A high-precision coordinate measuring system for super-smooth polishing

For the purpose of fabricating free-form optical elements to an accuracy of 80 nm PV, the CSSP (Canon Super-Smooth Polisher) has been developed. This device finishes workpieces by alternately repeating contour measuring and corrective polishing. In such a system, contour measuring is important because it limits the final accuracy of the workpiece. This paper focuses on the CSSP's on-machine contour measuring method. A contact probe is employed to ensure adaptability to free-form contours with maximum tilt angles of 35 degrees. A unique probe structure is proposed, by which both inclination and motion errors of the probe are simultaneously compensated. Th flaw problem is discussed from an experimental point of view; it was found that the major cause of flaws is dirt on the probe or the workpiece. Dirt also causes unstable irregular figure errors. Thus, the influence of dirt can be estimated from the results of contour measurement. A model of dirt is proposed, and the distribution on a width-height chart of irregular figure error predicted by the model agrees well with the experiment. A cleaning procedure was developed that is effective in reducing the problem of dirt. Using a ceramic air slide and a linear motor, the contact force was controlled to a constant of 2 mN, which is much smaller than the force of 260 mN that causes a yield stress on the CaF2 workpiece. The probe's scanning speed of 4 mm s(-1) was achieved by speeding up this force controller. The margin of force error during contour measurement was under 0.2 mN. A coordinate measuring method using fourteen-axis interferometers was also proposed for compensation of the major mechanical motion errors of the probe and the tables. Some methods of compensation for system errors are discussed. A new method was proposed by which three kinds of angle error, on the x, y, and z axes of measurement, are simultaneously compensated with a repeatability of 0.08 mu rad RMS. The measurement results showed good repeatabilities of 3 nm RMS for a 540 mm line measurement, and 9 nm RMS for a <empty set> 500 mm aspherical surface measurement.