Worked example, for planning only. A stamped bracket has a hole position toleranced at ±0.10 mm, a band 0.20 mm wide. With a 5-megapixel camera that is 2,448 pixels across, each pixel covers the field of view divided by 2,448. The table shows how the choice of field of view changes how many pixels span the tolerance band. These figures are arithmetic on the stated assumptions, not a promised accuracy: what the station actually achieves is shown by the study on your parts.
Example: field of view, pixel size and a ±0.10 mm tolerance (assumed 2,448-pixel-wide sensor)| Field of view across | Pixels across | Size of one pixel on the part | Pixels across a 0.20 mm band | What it means |
|---|
| 30 mm | 2448 | 0.012 mm | about 16 pixels | Comfortable margin; small parts or one feature |
|---|
| 60 mm | 2448 | 0.025 mm | about 8 pixels | Typical single-part view with room for position tolerance |
|---|
| 120 mm | 2448 | 0.049 mm | about 4 pixels | Relies heavily on sub-pixel edge fitting and very stable light; often split into two cameras |
|---|
Two things follow. First, the field of view is the main lever: halving it doubles the pixels across the tolerance. Second, edge-finding tools in vision software, such as the size measurement tools in HIKROBOT VisionMaster (HIKROBOT VisionMaster), fit edges to sub-pixel positions, but only on crisp, stable edges. That is why precision stations pair a parallel backlight with a telecentric lens, as HIKROBOT’s own catalogue describes (HIKROBOT light catalogue).
How the acceptance numbers are defined
The acceptance report uses the terms below. NIST’s handbook describes gauge R&R studies as characterising the errors that affect measurements made by production gauges (NIST e-Handbook, 2.4). The AIAG thresholds are quoted from BPI Consulting, which also argues they can mislead (BPI Consulting); uncertainty follows the GUM (JCGM 100:2008).
Acceptance metrics for a vision measuring station| Term | What it means on a vision station | How it is reported |
|---|
| Repeatability | Variation when the same part is measured again under the same conditions, with the part reloaded each time | Standard deviation of repeat measurements; NIST treats it as the basic precision of the gauge |
|---|
| Reproducibility | Variation when conditions change: shift, recipe reload, fixture, temperature | Extra spread between condition groups |
|---|
| %GRR | Repeatability and reproducibility together, as a share of the tolerance or of process variation | AIAG guideline: under 10 % adequate, 10 to 30 % may be acceptable, over 30 % unacceptable; criteria are agreed with the customer |
|---|
| Bias (agreement) | Average difference between the station and your reference (CMM or calibrated gauge) on the same parts | Corrected in calibration, then re-checked on reference parts at set intervals |
|---|
| Measurement uncertainty | The doubt that remains about a measured value, combining all known error sources | Evaluated with the GUM method; decides how wide the near-limit divert band is |
|---|
Traceability in South Africa
A station is only as right as the reference it is compared with. In South Africa, the Measurement Units and Measurement Standards Act provides for national measurement standards and establishes the National Metrology Institute, whose role includes disseminating traceability to industry (Act 18 of 2006). SANAS is recognised as the only national body that accredits calibration, testing and verification laboratories (Act 19 of 2006, s4). In practice: the reference parts used to accept a vision station are measured on a CMM or with gauges whose calibration certificates come from an accredited laboratory, and those certificates go into the acceptance file.