Fixtures are checked as often as parts
Why a fixture drifts, and what happens when nobody notices.
A fixture suffers the same wear, vibration and accidental knocks as anything else in a workshop, and because every part it locates is trusted simply for having come from it, a fixture that has quietly drifted out of true produces an entire run of parts that agree perfectly with each other and are all wrong in the identical way.
How a trusted fixture drifts
The earlier articles in this set built up a fixture as the true reference a delicate part could not provide for itself, and that argument holds only while the fixture stays true. A locating pin pressed against thousands of parts wears down microscopically with each cycle. A clamp tightened and released over and over can work its mounting screws looser than their original torque. A dropped tool or an over-forced part can dent or shift a reference surface without anyone noticing at the time. Each change is gradual and none looks like a failure when it happens, so a fixture can creep away from the condition it was verified in without any single event announcing it. Across many thousands of cycles, wear too small to feel on any one use can add up to a shift as large as the whole tolerance band the part is meant to hold.
A rip fence that goes out of square
A table saw's rip fence is set square to the blade once, carefully, and every cut afterwards is trusted to be parallel because the fence was correct when checked. Vibration, knocks from stock being loaded, and wear in the fence's locking mechanism nudge it slightly out of true over weeks of cutting. A woodworker who relies on the lock and never rechecks with a square can cut a whole run of boards with a barely perceptible taper and find out only when the boards refuse to sit flush at assembly. No single cut looked wrong compared with the one before it.
A fixture that breaks outright, with a cracked locating surface or a clamp that will not close, gets stopped the first time anyone uses it. A drifted fixture keeps working smoothly, and the parts coming off it still measure consistently against each other, which is easily mistaken for evidence that all is well. Consistency cannot show whether the reference itself has moved. The parts end up precise relative to each other and inaccurate relative to the drawing, the same trap a biased measuring instrument falls into, with the bias built into the tool making the parts instead of the tool measuring them.
Checking fixtures on a schedule
The fix is to inspect the fixture against an independent reference on a fixed schedule, not only when something already seems wrong, which catches the drift while it is small enough to correct instead of after a whole batch has to be scrapped together. How often depends on duty. A fixture used rarely, or one holding a part with tolerances loose enough to absorb any plausible drift, needs little attention, while the ones doing the heaviest duty on the tightest tolerances, the kind this set has spent five articles building up, cost the most when skipped and give the least warning.
The same holds for any tool trusted to define truth for something else, whether a gauge, a template or a reference standard. Each earns that trust by being checked, and having once been correct is not enough.