BGA線上檢測系統之開發

Abstract

近年來，隨著半導體業的迅速發展，IC的規格已逐漸朝著尺寸縮小，腳位間距縮短的構裝型態作轉變，因此使得球格陣列（Ball Grid Array,BGA）—高腳數、低腳距的構裝方式，逐漸取代傳統的四面平封裝方式（Quad Flat Package,QFP）。本論文便是利用電腦視覺來檢測BGA的品質，依據目前BGA生產線上的實際需求，整合一套線上BGA檢測系統。 檢測前必須先做硬體設備校正的工作，接著在側照式光源照明的環境下，透過CCD攝影機擷取影像，經由中間值濾波、影像二值化和型態學的處理將影像強化，接著進行教導學習的工作，利用強化過的二值影像建立比對用的樣本，此樣本可作多球、缺球、球位置偏移和球相連缺陷的檢測。另一方面，藉由區塊分析來計算次像素邊緣檢測的搜索路徑，由這路經便能夠找出錫球的邊緣位置座標，最後再將邊緣位置座標代入最佳逼近橢圓方程式，便能夠找出錫球的逼近橢圓，藉這逼近橢圓來作橢圓尺寸大小的檢測。論文最後以三套檢測系統來作BGA缺陷檢測的比較，本論文實驗中量測的錫球有兩種規格，分別是(1) 錫球直徑0.63mm，球間距1.27mm和(2) 錫球直徑0.3mm，球間距0.5mm。

Recently, with the quick growth of the semiconductor industry, the size of IC is getting more and more smaller, which makes the new package method BGA(Ball Grid Array) replaces the traditional package method QFP(Quad Flat Package).The objective of the thesis is to develop a machine vision inspection system, to validate product correctness and quality. And it also utilizes the rapid computation of PC to increase the efficiency of manufacturing. Before doing the inspection job, we need to calibrate the hardware facility. In the ring light illuminating environment, we grab the solder balls’ image of BGA IC and enhance the gray image via preprocessing, including following three steps: median filtering, image binarizing, and the morphology image processing. By the preprocessed binary image, we build up the sample that is used to do comparison job. With the sample we can detect defects like ball presence/or absence, ball position relative to the pad, and double balls. Besides, to calculate the scanning paths via blob analysis method, and then we can utilize the paths to find edge points of the solder ball’s contour. With the edge points we found, we can estimate the diameter of the solder ball by using least square method. There are two kind of specifications of solder balls used in the experiment, one is 0.63mm in diameter,1.27mm in ball pitch and the other is 0.3mm in diameter,0.5mm in ball pitch.