Positioning Reference Selection Methods for Machining Parts in a Machining Center
How you choose the datum decides whether a batch holds ±0.005 mm or drifts at the third setup. This guide covers the positioning reference selection methods we use on 5-axis, 4-axis and 3-axis work, with setup steps, error budgets and the cases where a method stops working.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
What positioning reference selection methods actually control
A positioning reference is the surface, hole or edge that locates the part in the machine coordinate system. It is not the same as the programming zero point, although the two are linked by a fixed geometric offset you define during setup. When the reference is wrong, no amount of machine accuracy saves the part: a 0.01 mm datum shift becomes a 0.01 mm shift on every feature cut from that setup.
The choice of reference controls three things at once. It sets the dimensional accuracy of each feature, the position accuracy between features, and how evenly the remaining stock is distributed before the next operation. Engineers often fix one of these and lose the other two. A datum that gives a clean Z zero may leave 0.3 mm of uneven stock on a side that must be finished later.
For machining center work, most parts are located from a combination of a primary plane, a secondary edge or hole, and a tertiary stop. That combination is what the 3-2-1 rule describes. The rule matters because it maps directly onto the six degrees of freedom: three translations, three rotations. Constrain fewer than six and the part can still move when the cutter pushes it.
- 1Datum vs program zeroThey can differ, but the offset between them must be known and measurable.
- 2Position accuracyHole-to-hole and face-to-face relationships depend more on the datum than on spindle accuracy.
- 3Stock distributionA good datum leaves uniform allowance on surfaces that will be finished in a later setup.
Design datum, process datum and the cost of converting between them
The design datum is the reference the drawing dimensions are measured from. The process datum is whatever the fixture can actually grip. These two are the same on maybe half the parts we see, and that is the cheap case. When they differ, the shop has to convert dimensions through a chain, and each link in that chain adds error.
Take a bracket with a Ø12 mm bore that must sit 85 mm from a machined face. If the fixture locates on a cast boss instead of that face, the machinist now depends on the cast boss position, which may vary ±0.4 mm as-cast. The finished bore can still be round and on size, yet the 85 mm dimension fails. This is the most common datum mistake in first-article inspection.
The practical rule: pick the process datum as close to the design datum as the fixture allows, then machine the surfaces that will serve as datums for later operations as early as possible. On a 3-axis machine with a 500 × 500 × 450 mm envelope, it is often worth a roughing pass just to create a clean datum face before any finished feature is cut.
- 1Cheap caseDesign datum equals process datum. No dimension chain, no added error.
- 2Expensive caseCasting or forging surface used as datum. Add the raw stock tolerance to the stack.
- 3FixMachine the datum face in the first operation, then locate from it in the second.
How much error each reference choice adds to the stack
Errors from datum selection fall into four groups: fixture location error, part deformation under clamping, thermal drift during the cycle, and machine positioning error. Machine positioning is usually the smallest of the four. On a well-maintained machining center it can sit inside ±0.005 mm, while an over-tightened vise can push a thin wall 0.03 mm out of position before the tool even touches it.
Clamping force is the one engineers forget. A 6061 aluminium plate 8 mm thick, held in a vise with 2,000 N of clamping force, will bow measurably. The part springs back after unclamping and the measured flatness is worse than the in-process reading. Lighter clamping with more contact points, or a vacuum fixture with 0.6–0.8 bar holding pressure, usually beats brute force.
Thermal drift matters on long cycles. A 40-minute roughing pass can move the part 0.01–0.02 mm as the bed warms. If the drawing calls for ±0.005 mm between features cut at the start and the end of that cycle, the datum plan has to include a warm-up or an in-process re-probe. We probe datums again after roughing on any part with tight position tolerances.
- 1Fixture error0.005–0.02 mm typical, driven by stop wear and chip nesting.
- 2Clamping distortionUp to 0.03 mm on thin walls if force is not controlled.
- 3Thermal drift0.01–0.02 mm over a 40-minute roughing cycle.
Cases where a single reference will not hold the tolerance
Some parts cannot be located from one datum and still meet the drawing. Long shafts, thin rings and parts with a tolerance across two opposite faces are the usual offenders. A 4,000 mm long extrusion, for example, cannot be held to a face-to-face tolerance over its full length without intermediate support. The reference has to be re-established along the part, not just at one end.
Another failing case is a part with two datums that are dimensioned from each other but machined in different setups. If OP10 locates from face A and OP20 locates from face B, the A-to-B relationship depends entirely on how well A was machined in OP10. Any error there is baked into the finished part and cannot be inspected away.
When this happens, the answer is usually to add a common reference that carries across setups. Options include a pair of reamed tooling holes, a turned spigot, or a ground pad. On 5-axis work with a Ø400 mm rotary table, a common bore located on the table center lets all five sides share one reference and removes the inter-setup stack entirely.
- 1Long partsNeed re-established datums along the length, not a single end reference.
- 2Two-datum partsRequire a common reference that survives both setups.
