When two laboratories disagree about a batch of caps, the instinct is to suspect one of them. More often both are right: their gauges sit at opposite ends of the permitted tolerance band, and the parts sit near the limit that separates them.
A go gauge is permitted to be made anywhere between the nominal and the wear limit. Two suppliers, both conforming, can differ by the full band width, which for some cap dimensions is 10 to 15 µm.
| Situation | Gauge A | Gauge B | Part at 50.005 mm |
| Both new, mid-band | 50.000 mm | 50.000 mm | Pass |
| A high, B low | 50.008 mm | 49.995 mm | Pass on A, fail on B |
| A worn to limit, B new | 50.012 mm | 50.000 mm | Pass on A, fail on B |
| Different sheet edition | Old revision | Current revision | Verdict depends on edition |
When a batch is contested, the gauge is not the referee. The drawing is. A CMM measurement of the actual feature, compared with the cap drawing in IEC 60061-1, gives an answer that both labs can accept.
Standards referenced: IEC 60061-3 for gauge sheets, IEC 60061-1 for cap dimensions, ISO/IEC 17025 for the calibration requirements.
When two labs disagree, the useful documents are the calibration certificates, not the gauges. A certificate that lists the measured dimensions and the uncertainty lets you calculate whether the two sets can produce different verdicts on the same part.
If gauge A measures 50.008 mm and gauge B measures 49.995 mm, a part at 50.000 mm passes both, but a part at 49.998 mm passes A and fails B. The certificates predict this before any argument starts.
The cleanest arrangement is a written rule between supplier and customer: parts are judged against the drawing with a stated measurement uncertainty, and gauges are used for screening, not for arbitration.
Where gauges must be the arbiter, both parties should use the same set, or sets calibrated by the same laboratory with stated uncertainties.
A disputed batch of a million caps can cost more in sorting and re-testing than the gauges cost to buy. The sorting is done with the same gauges that caused the dispute, which is the definition of an unsolvable problem.
When a batch is disputed, take ten parts from the batch at random, measure each with a CMM against the cap drawing, and record the values. Then run the same ten parts through both gauge sets.
A part that the CMM places inside the tolerance but a gauge rejects identifies the gauge as the conservative one. Ten parts are enough to show the pattern; a hundred only refine it.