Method for Selecting Positioning Data for Parts Processed by Machining Center
This guide is for process engineers and CAM programmers who must pick the datum, the workholding and the zero offset before a single tool enters the cut. Read it and you will be able to justify every positioning decision on the setup sheet, and know when a chosen datum will cost you tolerance.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
What positioning data actually controls
Positioning data is the set of numbers that tell the machine where the part is. That means the datum origin, the work coordinate system (G54 to G59), the tool length offsets, and the values your probe writes back after touching the blank. Get these wrong and no amount of toolpath polish will save the part.
For parts processed by machining center, the datum choice decides how tolerance stacks. If the drawing dimensions run from a bore to a face, and you clamp on the opposite face, you inherit the thickness variation of the blank. On a 60 mm aluminum plate with a ±0.5 mm saw cut, that variation alone can eat the entire ±0.05 mm positional callout.
The machine does not know your intent. It only knows the offset table. So the method has to be written down: which surface is primary, which is secondary, which is tertiary, and how each one is touched off. Write it on the setup sheet and the second shift will repeat it.
One more point. Positioning data is not a one-time decision. On a 16-part run it is fixed, but on a 500-part run with tool wear and thermal drift, you re-verify the offset at set intervals. That interval belongs in the method too.
Matching the datum to the drawing callout
Start from the feature control frame. If position tolerances are called from datum A, B, C, then A is your primary locating surface, full stop. Do not substitute a more convenient surface just because the vise grabs it better.
The 3-2-1 rule still works for prismatic parts. Three points on the primary plane kill one translation and two rotations. Two points on the secondary kill one translation and one rotation. One point on the tertiary kills the last translation. Six degrees of freedom, six contact points, done.
On parts processed by machining center in a five-axis setup, the rule extends. You still need six constraints, but they can come from a fixture plate plus a rotary table with a Ø400 mm face. The rotary centerline becomes datum, and the part must be dialed to it within 0.01 mm before you trust any offset.
Where engineers get this wrong: they choose a datum that is machined in the same setup. That datum does not exist yet when you probe. Use a raw surface for the first setup and move to the machined surface for the second operation. This is the whole logic behind op-10 and op-20 datum transfer.
- 1Primary datumUsually the largest flat face. Three contact points, no rocking.
- 2Secondary datumA long edge or a bore. Two points, controls rotation in the plane.
- 3Tertiary datumA short edge or pin. One point, locks the last translation.
- 4Do not mixNever locate from a surface with a draft angle or a cast skin.
Fixture choice and how it changes the numbers
A vise gives you two contact surfaces and a lot of clamping force. That force deforms thin walls. On a 2 mm wall in 6061, a standard vise at 30 bar will spring the part and the measured position after unclamping will be off by more than the tolerance.
Soft jaws machined in place solve most of this. Bore the jaw pocket at the actual part size, not nominal, and keep the clamping pressure under 15 bar for thin sections. The jaw then becomes a locating feature with repeatability around 0.02 mm across loads.
For parts processed by machining center that need five faces in one setup, a self-centering fixture on the rotary table is the practical answer. You lose some access at the bottom, but you gain positional consistency because the part never leaves the nest.
Magnetic chucks and vacuum plates are fast but they do not resist side load well. Use them for light finishing passes only. If you are taking a 6 mm radial cut in 4140, the part will move and the positioning data becomes fiction.
Setting and verifying the work coordinate system
Touch off the primary datum, then the secondary, then the tertiary. Record the machine coordinates in the offset page. Do not eyeball the edge finder; use a 3D probe or a coaxial indicator for anything tighter than 0.05 mm.
For a bore-based datum, indicate the bore in two axes and set X0 Y0 to the bore center. Check the runout at the top and bottom of the bore. If the two readings differ by more than 0.01 mm, the bore is tapered or the part is tilted, and your Z zero will drift as the tool wears.
Write down the probe results on the setup sheet with a timestamp. When the second shift loads the same job, they compare their numbers to yours. A shift of 0.03 mm on the same fixture means something moved, and it is cheaper to find that now than after 200 parts.
Set tool length offsets with the same reference surface used for the work offset. Mixing a tool setter with a spindle probe is a common source of Z error on parts processed by machining center.
Step by step: choosing positioning data
- 11. Read the feature control frameList every datum letter on the drawing. Rank them A, B, C. If the frame calls A as a flat face, that face is your primary. Write the ranking on the setup sheet before you touch the machine.
