CNC Milling Machine Precision Demonstration
A tolerance callout on a drawing is a promise. A precision demonstration is the evidence behind it. Here is what we measure, how the numbers hold on a real batch, and how to read a demo without being fooled by a single shiny part.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
What a demo really proves
What a CNC milling machine precision demonstration actually measures
A precision demonstration is a controlled test. You pick one or two features on a real part, machine them on the same setup that production will use, then measure them with a traceable instrument and publish the numbers. The point is not to show that a machine can cut metal. Any mill can do that. The point is to show that the same feature lands in the same place on part 1 and part 50.
Machine builders publish positioning accuracy and repeatability for a new machine in a temperature-controlled room. That figure is a starting condition, not a result. Once you bolt on a vise, load a 40 kg block of 4140, and run coolant for six hours, the number moves. A serious demonstration recreates those conditions rather than hiding them.
The output is usually a short report: the nominal dimension, the tolerance band, the measured spread, the gauge used, and the ambient temperature at the time of measurement. If a shop shows only a best-case part and a glossy photo, treat it as marketing, not evidence.
- 1Nominal and toleranceThe drawing value plus the allowed band, for example Ø25.000 ±0.005 mm.
- 2Measured spreadMax minus min across the sampled parts, not just the average.
- 3Gauge and traceabilityCMM, micrometer or bore gauge, with calibration status.
- 4ConditionsRoom temperature, spindle warm-up time, fixture type.
The five error sources behind any precision demonstration
Geometric error comes from the machine itself: squareness between axes, straightness of travel, spindle runout. A spindle with 3 μm of runout cannot hold a 5 μm bore consistently, no matter how careful the operator is. This is why we keep spindle taper and tool holder condition on a maintenance schedule rather than reacting after a bad batch.
Thermal error is the one that surprises engineers most. A spindle running at 12,000 rpm warms and grows. Ballscrews expand along their length. Over a two-hour run, the same program can drift 10 to 20 μm on a long part if the machine has no thermal compensation. Warm-up cycles and in-process probing are the practical countermeasures.
Tool deflection and wear are the third source. A Ø6 mm end mill at 3× diameter stick-out will push away from the wall under load, cutting a taper instead of a straight wall. Tool wear adds a slow drift across the batch. Keeping tool life logged and replacing on a count, not on a hunch, keeps the spread tight.
Fixturing and workholding decide whether the part is where you think it is. A three-point clamp on a thin wall bends it. The cut is accurate, then the part springs back when unclamped and the measurement is wrong. Finally, measurement uncertainty sets the floor: no gauge resolves better than its own calibration allows.
- 1GeometricAxis squareness, straightness, spindle runout.
- 2ThermalSpindle and ballscrew growth over the run.
- 3ToolDeflection from long stick-out, wear across the batch.
- 4Fixture and gaugeClamping distortion, and the limit of what you can measure.
Why 5-axis changes what a demonstration can show
On a 3-axis mill, the tool axis stays vertical. Every angled face, undercut or cross-hole needs a second setup, and every setup adds a datum shift. Two setups can easily add 20 to 30 μm of positional error before the cutter touches metal. For a bracket with three faces that is often fine.
A 5-axis center tilts and rotates the tool or the table, so those faces come off in one setup. Fewer setups means fewer datum shifts, and the accuracy of the first setup carries through the whole part. On a housing with bores on four sides, that is the difference between a stack-up you have to manage and a stack-up you avoid.
The trade is not free. Rotary axes add their own positioning error, and a tilting head is less rigid than a fixed one. Deep pockets in hardened steel still cut better on a stiff 3-axis machine. The demonstration should say which machine ran the parts and why that choice was made for that geometry.
At GreatLight, 16 simultaneous 5-axis machining centers handle the multi-face work, while 27 three-axis machines cover flat plates and simple pockets. Rotary tables go up to Ø400 mm, and the largest travel is 4,000 × 400 × 150 mm for long parts.
How to read a precision demonstration report
Start with the feature. A report on a Ø10 mm reamed bore in aluminum tells you very little about a 300 mm long bore in 17-4PH. Material, depth-to-diameter ratio and feature type all shift the achievable band. A demonstration that does not name the feature and material is not comparable to your job.
Then look for the sample size. One part proves capability on a good day. Ten parts across a shift proves repeatability. If the report gives an average but no max and min, ask for the spread. The average hides exactly the outliers that cause assembly problems.
Check the gauge. A caliper reads to 0.01 mm, so it cannot confirm a ±0.005 mm tolerance. Bore gauges, micrometers and CMMs can. We run raw material checks, in-process monitoring and final inspection, and inspection reports are available on request.
Finally, compare the tolerance to the function. A bearing seat needs a tight band because clearance decides noise and life. A cover plate mounting hole does not. Paying for ±0.005 mm on a non-critical hole is money spent on nothing. The report should help you separate the two.
- 1Feature and materialA bore in aluminum is not a bore in Inconel.
- 2Sample sizeTen parts across a shift, not one good part.
- 3Spread, not averageMax and min show the real band.
- 4Function firstTight only where the fit demands it.
