藉高速拍攝輔助晶圓檢測系統研究

Abstract

在半導體製程中，晶圓檢測的技術至關重要，在維持高精度的前提下，提升檢測平台的掃描速度有助於降低時間成本。然而，掃描速度的提升也意味著拍攝的影像可能出現動態模糊，本論文試圖從主動控制的角度，還原遭受動態模糊破壞的影像。 首先，將全域式快門相機架設至平台後，控制平台等速移動以拍攝動態模糊的影像，接著，對模糊影像進行光學變形校正，再使用Pan所提出的影像還原方法，還原得到清晰的影像。Pan的影像還原模型，包含影像灰階值和影像梯度資訊，其演算法是利用迭代的方式，從粗糙到精細輪流求出模糊核和潛在影像。 使用解析度測試圖作為拍攝對象，經由影像還原模型去除模糊後，可針對結果繪製MTF圖表，本論文設計多種實驗條件，以比較影像還原的成果。除了發展高速影像還原的技術，本論文也對晶圓檢測平台的馬達進行系統識別，再根據識別結果設計控制器，以降低平台移動的位置誤差。

Wafer inspection technology has its greatest importance in the semiconductor manufacturing. Under the condition of maintaining high precision, increasing the inspection stage’s scanning speed contributes to time and cost savings. However, the increase of scanning speed also implies that there might exist motion blur in captured images. By controlling the imaging setup, this thesis tries to recover motion-blur-degraded images. Firstly, the global shutter camera is installed on the stage, and the stage is controlled to move constantly in order to capture motion-blurred images. Secondly, the optical distortion correction is applied to the blurred image, and Pan’s image deblurring method is adopted to get the clear image. Pan’s image deblurring model includes image pixel values and image gradients information. In a coarse-to-fine manner, the algorithm iteratively solves the blur kernel and the latent image in turns. By selecting the resolution test chart and using the image deblurring model to remove the blur, modulation transfer function can be drawn from the result. This thesis designs several experimental conditions and compares their deblurred results. In addition to developing image restoration techniques, this thesis identifies motors’ system of the wafer inspection stage, and designs the controller based on the result to reduce the stage’s position error in motion.