O4.4 - Relibility Analysis of Indicator Functions in an Automated Multiscale Measuring System

SENSOR+TEST Conferences 2009
2009-05-26 - 2009-05-28
Congress Center Nürnberg
Proceedings OPTO 2009 & IRS² 2009
OPTO 4 - Optical Inspection and Quality Assurance
A. Burla, J. Regin, W. Lyda, J. Zimmermann, W. Osten, E. Westkaemper, O. Sawodny - Universität Stuttgart, Stuttgart, Germany
107 - 112


Modern micro systems are manufactured in batch processes at wafer scale with critical functional components in sub-micron dimensions. To ensure high quality of these components high resolution inspection systems are needed. Due to the limited area-related resolution of commercial available sensors a conflict between measurement time, object size and resolution arises.
In former publications we have presented the general design of an Automated Multiscale Measuring System (AMMS) based on an intelligent and hierarchical measurement strategy to balance this conflict. The system pursues a multiscale active exploration strategy, starting with a coarse scale sampling. To identify new regions of interest for more detailed measurements, task specific, coarse-scale indicator functions are used. Based on the infommation from the indicator analysis, sensory and positioning data are processed step-by-step and merged using intelligent data fusion. The system utilises high resolution and accuracy only where it is needed. By task specific sensor parameterisation and actuator control, the system automatically adapts to the given object and measurement task.
The concept was realised in a prototype based on a modified Mahr MFU 100. The prototype is armed with three different sensors for fast inspection of micro optical elements like micro lens arrays. Due to the fact that micro lens arrays are manufactured in batch processes, a wide bandwidth of defects from wide area pollution and missing lenses down to sub-λ wavefront defects can occur. Hence the micro lens arrays are suitable objects for the investigation of measurement certainty and defect detection reliability of the AMMS.
This contribution is focused on the investigation of indicator reliability, a crucial performance factor for the multiscale measurement system. Several indicator algorithms are developed to accurately detect flaws and defects on the surface of the micro lens arrays. For the reliability analysis of the algorithms microlens arrays with different surface and shape defects are simulated using synthetic data based on mathematical as well as statistical methods.
The feature detection algorithms include Fourier self-filtering, two-point statistical texture featuring, scratch detection and normalized cross correlation. These algorithms are parameterised according to the type of defect. Reliability analysis of the detection algorithms was carried out to evaluate the amount of false positives or false negatives detected, to find the detection certainty of the algorithms with respect to the changes in the defect properties. The results of this reliability analysis are shown in the following sections.