Chapter iii · Datums & Reference Frames

Set a part on a table and it can still slide, slide, and spin. A datum is how a drawing nails a thing down — without ever touching it.

Here is the consequence, before the theory. Take one bracket and dimension it from datums A then B then C. Take the identical bracket and dimension it B then A then C. Inspect both against their drawings and you will get different measurements — sometimes a good part on one becomes a reject on the other. Same part. Same surfaces. Different order. This isn't academic: a part seated B-then-A when the drawing meant A-then-B is the same metal built to the wrong recipe — good parts that fail in Ohio and pass in Osaka.

That is the whole secret of this chapter: a datum reference frame is not three faces you happen to touch. It is an ordered recipe for seating a part, and the order is the meaning.

Start from physics. A rigid body floating in space has six degrees of freedom — three translations (slide along X, Y, Z) and three rotations (spin about X, Y, Z). To inspect a part you must first kill all six, so it can sit exactly one way. GD&T does this with the 3-2-1 rule: a primary planar datum seats on three high points and removes 3 degrees of freedom; a secondary touches two and removes 2; a tertiary touches one and removes the last 1. Three, two, one — fully constrained.

The datum reference frame is three mutually perpendicular planes meeting at a corner: floor A, walls B and C. The part seats against A first, then B, then C — each plane removing the freedom the one before it left behind.

Click the planes below in order and watch the six freedoms fall away. Try clicking out of order — the gauge will refuse, because a secondary datum cannot constrain anything until the primary has seated.

Click the three planes in order — A, then B, then C. Each one nails down more of the part's freedom. Click out of order and the gauge refuses: precedence is the meaning.

One distinction does almost all the work in practice — three things that beginners blur into one:

The datum feature is the real, imperfect surface on the part — the casting skin, the milled face, with all its bumps and waviness.
The datum is the perfect theoretical plane (or axis, or point) that the drawing refers to. It has no error because it does not physically exist.
The datum feature simulator is the gauge — the surface plate, the chuck, the precision pin — that contacts the imperfect feature and establishes the perfect datum from it.

Think of a foot, a footprint, and a shoe last: the datum feature is your bumpy foot, the datum is the ideal shape the shoemaker means, and the simulator is the last that presses against the foot to capture it. So when a drawing says "perpendicular to A," it never means "perpendicular to that exact bumpy face." It means perpendicular to the ideal plane the high points of that face define when the part is set on a surface plate. The datum is an abstraction the drawing nails down — without ever touching the part.

Datum targets — when the whole surface can't be trusted

A raw casting or forging has a surface too rough or wavy to seat reliably on a flat gauge: set it down twice and it rocks to a different rest each time. For these, the drawing specifies datum targets — designated points, lines, or small areas (drawn as targeted circles with a dimension-line leader) where the gauge is allowed to contact. Three target points define the primary plane, two define the secondary, one the tertiary — the 3-2-1 rule made explicit and repeatable, so the part seats the same way every time, in design, in machining, and in inspection.

For the advanced reader → A cylinder is not a face — the primary axis removes four

Three datums do not have to be three flat faces. A cylindrical datum feature — a shaft, a bore — establishes a datum axis, not a plane. When that cylinder is the primary datum, the simulator (a precision chuck or a circumscribing/inscribing cylinder) seats on it and removes 4 degrees of freedom: 2 translations and 2 rotations, leaving only the spin about the axis and the slide along it.

That is why "three datums = three flat faces" is a misconception. A typical shaft frame is a primary cylindrical datum (axis, 4 DOF) plus a secondary planar face (an end, 1 translation) — five gone, one rotation about the axis still free until a tertiary clocks it. A cylinder used as a secondary or tertiary datum removes fewer, because the planes that came before it have already used up the translations and rotations it would otherwise control.

Try it

Seat the Part

You set a part on its primary planar datum first. How many of the six degrees of freedom does that single plane remove?

A drawing says a feature is ‘perpendicular to datum A’, where A is a rough milled face. Perpendicular to what, exactly?

A shaft is used as the PRIMARY datum feature. How many degrees of freedom does its datum axis remove?

A datum is the most quietly radical idea in the language. It says: the surface you can touch is not the thing you mean. The thing you mean is a perfect plane, an exact axis — a fiction the drawing agrees to, summoned from an imperfect part by the order in which you set it down.

Get the order right and a stranger half a world away seats the part exactly as you imagined it. That is the whole promise of a reference frame.

Now we can say square to A, located from A·B·C. But first: is the surface even true to itself? That's where form begins — chapter iv.