When a tight tolerance demonstration is worth the cost
Tight tolerances cost money in three places: slower feeds, more inspection and higher scrap risk. A ±0.005 mm band usually means a finishing pass with a small stepover, a temperature-stable room and a CMM check on every part. On a 10,000-piece run that adds up fast. The question is whether the assembly actually needs it.
It usually does when the feature is a fit. Bearing bores, dowel pin holes, spigot diameters and sealing faces all convert tolerance directly into function. A 0.01 mm oversize bore changes the interference on a press fit and the joint may come loose in service. Here the tight band is not vanity, it is the design.
It usually does not when the feature is clearance. Bolt clearance holes, cable pass-throughs, cosmetic edges and mounting slots have wide functional windows. Holding them at ±0.05 mm is cheap and reliable. Pushing them to ±0.005 mm raises the price with no gain at the assembly bench.
Geometry also sets a limit. A 400 mm long bore with a 15:1 depth-to-diameter ratio will deflect, and no machine setting fixes that. A 0.5 mm thin wall will move when you clamp it. In those cases the honest answer is to change the design, add a support feature, or split the part, rather than promise a band the geometry cannot hold.
- 1Spend tight on fitsBearing seats, dowel holes, sealing faces.
- 2Stay loose on clearanceBolt holes, slots, cosmetic edges.
- 3Respect the geometryDeep bores and thin walls have real limits.
Step by step: how we run a precision demonstration
- 1Fix the feature and datumPick one or two features tied to a clear datum. Write down the nominal, the tolerance band and the material.
- 2Warm up the machineRun the spindle 20 to 30 minutes at the production speed so thermal growth is already in the system before the first cut.
- 3Cut on the production setupSame fixture, same tool holders, same program. No special one-off setup that production cannot repeat.
- 4Probe in processUse the touch probe to check the feature after roughing, then apply the offset before the finishing pass.
- 5Measure with the right gaugeBore gauge or CMM for tight bands, micrometer for outside diameters. Record the room temperature.
- 6Run a batch, not a partMachine 10 or more parts across the shift and report max, min and average for each feature.
- 7Publish the spreadSend the report with the nominal, band, measured values, gauge used and conditions. Note any feature that fell outside.
Which milling setup fits which tolerance demand
Bands reflect typical production results on the machines listed, not a guarantee on every geometry.
| Setup | Typical band | Best for | Watch out for |
|---|---|---|---|
| 3-axis, one setup | ±0.02 mm | Flat plates, open pockets, simple profiles | Angled faces need a second setup |
| 4-axis with rotary | ±0.01 mm | Parts with features on several sides | Rotary backlash on reversal |
| 5-axis simultaneous | ±0.005 mm | Contoured surfaces, multi-face housings | Lower rigidity than a fixed head |
| Mill-turn | ±0.01 mm | Shafts with milled flats and cross-holes | Long parts can whip if unsupported |
| 3-axis plus hand polish | ±0.01 mm | Cosmetic surfaces, Ra 0.2–0.8 μm | Hand work is hard to audit |
| Wire EDM for the last cut | ±0.005 mm | Hardened steel, sharp internal corners | Slow, and not a milling process |
| Grinding after milling | ±0.003 mm | Bearing bores, sealing faces | Adds a process step and a setup |
The honest trade
If the feature is a fit, chase the tight band and pay for the inspection. If it is clearance, hold ±0.05 mm and put the money into the parts that mate. A demonstration that shows both, with real spread numbers, is the one worth trusting.
Precision demonstration questions
Can you hold ±0.005 mm on every feature?
No, and no shop can. ±0.005 mm is the floor we reach on suitable features: reamed bores, ground faces, short turned diameters on rigid setups.
Long bores, thin walls and deep pockets have their own limits set by deflection and thermal drift. We will tell you which features can hold the band and which cannot before the job starts.
What instruments do you use to verify a tight tolerance?
Bore gauges and micrometers for diameters, a CMM for position and form, surface roughness testers for Ra values. Calipers read to 0.01 mm and are not used to sign off a ±0.005 mm feature.
Inspection reports are available on request, covering raw material, in-process checks and final inspection.
How does temperature affect the result?
Aluminum grows about 23 μm per meter per degree Celsius. A 300 mm part measured 5 °C warmer than the reference reads roughly 35 μm longer.
We let parts stabilize before final measurement and record the room temperature in the report so the numbers can be compared fairly.
Do you need a specific quantity to run a demonstration?
No minimum order quantity applies. We run from one prototype to 10,000+ part runs.
For a meaningful demonstration we prefer at least 10 parts so the spread is visible, but a single part can still be measured against the drawing.
Will a 5-axis machine always beat a 3-axis machine on accuracy?
Not always. A 5-axis center wins when fewer setups remove datum shift, which is common on housings with features on several faces.
On a flat plate with simple pockets, a rigid 3-axis machine can hold the same band with less setup time. The geometry decides, not the axis count.
What file formats do you need to review a tolerance stack?
STEP and IGES for the 3D model, plus a 2D drawing with the tolerance callouts and datums. The drawing is what we measure against.
Uploads are handled as confidential, and a non-disclosure agreement is available on request.
Send a drawing, get a real tolerance answer
Tell us which features matter and we will say which band they can hold on our machines, before you commit to a production order.
12-hour quoteFree DFM analysis100% inspection
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
