Reducing Datums

[[Ja...]]

Hey guys, is it best practice to reduce datums down to individual points? We know datums are perfect, and exist only in theory, while datum FEATURES are real surfaces, holes, etc. For example, scanning a plane and extrapolating the 3 highest points - scanning a circle and grabbing the 3 highest, so on and so forth. This will reduce the sigma and form error to zero, casting any inherent error in the datum feature onto the measured feature being reported to it, right? This "reducing datums" idea is something that has been proposed to me by engineering, but I'm not so sure. Any help is appreciated!

[[Ri...]]

On a casting, this is typically what is done. Most commonly known as a 3-2-1 alignment.

On a machined part, I'm not certain how you would achieve this unless you know there is absolutely no form error.

Some people do this with Datum simulation with things like a fixture, but that's something completely different, and poses its own challenges.

[[Ja...]]

Please sign in to view this quote.

Thank you, Richard. We measure very few rough castings; the majority of our datums are machined, though there is the occasional plasma-cut hole. Repeatability of our programs is already very good, so I'm not sure what they're wanting to achieve.