Multi-axis positioning system for optical inspection: stacked, integrated or planar XY axes under the camera, corrected with a 2D error map

01

Stacked, integrated or planar XY: what each does to flatness across the inspection field

PI explains that XY motion can come from stacking two linear stages, which is modular and flexible, but over long travel flatness and pitch/roll errors grow towards the end of travel because the base of the upper axis is unsupported; its flatness measurement on a stacked stage shows the cantilevered, loaded Y axis performing worse than the lower X axis. Integrated XY stages, which it says are at times wrongly called planar, share a central plate as the top of the lower axis and the base of the upper one, giving a lower profile and a wider bearing base that reduces cantilever errors, with the same bearings and drives as stacked stages. Planar XY stages use one reference plane for both axes, based on air bearings, magnetic levitation or flexures, and PI shows air-bearing planar stages reaching flatness of 1 µm and better over hundreds of millimetres. [1]

PI calls flatness especially critical where an optical process must stay in focus, since vertical deviation turns into focus error. [1] Editorial calculation: an orthogonality error of 10 arcseconds between X and Y shifts a point by about 4.8 µm over 100 mm of Y travel, which can exceed the pixel size of the inspection camera; it is one of the errors a 2D map can correct.

Module-built systems sit at the stacked end of this range: Fuyu lists XY tables and XYZ stages from single to three axes and gantry systems, with visual inspection among its application fields. [4]

Give the supplier the part or wafer size, the inspection field and the camera's pixel size, the axes needed (X, Y, theta, focus), the load, the move-and-settle time per image, the accuracy needed across the field, and the base and environment.

02

A 2D error map from a glass grid plate and the camera, fitted point by point

Dover Motion lists the errors that mapping can correct: leadscrew cumulative and periodic error, Abbe error, nut backlash, cosine error and deviation from orthogonality in multi-axis systems. Its mapping station places glass grid plates with 12 µm wide chrome lines every 10 mm, the smaller ones on near-zero-expansion Zerodur, in a tip-tilt-yaw mount on the XY stage under test, which sits on a granite surface plate under a bridge carrying a Z focusing stage, a long-working-distance 100× objective and a machine vision camera; it reports X-Y accuracy improved ten to twenty times over an unmapped stage. [2]

Dover's worked example: a screw stage with 1 µm resolution that moves 100.013 mm for 100,000 steps is commanded 99,987 steps for a true 100 mm, and, assuming linear screw error, 49,994 steps for 50 mm; it notes that thermal expansion limits how much accuracy mapping can add to leadscrew systems. [2] Editorial calculation: 0.013 mm over 100 mm is a scale error of 130 ppm.

Zaber's 2D calibration note, for calibrating XY systems for orthogonality, distortion and stretch of each axis, fits expected against actual grid positions: a biquadratic fit needs at least 3 × 3 = 9 points, and with 5 × 5 = 25 points the code makes a least-squares fit that reduces the error further. [3]

03

Send the inspection field, axes and accuracy together

Paste it into your enquiry below or into an email to a maker, then replace every bracket.

Field: [part or wafer size; camera pixel size]
Axes: [X; Y; theta; focus]
Construction: [stacked; integrated; planar]
Accuracy: [across field; repeatability]
Mapping: [grid plate; interferometer; points]
Cycle: [move-and-settle per image]

See manufacturer records Send an enquiry through CNFX