Pitfalls of using the granite (NOT THE PART) as a Datum Alignment Feature

I can't point you towards a specific article. I can tell you exactly why its bad, though.

If you're in an industry that procedurally requires the use of calibrated measurement equipment (AS9100, ISO 13485, ISO 9001, etc...) then you can't use the CMM's granite table as a datum simulator because it isn't calibrated as a part of the machine (or at all). The table is just that. A table. By design, it is a reference only surface that you are ONLY to put stuff on for the machine to then measure. It is pretty flat (my table is 0.0025 across the entire surface and I have a 7107) but without that precision lapping and certification from a 17025 facility, its just a rock.

What I advise in this situation:
Depends on your industry. I have situations where my tolerance is ± 0.015 or more, it isn't a finished part dimension (more would be cut off later)....THEN I will beep off of the table. For FINISH DIMENSIONS on QMS regulated stuff, if the bottom surface of your part is a datum and you'd like to use the table as a datum simulator, don't do that. Instead, use calibrated inspection risers or gage-blocks or something like that. Have the machine probe those surfaces in DCC mode, create a plane, load your part on to them once you're done making the plane, then measure your part. Handle the measurement of the blocks with readpoints & comments that have pictures and the loading/unloading of the part with comments that have pictures. You could also have your company make an inspection fixture, measure the fixture itself & create an origin, THEN load your part onto it and go to town.

I can't point you towards a specific article. I can tell you exactly why its bad, though.

If you're in an industry that procedurally requires the use of calibrated measurement equipment (AS9100, ISO 13485, ISO 9001, etc...) then you can't use the CMM's granite table as a datum simulator because it isn't calibrated as a part of the machine (or at all). The table is just that. A table. By design, it is a reference only surface that you are ONLY to put stuff on for the machine to then measure. It is pretty flat (my table is 0.0025 across the entire surface and I have a 7107) but without that precision lapping and certification from a 17025 facility, its just a rock.

What I advise in this situation:
You could also have your company make an inspection fixture, measure the fixture itself & create an origin, THEN load your part onto it and go to town.

isn't using an un-calibrated fixture the same as using the rock?

you would use a calibrated fixture

You could also lift the part off of the granite high enough to probe the underside of the part.
Or, if it's not a very big part, a machined block, like a 1-2-3/4-5-6 block.
Or, if you can rotate the part onto the side, or flip it.
Or, if you can't do any of that, probe the granite enough to verify the flatness of the area you need to use.

I have worked on some small 5-5-5 and 5-7-5 Brown & Sharp machines that had granite tables that were not part of the mechanicals of the machine. Those were actually calibrated and lapped as needed on an annual basis by our local surface plate calibration service. The annual CMM calibration was scheduled shortly after, just to make sure nothing 'moved' on the machine. It was nice because I could reliably use it as a datum simulator.

I have also heard of some large gantry CMMs with plates that are moved in and out of the machine on rails like train cars. Those too could be calibrated. Really, it is just a bigger version of Wes' suggestion.