Automated probe-sample alignment for the evaluation of integrated circuits

Abstract

By measuring and testing integrated circuits, quality and functionality can already be determined during and after the production. But with ever increasing demands, such as cost or efficiency, the complexity and density of the circuits is also increasing. This leads to ever smaller structures and contact pads, that pose challenges to the measurements. As a result, a more accurate alignment between the measurement probe to the investigated circuit is necessary. Currently, testing is achieved utilizing manual probing stations, where the alignment is usually done by a human. Due to bad reproducibility and slow execution, automation of this process is proposed in this work. To this end, a vision guided alignment system is implemented and its accuracy is evaluated. Various machine-vision techniques for spatial detection of the objects position, such as the probes tip and the samples contact pad, are implemented and compared. The images are acquired utilizing a camera and an objective lens. To measure the vertical positions, the focus of this optical system is adjusted with the usage of focusing algorithms. The method "Tenenbaum variance" shows the best performance. The horizontal position of the objects is measured using "Template matching". After the measurement of the spatial positions, alignment is achieved by controlling positioning stages. Algorithms to accelerate the alignment process, as well as methods to increase the robustness are presented and compared. Since the determination of the positioning accuracy with usual probes is not possible, an atomic force microscope utilizing a piezoresistive cantilever is implemented. This offers the possibility of performing a scan after the completed alignment, thereby determining the spatial offset. The achieved positioning accuracy in vertical and horizontal direction are 880nm and 490 nm, respectively. The average duration for the alignment process is 48 s. Limits, such as the objective lenses depth of field and magnification as well as the cameras resolution, result from a trade-off between accuracy, execution speed and ease of implementation.