In sensing applications, performance can be improved by combining accuracy and precision, but what’s the difference between them?
Consider accuracy and precision in both two-dimensional and three-dimensional spaces, advises David Perkon, vice president of advanced technology, AeroSpec. “In its simplest form, accuracy is measured everywhere in space, and it is measured to national standards,” he says. “It is exact everywhere in the 3D space around the equipment work envelope, which is important in many applications. Precision, or repeatability, may vary depending on where the tool or part is positioned. A noted difference between accuracy and precision is that, in the 3D space, repeatability is more consistent in some positions than others.”
In machine vision, accuracy means how close a measurement is to its true value relative to a standard, explains Ben Dawson, director of strategic development, Teledyne Dalsa. “Precision is the fineness to which the measure can be made, often limited by measurement noise,” he says. “For example, when a machine vision system measures a dimension, the number of valid—noise-free—digits in the measure is the precision, and the accuracy is how close the measure is to a reference dimension in, say, millimeters.”
You can imagine measurement as shooting arrows at a bull’s-eye target, suggests Dawson (Figure 1). “The center of the target is the desired true value of the measurement,” he explains. “Then accuracy is how close to the center your measurement arrows are, and precision is how closely your arrows are clustered, assuming that arrows have a finite diameter.”
Precision is a measure of the sensor’s ability to repeatedly report the same value when moving to the same position, says Matt Hankinson, senior technical marketing manager, MTS Sensors. “It’s a statistical measure of the random spread or variability of values at the same point on the sensor scale,” he explains.
“Accuracy is a measure of the closeness to the true value across the full measuring range. In many applications, it is important to have good repeatability, or precision, from a control perspective, so the movement is consistent, but it isn’t necessarily important to have absolute accuracy with the true value. The term accuracy is sometimes used to mean trueness, which is technically a different parameter. Trueness is the closeness to the true value over an average of samples and represents a systematic bias in the measurement. It is possible to have a large spread of measurements—poor precision—that average out to the true value, which would have a high degree of trueness. Accuracy is technically a combination of the multiple errors source for the overall closeness to the true value for each measurement. Resolution, not to be confused with precision, is the underlying smallest change that can be measured and doesn’t factor the repeatability of precision.”
Knowing the precision and accuracy of readings is key to effective measurement and sensing, advises Peter Thorne, director of the research analyst and consulting group at Cambashi. “Take repeated readings, and, even when the item being measured is not changing, you will get a range of values,” he warns. “The differences are caused by random reasons like electrical noise, or perhaps vibration of the equipment you are using. The precision of the measurement quantifies variations between repeated readings. Accuracy is different. It quantifies the gap between the true value and the measurement. Accuracy is often more difficult to determine than precision is, because systematic errors can impact the accuracy without changing the precision.”
If you're carrying out a set of repeated measurements, precision describes the ability of your process to get the same result each time for multiple measurements of the same item, but it doesn't take into account whether the results are close to the true value, explains Colin Macqueen, director of technology, Trelleborg Sealing Solutions. “Accuracy describes how close your results are to the true value, but it doesn't take into account the consistency of the results,” he says, “so precision tells you how far your individual results are from the mean, whereas accuracy tells you how far your mean result is from the true value.”