Every measurement carries an associated doubt, and uncertainty is the way to express it rigorously. Instead of saying “the length is such-and-such a value”, you say that the value is that one, with an interval within which the true value is expected to lie with a certain probability. Uncertainty does not measure how wrong the instrument was, because the actual error is unknown; it describes how well the result is known.
Where it comes from
All the sources that influence the result are combined, each converted to a comparable deviation:
- The repeatability of the readings (random effect).
- The uncertainty of the standard with which the instrument was calibrated.
- The resolution of the indicator.
- Environmental effects such as temperature or humidity.
- The drift of the instrument since its last calibration.
- The method and the operator.
Standard and expanded uncertainty
Each contribution is first expressed as a standard uncertainty, equivalent to a standard deviation. Combining them, normally in quadrature when they are independent, gives the combined uncertainty. This is then multiplied by a coverage factor to obtain the expanded uncertainty, which corresponds to a stated confidence level. A result without its coverage factor or confidence level cannot be properly interpreted.
What the engineer sees in practice
On a calibration certificate it appears next to each measured point. It is used to decide whether the instrument is suitable: if the uncertainty is comparable to the tolerance you want to control, the decision to accept or reject a part becomes doubtful. It also makes it possible to compare results from different laboratories: two different values may be compatible if their intervals overlap.
How to interpret the result
A small uncertainty does not by itself imply a better instrument, but rather that the combination of method, standard and conditions was better controlled. And a large uncertainty is not a hidden defect: it is honest information that you must take into account when making decisions.