A scanner can have very dense sampling and poor dimensional truth, or modest point spacing and excellent length measurement. The specification terms describe different things and should not be collapsed into one “precision” number.
Resolution tells you how finely the system can sample or represent geometry. Accuracy tells you how close a result is to the reference. Precision describes repeatability. Volumetric accuracy describes behaviour across a measurement length or volume. Measurement uncertainty describes how much doubt remains around the reported result. You need the combination that matches the characteristic you are measuring.
Scanner marketing often compresses several technical quantities into one line. Before comparing devices, ask exactly what the number represents: a single frame, repeated scans of one artefact, point spacing after fusion, a length error, or a maximum permissible error under a defined standard. Those are not interchangeable.
Imagine a scanner producing points every 0.05 mm while the entire reconstruction has a 0.30 mm scale error. The mesh will look extremely detailed, but the part will still be wrong. Conversely, a system with coarser point spacing can measure a 200 mm centre distance well if scale, calibration and registration are controlled.
The right analogy is a digital photograph: more pixels can reveal smaller texture, but pixel count does not guarantee that a ruler photographed in the image has the correct scale.
When manufacturers publish a relationship such as 0.02 mm + 0.04 mm × L(m), the fixed term and the length-dependent term acknowledge that error can grow across larger objects. That is the right direction of thinking for reverse engineering because registration, camera geometry and calibration have spatial effects.
For a 0.50 m measurement length, a specification written as 0.02 mm + 0.04 mm × L(m) evaluates to 0.04 mm under the manufacturer's stated model. That arithmetic is useful, but it is still not a complete uncertainty budget.
A real measurement result depends on the scanner, calibration artefact, environment, operator, surface, registration, software settings, alignment method and the characteristic being extracted. If the tolerance is ±0.10 mm and the expected measurement uncertainty is of similar magnitude, the scanner cannot support a confident pass/fail decision simply because a datasheet says “0.02 mm accuracy”.
For inspection work, think in terms of a measurement capability ratio: the measurement system should consume only a controlled fraction of the tolerance. The tighter the decision, the more important traceable artefacts, repeatability studies and independent verification become.
Structured-light and laser systems depend on observing projected or reflected light. Dark absorption, specular reflection, translucency and subsurface scattering can shift or remove data. A “no spray required” capability can improve workflow dramatically, but it should be validated on the actual material and finish. Black anodised aluminium, chrome, translucent mouldings and carbon-fibre laminates are not optically equivalent.
The smallest visible groove is influenced by point spacing, optical blur, triangulation angle, exposure, surface contrast, meshing and the orientation of the feature relative to the cameras. If a critical edge is only two or three sample intervals wide, dimensional extraction will be unstable even if the feature is visible on screen.
For design capture, aim for substantially more sampling density than the smallest feature you need to reconstruct. For inspection, validate the actual characteristic with an artefact or a known part rather than relying on a theoretical sampling ratio.
| Question | Why it matters |
|---|---|
| Is the accuracy single-frame, length measurement or volumetric? | Determines whether the number applies to local geometry or a whole object. |
| What object size and working distance were used? | Performance often changes with field of view and distance. |
| What tracking mode? | Feature, marker and photogrammetric tracking have different drift behaviour. |
| What surface and preparation? | Optically cooperative artefacts are easier than real parts. |
| Can the user recalibrate? | Reverification matters after transport, impacts or thermal change. |
| Is a standard referenced? | A defined acceptance method is stronger evidence than an isolated marketing claim. |
For serious reverse engineering, keep a simple artefact or known component with traceable dimensions in the workflow. Scan it using the same mode, working distance and processing settings as the job. Trend the result over time. This will not turn a consumer scanner into a CMM, but it will tell you whether the process has drifted and whether a claimed capability is reproducible in your environment.
Sources and standards checked at the review date.
