8th International Workshop on Semiconductor Pixel Detectors for Particles and Imaging.

Development of Visual Inspection System for Detecting Surface Defects on Sensor Chip

Not scheduled

Sestri Levante

poster

Speaker

Mr Adi Nurhadiyatna (Indonesian Institute of Sciences)

Description

This study presents a visual inspection method based on image processing to detect surface defects found on pixel chip pad. Pixel chip is a particle detector used in Inner Tracking System (ITS) which is part of ALICE experiment (A Large Ion Collider Experiment). The quality of chip including the surface pad quality has to well assessed in order to guarantee the successfulness of the overall experiments. A large number of pixel chips are used in ITS which demands both quick and accurate tests of the surface pad. Therefore, an effective inspection technique is needed to address such a problem. This paper proposes a design approach to assess the quality of surface pad by using image processing techniques. The design approach consists of three steps. Firstly, K- Means is employed to divide the surface pad into clean and defect areas. Secondly, defect area is extracted by using Gabor filter. Lastly, Canny Edge filter is executed to detect the defects of the defect area. Experimental results show that the proposed approach can achieve significantly high accuracy and recall probability which are by 84.9\% and 77.9\% respectively.

Summary

ALICE (A Large Ion Collider Experiment) is an experiment at CERN(Conseil Europeen pour la Recherche Nucleaire), Switzerland that is designed to address the physics of strongly interacting matter and properties of the Quark-Gluon Plasma (QGP) using proton–proton, proton–nucleus and nucleus–nucleus collisions. In this experiment, a Large Hadron Collider (LHC) is used to accelerate two beams in opposite direction.The collision produces billion of particles with scattered trajectories, in which those particles need to be tracked and then identified..

The LHC apparatus consist of a central barrel, a forward muon spectrometer and a set of small detectors for event characterisation. Inner Tracking System (ITS) is basically an assembly of multi layers detector which serves as a first stage detection after collision process. One of the main components of ITS is a Pixel Chip which is a pixel size particle detector . There are a large number of Pixel chips which is about 24000 are required to build a complete ITS. For each chip, there is 103 pads, which makes 103 times 24000 pads for overall Pixel chips. Pixel Chips from mass production may have damages on the surface pads, while manual inspection by using human eyes will be very ineffective  due to small scale of pads and the large number of pads to be tested. Moreover, manual inspection by using human eyes is costly, and dependent on its investigator (mood,eye fatigue, experience, etc). Therefore, a visual inspection based on image processing is needed to quickly and accurately test the quality of chip pads.

The pALPIDEfs design is used for particles detector. This design has dimension 15x30 mm/-2 for a single chip and 103 contact region with 200$\mu$m of diameter. As we can see in Figure, 103 contact area is aligned on the chip. According to its thickness, manual inspection is very hard to be done, because of light in weight, and small in size. These contact area are gold plated, and coated with other material (silicon). As mentioned before, during the production process damages may happen with unexpected condition. Figure shows the contamination process during placement activities using vacuum pick up tool.

In accordance with the pad area diameter, microscope used for visual inspection system needs to reach an appropriate view with certain magnification (500x). This system will be applied in mass production process to avoid several damages. Visual inspection for chip quality assessment is widely applied in industrial application by automated inspection system.

The study is focused in visual inspection for chip pad area.  It aims to detect contaminated or defect area that may happen during the production and distribution stages. A series of methods is applied to detect the crack existence. Firstly,  K-Means clustering is used to cluster the pad and defected regions. Gabor wavelet algorithm is then used to extract the defect area.  The method is ended by finding defect area through Canny edge detection.