How to select a linear stage

In short:Pick the drive from speed, stroke and settling (ball screw for stiffness and moderate speed, belt for long and fast, linear motor for the fastest settling with no backlash), then build an error budget at the work point: repeatability, accuracy, straightness and flatness, Abbe error from pitch and yaw times the offset, orthogonality between stacked axes, thermal expansion and the flatness of the surface it bolts to.

What makes a stage different from an actuator

A linear stage is a guided carriage built for precision: flat mounting surfaces, preloaded profiled rails or crossed rollers, a defined drive and usually a specified accuracy, straightness and flatness. An actuator moves a load; a stage locates it. If your tolerance is tens of microns or tighter, or you are stacking axes into XY or XYZ, you are buying a stage and should specify it as one.

Work in this order:

1. Drive type: ball screw, belt or linear motor.
2. Load and moments: payload, centre of mass, process forces.
3. Precision specs: repeatability, accuracy, straightness, flatness, angular errors.
4. Error budget at the work point: Abbe, orthogonality, thermal, mounting.
5. Feedback: rotary encoder on the motor, or linear encoder on the carriage.
6. Environment: dust, coolant, cleanroom class, vacuum.

Screw vs belt vs linear motor stages

Aerotech describes linear motors as giving the highest acceleration and speed with no backlash, wind-up or wear, and notes that linear-motor and belt-driven stages cannot be used as the vertical axis of its XZ assemblies (Aerotech). Tolomatic’s figures for belt and screw rodless drives give ±0.05 mm and ±0.013 mm unidirectional repeatability respectively (Tolomatic). Within our range, the BLM ballscrew module quotes ±0.01 mm and the LE module ±0.02 mm.

For screw stages on long travel, check critical speed: it falls with the square of the unsupported length, and Linear Motion Tips recommends staying below 80% of it (Linear Motion Tips). Our belt vs screw comparison works through an example. For linear motor stage sizing, see how to select a linear motor.

The precision specs, and what they really mean

Newport’s definitions are a good common language (Newport):

• Resolution: the smallest increment the system can command or detect. It is not the same as the smallest move it can reliably make.
• Minimum incremental motion: the smallest step the stage delivers consistently. Friction and stiction usually make this larger than resolution.
• Repeatability: unidirectional (always approaching from one side) or bidirectional (unidirectional plus or minus half the reversal error). Newport quotes it at three standard deviations; a figure quoted at one sigma (RMS) looks three times better.
• Accuracy: closeness to the true position along travel. Affected by lead error, scale error, thermal expansion, cosine and Abbe errors.
• Straightness and flatness: deviation from a straight line in the horizontal and vertical planes.
• Pitch, yaw and roll: angular errors about the three axes as the carriage moves.

Always ask on what basis a figure is quoted (sigma level, measured where, under what load). Aerotech, for example, measures straightness and flatness by interferometer about 44 mm above the stage, under no load, mounted on granite (Aerotech). Your tool may be higher and heavier.

The test standard for positioning accuracy and repeatability of numerically controlled axes is ISO 230-2; ASME B5.57 is the US equivalent. Asking which one a figure was measured to tells you whether two catalogues can be compared.

Abbe error: why the height of your tool matters

Angular errors act through the distance between where position is measured and where the work happens:

Abbe error ≈ offset × angle (rad), with 1 arcsecond = 4.85 × 10⁻⁶ rad.

Aerotech’s example: 10 arcseconds of pitch with a 25 mm offset gives about 1.2 µm in Z. The same angle at 150 mm gives about 7 µm. Two ways to reduce it: keep the work point low and close to the encoder line, or buy a stage with tighter angular specifications. Putting a linear encoder at the height of the work point removes most of it along that axis.

XY and XYZ stacking

Stacking introduces errors that no single-axis specification shows:

• Orthogonality: the angle between the two axes. Error = travel × sin(angle). Aerotech’s example: 5 arcseconds over 400 mm gives about 9.8 µm (Aerotech). Aerotech offers assemblies aligned to 5 or 10 arcseconds as options; ask for the figure.
• Crosstalk: the lower axis’s straightness appears as position error in the upper axis’s direction.
• Lower axis angular errors act through the full height of the upper axis and payload, amplifying Abbe error.
• Load and moment on the lower axis: it carries the upper axis, its motor, cable and payload, often with a large offset.
• Cable management: the upper axis’s cables flex with every lower-axis move.

Catalogue accuracy applies only along each axis’s travel; in 3D, stacked-axis errors add. If XY accuracy matters, specify it as a 2D figure across the working area and ask whether the controller can apply a 2D error map.

Thermal effects and the mounting surface

Thermal expansion. Aerotech uses 11.7 ppm/°C (11.7 µm per metre per °C) and tests at 20 ± 1 °C. A 300 mm steel scale or screw grows about 3.5 µm for each 1 °C. Motors, screw nuts and nearby process heat all warm a stage during a shift; if accuracy is critical, specify the operating temperature and allow warm-up time, or use a linear encoder with a scale that expands like the machine base.

Mounting surface. A stage is only as straight as what it bolts to. Aerotech asks for local mounting surface flatness better than 5 µm; more can distort the stage and push pitch, roll, yaw, straightness and flatness beyond specification (Aerotech). A welded frame without a machined datum will not give you a precision stage’s catalogue figures.

Speed limited by encoder data rate. Maximum speed = encoder output rate × resolution. Aerotech’s example: 4 MHz with 0.1 µm resolution allows 400 mm/s, and it recommends commanding 10% less (Aerotech). Check the drive’s encoder input frequency as well as the encoder’s.

