Datum Scheme: How the Datum You Choose Changes the Price of the Part
A datum scheme is a locating plan written in symbols. It tells the shop which surfaces define zero, in what order, and how much the part is allowed to move relative to them. Two engineers can tolerance the same bracket to the same numbers and get quotes 40 percent apart, because one chose a datum set reachable in a single setup and the other chose one that needs three.
It is a cost decision as much as a quality decision, usually made last, by whoever fills in the title block.
3-2-1, and what it means at the machine
The 3-2-1 principle says a rigid body needs six constraints: three points on a primary plane, two on a secondary, one on a tertiary — datum A, B and C. Every vise and soft jaw is an attempt to implement it.
Over-constraint is the common error. Four pads on a plane fight each other if the surface is not flat, and a casting or welded frame bends to comply. The part is machined in a stressed state and springs when released. That is why a three-point primary datum on a rough casting separates a part that measures flat free-standing from one that rocks.
At the machine, the scheme becomes the setup sheet: A is the face on the fixture, B the edge against the stop, C the pin in the hole. If the fixture cannot present the datums in that order, the shop either builds a special fixture or silently picks different datums.
Functional versus manufacturing datums
A functional datum is the surface that locates the part in the assembly. A manufacturing datum is the surface that is easiest to hold in the machine. They are often not the same, and the difference is where tolerance stacks appear.
A gearbox cover illustrates it. Functionally, the spigot that centres the cover in the housing matters most, so a designer makes it datum A and hangs the bolt pattern and seal face off it. Correct for function — but the spigot is an interrupted cylinder a three-jaw cannot hold without distorting, so the shop would rather machine from the flat face and bolt circle and let the spigot be a result.
Neither is wrong. The functional scheme needs a fixture that holds by the face and indicates the spigot; the manufacturing scheme needs a note that the spigot is cut in the same setup as the face. The mistake is specifying functional datums and expecting manufacturing-datum pricing.
A workable rule: use functional datums, and say which features must be cut in one setup to satisfy them. That sentence is often worth more than a tighter tolerance.
See also: Deep Pocket Machining: Depth-to-Diameter Limits, Corners and Deflection and Drone Frame Machining: Billet Center Plates, Ribs and Motor Patterns.
Datum shift and the bonus you did not plan
When a datum feature is referenced at maximum material boundary, the reference frame can shift by the feature's departure from that boundary. A Ø40 H7 bore used as datum B at MMB gives a 0.025 mm shift at the top of the band, and that shift is legal.
So a part can be "in tolerance" and still fail to assemble, because every feature was evaluated against a frame that moved. On a sensitive joint, reference the datum at regardless of feature size (RFS), which removes the shift — and accept that the shop must hold the datum tighter or align on a CMM instead of a pin.
Production shops prefer MMB because it simplifies fixturing; assemblies that bind at the customer prefer RFS.
Why re-clamping breaks the scheme
Every time a part leaves a fixture and returns, it gains an error. Re-chuck error on a good three-jaw is 0.01 to 0.02 mm; with hand-tightened clamps it reaches 0.03 mm. Cut the datums in setup one and the critical bore in setup two, and that error lands directly in the bore's position.
There is a second effect. A part clamped in a distorted shape is machined round in that shape and springs back when released: round on the machine, out-of-round on the CMM, and tilted relative to its datum face.
The fixes are all sequence:
- Machine all datums in the first setup, and mark them on the drawing.
- Keep functional features in the same setup as their datum wherever the machine envelope allows.
- If a second setup is unavoidable, never re-clamp on a finished datum face.
- Finish the datums last only when they are non-functional and will be covered.
Worked example: two datum options on one bracket
A 120 × 80 × 30 mm 6061 bracket carries a Ø40 H7 bearing bore through the 30 mm dimension, a flat mounting face, and two Ø8 dowel holes in the mounting face, 90 mm apart. The bore must be perpendicular to the mounting face within 0.02 mm.
Option A — face-based datums. Datum A is the mounting face, B is the long side, C is the short end. The face and dowel holes are cut in setup one from bar stock. The part is then flipped into a second setup and the bore is bored perpendicular to A.
Option B — bore-based datums. Datum A is the bore axis, B is the mounting face. The bore and the face are cut in the same 5-axis setup, with the part held in soft jaws on the as-sawn stock, so perpendicularity comes from machine geometry rather than from a re-clamp.
| Item | Option A | Option B |
|---|---|---|
| Setups | 2 | 1 |
| Fixture | Vise plus angle plate, stops | Soft jaws, stock-held |
| Bore to face perpendicularity | 0.03 mm achievable, 0.05 mm worst case | 0.01 mm from machine geometry |
| Stack-up to dowel position | 0.04 to 0.05 mm | 0.015 mm |
| Cycle time | 22 minutes | 28 minutes |
| Relative unit cost, 200 pieces | 1.00 | 1.15 |
| Risk | Re-clamp error on every part | Fixture must be dialled in once |
Option A is cheaper and adequate if the bore locates a low-speed shaft. Option B costs 15 percent more and removes the stack-up, which matters if the bore carries a gear meshing with a shaft across the bracket, where angular error becomes noise.
The drawing's datum choice decides which of these you get. That 15 percent should be a decision, not a surprise.
Writing a scheme that survives the shop floor
Four habits keep a scheme workable. Put datums on machined surfaces, not as-cast, unless the casting is stable and marked. Keep the count at three unless a fourth does real work. State the setup intent — "bore and face in one operation" — beside the tolerance it protects. And check that every datum is reachable by a clamp.
If a tolerance is tighter than the process can hold, the datum scheme is usually the reason, not the machine. Send the drawing and the assembly it goes into, and we will tell you which features drive the price before you commit to a scheme — send the file, or read how we sequence how we work.
Frequently asked questions
How many datums should a part have?
Three for a rigid part: a primary plane, a secondary and a tertiary. More datums are only justified when a specific function needs them, and each one adds fixturing cost and inspection time.
Can I use a bore as a primary datum?
Yes, but it forces the part to be held on the bore, which usually means a expanding mandrel or a 5-axis setup. It is the right choice when the bore is the functional centre, and an expensive one when it is not.
What is datum shift and why does it matter?
When a datum feature is referenced at MMB, the frame can move by the feature's departure from that boundary — up to 0.025 mm on a Ø40 H7 bore. That shift is legal but can consume the clearance in a mating assembly, so use RFS on sensitive joints.
Does a tighter tolerance fix a bad datum scheme?
Rarely. Tightening a bore position from 0.05 to 0.02 mm does nothing if the datums are cut in a different setup, because the re-clamp error dominates. Fixing the sequence usually costs less than tightening the number.
Related reading
Send a drawing for a quote and DFM review.
STEP, IGES, X_T, PDF or DWG up to 50 MB. Tell us quantity and finish, and the quote comes back with tolerance and process notes, not just a price.