- 22. Check the blank for usable surfacesMeasure the raw stock. A saw-cut face with 0.5 mm taper is not a datum. If no raw surface is clean enough, plan a first operation that machines the datum, then re-clamp for op-20.
- 33. Pick the fixture from the datum, not the other way roundThree-point contact on the primary. Two on the secondary. One on the tertiary. For thin walls under 3 mm, drop clamping pressure below 15 bar or switch to soft jaws machined to size.
- 44. Set the work coordinate systemProbe or indicate each datum in order. Store in G54 for op-10 and G55 for op-20 so the two setups never share an origin. Record machine coordinates to 0.001 mm.
- 55. Verify against a known featureCut a test feature, or probe a pre-machined surface, and compare the measured value to the offset. Anything over 0.02 mm disagreement means the datum or the offset is wrong. Fix it before the run.
- 66. Set re-verification intervalsOn runs over 200 parts, re-probe the primary datum every 50 parts. On long-cycle parts over 30 minutes, re-probe every 5 parts to catch thermal drift in the spindle and the fixture.
- 77. Freeze the methodOnce the first article passes, lock the offset values and the datum sequence into the program header. Any change needs a new first-article check, not a verbal note at the machine.
Which positioning approach fits your part
Use part geometry and tolerance to pick the row. The right column tells you what to watch.
| Part condition | Recommended datum | Fixture | Main risk |
|---|---|---|---|
| Prismatic, tolerance ±0.05 mm | Machined flat face, two edges | Precision vise with soft jaws | Clamp distortion on thin walls |
| Thin wall under 3 mm | Machined face, bore center | Soft jaws, pressure under 15 bar | Spring-back after unclamping |
| Five faces in one setup | Rotary centerline plus face | Self-centering fixture on Ø400 mm table | Dial-in error on the rotary axis |
| Cast or forged blank | Machined datum from op-10 | Two-operation setup | Datum transfer stack-up |
| Round part, tight runout | Bore center, indicated top and bottom | Collet or expanding mandrel | Bore taper throwing off Z zero |
| Large plate, 4,000 mm class | Machined edge and face | Vacuum plate plus side stops | Part shift under side load |
When the datum choice is still wrong
If the first article passes but the 50th part drifts, the problem is fixture repeatability, not the toolpath. Re-check the contact points and the clamping pressure before you change any offset value.
Positioning data questions engineers ask
Can I use the same work offset for op-10 and op-20?
Only if the part returns to exactly the same nest with the same contact points. In practice that almost never happens, because op-20 usually removes the surface you clamped on in op-10.
Use separate offsets, G54 and G55, and record both machine coordinate sets. If a part is re-loaded into G54 for op-20, any fixture wear shows up directly in the positional tolerance.
How often should I re-verify the zero offset during a run?
Every 50 parts on a run over 200 pieces, and every 5 parts on cycles longer than 30 minutes. Thermal growth in the spindle and the fixture moves the effective zero by 0.01 to 0.03 mm over a few hours.
If the machine runs lights-out, put a probe cycle in the program at fixed intervals and let the control write the corrected value back to the offset table.
What tolerance should the fixture itself hold?
The fixture should be at least three times tighter than the part tolerance. For a ±0.05 mm positional callout, aim for fixture repeatability within 0.015 mm across loads.
Measure repeatability by loading and unloading the same part ten times and recording the probe result. The spread, not the average, is what matters.
Does a probe replace a dial indicator for setting the datum?
For most work, yes. A spindle probe reads to 0.001 mm and writes the offset automatically, which removes transcription errors.
For a bore datum with a tight runout callout, indicate manually top and bottom. A probe touch at one Z height cannot tell you whether the bore is tapered.
How does five-axis work change datum selection?
The rotary centerline becomes a datum in itself. You must dial the part to that centerline within 0.01 mm, or every rotated position carries an error that grows with distance from the axis.
Set the rotary zero once, verify it with a test cut, and keep the part in the same nest for all five faces. Re-clamping between faces defeats the purpose of a five-axis setup.
What is the most common mistake in positioning data?
Locating from a surface that is machined in the same setup. The datum does not exist at probe time, so the offset is set from a raw surface and then the part grows or springs when that surface is cut.
The fix is simple: machine the datum in op-10, then use it in op-20. It costs one extra setup and removes a whole class of positional errors.
Send us your drawing and datum callouts
We review the feature control frame and return a quotation with free DFM analysis within 12 hours. Uploads stay confidential, and an NDA is available on request.
12-hour quote100% inspection±0.005 mm toleranceNo minimum order quantity
More CNC process notes
We publish setup notes, tooling trials and inspection data from the factory floor.
