What it does
Instruments drift. A pressure transmitter, a balance, a thermometer or a torque wrench reads a little further from the truth every month, and the calibration interval decides how far it can wander before someone checks it. Too short wastes money; too long risks measuring with an out-of-tolerance instrument and having to recall the work done with it.
This tool takes the calibration history of one instrument and:
- fits a straight-line drift of the as-found error against time, with the drift rate’s confidence interval and a test of whether the drift is real;
- draws the prediction band for the next as-found error;
- finds the date when the drift line, and the prediction band, reach the tolerance limit;
- suggests the next calibration date that keeps the chance of still being in tolerance at your reliability target (for example 95 %), using the t distribution of the prediction;
- optionally applies a guard band, pulling the limits in by the latest calibration uncertainty.
Nothing is stored or sent: the history stays on this page and is gone when you close it.
How to use it
- Paste or open the history: date, as-found error (reading minus reference, before adjustment) and the expanded uncertainty U of each calibration (optional, drawn as error bars and used for guard banding).
- Enter the tolerance limits (one is enough if the error can only go one way).
- Choose the reliability target: 95 % is common for general instruments; critical ones often use 99 %.
- Read the suggested date and the facts. “Check an interval” shows the chance of being in tolerance after any interval you are considering.
Worked example
The made-up history of a pressure transmitter has 9 calibrations from 2019-03-12 to 2024-02-21, with as-found errors rising from 0.01 to 0.19 kPa and U = 0.03 kPa; the tolerance is ±0.25 kPa. The drift is +0.0361 kPa per year (95 % CI +0.0308 to +0.0413), clearly significant, with a scatter of s = 0.0106 kPa about the line. The line itself reaches +0.25 kPa on 2025-11-08, but the 95 % prediction band reaches it on 2024-12-20. For 95 % reliability the next calibration is due by 2025-02-18, 11.9 months after the last. After 12 months the chance of being in tolerance is 94.9 %, with a predicted error of 0.224 kPa.
Tips
- Use as-found values only, and only since the last adjustment; an adjustment resets the drift.
- With fewer than 5 calibrations, the prediction band is wide and the interval short. That is the honest answer: the data do not yet support a long interval.
- If the drift is not significant, the interval is limited by the scatter. Consider a longer interval only if many instruments of the same type behave the same way.
FAQ
Why does the prediction band reach the limit before the line? The band includes the uncertainty of the fitted line and the scatter of single calibrations; the reliability target is applied to that, not to the line alone.
Does it save my instruments? No. Nothing is saved in your browser or sent anywhere. Download the CSV or report to keep a record.
Is the interval method official? It follows the reliability-target idea used in calibration interval guidance, but your quality system decides. For study and planning, not a substitute for professional judgement.
Where the numbers come from
- Least-squares straight line with its prediction interval and the t distribution (standard regression results, e.g. Montgomery and Runger, Applied Statistics and Probability for Engineers).
- The reliability-target approach to calibration intervals follows the ideas in NCSL International RP-1 and ILAC-G24 / OIML D 10 (guidance on the determination of calibration intervals), described in fibub's own words; no text or tables copied.
- Checked against scipy.stats.linregress, scipy.stats.t and scipy.optimize.brentq; no code copied. The example history is made up by fibub.
Standard scientific values, such as constants, atomic weights and melting points, are facts found in every textbook; we name where ours were checked. If you are an author or publisher and want a value, a credit or a table changed or removed, tell us and we will act quickly.
Last updated 2026-11-09.