、通訊產品、消費性產品、工業產品、汽車產品、航太產品的應用均已與人們的生活密不可分。CMOS影像感測器製程在最終檢查之前必須到清洗站進行清洗，清洗目的是清除附著於CMOS產品上的微塵物，因此微塵物對於晶圓封測廠而言，它就是品質的殺手，因為在CMOS晶片的表面上若有微塵物殘留，就會影響拍照的成像。因而造成CMOS晶片重工時間、品質成本的增加，嚴重時也會影響交期與產能利用率。為協助T公司改善晶圓微塵附著問題，本研究針對晶粒清洗過程找出影響微塵附著的因素，並利用2k-1部份因子實驗設計法規劃實驗，以變異數分析分析數據，加以得到最佳因子水準組合。最後將此最佳因子水準組合進行驗證實驗，發現藉由最佳因子水準組合設定，能有效降低微塵附著問題的發生。此研究成果除了改善CMOS晶片微塵附著問題以及降低重工時間、品質成本外，也希望提供給晶圓封裝廠在追求提升品質與降低成本的策略面向作為參考。

Shipments of smart phones and the tablet PCs are expected to exceed 900 million units and 200 million units, respectively. This situation has brought huge demand for commodities and CMOS chips. Now a days, most mobile phones have a camera function. It reflects the fact that the image sensor applications in the information products, communication products, consumer products, industrial products, automotive products andaerospace products have been inextricably linked with people's lives. CMOS chips play an important role in these products. Before the final inspection, CMOS image sensors must be cleaned in the cleaning station. The purpose of the cleaning process is to remove the adhering particles on the CMOS productions. As far as a wafer packaging and testing plants concern, particle is the killer of the product quality. If the surface of the CMOS chip has residual particles, it will affect the image of the picture. Consequently, Resulting in increasing the rework time, cost of quality of CMOS chips. Moreover, the residual particles on CMOS chips may further affect the delivery and capacity utilization. In order to improve the problem of particle-adhesion chips, the significant factors that affect particle adhesion must be identified first in the cleaning process.. The fractional factorial design is then utized to find the optimal factor-level combination to effectively improve the particle adhesion problem. The CMOS production process of a Taiwanese wafer packaging company is utilized as a case study to demonstrate the effectiveness of the proposed method. The results of indicate that the optimal factor-level combination obtained by the designed experiments can actually improve the CMOS chip dust adhesion problem and reduce rework time and cost of quality. Furthermore, the proposed method can be used as a reference for wafer packaging plants in making their quality improvement and cost reduction strategies.