Machine vision: the eye of semiconductor manufacturing
Machine vision software is becoming more prevalent in the continually diverse process of semiconductor manufacturing.
Global semiconductor manufacturing capacities are continuously expanding. However, the production processes in this industry are complex and include more than 1,000 different steps.
As a result, the development of production capacities is equally complicated and delicate. Thus, there is a need for technologies that can be rapidly implemented and adapted, while simultaneously enhancing production efficiency. A key technology here is machine vision. Its main advantage lies in the fact that in the high-precision manufacturing of semiconductors, the numerous necessary inspection and alignment processes can be automated and carried out with high precision. With its software products HALCON and MERLIC, the Munich-based company MVTec Software supplies machine vision software which benefits the diverse process steps of semiconductor manufacturing.
Powerful machine vision technologies for semiconductor manufacturing
In nearly all semiconductor production scenarios, there is at least one step where the product is inspected for functional or cosmetic defects. In a highly automated production environment, quality inspection using machine vision offers numerous advantages over manual inspection: machine vision is much faster, delivers objective and reproducible results, and the quality of the inspection does not run the risk of degrading due to fatigue or the monotony of the task. Deep learning technologies like anomaly detection are also suitable for this purpose. For example, it allows automated surface inspection for detecting and segmenting defects.
When it comes to quality inspection, it is essential to inspect for both dimensional accuracy and defects. Machine vision allows for the measurement of edges with subpixel accuracy, along lines or circle segments, in a few milliseconds. With 2D measurement, it is also possible to inspect objects of a specific geometric shape. In addition, there are methods available for 3D measurement, including the use of disparity images, distance images, or the reconstruction of 3D surface coordinates by various methods.
Packaging: the last steps to the finished chip
Apart from quality inspection, finding and aligning wafers and chips is another application where machine vision can provide very effective support. This is where subpixel-accurate shape-based matching comes into play. The technology accurately and robustly detects objects in real-time. This works even when objects are rotated, scaled, perspectively distorted, locally deformed, partially covered or outside the image.
The added value offered by the technology is illustrated by the example of the packaging production step, where the chip is encapsulated in a plastic housing. The single dies are first inserted into the housing. After that, they are connected to the relevant contacts of the corresponding package in several steps. The exact positioning of dies within the package is a crucial step for enabling subsequent processing. For this purpose, shape-based matching methods are suitable. In practice, the model of a die is trained, for example via CAD data. By matching the training data with the original image, the die is located and its position in the plastic housing is calculated, enabling further processing.
The listed production steps were selected as examples to illustrate the added value of machine vision. Moreover, as “the eye of semiconductor manufacturing”, the technology along the complete process chain. Especially in quality assurance and packaging, high-performance machine vision software makes a valuable contribution to efficient production.