應用類神經網路於以電腦影像為基礎的BGA封裝的量測與檢測

Applying Computer Image Based on Neural Networks to Measurement and Inspection for BGA Packages

林家興(Chia-Hsing Lin)

指導教授 ： 戴兢志

大同大學/工程學院/機械工程學系所/碩士(2004年)

摘要

近年來，隨著電子資訊產品走向輕薄短小及功能多樣化之趨勢，IC元件亦朝高密度化、小型化、高腳數等方向發展。在此趨勢之下，IC構裝亦逐漸邁向腳距微細化、腳數多腳化、構裝尺寸小型化發展，其中最引人注目且逐漸量產的構裝方式則非球柵陣列BGA（Ball Grid Array）莫屬。目前球柵陣列BGA晶片已漸漸成為今日半導體構裝技術的重要製程之一。BGA是以錫球（Solder Ball）代替傳統金屬導線架作為IC與印刷電路板間的接腳，且在底部是以陣列方式佈置錫球。在安裝時要求錫球需具有很高的位置精度，如果錫球的三維尺寸誤差較大，尤其是在高度方向，將會造成錫球頂點不共面。安裝時個別錫球和線路板接觸不良，將會導致漏接、虛接，影響IC晶片的可靠度，因此錫球的高度及共面檢測就顯得相當重要。
非接觸三次元畫像測定儀是最常見的量測設備，量測者以固定間距對BGA表面的錫球進行量測，此量測策略是最常見的一種量測方法。但因BGA的製作過程中，常會有漏植錫球、錫球相連、錫球大小不同、或錫球間距不一等等的情形發生。所以此量測策略無法真正量測到錫球最高點，因此使得BGA錫球高度與共面度的檢測愈來愈困難。
本研究擬整合光度立體攝影技術與電腦影像處理技術來改善非接觸式量床之量測策略，以期獲得較佳的量測點位置。光度立體攝影技術主要是根據在三個不同位置與方向之光源照射在待測錫球上，可以產生一組照度方程式系統。此組照度方程式系統配合類神經網路可以求得錫球各量測點之法線向量，所得的法線向量可運用來規劃錫球最高量測位置。

關鍵字

球柵陣列 ； 影像處理 ； 類神經網路

並列摘要

In the last few years, the electronic products have a tendency towards light-thin-short-small and a variety of functions. The IC components become more concentrated, finer and more pins. Under this trend, the IC component packaging will go toward a tendency of the closer pin’s pitch, more number of pin, and packaging miniaturization. The most popular method is BGA (Ball Grid Array). At the present day, the BGA packaging technology becomes one of the most important processes in the semiconductor fabrication gradually. In the BGA technology the solder ball instead of the metal wire is used between the IC chip and PCB. They are set on the bottom by the array method. When it is setup, the solder balls need highly positional precision, especially in the height. If not, the solder ball’s top will be in-coplanar. The bad contact between the solder ball and a circuit board will bring about a leakage contact and a virtual contact, and these will affect the reliability of the IC chip. For this reason, it is very important to check the co-planarity and the height of the solder ball.
The 3D non-contact image measuring system is the most familiar measuring equipment. The measurer usually measures the solder ball at a constant pitch on the substrate and that is the most familiar measurement. But in the process of produce BGA there is often some situation such as left-out solder ball, contact between two balls, different size among the the balls, or the different distant at solder ball. So the measuring strategy is incapable to measure the highest of solder ball. Therefore, it is more and more difficult to measure the solder ball’s height and co-planarity on BGA.
In this research photometric stereo technique and computer vision processing technique are integrated to improve the measuring strategy of the non-contact measuring system so that the better position of the measuring point can be obtained. Photometric stereo technique is based on the solder ball that is illuminated by a light source of three different positions and directions to illuminate the solder ball. It can produce a system of illumination equations. The normal vectors of the solder ball can be obtained by using this system of equations with neural networks; all normal vectors can be used to plan the highest measurement of the solder ball location now.

並列關鍵字

ANN ； BGA ； Image processing

參考文獻

1.R. Zhang, P. Tsai, J. Cryer, M. Shan, “Shape from shading: a survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence 21 (8) pp.690-706, 1999.

2. M. Zerrog, R. Nevatia, “Part-based 3D Descriptions of Complex Objects from a single image,” IEEE Transactions on Pattern Analysis and Machine Intelligence 21 (9) pp.835-848, 1999.

3. Christian Heipke, Christian Piechullek, Heinrich Ebner, “Simulation studies and practical tests using multi-image shape from shading,” ISPRS Journal of Photogrammetry & Remote Sensing, Vol. 56, pp. 139-148, 2001.

4. Q. J. Peng, M. Loftus, “A new approach to reverse engineering based on vision information,” International Journal of Machine Tools & Manufacture, Vol.38, pp. 881-899, 1998.

5. Q. J. Peng, M. Loftus, “Using image processing based on neural networks in reverse engineering,” International Journal of Machine Tools & Manufacture, Vol. 41, pp. 625-640, 2001.