What Does 0.02 mm Accuracy Really Mean for a Small Aerospace Part? A Verification Protocol

A scanner spec sheet lists two numbers for the same handheld device: 0.015 mm accuracy and 0.020 mm resolution. In the incoming-inspection procedure for a small titanium bracket with a ±0.05 mm profile tolerance on its mounting flange, someone writes "0.02 mm accuracy" and moves on. Two different specifications just got folded into one number, and the part's actual tolerance never entered the conversation.

This happens constantly with small aerospace hardware. Brackets, clevises, fittings, and turbine components carry tight profile and position tolerances, and the temptation is to look at a single headline accuracy figure and assume it covers the part. It doesn't, not by itself. Accuracy, resolution, repeatability, volumetric accuracy, and measurement uncertainty are five different things, and mixing them up is how a "compliant" inspection report ends up backed by less margin than anyone realizes.

This article walks through what each term actually measures, how a stated accuracy claim like 0.02 mm should be verified against the relevant standards, and why device accuracy and part tolerance are never the same comparison.

Five Terms That Get Flattened Into One Number

Vendors, and sometimes internal procedures, tend to collapse a handful of distinct metrology terms into a single adjective: "accurate." Each of the five terms below answers a different question, and a quality engineer needs all five before deciding whether a device is right for a job.

Accuracy describes how close a measurement comes to a true or reference value, usually a certified artifact like a calibrated sphere or step gauge. It's a single number that describes closeness, not a guarantee that every measurement on every part will land that close.

Resolution is the smallest change in the measured quantity that produces a detectable change in the device's output, a question of instrument discrimination, not of how close a reading lands to the truth. A scanner can have excellent accuracy against a reference sphere and still have a coarser resolution limit that makes it harder to pick up a shallow scratch or a fine radius. Accuracy and resolution are reported separately on a spec sheet for exactly this reason. In the scenario above, the 0.020 mm figure is resolution, not accuracy, and reading it as the accuracy number understates the actual accuracy (which is finer, at 0.015 mm) while also skipping past what resolution is actually telling you.

Repeatability is what you get from measuring the same feature on the same part with the same device and operator, over and over, in a short window. It's reported as a standard deviation or a spread, and it answers "how consistent is this device with itself," which is a different question from "how close is it to the truth." A device can be repeatable and still be off from the reference value in a consistent, biased way.

Volumetric accuracy accounts for the fact that error tends to grow as the measurement volume gets larger. It's typically expressed as two terms added together, a fixed value plus a length-dependent term in mm per meter, because a device's error at 100 mm and its error at 1 meter are not the same number. For a small aerospace bracket that fits in two hands, the length-dependent term barely moves the needle. For a fuselage panel or a long tooling fixture, it matters a great deal.

Measurement uncertainty is the broadest of the five. Per the internationally recognized Guide to the Expression of Uncertainty in Measurement (GUM), it's expressed as a dispersion parameter, often reported as an expanded interval around the result, and it can draw on contributions from the device, the operator, the part's material and surface finish, fixturing, alignment strategy, and the environment. A device's published accuracy is one possible input into that budget, not the whole budget.

How a Stated Accuracy Number Is Supposed to Be Verified

A single accuracy figure on a spec sheet is a starting point, not proof. Three things separate a claim that holds up from one that doesn't: the test standard behind it, whether the number is expressed as Maximum Permissible Error (MPE), and whether an accredited lab produced the certificate.

ISO 10360-13 is the acceptance and re-verification test standard for optical 3D coordinate measuring systems. It extends the broader ISO 10360 family, originally built around tactile coordinate measuring machines, to scanner-based systems, defining how probing error and other performance metrics are measured against certified reference artifacts.

VDI/VDE 2634 Part 3 was the German guideline covering multiple-view optical systems based on area scanning (as opposed to single-point probing), historically the standard most often cited for handheld and structured-light 3D scanners specifically, with quality parameters built around sphere and length-related checks. It was formally withdrawn in favor of the international ISO 10360-13:2021 standard (adopted in Germany as DIN EN ISO 10360-13:2023-11), though the transition is recent enough that many manufacturers, SCANOLOGY included, still cite VDI/VDE 2634 Part 3 alongside ISO 10360-13 on current spec sheets. A certificate referencing VDI/VDE 2634 Part 3 on its own isn't automatically a red flag, but for a new purchase or a re-verification cycle going forward, ask for ISO 10360-13 as the primary reference.

