Software testing in roughness calculation

Abstract

A test method to determine the function quality provided by the software for roughness measurement is presented in this study. The function quality of the software requirements should be part of and assessed through the entire life cycle of the software package. The specific function, or output accuracy, is crucial for the analysis of the experimental data. For scientific applications, however, commercial software is usually embedded with specific instrument, which is used for measurement or analysis during the manufacture process. In general, the error ratio caused by the software would be more apparent especially when dealing with relatively small quantities, like the measurements in the nanometer-scale range. The model of 'using a data generator' proposed by NPL of UK was applied in this study. An example of the roughness software is tested and analyzed by the abovementioned process. After selecting the 'reference results', the 'reference data' was generated by a programmable 'data generator'. The filter function of 0.8 mm long cutoff value, defined in ISO 11562 was tested with 66 sinusoid data at different wavelengths. Test results from commercial software and CMS written program were compared to the theoretical data calculated from ISO standards. As for the filter function in this software, the result showed a significant disagreement between the reference and test results. The short cutoff feature for filtering at the high frequencies does not function properly, while the long cutoff feature has the maximum difference in the filtering ratio, which is more than 70% between the wavelength of 300 νm and 500 νm. Conclusively, the commercial software needs to be tested more extensively for specific application by appropriate design of reference dataset to ensure its function quality. © 2005 IOP Publishing Ltd.