A comparison of reversal and multiprobe error separation

This paper demonstrates the application of two methods of separating spindle error motion from artifact roundness on a spindle with less than five nanometers radial error. Two error separation methods, reversal and multiprobe, were each applied to data taken on two different test stands allowing direct comparison of the four combinations of hardware and separation algorithm. Because the theory of both separation methods is well documented, this work focuses on their implementation for nanometer-level measurements. As will be seen, a number of issues must be addressed to obtain repeatable results at this level of precision in spindle metrology. Ultimately, the results show that sub-nanometer features in both spindle error and artifact form can be reliably and repeatably resolved by both techniques. Introduction This work seeks to characterize the radial error motion of a precision air bearing spindle with nanometer-level accuracy. This task is complicated by the requirement that the necessary measurements are usually made using a lapped spherical artifact whose out-of-roundness is too large to be ignored; a research-grade lapped spherical artifact may have an out-of-roundness at its equator of between 10 and 25 nm. The artifact out-of-roundness necessitates the use of an error separation method to extract the spindle error and artifact form from the raw measurements. Although the theory of error separation is widely known and well documented, there is significant exposure to mistakes made in hardware design (particularly when testing at high