Reliability in manufacturing requires traceable measurements. These in turn require that not only a value is given for the measured quantity, but also a statement of the uncertainty. The uncertainty should cover all values that can reasonably be assigned to the measured quantity, but no more. For simple measurements, the uncertainty can be estimated by formulating an analytical model of the measurement process and determining the sensitivity for the different influence quantities. For more complex measurements, this is no longer practical, or even possible. In the field of surface topography measurements, the measurement quantity is typically some global parameter extracted from a 2.5D surface scan, such as roughness, step height or pitch. The evaluation algorithm and the necessary filtering operations are often non-analytical. Furthermore, the scan itself consists of a large number of coordinates, which are all correlated as a result of the systematic errors of the measuring instrument. The solution is to model the measurement process in software, resulting in a so-called 'virtual instrument'. This model is than used in combination with input parameters which are randomized according to some specified statistical distribution, in order to simulate a distribution for the measurement result. From this, the uncertainty can be determined. At NMi VSL we have developed a software package for this type of calculation, which can be adapted for different types of surface measuring instruments and measurement tasks. We will present the framework and examples of its application to stylus instruments and AFMs.
展开▼
