What stability means
A reference standard does not keep its value forever. Components age, reference elements drift, and impacts or thermal cycles can shift the value. Stability describes how much the value of the standard changes over time, and it is one of the reasons the interval between calibrations is not set at random.
The core method: the history
The most robust way to evaluate stability is to accumulate the calibration values of the same standard in each cycle and observe how they evolve. With that history you can:
- Plot the deviation from the nominal value at each calibration.
- Estimate the drift rate with a trend over time.
- Decide whether the drift is small compared with the required uncertainty, or whether it calls for a shorter interval.
- Correct the value of the standard using the trend, if technically justified.
Intermediate checks
Between calibrations, a check standard (or control standard) makes it possible to detect sudden changes. It is measured periodically against the reference standard, with the same procedure and conditions, and the results are recorded on a control chart with defined limits. If a point falls outside the limits, or there is a sustained trend, it is investigated before the standard continues in use.
An example of what you see
A resistance standard with three consecutive calibrations shows a small variation, but always in the same direction. That systematic behavior suggests real drift, not measurement noise. If the drift per interval is comparable to the calibration uncertainty, it cannot be distinguished from the measurement itself with few data; more cycles are needed.
Which conditions must be controlled
- Temperature and other influence quantities, the same in each comparison.
- A stabilization time before measuring.
- The same measurement configuration, cables and procedure.
- Careful transport and handling between measurements.
Common mistakes
- Basing the judgment on a single calibration.
- Comparing results from laboratories with very different uncertainties without taking them into account.
- Confusing a variation within the uncertainty with demonstrated drift.
- Extending the calibration interval merely for convenience, without supporting data.
- Failing to record events such as impacts, repairs or environmental changes that explain jumps.