全自動非穿透式面對面精密定位設備開發之研究

Study on Developing Automatic Non-penetration Face-to-Face Precise Alignment Equipment

指導教授 : 黃榮堂    研究生 : 許哲豪  自動化科技研究所 91年


摘要

  當兩物體需要精密定位結合在一起時,一般通常會令兩物體擁有特定的對位標記,再將兩物體以非常靠近的距離,利用穿透式對位來完成,由於機構位移誤差非常小,因此可得到極佳的對位效果。但是如果遇到像覆晶封裝或微機電封裝這類結合物體皆為不透明物時,常見方式會利用兩組攝影機分別取像再計算機構所需運動行程,再將物體置放於定點。如此雖可達到定位目標,但通常經過大行程的運動後,其誤差也相對提高,因此通常需要操作者再進行調校才能達到可接受範圍。
本研究為解決此一問題,自行設計開發一部五軸全自動覆晶對準機,具有XYθ方向移動的工作物置放平台、Z軸真空取放晶片機構及可做X方向移動的視覺取像裝置,配合適當光源及立體分光鏡即可以一組視覺取像裝置分別取得上、下物體的影像,進行相對位移分析。在軟體部份,本研究以Boland C++ Builder & Matrox MIL & PC Based 運動控制卡函式做為開發基礎,發展出一套可自動依程序取像、計算、運動之程式。工作時,首先取得晶片錫球面之錫球位置及邊界相關尺寸,再依錫球位置和銲墊位置計算補正量,待移位補正後再令機構以非常靠近基板的位置向取得晶片正面邊界尺寸進行錫球位置定位,接著取得銲墊位置影像,用已計算出的下放置晶片。如此將可減少機構移動誤差及得到和穿透式對位一樣的精準結果。
運算放大器、訊號處理

關鍵字:非穿透式面對面對位、機器視覺、覆晶、微機電封裝、自動化設備

ABSTRACT

  When two objects need precise alignment and bonding together, usually they must have the same alignment marks. Then these two objects are moving closely and use penetration alignment to finish positioning. Due to the displacement offset is very small, this method can obtain quite accurate. However, if bonding objects are opaque such as flip-chips (FC) and MEMS components, in general, it may need two CCD cameras to grab the image and compute the displacement, then move to allocating point. Although, this method can achieve the goal of alignment, but position error may be to large after moving a long distance. So, the operator may need much calibration works to reach acceptable range.
This thesis is to design a machine for solving such problem occurred in current flip chip mounting requirements. This machine uses five servo axes for pick & place and alignment. Turn table can hold substrate and move in X, Y, and θ axis. Use Z-axis to pick (suck) and place the chip. CCD camera can move along X direction. Assisted by appropriate illumination and beamsplitter cube the camera can separately grab chip image upward and substrate image downward for alignment analysis. As to software development, this thesis employs Borland C++ Builder, Matrox MIL, and PC based motion control library to establish program for automatically grabbing image, computing and positioning. The proposed process of alignment is: First, grab solder ball surface image of the chip, then compute chip edge and ball center . Second, grab top image of chip, compute its edge and rebuild its ball position. Third grab the pad image of substrate, compute X offset, Y offset and tilt angle, then move it to correct position. Finally put down chip. By using proposed system and method, it can really reduce movement offset of mechanism and get precision as well as penetration alignment method.

Keyword: non-penetration face-to-face alignment, machine vision, flip chip, MEMS package, automatic equipment