MPE language matters. A spec that states a Maximum Permissible Error, a conformity limit defined for a specific test method and set of conditions, is a bounded, checkable claim. An open-ended "up to" figure, with no stated test method, measurement volume, or conformity rule attached, is much harder to hold a vendor to.

Accreditation closes part of the loop. A calibration certificate is only as trustworthy as the lab that issued it and the scope that lab is accredited for. ISO/IEC 17025 accreditation confirms a lab's competence for the specific tests within its accredited scope, so check that the test in question actually falls inside that scope rather than treating a general accreditation as automatic coverage. Then check that the certificate itself states the artifact used, the measurement volume it covers, the environmental conditions during the test, and an associated uncertainty value, not just a pass/fail stamp.

Two separate tests matter here: an acceptance test, run when a device is new or right after a repair, and periodic re-verification tests, run on a schedule to confirm the device hasn't drifted. A three-year-old test result on a device that's been dropped, serviced, or moved between shop floors doesn't tell you much about its current state.

A Practical Verification Protocol for a Small Part

For a quality engineer receiving a scanner, or receiving scan data from a supplier, the following sequence turns a spec sheet number into something defensible on paper.

1. Pull the calibration certificate first. Confirm it comes from an ISO/IEC 17025 accredited lab, states an MPE (not an "up to" claim), and references ISO 10360-13 or VDI/VDE 2634 Part 3 by name. If none of that is on the certificate, treat the headline accuracy number as unverified.

2. Check what the certificate actually covers. A certificate built around a large reference artifact tells you about performance at that artifact's scale, not necessarily at the scale of a 150 mm bracket. Match the reference volume to your part's actual envelope so the number you're relying on transfers directly to the job in front of you.

3. Run a short acceptance check of your own. A certified reference sphere or step gauge sized to your part, measured on receipt and periodically afterward, gives you a live data point instead of relying entirely on a certificate that may be months old.

4. Run repeats separately from the accuracy check in step 3. Measure the same feature on your actual part five to ten times under your real shop conditions, lighting, temperature, and fixturing included. Step 3 tells you how close the device lands to a certified reference value; this step tells you how consistent it is with itself, which is a different (and often larger) number.

5. Build an uncertainty budget for the actual job, not just the device. Add in the part's surface finish and material (reflective or dark surfaces change how reliably a scanner captures data), the fixturing and alignment strategy, and the environment. The device's MPE is one line item in that budget, not the final answer.

Device Accuracy Is Not the Same as Part Tolerance

This is the step that gets skipped most often, and it's the one that actually determines whether a device belongs on a given job.

Measurement systems analysis practice offers two separate guidelines here, and it's worth not blending them into one. The AIAG "Rule of Tens" calls for a gauge's resolution to divide the tolerance band into at least ten parts, a starting point for picking an instrument, not a statement about total measurement uncertainty. A separate, more rigorous convention used in calibration work, the Test Uncertainty Ratio (TUR), calls for a measurement system's expanded uncertainty for that specific application to sit at roughly one-quarter of the tolerance band (a 4:1 ratio), which is a different number from either resolution or the device's headline accuracy figure.

Applied to a ±0.05 mm profile tolerance (a 0.10 mm total band): the Rule of Tens points toward an instrument with resolution at or below roughly 0.01 mm, using the full 0.10 mm band. TUR is conventionally calculated against the tolerance half-width instead, so a 4:1 TUR here means the expanded measurement uncertainty for this specific application needs to sit at or below roughly 0.0125 mm (the ±0.05 mm half-width divided by 4), not simply a device accuracy spec of 0.02 mm sitting next to a 0.05 mm tolerance. Neither guideline is satisfied just because two numbers on paper happen to look close to each other. Stack repeatability, the part's volumetric growth term (small for a bracket, but not zero), and alignment uncertainty on top of the single-point accuracy figure, and the margin that looked comfortable on a spec sheet gets thinner fast.

