Drift and stability
The change in an instrument’s reading over time, with no corresponding change in the quantity being measured.
Drift is the continuous or gradual change in an instrument’s reading over time while the value of the measured quantity stays put. The instrument can therefore give a different reading today from the one it gave at its last calibration, even though it is measuring the same standard under the same conditions.
Stability is the instrument’s ability to hold its metrological properties constant as time passes. The smaller the drift over a given interval, the more stable the instrument is considered to be.
How it shows up
Drift can take more than one form.
The zero reading itself moves. An unloaded balance, for instance, may read 0.3 g instead of zero after some time has passed.
The sensitivity or the calibration curve changes, so the error grows by a different amount at each point of the measuring range.
A micrometer is checked at intervals against the same 50.000 mm gauge block, under the same conditions:
At calibration 50.002 mm
Three months 50.004 mm
Six months 50.006 mm
The total change of 0.004 mm is evidence of drift, provided the standard and the measuring conditions stayed the same. The three successive readings also reveal its character: the deviation does not appear suddenly, it grows steadily.
Where it comes from
Drift may come from ageing components, wear, mechanical parts working loose, changes in electronic components, or the release of internal stresses. Changes in temperature and humidity can shift the reading too. To separate genuine drift from an environmental effect, though, the comparisons have to be made under the same specified conditions.
How it is caught
Drift is not the same thing as random error. Random error scatters the readings· drift shows up as a trend over time. It is caught through periodic calibration, intermediate checks, and plotting the results to watch where they are heading.
Re-zeroing can hide a zero drift for a while, but it proves nothing about whether the instrument is still correct across its whole range. That is why stability is assessed with standards at more than one point of the range.