The most useful number in a calibration record is the one taken before anyone touches the instrument.
Take a room differential pressure transmitter holding a cascade with a 10 Pa alarm limit. If it reads 2 Pa high, the room could have sat at 9 Pa for months while the screen showed 11. Only the as-found reading can tell you that.
The order matters:
- As-found first. Apply the reference at each test point and record what the transmitter reads, before any adjustment. Do not zero it "just to get started".
- Compare against a traceable reference, with a certificate that is in date and an uncertainty small enough for the tolerance you are checking.
- Adjust only if needed, then record the as-left readings at the same points.
- If the as-found was out of tolerance, raise it. Every record that instrument produced since its last good calibration is now in question, and QA has to assess whether any batch or room release relied on it.
A technician who zeroes first and records afterwards produces a clean certificate and erases the months of drift that QA needed to see.
Where the tolerance comes from matters too. The transmitter's datasheet accuracy says what the instrument can do; the process limit says what you need. A transmitter inside its own accuracy class can still be too coarse for a 10 Pa margin.
Humiley's Maintenance Service includes periodic calibration and requalification.




