CNC accuracy in Belgium: what the number on your drawing actually buys
This page is for design engineers and buyers in Belgium who have to specify a tolerance and then prove it was met. It covers how accuracy is quoted, where it is lost on the shop floor, and which measurements belong in a first-article report. By the end you can tell whether a supplier's accuracy claim is checkable or just a word.
Accuracy is a system, not a machine spec
A tolerance is only real when the machine, the fixture, the tool and the temperature all agree at the same moment.
What Belgian buyers mean when they say accuracy
Two words get mixed up in most RFQs. Accuracy is how close a machined feature lands to the nominal value on the drawing. Repeatability is how close a batch stays to itself after the machine has run for six hours. A shop can be repeatable without being accurate, and that is the worst case for you: every part matches its neighbour and every part is wrong.
The second split is dimensional versus geometric. A bore can measure Ø25.000 mm on a micrometer and still fail because it is oval, tapered or off-axis. Position, concentricity, perpendicularity and profile tolerances come from the same cut but are measured a different way, usually on a coordinate measuring machine rather than with hand tools.
On our floor the working limit for a normal production run is ±0.005 mm (±0.0002 in). That figure is not a sales number. It is what the process holds when the machine is warm, the fixture is rigid and the drawing allows a single setup. Ask for it on a thin wall in a soft aluminium and the answer changes, which is why we read the drawing before quoting.
- 1AccuracyDeviation from nominal on one feature
- 2RepeatabilitySpread across parts in the same run
- 3Geometric toleranceForm, orientation and location, not size
- 4CapabilityWhat the process holds at volume, not once
Six places accuracy disappears before the cutter touches metal
Thermal growth is the first. A spindle that has run for twenty minutes is longer than a cold one. Castings, ballscrews and the part itself all move at different rates. On a 300 mm aluminium part, a 5 °C shift across the length is already worth more than the tolerance we are chasing. That is why preheating cycles and temperature-controlled rooms are not optional for tight work.
Setup count is the second. Every time a part is unclamped and re-fixtured, the relationship between the two datums is rebuilt from scratch. A feature machined in a second setup inherits the error of the fixture, the vise jaw and the operator's dial-in. Five-axis machines exist largely to remove that step: more faces cut in one clamping, fewer stacked errors.
The third is tool deflection. A long reach tool pushed hard in a deep pocket bends. The cut looks fine on the machine and the wall is tapered when you measure it. Roughing strategies, reduced radial engagement and shorter gauge length are the fixes, not a slower feed alone.
Fixtures matter more than people expect. A part clamped on four points and supported in the middle will spring back after unclamping, so the measurement after release does not match the measurement in the fixture. The fourth is material instability: 7075 and 17-4PH move after stress relief is broken by heavy roughing. Rough, stress-relieve, finish. Skip the middle step and the part drifts for a week.
The fifth is probing and offsets. A worn probe tip or a stale tool offset quietly shifts every feature on the part. The sixth is the drawing itself: a tolerance applied to a datum that does not exist on the finished part cannot be measured, and will be argued about at incoming inspection.
- 1Thermal driftPreheat the spindle, control the room
- 2Setup countFewer clamps, fewer stacked errors
- 3Tool deflectionShorten gauge length, reduce radial load
- 4Residual stressRough, relieve, then finish
Which tolerance class fits which feature
Pick the loosest tolerance the function allows, then let the shop hold it comfortably.
| Feature | Typical tolerance | Process route | Note |
|---|---|---|---|
| Bearing bore, aluminium | ±0.005 mm | 5-axis, single setup | Hold after thermal soak |
| Bolt hole pattern | ±0.05 mm | 3-axis, drill and ream | Position, not size, drives cost |
| Sealing face | ±0.01 mm, Ra 0.8 μm | Mill, then fine finish | Flatness usually matters more |
| Thin wall, 1 mm | ±0.05 mm | 4-axis, light passes | Springback after unclamping |
| Long shaft, 1,000 mm | ±0.02 mm | Mill-turn | Support the free end |
| Prototype, one-off | ±0.05 mm | 3-axis | Cheaper to iterate, not to gauge |
How to check the claim without owning a CMM
A tolerance you cannot measure is a promise, not a specification. If your team has no coordinate measuring machine, ask for the numbers instead of the part. A first-article report with the nominal, the actual and the deviation for each controlled feature is enough to see whether the process is centred or running at the edge of the band.
