In our feature article in the March 2009 issue (“Heat & Motion Duke It Out,” p. 60), we pointed out that the dominant cause of error in most motion applications is temperature, or, more precisely, temperature changes. Temperature changes hold the key to high-precision positioning applications. We discussed the fact that various materials (e.g., aluminum, granite or steel) used in a lab setup have different linear thermal expansion coefficients. For example, the changing temperatures during an experiment can create significant drift and affect repeatability or reproducibility in a motion control process due to the expansion of the optical table, the aluminum base of a linear stage or its steel lead screw, or even an overhead granite bridge. Even though temperature change is the main source of error in motion systems, in this article we will direct our attention to other error sources that must be addressed in positioning applications. Consideration must be given to the effect of the flatness of mounting surfaces and overhanging loads on the performance of precision positioning systems. Pitch error Precision stages are typically machined to high tolerances. The smoothest surface profiles that maintain flatness and straightness lead to higher accuracy in motion or positioning tasks. The stages are generally tested under ideal conditions – in rooms that are temperature-controlled and on very flat mounting surfaces. Typical mounting surfaces used in metrology are granite flats with surface flatness ranging anywhere from 1 to 5 μm. In reality, as opposed to the controlled testing environment, the most common mounting surface is a vibration isolation table with stainless steel skins. The table’s surfaces are usually ground with a pattern of drilled and tapped holes spaced 25 mm along the length of the table. A honeycomb network of ribs separates the top and bottom steel skins, providing stability and rigidity to the table. The honeycomb is glued into place as the skins are squeezed between flat platens. The overall flatness of the table depends on the flatness of the platens, and the local flatness depends on the quality of the drilling and tapping operation. Typical flatness of honeycomb-structured tables is on the order of ±0.10160 mm over 0.60960 m, or about 16 μm per 100 mm. Figure 1 illustrates a linear stage mounted directly onto a surface with a flatness of 5 μm over 100 mm of travel. The resulting deformation of the stage bearings introduces a pitch error on the stage as an object moves from left to right. Based on the geometry shown, the resulting pitch is 100 μrad over 100-mm travel. Figure 1. Depicted is a 200-mm stage on a nonflat surface. For sectors and chords, the angle (α) can be derived with the following formula: α = 8 x h/L = 8 x 0.000005/0.2 = 200 μrad = >100 μrad for 100-mm travel. Referring to Figure 1 again, the sample is either 100 mm above the theoretical plane of the drive screw that may have a motor-mounted rotary encoder or above the plane of the linear encoder. Because of the induced pitch error, the point of interest is now ahead by 10 μm (100 μrad x 100 mm). Of course, the closer the sample is to the measurement plane, the less pitch error effect there is. In this case, 5-μm flatness introduces a pitch error of 100 μm. For a less flat table, the pitch error can only become worse. To eliminate the effect of nonflat surfaces, a number of solutions are available, if one has a steel table. Mounting stages on granite bases with flatness of 5 μm or less (available from Newport Corp.) and placing the granite base on the table will reduce the effect of the lack of table flatness. Another method is to use three-point kinematic supports under the stage, which defines a perfect plane, thereby eliminating distortions. However, the precision stage in this condition must be very stiff, with enough rigidity to handle the unsupported sections of the stage. Figure 2. If a stage is not fully supported, pitch errors can occur when a load moves toward the unsupported side of the stage. Base support and load distribution Another source of error is the amount of support underneath the precision stage. Ideally, the base of the stage is fully supported, so the load is distributed through the bearing elements into the base and evenly over the mounting surface. In Figure 2, a 600-mm stage is supported only at the middle of travel. As the 60-kg load is moved to the extreme ends, the load induces a pitch error. The left graph in Figure 3 shows a pitch error of 450 μrad. Using simple geometry with a sample at 100 mm from the measurement plane, this equates to a 45-μm linear error at each end of travel, as shown in the right graph. Figure 3. In this example, a 450-μrad pitch error equals a 45-μm linear error. A specific case of load distribution is on X-Y stacks where the load can cause angular deviations as the center of gravity moves beyond the bearing support of the lower X-stage, in the overhang region (Figure 4). Usually for high-precision products, stiffness constants are provided by experienced suppliers of motion products. In this example, the angular deviation constant around the X-axis, kαx, equals 1 μrad/N-m. When the 5-kg load is moved to the extreme position of the Y-axis, the resulting roll angle deviation to the X-axis is 12.5 μrad. Similar to the effect in Figure 1, this angular deviation acts as a pitch error effect on the Y-axis. Also, this roll angle equates to a flatness of 3.13 μm. With an X-stage that has a wider base, the angular deviation constant will be smaller, so the effect of overhanging loads is less. Figure 4. In this example, travel is ±250 mm, the load on the Y-stage is 5 kg, the stiffness constant (kαx) = 1 μrad/N-m, and the roll angular deviation due to load overhang is 12.5 μrad (1 μrad/N-m x 50 N x 0.25 m). At the extreme end of the Y-stage, the roll is ±12.5 μrad, and the flatness is 3.13 μm (12.5 μrad x 0.25). It is clear from our discussion that attention must be paid to the quality of the surface to which a precision stage is mounted, preferably 5 μm or better. Consider also fully supporting the stage and the width of the bottom stage to reduce the angular deviation effect. All these considerations will ensure achievement of the performance specified by the precision stage manufacturer. Meet the authors Thomas Bartholomäus is director of European sales at Newport Motion; e-mail: thomas. bartholomaus@newport.com. Beda Espinoza is senior manager for product marketing at Newport Motion; e-mail: beda.espinoza@newport.com.