- 35-axis fixUse a table-center bore so all sides share one datum.
Step-by-step datum selection for a new part
- 1Read the drawing for datums firstMark every GD&T datum symbol before looking at the fixture. The drawing already tells you which surfaces carry the tolerance. List them in order of importance.
- 2Check which datums the fixture can reachCompare the datum list against available contact surfaces. If a datum is on a curved or as-cast surface, plan to machine it first rather than locate from it.
- 3Apply the 3-2-1 rule and count constraintsPrimary plane takes three points, secondary edge takes two, tertiary stop takes one. Count to six. If you have fewer, add a stop or use a self-centering vise.
- 4Choose the locating feature on the fixturePrefer fixed stops over adjustable ones; a worn adjustable stop drifts 0.01–0.02 mm over a few hundred cycles. Hardened dowel pins and ground pads hold better.
- 5Set the offset between datum and program zeroProbe the datum and record the offset. Keep it under 0.005 mm by averaging three touch points on the primary face rather than a single touch.
- 6Test with a first-article cutMachine one part, measure the datum-to-feature dimensions, and compare against the error budget. If a dimension fails, check clamping distortion before blaming the machine.
- 7Lock the setup and document itRecord stop positions, clamping torque and probe offsets in the setup sheet. Repeat setups drift most when the sheet says 'same as last time'.
Which reference method fits which part
Pick the row that matches the part geometry, not the machine.
| Part type | Best reference | Typical accuracy | Watch out for |
|---|---|---|---|
| Prismatic block, 3-axis | Machined face + two edges | ±0.01 mm | Chip nesting under the stop |
| Plate with holes, 3-axis | Two reamed tooling holes | ±0.005 mm | Hole wear after 500+ cycles |
| Shaft or long part | Center bore + tailstock support | ±0.02 mm over length | Sag in the middle |
| Thin-wall housing | Vacuum fixture, full contact | ±0.01 mm | Clamping bow and springback |
| 5-axis multi-side part | Rotary table center bore | ±0.005 mm | Table runout before setup |
| Cast or forged blank | Machined datum pad first | ±0.02 mm | Raw stock variation ±0.4 mm |
| Welded assembly | Machined face after stress relief | ±0.03 mm | Weld distortion after cutting |
Pick the datum the drawing already gave you
Choose the design datum as the positioning reference whenever the fixture can reach it, machine the datum faces early, and constrain all six degrees of freedom. When the part cannot be located from one reference, add a common feature instead of stacking tolerances.
Positioning reference questions we get from engineers
Can the positioning reference and the program zero be different?
Yes, and they often are. The reference is the physical surface that locates the part; the program zero is the point the CAM toolpath is written from. What matters is that the offset between them is known, measured, and stable across the batch.
If the offset drifts because a stop wears, the part moves even though the program never changed. Probing the datum at the start of each cycle keeps the offset honest.
How many locating points do I actually need on a machining center?
Six, in a 3-2-1 pattern, for a fully constrained rigid part. Three points on the primary plane, two on the secondary, one on the tertiary.
Fewer points work only when the part is heavy enough that its own weight holds it, or when a self-centering vise constrains the remaining freedoms. On thin or light parts, missing constraints show up as chatter and size drift.
Why does my part measure good on the machine but fail after unclamping?
That is clamping distortion, not a datum error. The part is being pushed into shape by the vise and springs back when released. On aluminium walls under 5 mm thick, 2,000 N of clamping force can move material 0.03 mm.
Reduce clamping force, spread it over more contact area, or switch to a vacuum fixture. Measure the part after unclamping, never in the vise, when the tolerance is tighter than ±0.02 mm.
Should I machine a datum surface before cutting finished features?
On castings, forgings and welded parts, yes. The raw surface can vary ±0.4 mm, which is far outside any finished tolerance. Skim the datum face in the first operation, then locate from it in the second.
On a part cut from pre-ground stock, this extra pass is often unnecessary. Check the incoming stock tolerance before adding cycle time.
How do I hold one reference across five sides of a part?
Use a common feature that stays with the part through every setup, such as a pair of reamed holes or a turned spigot. Locate on that feature for each rotation instead of re-datuming from a freshly cut face.
On a 5-axis machine with a Ø400 mm rotary table, a bore on the table center line lets all sides share one reference and removes the inter-setup tolerance stack.
What tolerance can a well-planned datum setup realistically hold?
On rigid prismatic parts located from machined surfaces, ±0.005 mm is achievable and is what we quote for critical features. Position between features cut in the same setup is usually tighter than between setups.
Cross-setup dimensions depend on the datum transfer, so budget ±0.01–0.02 mm there unless a common reference ties the setups together.
Send us the drawing and the datum callouts
We review your datum plan with the fixture in mind and return a quote with free DFM analysis within 12 hours.
12-hour quote100% inspection±0.005 mmNo MOQ
More machining setup notes
We publish setup notes, tooling trials and inspection data from the factory floor.
