Enhanced Surface Metrology

Publication Details

Abstract

The constant search for more accurate measurement generally leads to higher cost, greater complexity, and/or devices that do not lend themselves to manufacturing environments. For example, surface metrology can be accomplished by a number of methods, ranging from rulers and visual estimation (cheap, fast, and inaccurate) up to fixed coordinate measuring machines (expensive, slow, and accurate). The tradeoffs involved in selecting metrology methods generally involve these three parameters of cost (initial and operation), measurement speed, and accuracy. We present a method of adding one more tradeoff: measurement precision (perpendicular to the surface) vs. sampling resolution (along the surface). Through application of statistical sampling and curve fitting, we can improve precision by approximately the square root of the amount that we decrease resolution. We applied this technique to a number of known and unknown targets, using the Cognitens WLS400 white-light stereovision system from Hexagon Metrology and a custom laser interferometry measurement system. Using the enhancements described in this article, we were able to improve the measurements significantly. Measurement of a flat reference surface, for example, was enhanced by reducing noise by a factor of 11 and improving surface measurement accuracy 2× (limited by the actual surface figure). An application of this technology to a known sphere reduced noise by a factor of eight and demonstrated that the sphere was within its surface and diameter specifications. We used this statistical technique to reduce noise in an interferometry system 15×, and demonstrated that the supposedly flat surface had deviations exceeding 16 percent over a square region 1 cm on a side. Finally, we modeled the white-light scanner to determine its probability of identifying small surface features. Based on this model, we found that statistical noise reduction can improve the minimum resolvable feature height by a factor of five without significant difficulty.