Sunteți pe pagina 1din 4

Proceedings of the 9 euspen International Conference San Sebastian June 2009

th

Nanometer-level environment

spindle

metrology

in

production

Byron R. Knapp1, Eric R. Marsh2, David A. Arneson1 and Donald L. Martin3


1 2 3

Professional Instruments Company, Hopkins, MN, USA The Pennsylvania State University, University Park, PA, USA Lion Precision, St. Paul, MN, USA

bknapp@airbearings.com Abstract This paper demonstrates precision spindle metrology in a production environment through the application of two methods of separating residual synchronous spindle face error motion from artifact circular flatness. Face error measurement results obtained using Estlers face error reversal and multiprobe face error separation are demonstrated for an ultra-precision air bearing spindle produced for diamond-turning applications. It is demonstrated that the tilt error motion obtained via radial error measurements agree with tilt error motion obtained via face measurements. Although the theory and application of spindle metrology is widely known, there are problems with inadequate hardware design (when testing at high speed), data acquisition, data analysis, and environmental control. This work documents the implementation of test hardware and analysis software to measure nanometer-level error motions in a production environment. 1 Introduction The spindle shown in Figure 1 features two long radial journal bearings with a central thrust captured on both sides by air bearing surfaces. The demanding accuracy requirements for the spindle could only be certified through thoughtful design of a new generation of metrology tooling. Final testing of the motorized spindle customer would like our spindles to be better than 12 nm. Figure 1: A Professional Instruments Company 5.5 ISO motorized air bearing spindle.

must conform to the 25 nm radial and 25 nm axial error specifications but the

Proceedings of the 9 euspen International Conference San Sebastian June 2009

th

Spindle measurement techniques rely upon a suitable artifact that is mounted on the spindle and measured during rotation. We have produced artifacts with form error less than 5 nm measured with a capacitance sensor, but 15-20 nm form error is more typical. As a result, the contribution of artifact form error to the total measurement can easily exceed the specification and must be separated from spindle error motion. The reversal techniques described by Donaldson and Estler theoretically provide the simplest method for separation [1,2]. However, the multiprobe method can be more convenient to perform while providing accurate results. Spindle error is characterized by five components of off-axis motionone axial (Z), two radial (RX and RY ), plus either two face (F X and FY ) or two tilt (X and Y ). The axial component of spindle error does not require error separationan artifact's imperfections do not influence the axial measurement except by second-order effects. In practice, the tilt error is usually calculated from radial measurements taken at some known axial spacing. Alternatively, tilt can also be obtained by face error separation. The multiprobe method estimates a single component of spindle error motion from three measurements. With enough capacitive sensors and appropriate fixturing, it is possible to completely characterize the error motion of a spindle in a single test. Only multiprobe offers this possibility as reversal methods require indexing of the artifact and the sensor during the testwhich adds a time-consuming step that also introduces a significant potential error source. Asynchronous error can be obtained using multiprobe making it suitable for rolling element spindlesa limitation of reversal techniques. These characteristics are consistent with Moores requirements of an ideal inspection toolit is accurate, it requires a minimum of operator skill, it inspects a specific type of error, it is fast to use, and it is self-checking [3]. 2 Approach The spindle used in this work (Professional Instruments 5.5 ISO) features a 1024count rotary encoder (Renco R22i) and a brushless, frameless motor and amplifier (MCS custom-wound motor and SA-2000 amplifier). The capacitive sensor (Lion Precision C23-C, 0.5 nm/mV) targets a 150 mm lapped flat target ring for face measurements and a 25 mm lapped sphere for axial and radial measurements. The amplifier (Lion Precision CPL290) incorporates a 1,000 Hz first-order, low-pass

Proceedings of the 9 euspen International Conference San Sebastian June 2009

th

analog filter with linear phase response. The data acquisition system (Lion SEA V8.3) is triggered by the encoder index pulse, providing immunity to synchronization errors caused by speed variation. Further low-pass digital filtering is done in software (typically filtered to 100 undulations per revolution). Additionally, the frequency components caused by thermal drift and fluctuations in air bearing supply pressure are attenuated by high-pass digital filtering (0.1 Hz cutoff). In this implementation of multiprobe separation, the three sensor angles are chosen to be 0, 99.84375 and 202.5. These angles were chosen for their asymmetry and because they coincide with encoder pulses. Several potential contributors to measurement uncertainty were eliminated by sequentially indexing a single sensor rather than using three different sensors. 2.1 Face Error Separation Face error motions are often of greater concern than radial error motions when diamond-turning optics. Using the tooling shown in Figure 2, we can perform Estlers reversal and accurately separate artifact form error from residual face error motion. With Estlers reversal, the artifact and sensor must be rotated 180 relative to the spindle. Using the same tooling, we can also perform multiprobe error separation which does not require indexing of the artifact. The tooling permits accurate indexing of the sensor for multiprobe separation. Figure 3 shows three representative face error measurements and Figure 3: Face error measurement results using multiprobe error separation. the results of the error separation. The artifact form error and residual synchronous face error agree with the reversal method to better than 5 nm. 2.2 Tilt Error Measurement Previous work has described the various iterations and improvements made to the radial error motion tooling [4, 5]. This tooling was designed to allow separation of artifact and spindle errors with multiprobe separation and Donaldsons reversal. This test is repeated here for comparison of the computed tilt error with the face Figure 2: Sectioned view of face error motion tooling.

Proceedings of the 9 euspen International Conference San Sebastian June 2009

th

measurements. Multiprobe separation is again used and the indexing plate is then rotated 90 to measure the orthogonal component of radial and tilt error. These tests are repeated with a 150 mm riser. The calculated tilt from the radial test is then compared to the calculated tilt from the face test. As expected, the tilt error motion obtained via radial error motion measurements agrees with tilt error motion obtained via face error measurements. 3 Conclusion The results show that the residual synchronous face error motions and artifact form error obtained by multiprobe error separation are comparable to results obtained by Estlers reversal. The multiprobe technique is performed without indexing the artifact which simplifies the testing procedure. Using four capacitive sensors simultaneously to measure the axial and face error motions of the spindle yields synchronous and asynchronous axial and tilt error motions. Measurements are made sequentially with a single probe to reduce error sources. Independent measurements of the tilt error calculated from multiprobe radial and multiprobe face error measurements are shown to agree. In this situation, calculating the tilt error from the multiprobe face separation result is more convenient than the usual approach of measuring the radial error at two axial distances. These results confirm that multistep error separation may be a reasonable alternative to reversal methods in a production environment requiring rapid testing at high speeds. Our next generation of spindle testing tooling will allow measurement of rotating sensitive face and radial error motions without rotating the the tooling. References: [1] R.R. Donaldson. A simple method for separating spindle error from test ball roundness error. Annals of CIRP, 21(1):125126, 1972. [2] C.J. Evans, R.J. Hocken, and W.T. Estler. Self-calibration: reversal, redundancy, error separation, and absolute testing. Annals of CIRP, 45(2):617634, 1996. [3] W.R. Moore. Foundations of Mechanical Accuracy. The Moore Special Tool Company, 1970. [4] E.R. Marsh. Precision Spindle Metrology. DEStech Publications, 2007. [5] E.R. Marsh, J.A. Couey, and R.R. Vallance. Nanometer-level comparison of three spindle error motion separation techniques. ASME Journal of Manufacturing Science, 128(1):180-187, 2006.

S-ar putea să vă placă și