Small aerospace parts add their own complications on top of the arithmetic. Thin-walled brackets, deep holes, blended fillets, and tight datum features all introduce alignment and feature-extraction uncertainty that a certified-sphere accuracy test never touches, because a sphere doesn't have a fillet radius or a datum scheme. A device that performs well against a reference artifact still needs to be evaluated against the specific feature types on the part in front of you.

The takeaway isn't that a 0.02 mm-class device is unsuitable for tight-tolerance aerospace work. It's that "the device is accurate to 0.02 mm" and "the device can verify a ±0.05 mm tolerance on this part" are two separate claims, and only the second one is the question that matters. Matching device class to tolerance class has to be a deliberate step, not an assumption based on two numbers that happen to look similar.

Where SIMSCAN-S Gen2 and DefinSight Fit Into This Protocol

SCANOLOGY's SIMSCAN-S Gen2 is a palm-sized handheld 3D scanner built for exactly the small, tight-tolerance parts this protocol is meant to evaluate. Its published specifications are 0.015 mm accuracy and 0.020 mm resolution, with volumetric accuracy of 0.015 mm + 0.03 mm/m in standard mode, improving to 0.015 mm + 0.012 mm/m when paired with a high-precision scale bar using Adaptive Photogrammetry. Performance is verified under ISO 10360-13, VDI/VDE 2634 Part 3, and JJF 1951, with testing through ISO/IEC 17025 accredited labs, the same standards this article walks through above.

SCANOLOGY positions SIMSCAN-S Gen2 for high-end manufacturing environments including aerospace and advanced machining, where tolerance integrity drives fit and quality decisions. That framing lines up with the small-part scenario at the top of this article: a compact, short-baseline scanner designed to reach into deep holes, slots, and gaps on a bracket-sized part while holding its published accuracy on complex geometry, not just on a smooth reference sphere.

Running the protocol above still requires software that turns raw scan data into a documented result. DefinSight, SCANOLOGY's own metrology software platform, handles the alignment, full-field deviation analysis, GD&T feature evaluation, and report generation that connect a scan to a written inspection record. That's the piece that makes steps three and four of the protocol, the acceptance check and the repeatability run, something you can actually capture and hand to an auditor as a report, rather than something that lives only in a spec sheet claim. Building the uncertainty budget in step five is still an engineering judgment call on the quality engineer's side; the software's role is producing the aligned, documented measurement data that budget is built on.

None of this replaces running your own acceptance check on receipt. A published accuracy figure, even one backed by ISO 10360-13 and an accredited lab certificate, is the starting point for the verification protocol above, not a substitute for it.

FAQ

Is a 0.02 mm accuracy scanner good enough for a ±0.02 mm part tolerance? Generally no, or at best with almost no margin. A tolerance band that matches the device's accuracy figure one-to-one leaves no room for repeatability, volumetric growth, or alignment uncertainty stacking on top. Measurement systems analysis guidance calls for the measurement system to resolve to roughly a tenth of the tolerance band, which for a ±0.02 mm tolerance points toward a device rated well finer than 0.02 mm, not equal to it.

What's the difference between accuracy and repeatability? Accuracy measures closeness to a true, certified reference value. Repeatability measures how consistent the device is with itself across repeat readings on the same feature, same operator, same conditions. A device can be highly repeatable and still carry a consistent bias away from the true value, which is why both numbers get reported separately rather than folded into one.

Does volumetric accuracy matter for a small bracket if it barely matters for a 100 mm part? It matters less in absolute terms, since the length-dependent term in a volumetric accuracy figure (the mm/m portion) has little distance to accumulate over on a small part. But it's still worth checking, because "small" and "zero" aren't the same thing, and multi-position scans that stitch several small features together across a larger fixture can reintroduce the volumetric term.

What should a calibration certificate actually include? At minimum: the accredited lab's ISO/IEC 17025 status, the test standard used (ISO 10360-13 or VDI/VDE 2634 Part 3), the reference artifact and measurement volume tested, the environmental conditions during the test, an MPE or measured error value, and an associated uncertainty figure. A certificate that only states a pass/fail result without these details doesn't give you enough to build your own uncertainty budget.