【2026 Instrument Selection】Instrument accuracy, repeatability, and uncertainty
Thread Content
This post was last edited by The one on 2026-6-11 13:44. Accuracy (level of precision) is the maximum amount of error that the manufacturer guarantees. Repeatability refers to how much the values vary when the same measurement is taken several times on the same table. Uncertainty represents the extent to which a measurement result can be trusted, after taking all factors into account. When selecting a model: first determine how much tolerance the process allows, and then work backwards to decide what grade of gauge to purchase and what range to set!
1 What exactly do these three words mean? Let’s use target shooting as an example first. The bull’s eye represents the \"true value\", and each shot constitutes a measurement: ▪ Accuracy: The impact point is generally around the bull’s eye. If it misses a portion of the target but remains centered on it, this is considered to be good accuracy – that is, there is little systematic error. To use an analogy: if a household scale is programmed to show 2 kg more than the actual weight, then no matter how carefully you weigh yourself, it will still give a value that is 2 kg higher – this is a systematic error.
▪Precision/repeatability: The impact points stay closely clustered together, even if the entire group shifts to one side. The corresponding random error is small – when measuring several times in a short period of time under the same table, with the same person and under the same conditions, the readings remain fairly consistent.
Note: Hugging tightly together does not equal hitting accurately. It is a gun with a misaligned sight; every shot lands in the same incorrect position – it has good repeatability but is inaccurate, as shown in Figure 1(b). This is why calibration (adjusting the sight) is never optional.
「\"Instrument accuracy of 0.5 grade\": What this really means is that the manufacturer guarantees that the maximum error will not exceed ±0.5% of the range (this term is known as the maximum permissible error, or MPE, expressed as a percentage of the range). Pay attention to two key terms: one is maximum—this is a commitment regarding the upper limit of the error, not the actual error. Second is the range – the denominator is the range rather than the current reading, and this directly affects the selection process.
2 Repeatability: It shows whether one’s hands are steady. The question related to repeatability is: with all conditions remaining unchanged, how much will this watch shake on its own. In the specification sheet, it is usually defined as follows: for the same input, when a voltage is gradually applied from the same direction, the difference between multiple readings (expressed as a percentage of the full scale). Note that it generally does not include hysteresis – hysteresis refers to the mismatch between the readings on the upward and downward strokes, and it is another separate parameter.
Many transmitter data sheets combine linearity, hysteresis, and repeatability into a single \"reference accuracy\", such as ±0.075%. When reviewing the specifications, it is essential to understand what components are included in this value and under what conditions it was measured (usually in laboratory conditions at around 25 °C). Effects of temperature and static pressure, as well as long-term drift, are often listed separately; these values need to be added manually when using the device in actual conditions.
There is another word that is easily confused: reproducibility. Repeatability means \"using the same table, the same person, and conducting the test again immediately\"」 ； Reproducibility means \"testing again in a different location, with different people, and using different equipment.\" Testing whether the two laboratories obtain the same result when weighing the same standard weight once is a way of assessing reproducibility.
3 Uncertainty 3.1 Why having only a \"precision grade\" is not enough Error = Measured value − True value, but the true value can never be known with certainty; therefore, it is strictly speaking impossible to determine what the error of a particular measurement is. Rephrased, the answer becomes clear: what is the probability that the truth value falls within a certain range around the measured value? This \"half-width of the range\" is the uncertainty. It does not need to know the truth value; it is a quantitative answer to the degree of credibility of the result. The internationally recognized evaluation method is called GUM, while the corresponding domestic standard is JJF 1059.1.
3.2 Two types of sources: those measured by oneself + those obtained from research. 1) Type A (statistical method): Perform the measurement repeatedly n times by oneself to calculate the standard deviation s. When calculating the average value, the uncertainty of that average is u = s/√n – the more times it is measured, the more reliable the average value becomes.