X-ray meshed tomography for electronic packaging inspection

Abstract

Due to the growing levels of integration of semiconductor products, requirements for the quality assurance are continuously increasing. X ray imaging technique is widely used, because it can provide a way to inspect not only visible defects but also invisible ones non-destructively. In case that several patterns are mixed with each other, however, it is very difficult to recognize each pattern with an X ray projection image. Therefore, three dimensional imaging technique, such as CT (Computerized Tomography for medical purpose) and laminography (for industrial use), has been studied. Result of laminography, however, is severely transmuted by the superposition of the blurred images of the out-of-focus structures. Conventional tomography requires too much time to use it in real time applications. This thesis proposes a new reconstruction algorithm which has better precision than laminography and shorter reconstruction time than tomography. Proposed algorithm saves a lot of reconstruction time, in a way that it keeps resolution high in the region of interest and reduces it in the other region. Introduction of reconstruction matrix also increases the speed of reconstruction. Simulation result shows that the proposed algorithm is about 1,000 times faster or more than conventional tomography in a particular application, whereas image quality is superior than that of conventional laminography and comparable to that of full resolution tomography. Mathematical modelling of the system and the calibration of model parameters are very important, because it has a direct effect on the reconstruction result. Until now, there are few approaches to calibrate the parameters of overall system including X ray source, image intensifier, camera, image processor, and positioning system. Phenomenon of image distortion and the estimation of distortion center are generally ignored in conventional calibration technique. Although some algorithms deal with those, they need complicated nonl...