We inspect 100% of parts before shipment, with raw material verification, in-process checks and a final inspection. Reports are issued on request. For tight features the report names the instrument: micrometer, bore gauge, height gauge, profile projector or CMM. If a supplier cannot say which instrument produced a number, treat the number as marketing.
For Belgian buyers importing from Asia, the practical question is not only whether the first article passes. It is whether part 500 still passes. Ask for the inspection method and the sampling plan for the production run, and ask what happens to a part that measures 0.006 mm out on a ±0.005 mm callout. A shop with a real process knows its scrap rule before you ask.
One more habit worth building: measure the part at 20 °C, or state the temperature. A steel part measured on a warm bench in a cold room can read outside tolerance purely from thermal expansion. It happens often enough that we write the measuring temperature into the report.
- 1First articleNominal, actual, deviation per feature
- 2InstrumentNamed in the report, not implied
- 3TemperatureMeasure at 20 °C or record it
What our machines and quality system can hold
GreatLight has run since 2011 and now operates 127 high-precision CNC machines across three wholly-owned plants covering 7,600 m², with 150 technicians. The five-axis fleet is 16 simultaneous 5-axis machining centers, plus 12 four-axis mills, 27 three-axis machines and 16 mill-turn centers. Maximum processing size is 4,000 mm, with travels from 500 × 500 × 450 mm up to 4,000 × 400 × 150 mm and a Ø400 mm rotary table.
The quality system is audited to ISO 9001:2015, IATF 16949:2016, ISO 13485:2016 and ISO 27001:2022. That last one matters for European customers who send drawings and CAD data: it covers how files are stored and who can open them. NDAs are available on request and uploads stay confidential.
Materials cover aluminium 6061, 7075 and 6082, stainless 303, 304, 316L and 17-4PH, steels including 4140 and 4340, titanium TC4 (Ti-6Al-4V), Inconel and engineering plastics such as POM, PEEK and PC. Finishes run in-house: anodizing, plating, powder coating, bead blasting and laser marking. Keeping post-processing under one roof avoids the accuracy loss that comes from shipping parts between vendors.
Lead time is usually the constraint Belgian buyers feel most. Quotation with a free DFM analysis comes back within 12 hours, production can start within 24 hours, and parts ship in 3–5 days. There is no minimum order quantity, so a single prototype and a 10,000-part run go through the same inspection route.
- 1±0.005 mmWorking production tolerance
- 2Ra 0.2–0.8 μmFine finish when the drawing calls for it
- 399.99%Qualification rate across inspected parts
Questions engineers ask before they send a drawing
Can you hold ±0.005 mm on every feature?
No, and no shop can. ±0.005 mm is what the process holds on rigid features cut in a single setup with a thermally stable machine.
A thin wall, a deep pocket with a long tool, or a feature that needs three setups will sit looser. We tell you which features we can hold tight and which we cannot, before you place the order.
Do I need a CMM report for a prototype?
For a fit check, usually not. For a prototype that will become a production part, yes. The first-article numbers tell you whether the design is manufacturable at the tolerance you drew.
Catching a 0.01 mm stack-up problem on part one is far cheaper than catching it after tooling is committed.
How does shipping to Belgium affect accuracy?
Transport does not change a machined dimension, but it can change a part that was under residual stress. Parts that move after machining move in the crate.
For 7075 and 17-4PH we rough, stress-relieve and finish, and we inspect after the final operation rather than before.
What file formats do you need for a quote?
STEP and IGES for the model, PDF for the drawing with tolerances and datums marked, and a material and finish callout.
If the drawing has a tolerance table, include it. A model alone does not carry the tolerance intent, and guessing it is how quotes come back wrong.
Is the 12-hour quote a real number?
It is the target for a quotation plus free DFM feedback once we have the drawing and the quantity. Complex parts with many setups take longer, and we say so rather than sit on the file.
The DFM notes list the features we would change to make the part cheaper or more repeatable.
Can you keep the design confidential?
Uploads are handled as confidential and an NDA is available on request. Our information security management is certified to ISO 27001:2022.
If you need a signed NDA before releasing files, send it first and we will return it before quoting.
Send the drawing, get the tolerance answer back
Upload a STEP file and drawing and we return a quotation with free DFM notes within 12 hours. Tell us which features are critical and we will say plainly which ones we can hold.
12-hour quote100% inspectionNDA on request
