How to Display the Image Magnification on a Fully Automatic CNC Image Measurement Instrument

• 1
  Key takeaways
• 2
  What the magnification readout actually tells you
• 3
  Choose the right magnification for the feature
• 4
  Calibrate the displayed magnification before you inspect
• 5
  Set lighting and edge detection so the readout stays honest
• 6
  Common mistakes that break the magnification reading
• 7
  Step by step: display and lock the magnification
• 8
  Magnification settings and what they suit

On a fully automatic CNC image measurement instrument, the magnification shown on screen is a calculated value. The software takes the objective or zoom lens position, the camera sensor pixel pitch, and the current field of view, then converts them into a single number. That number tells the operator how many micrometres of the part sit on one camera pixel.

A 2× setting on a 5 MP camera with 3.45 μm pixels gives roughly 1.7 μm per pixel. A 10× setting on the same camera drops that to about 0.35 μm per pixel. The second number is more useful for measuring a 0.1 mm chamfer, but it also shrinks the field of view, so you may only see 1.5 mm of the part at once.

The instrument does not measure the part directly. It measures pixels, then multiplies by the calibration factor. That is why the displayed magnification and the calibration factor must always be paired. If you change the lens and forget to reload the calibration file, every dimension on the report is wrong by the ratio between the two settings.

• 1
  Pixel size matters more than the round numberA 10× label is marketing; μm per pixel is what limits your measurement.
• 2
  Field of view shrinks as magnification risesCheck that the whole feature fits before you commit to a zoom level.

A common rule on the floor is to spend at least 10 camera pixels on the smallest tolerance you need to verify. If the drawing calls for ±0.02 mm, you want about 0.002 mm per pixel or better. That usually means a 5× to 10× objective on a standard 5 MP sensor, or a zoom setting in the upper half of the range.

Going higher is not free. At 20× or 30×, depth of field falls to a few micrometres. A stamped bracket with 0.15 mm of bow will drift out of focus across the field, and the edge detection starts wandering. For those parts, drop to 5× and check flatness separately with a touch probe or a height gauge.

Zoom lenses are convenient but their magnification drifts with temperature. A machine that sits near a loading door can shift 0.1% to 0.3% over a day. For work inside ±0.005 mm, use a fixed objective and re-run the calibration check at the start of each shift. It takes about two minutes and it catches most drift before it reaches a report.

• 1
  Minimum 10 pixels per tolerance bandFewer pixels means the edge algorithm has too little data to average.
• 2
  Fixed objectives for tight workZoom lenses are fine for screening; fixed glass holds calibration better.

Calibration starts with a certified stage micrometer, typically a 1 mm scale with 10 μm divisions, or a glass grid with 0.1 mm squares. Place it on the stage, bring the scale into focus at the magnification you plan to use, and let the software measure the known distance. The instrument then back-calculates μm per pixel for that exact optical setup.

Do not calibrate at 2× and then measure at 10×. Each objective or zoom position needs its own calibration file. The same applies after any of these events: a lens change, a camera swap, a ring light replacement that changes the working distance, or a software update that touches the image pipeline.

After calibration, verify with a second artefact. A 5.000 mm gauge block or a certified pin gauge works well. If the check reads 4.998 mm or 5.002 mm, the setup is usable. A 0.05 mm error means something is wrong with the scale entry, the lens selection, or the pixel pitch value in the configuration file.

• 1
  One calibration file per lensName the file with the objective and zoom position so nobody loads the wrong one.
• 2
  Verify against an independent artefactThe calibration standard and the check standard should not be the same object.

Edge detection on a CNC image measurement instrument uses contrast, not geometry. A sharp step in grey value is treated as the part edge. That means the lighting setup directly controls the number the instrument reports, even though the magnification readout has not changed.

Coaxial light works well for flat, reflective surfaces such as ground steel or anodized aluminium. It reveals scratches and burrs that a ring light hides. Ring light at a low angle, around 20° to 30°, is better for turned diameters and drilled holes because it throws a shadow into the edge. Backlight is the most repeatable option for through-holes and thin sections, but it needs a translucent fixturing plate.

Set the exposure so the background sits near 200 to 230 grey levels and the part edge falls to below 50. Then fix the threshold. If you let the software auto-set the threshold on every image, the reported edge moves with surface finish. A single threshold value stored in the program keeps the measurement repeatable between operators.

• 1
  Lock the exposureAuto exposure turns a surface finish change into a dimensional change.
• 2
  Store the threshold in the programOne value per feature type, checked on a master part each shift.

The most frequent failure is calibrating at one magnification and measuring at another. The screen still shows a sensible number, so nobody questions it, and the report ships with a 5× error on every dimension. Add a line to the setup sheet that names the calibration file and the objective, and have the operator sign it.

The second failure is measuring a feature that does not fit the field of view. Software can stitch images, but the stitch introduces its own error, often 2 to 5 μm at the seam. If a 12 mm bore needs a 10× setting, measure it at 5× or use a touch probe instead.

The third is a dirty or scratched protective window on the lens. It lowers contrast, the edge detection widens, and the reported dimension creeps. Check the window at the start of every shift and replace it when scratches are visible under a torch.

• 1
  Wrong calibration fileSymptom: all dimensions scale by the same ratio.
• 2
  Feature larger than the fieldSymptom: repeatability drops on the biggest dimension only.
• 3
  Contaminated lens windowSymptom: slow drift over a shift, worse on reflective parts.