Your first lab (30 minutes)
Run a complete pixel size calibration with no hardware attached, and see how far your result sits from ground truth.
One goal here: get you to a deviation number within 30 minutes. Not "the software opens" — "my mm/px is 0.018 mm off the truth".
Why pixel size first
Pixel size (mm/px) is the base of 2D vision work: it converts distance in the image into distance on the stage. Positioning, alignment and measurement all sit on top of it. It is also easy to get wrong, and getting it wrong does not raise an error — it shows up on the line as "sometimes it's a bit off".
Steps
1. Open the virtual machine
File → Open machine → demo-2axis-vision. Two axes and one camera:
X/Y stage 200 × 200 mm travel, ±3 µm repeatability, 2.4 µm backlash
Camera 2448 × 2048, 3.45 µm pixels, 25 mm lens, 200 mm working distance
Target Circle grid, 5.000 mm spacing (ignore this value for now — it is the answer key)
2. Start the wizard
Calibration → Pixel size. Five steps, exactly as on a real machine:
| Step | What it does | What to watch |
|---|---|---|
| prepare | Pick camera, target type, number of poses | At least 9 whole circles visible |
| acquire | Move to N poses, grab a frame at each | Progress, circles found per pose |
| solve | Least squares for mm/px, rotation, non-orthogonality | Residual RMS |
| verify | Re-project poses that were not used in the solve | the deviation number |
| persist | Versioned write to disk | Record ID |
Press Start, keep the defaults, wait about 40 seconds.
3. Read the result
solve: mm/px = 0.019823 theta = 0.0041 rad skew = 1.6e-4
truth: mm/px = 0.019841 theta = 0.0039 rad
verify: deviation 0.018 mm over 20 mm tolerance 0.05 mm → PASS
The second line is the point. The virtual device knows the true value, so it can tell you whether your calibration is actually right. On real hardware that line does not exist — which is exactly why practising here is worth the time.
4. Now break it on purpose
Do not skip this step; it is worth more than the first three.
Devices → Virtual devices → Camera → Error model: change lens distortion k1 from
-0.0214 to -0.08, then re-run only acquire and solve (the wizard re-runs single
steps; you never start over).
You will see:
- residual RMS grows — but not obviously enough to look wrong;
- the verify deviation goes from 0.018 mm to roughly 0.15 mm, a clear FAIL.
That is the classic signature of calibrating pixel size before distortion: fine near the centre, wrong at the edges. Crowding all your poses into the middle of the field hides it completely, which is the single most common beginner mistake.
5. Do it right
Run Calibration → Intrinsics and distortion first, then pixel size again. The deviation
returns to the 0.02 mm range.
That is what "chain" means: these calibrations have an order, and if an earlier one is wrong, the later one cannot be right.
Where to go next
- Full pass criteria, common mistakes and review questions: Lab 1;
- How the virtual camera produced that image: World and imaging;
- Driving this from your own code: vmc.dll API.