主动红外焊球缺陷检测方法研究
摘要随着电子信息产业的快速发展，倒装芯片技术因其符合封装技术小型化、轻 薄化的发展趋势而得到了广泛的应用。倒装芯片是将芯片倒扣在基板上，使芯片 的电气面朝向基板，然后通过焊点将芯片与基板的焊区形成机械连接，从而实现 信号的传输。但是随着封装密度的增加，焊球尺寸和间距的不断减小，加上工作 时容易受到应力的影响，焊球容易产生缺陷，从而影响封装质量。
本文基于主动红外热成像无损检测技术，对焊球直径为 500μm，间距为 1000μm 的一个 5×5 阵列式倒装芯片模板进行实验，对检测到的红外热像图进行 自适应滤波、边缘检测、伪边缘和不封闭边缘、图像分割等处理后提取焊球的特 征，利用 BP 神经网络实现焊球的自动分类。针对 BP 神经网络对大样本进行识 别和分类存在的不足，又提出了利用概率神经网络对焊球进行缺陷识别和分类。73966
毕业论文关键词: 倒装芯片 主动红外检测 图像处理 BP 神经网络 概率神经网络
The research on defect inspection of flip chip based on active thermography
Abstract With the rapid development of electronic information industry, flip-chip packaging technology in line with its miniaturized, lightweight development trend and has been widely used。Flip-chip is a chip upside down on the substrate, so that the electrical side of the chip to the substrate, then solder the chip pad and the substrate pad forming a mechanical connection to achieve signal transmission。 But with the increase packaging density and the decrease of solder bumps in dimension and pitch, what's more,it can be impacted by work stress,solder bumps are easily to have defects and affecting the quality of the package。
Based on active infrared thermal imaging nondestructive testing technology, solder bumps having a diameter of 500μm, 1000μm pitch of a 5×5 array flip-chip template experiment to detect the infrared thermal image adaptive filtering, edge detection, extraction ball after pseudo-edge and edge is not closed, image segmentation process is characterized by using BP neural network automatic clustering balls。For the disadvantages of BP neural network in identifing and classifing deficiencies on a large scale, but also proposed the use of probabilistic neural network to identify and classify defect solder bumps。
Key Words:Flip chip active infrared detection image processing BP neural network probabilistic neural network。
目 录
摘要 I
Abstract II
1 绪论 1
1。1 研究背景 1
1。3 主动红外检测焊球缺陷的基本理论概述 4
1。4 本论文研究工作 5
2 基于主动红外焊球缺陷检测方法研究 6
2。1 红外无损检测原理 6
2。2 主动红外检测系统组成 7
2。3 基于主动红外的焊球缺陷检测 8
2。4 本章小结 9
3 主动红外焊球缺陷检测实验研究