In Situ Warpage Measurement of Electronic Packages With Projected Speckle DIC and Deep Learning-Based ROI Identification
Yuhan Gao, Chuanguo Xiong, Weishan Lv, Kezhong Xu, Xin Lei, Fulong Zhu
Abstract
The integration of projected speckle 3-D digital image correlation (3D-DIC) into the field of electronic packaging reliability assessment represents a notable advancement in noncontact measurement techniques. However, the precision of these measurements is frequently compromised by specular reflections from solder balls, particularly when measuring substrate warpage with solder balls attached. To address this issue, this study proposes a method using a U-net-based network for automated identification of substrate areas. This method strategically excludes the solder ball regions from the region of interest (ROI) in digital image correlation (DIC) calculations, effectively mitigating the effects of reflective interference and thereby enhancing the precision of substrate warpage measurements. In addition, this study introduces a simple optical method for identifying solder ball regions and compares its error metrics with the proposed U-net-based network enhanced DIC (U-DIC) method, yielding insights into the measurement accuracy and applicable scenarios for each method. Experimental outcomes demonstrate the effectiveness of both methods, as evidenced by a decrease in the mean absolute error (MAE) of the warpage measurements from approximately <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$7~\mu $ </tex-math></inline-formula>m to approximately <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$2.6~\mu $ </tex-math></inline-formula>m with U-DIC and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$4~\mu $ </tex-math></inline-formula>m with the simple method. This development underscores the potential utility of both methods in enhancing the accuracy of reliability assessments in electronic packaging structures.