Cleanroom and contamination-sensitive stages

Specify the ISO 14644-1 class of the space. Each step is a factor of ten in particle count: ISO Class 5 allows 3,520 particles ≥0.5 µm per m³, Class 6 allows 35,200 and Class 7 allows 352,000 (Engineering ToolBox).

Aerotech’s cleanroom practice includes cleanroom-compatible lubricants, stainless hardware, anodised or painted surfaces, cleanroom cable management with little or no plastic cable carrier, cleaning in reagent-grade isopropyl alcohol with nitrogen drying, and no stage sealing belts, which can generate particles (Aerotech).

That last point is worth noting: a fully sealed module, such as our LEC and LBC, keeps factory dust out of the mechanism, which suits dusty or contamination-sensitive production. Whether it suits a given cleanroom class depends on the seal design and the supplier’s testing. Ask for the class the stage was tested to.

Worked example: XY vision inspection stage

Requirement. Inspect parts on a fixture with a fixed overhead camera. X travel 300 mm, Y travel 200 mm, Y stacked on X. Payload (fixture and part) 4 kg. Inspection point 150 mm above the X carriage’s guide line. Required position accuracy at the inspection point ±10 µm. Moves of 50 mm in about 0.3 s, settle, image. Factory floor at 20 °C ± 2 °C.

Drive type. Short, fast, settle-dominated moves with no process force: a linear motor stage gives the fastest settling, a ball screw stage with a linear encoder is the lower-cost option. Both need a linear encoder to hold ±10 µm.

Error budget at the inspection point (illustrative stage figures; use the datasheet).

• Abbe from X pitch. Assume a pitch specification of 20 arcseconds over travel: 150 mm × 20 × 4.85 × 10⁻⁶ ≈ 14.5 µm. Already over budget on its own.
• Orthogonality. At 10 arcseconds over 200 mm of Y: 200 × 10 × 4.85 × 10⁻⁶ ≈ 9.7 µm.
• Thermal. 11.7 µm/m/°C × 0.3 m × 2 °C ≈ 7 µm on X.
• Repeatability and encoder accuracy. A few µm each.

What to change.

1. Lower the work point. A fixture that puts the inspection point 50 mm above the guide line cuts the Abbe term to about 4.8 µm.
2. Buy the angle. A stage with 10 arcseconds of pitch halves it again.
3. Specify orthogonality at 5 arcseconds (about 4.9 µm over 200 mm) or calibrate it out with a 2D error map in the controller.
4. Control temperature or compensate. Warm the machine up before inspection runs.

Result. A linear-motor X axis with a precision linear encoder, a ball screw or linear-motor Y axis on top with the encoder near the work height, assembled with a stated orthogonality, mounted on a machined plate flat to better than 5 µm, and error mapped in 2D. The ±10 µm requirement is achievable, but only by budgeting the whole stack, not by choosing a stage with a “±5 µm accuracy” headline.

Products in our range to consider

• Tolomatic TRS: twin profile rail screw stage with a wide carrier, loads to 1,960 kg, roller screw lead accuracy 0.010 mm/300 mm, multi-axis brackets.
• BLM built-in linear module: compact C7 ballscrew module with built-in guide, repeatability ±0.01 mm.
• LE ballscrew linear module: ballscrew module for XY and XYZ systems, repeatability ±0.02 mm, strokes to 1,450 mm.
• LEC ballscrew linear module: fully enclosed, dust-resistant version of the LE for contamination-sensitive work.
• LBC belt-drive linear module: sealed belt module, 2,000 mm/s, repeatability ±0.04 mm.
• Akribis AUM: U-channel ironless linear motors with zero cogging, up to 1,980 N continuous, for fast-settling stages.
• Akribis AJM: compact iron-core linear motors with high force density for point-to-point stages.
• KappaSense FREEe: capacitive absolute linear encoder with a passive read head, ±5 µm accuracy and ±1 µm repeatability.
• Bogen ALS21: absolute magnetic linear read head, 1 µm resolution, IP67.

What else the build needs

• Servo drive that supports the linear encoder and the motor type: see servo drives and how to select a servo drive.
• Linear encoder for accuracy at the carriage: see encoders and how to select an encoder.
• Motion controller with error mapping (1D and 2D) and coordinated XY moves: see motion controllers.
• Motor and gearbox for screw stages: see servo and stepper motors.
• Brake or counterbalance on any vertical axis.
• Cable management sized for the stacked axes and cleanroom class.
• Machined base or granite to the stage maker’s flatness requirement.

Things that catch engineers out

• Comparing repeatability figures at different sigma levels. A 1σ figure looks three times better than a 3σ one.
• Treating resolution as accuracy. A 0.1 µm encoder does not make a 0.1 µm stage.
• Ignoring Abbe error. Tall fixtures turn arcseconds into tens of microns.
• Assuming single-axis accuracy holds in XY. Orthogonality and crosstalk add.
• Bolting a precision stage to an unmachined frame. Distortion wrecks straightness and flatness.
• Belt or linear-motor axes in vertical service without a brake or counterbalance.
• Encoder data rate capping speed. Check both encoder and drive input frequency.
• “Cleanroom version” without a class. Ask what it was tested to.

Checklist

Send these to our engineers, or enter them in the spec builder:

• Number of axes, travels and how they are stacked
• Orientation of each axis
• Payload mass, centre of mass and height of the work point above the carriage
• Process forces, if any
• Move profile: distance, time, settling window, duty
• Required repeatability and accuracy at the work point, and the sigma basis
• Straightness, flatness, angular and orthogonality limits if known
• Feedback preference: motor encoder or linear encoder, interface
• Mounting surface and its flatness
• Temperature range and stability
• Cleanroom class, vacuum or other contamination limits
• Controller and drive preferences

For the wider family, see our linear actuators overview.