What makes datum features good or bad?

Bad datum features lead to a bad datum reference frame.

We’ve talked about basic dimensions, limit tolerancing, and datum features versus datums. I have since fielded some questions regarding designers who choose ill-suited placements for their suction cups (datum feature identifiers) on their drawings. So, what makes a good datum feature?

Datum features exist on a part. They can be touched and should always be toleranced, or qualified.

Datums don’t exist until a datum feature (on a component) mates to a datum simulator (on production or inspection equipment). And even then, datums are only as substantial as the points, lines, and planes of a high school geometry class.

Bad datums will almost always result in bad measurements, even of good parts. This situation can add significant costs or delays to production. The problem may be hidden by large tolerances but becomes obvious as tolerances shrink. Ever measure the same part twice and get different answers—and then have the customer measure the part and get yet another different answer? Bad datum features are the best way to establish bad datums. And bad datums cost good money.

There are two major types of bad datum features:

1. Datum features that aren’t flat, aren’t straight, or aren’t oriented or located properly with respect to precedent datums. The resulting part might rock on a surface plate or rattle in its fixture, or its measurement systems might fail repeatability studies. (You do perform repeatability studies, don’t you? ISO, AAIG, and many customers require them.)
2. Datum features that establish (or try to establish) datums and datum reference frames (part coordinate systems) that are irrelevant to component function. When you try to orient and locate the part coordinate system off of a tiny surface or from edges that have no particular function, measurements in that coordinate system can become a roll of the dice.

It is important to remember that the first function of almost all parts is to be located in an assembly. Whether robots hold a stamping against an auto body-in-white for welding, a worker slides a pulley onto a shaft (and keyway), a tech pins a board in place for circuit testing, or even if parts are lined up “by eye,” part features do the locating. There is rarely a good reason for these part features not to be designated to establish component datums.

Designers: These are the features that should wear the suction cups. You should know this as you are the first authority on your part’s function.

Drafters and CAD operators: Don’t let designers pass the responsibility of designating datum features to you (unless you know better than they do how the component is located).

And dimensional management engineers? We know how the parts fit, know how the various processes make an assembly from the parts, and understand the struggles of the metrologist to represent parts truthfully in measurement reports. We speak fluent GD&T and know the six degrees of freedom and the right-hand rule to settle the X, Y, and Z directions and u, v, and w axes of rotation.

We know that some CMM routines use Y+ = outboard on both sides and some use Y+ = to the left—and that it’s critical to spot the difference. We are loyal to the parts, processes, and product from conception to final delivery to the customer.

We may use analysis software to tell if part tolerances are too big (a percentage of parts won’t function or assemble) or too small (increasing part cost). But it’s also a bit of voodoo because the parts’ tolerances on paper don’t always represent the parts’ variation in reality. But under the right assumptions, useful analysis can be done before parts are even made.