A coordinate measuring machine touches a feature, records a coordinate and moves to the next. It is accurate, it is trusted, and it is slow enough that very few plants measure every feature on every part. What gets checked is a sample of parts against a sample of features, on the assumption that nothing important is happening in the gaps.
That assumption holds until it does not. A fixture shifts, a tool wears, a supplier changes a process, and the deviation appears where nobody was probing. It surfaces at assembly or at the customer, once the cost has multiplied. So the question before comparing scanners is not which instrument is more accurate, but how much of the part your current method measures and how fast its answer arrives.
What full-surface scanning changes about the measurement
Optical scanning inverts the logic of probing. Instead of choosing measurement points in advance, the system captures the entire outer surface and compares all of it against the CAD model. The output is a deviation map rather than a list of readings, so a distortion nobody predicted still shows up.
The Polyrix fast 3D scanner does this with a 360 degree camera array that images the complete outer surface at once, rather than scanning in passes and stitching them together. A part is captured in about 40 seconds. Nothing touches the surface and nothing is repositioned between views, so the measurement repeats in a way hand-guided scanning does not. That repeatability is what lets a result be trusted by someone who did not take it.
Sampled inspection, or every part?
This decision usually determines whether the purchase makes sense, and it is a production question more than a dimensional metrology one. Inspection that reports days later documents scrap. Inspection that reports in seconds can prevent it, while the process is still running and can still be corrected.
That difference is not worth the same everywhere. Where parts are simple, the process is stable and rejects are rare, faster full coverage is a modest improvement. The case is strongest where a process drifts across a shift, where parts carry real value by the time they reach inspection, or where defects surface at assembly rather than at the station that made them.
Where the scanner will physically sit
These systems run as a standalone station or scan parts in place on an assembly line, and that choice changes what the data is worth. In a metrology room, results arrive after the batch has moved on, which is useful for validation and weak for control. At the line, the same measurement becomes a feedback signal.
So settle the practical questions early: who loads the part, who reads the report, and who can change a setting when a deviation appears.
Part size, and the resolution that applies at that size
The size range is wide, from parts a few centimetres across up to objects of 6.5 metres, and resolution is tied to the volume covered. Around 40 micron resolution is available on parts up to roughly 1.7 metres. Ask any supplier for the resolution at your part's size, not the best number in the range.
Then compare that to the tolerance you have to prove, not to a competitor's specification sheet. A tolerance band that is a comfortable multiple of the measurement uncertainty makes for calm inspection. One that sits close to it produces weekly arguments about whether the part or the measurement is at fault.
Parts with no model, and models that no longer match the part
Reverse engineering is the other common reason these systems get bought: legacy tooling, a supplier component that arrived without drawings, a die reworked on the floor until it stopped matching its file. Capturing the geometry that exists beats reconstructing it from measurements. It also serves first-article inspection and tooling validation, where the question is whether what you built matches what you designed across the whole surface, not at the sampled points.
The limits of measuring only the outside of a part
A 3D scanner measures the outside of a part: geometry, form and deviation from the model, and nothing else. No amount of resolution changes that, and the limit is worth stating plainly, because it is usually discovered after the system arrives.
Optical scanning is also line of sight. What the cameras cannot see is not measured, so deep bores, undercuts, internal channels and anything already closed inside a housing sit outside the data. If the question is internal, porosity in a casting, a void in a weld, a component misplaced in a sealed assembly, it belongs to industrial micro-CT, or at production volumes to a conveyor-integrated system such as Lumafield. Delamination and disbonds in composites belong to ultrasound, where XARION's optical microphone works without couplant gel and without contact.
The techniques are complementary rather than competing: scanning verifies the shape you can see, and non-destructive testing methods verify the structure you cannot. Two further limits get less attention. A deviation map says the part is wrong, not why it is wrong, and the cause is a process question no measurement system answers for you. And scanning does not retire the probe everywhere: where a customer specification names a probing method and a datum scheme, full-surface data supplements the CMM report rather than replacing it. Plenty of plants run both.
Scan the part you argue about most
One part usually generates the arguments: the housing that passes inspection and fails at assembly, the bracket whose flatness nobody can agree on. Scan that one, lay the deviation map against the tolerance you have to defend, and see whether it explains the failures the probe missed. If the surface is inside tolerance and the parts still fail, the fault was never on the outside.
Merkel Technologies represents Polyrix in Israel, so the demonstration runs on your own samples, with training and service handled locally. Bring the awkward part too: the reflective finish, the black polymer, the piece nobody can fixture cleanly.