“Get the stage below 1 μm” has to become a measurable condition before acceptance: under the specified travel and conditions, is the work-point positioning error along the motion axis under ±1 μm? If the spec demands “total error band under 1 μm,” that is a tighter bar — for example, Newport lists the XMS axial guaranteed accuracy as 1.5 μm (i.e., ±0.75 μm); the latter compares against ±1 μm but does not meet “total band under 1 μm”[2].
This article answers one question: how do you get a motion stage below 1 μm? The answer has four steps: write the acceptance definition clearly (± band or total band), make the position feedback measure the carriage, control heat and hold the work-point geometry (Abbe error), and finally accept at the work point under real conditions.
1. Write “below 1 μm” as an acceptance condition
The three numbers on the XMS100-S product page answer different questions[2]:
| Item | Spec | What it shows |
|---|---|---|
| Axial positioning accuracy, guaranteed | ±0.75 μm | The positioning error band under specified conditions |
| Bidirectional repeatability, guaranteed | ±0.040 μm | Scatter returning to the same position from two directions |
| Minimum increment (MIM) | 1 nm | The smallest reliably achievable displacement, not positioning accuracy |
The same page’s ±0.3 μm is typical accuracy and cannot substitute for the guaranteed value in acceptance — typical values are for comparing models during selection; guaranteed values go into the acceptance spec. The XM-S manual defines accuracy as the result after linear-error compensation — the specs above cannot be read directly as uncompensated absolute position error[1]. The manual lists a controlled test environment of 20 ±1 °C, a 5 μm mounting-plane flatness, and requires a controlled environment to meet the specs; the 1 nm MIM’s listed configuration is XPS-DRV11[1]. The 5 μm is a mounting condition, not an additional positioning error added to ±0.75 μm. Resolution is the tick of the command or reading, MIM is the smallest reliably achievable displacement, and positioning accuracy is the difference between command and actual position — more stepper subdivision or finer encoder counts cannot by themselves prove the work point is in place.
2. The position feedback must measure the carriage
A motor-shaft encoder measures the rotating part; lead error, backlash, and deformation behind the screw can still affect the carriage. To shrink the landing error, use a linear scale to measure the carriage’s position relative to the base and close the loop on that feedback[1].
Figure 1 | The position the scale reads vs. the work point’s position. Left: a generic side configuration (scale fixed, read head moves with the carriage, sensing gap kept); right: the offset-angle relationship, angle deliberately exaggerated. Click for full size.
This choice is separate from the transmission: a screw stage can also use carriage-position feedback — after compensation, the positioning error can approach the scale’s error and the compensation residual; a linear motor drives directly but still needs an accurate scale, compensation, and proper control. Motor and transmission classes are covered in “Which Motor for a Motion Stage, and How to Read the Numbers?”
3. Thermal management and work-point geometry
Heat: the first-order thermal-elongation formula is ΔL = αLΔT. Using the 20–100 °C average coefficient of 23.4×10⁻⁶/K from Kaiser’s 7075 tube data as an order-of-magnitude example, a 100 mm free length rising 1 °C elongates 2.34 μm[5]. This is a free-elongation estimate, not the whole-machine positioning error, and not a measured temperature coefficient — how the scale and structure are fixed, where the feedback measures, and the relative thermal changes of fixture, workpiece, and measurement reference decide how much elongation lands on the work point; the scale’s material and mounting both change the thermal behavior[4]. You cannot convert “20 ±1 °C test” directly into a whole-machine ±2.34 μm, nor derive a universal room-temperature allowance from this example. In practice: warm the system, control heat sources and gradients, and measure the drift under real conditions — warming up lets the spindle, structure, and scale reach thermal equilibrium, and the time is set by the measured drift curve.
Geometry: a linear scale reading the carriage does not mean every point on the workpiece has the same position error. Renishaw’s RLE guide explains: when the work axis is offset from the encoder axis, pitch or yaw produces Abbe error; at small angles, the error component for a single relevant offset is δx ≈ hθ[3]. Using the XMS’s guaranteed pitch amplitude of 25 μrad as a calculation example (manual: middle 80% of travel, controlled environment)[1]:
| Work-point offset h | Extra displacement hθ |
|---|---|
| 10 mm | 0.25 μm |
| 40 mm | 1.00 μm |
These are geometric calculations, not manufacturer-published work-point measurements. With ±0.75 μm as the baseline, an extra 0.25 μm can consume the remaining space under the ±1 μm bar — lower the fixture height and overhang, bring the work point close to the measurement axis, and check the tilt caused by load; yaw and lateral offsets need the corresponding geometry.
4. Accept at the work point, under real conditions
The error budget must list what is already included in the whole-machine spec and what the added fixture, workpiece, or environment brings — do not double-count included items, and do not merge different errors by root-sum-square from the catalog alone. Acceptance follows this order:
- State the measurement reference, work point, travel, and whether the bar is a ± band or a total band
- Install the real fixture and load; test at the planned speed, warm-up, and temperature conditions
- Use an independent, traceable measurement to correct the position bias and confirm the compensated residual — do not trust the controller’s own reading alone
- At multiple points along the travel, approach the target repeatedly from two directions, and record wait times, drift, and measurement uncertainty
When acceptance fails, go back to the error budget, find the term that exceeded, and fix that term — not retest the same conditions.
Back to the opening question: getting a stage below 1 μm takes four steps — a clear acceptance definition, feedback that measures the carriage, heat control plus work-point geometry, and final acceptance at the work point under real conditions. Scope: this article uses the XMS linear-motor stage as the example; the numbers are that series’ catalog guaranteed values under controlled-environment conditions. Unresolved limits: the scale, compensation, and mounting establish the measurement conditions; whether the 1 μm bar is met is judged by the work-point results. This is also not the same as post-cut part dimensional tolerance (see “What Do the 10 μm and 1 μm on a Drawing Actually Measure?”).