Measurement uncertainty in plain language
Every result on a calibration certificate comes with a ± figure. What does it mean, where does it come from, and why should you care?
Look at any line in the results table of a proper calibration certificate and you will see a number like ± 38 µV alongside the measured value. This is the measurement uncertainty, and it is arguably the most important number on the page. It is also the one most often ignored.
What uncertainty is not
Uncertainty is not the same as error. Error is the difference between what your instrument read and the reference value. Uncertainty describes how well we know that reference value and the measurement process in the first place.
If your multimeter reads 10.000 12 V when we apply 10.000 00 V, the error is +120 µV. If our uncertainty is ±38 µV, we are saying the true error lies somewhere between about +82 µV and +158 µV, with roughly 95 % confidence.
Where it comes from
No measurement is perfect. An uncertainty budget lists every factor that could influence the result and estimates the size of each one. For a DC voltage calibration, typical contributors include:
- the uncertainty of our reference standard, taken from its own calibration certificate,
- the drift of that standard since it was last calibrated,
- the resolution of the instrument under test,
- short-term noise, estimated from repeated readings,
- temperature effects on both instruments,
- thermal EMFs in connections and leads.
These are combined following the method set out in the Guide to the Expression of Uncertainty in Measurement (the GUM, JCGM 100:2008). The combined standard uncertainty is then multiplied by a coverage factor, usually k = 2, to give an expanded uncertainty with a coverage probability of about 95 %.
Why you should care
There are three practical reasons.
First, it tells you whether the calibration was meaningful. If the uncertainty is as large as the tolerance you are checking against, the calibration cannot reliably tell a good instrument from a bad one. A common rule of thumb is that the uncertainty should be no more than a quarter of the tolerance, known as a test uncertainty ratio (TUR) of 4:1.
Second, it affects pass and fail decisions. If a reading sits very close to a tolerance limit, the uncertainty may straddle the limit. In that case the lab cannot honestly say whether the instrument passes or fails, and a good certificate will say so.
Third, it feeds into your own uncertainty budgets. If you use a calibrated instrument to make measurements for your customers, the uncertainty on its certificate becomes one of the inputs to your own uncertainty.
A quick sanity check
Next time you receive a certificate, pick a test point and compare the stated uncertainty with the instrument’s specification at that point. If the uncertainty is small relative to the specification, the calibration has resolving power. If it is not, it is worth a conversation with your calibration provider.
