Understanding Measurement Uncertainty

In calibration laboratories, measurements need to be as accurate as possible. However, any measurement is potentially affected by variations and sources of errors that are inherently involved in any measuring process. In most cases, when measuring any piece of equipment with a high level of precision, achieving the exact same measurement every time is rarely possible.

It is impossible to completely control every factor that creates an error in a measurement. While the goal of any calibration laboratory is to keep these errors small, they cannot be completely eliminated. The challenge for any calibration technician is to understand the potential errors that exist and to determine the quantity of these errors for every measurement.

Measurement Uncertainty was developed as a way to quantify and unify all of the influences that could affect the process of making a measurement. Measurement Uncertainty can be described as a confidence level in the measurement. A smaller or tighter uncertainty value would represent a high confidence in the measurement result, where a larger or wider uncertainty value would represent a lower confidence in the result.

The total uncertainty of a measurement must always be considered when performing any type of measurement. A measurement result is meaningless without the measurement uncertainty value for that specific measurement. Only once the uncertainty value has been established can it be recognized as an accurate and traceable measurement result.

Contributors to Measurement Uncertainty

Measurement Uncertainty can be a confusing concept. Equipment users often assume that the reported manufacturer’s specifications are the only influence in their measurements, however, many other factors contribute to the measurement results. The total uncertainty is not limited to just the inherent calibration equipment error. Calculations for the total error in any measurement must consider all of the possible factors plus any other known sources of error in the system.

While it is always a challenge to know exactly which types of errors the measurement system has, there are certain identifiable factors that should be taken into consideration when estimating total measurement uncertainty. The most common factors are the influences of the reference standards such as specification, drift and resolution, the auxiliary items used in the measurement setup such as cables, probes and connectors, and the environmental conditions during the measurement such as temperature variations and electromagnetic interference.

Measurement Uncertainty Classifications

Measurement Uncertainty factors are usually divided into two classifications.

The Type A uncertainty classification is used to determine the random errors in the measurement which are typically caused by unknown variations in the measurement process. They are best detected by making a series of measurements and performing a statistical analysis to evaluate the measurement results. An example would the determination of the standard deviation of a series of repeatability and reproducibility measurements taken by several laboratory technicians. For dimensional measurements, where measurement technique is highly important, Type A uncertainty values usually represent the largest of the uncertainty contributions.

The Type B uncertainty classification is used to categorize the systematic errors that have a known and consistent effect on the measurement process. Systematic errors pertain to the irregularities in the equipment used in the measurement process or external influences on the equipment such as the reference standard’s specifications and uncertainties, equipment resolutions, temperature coefficients, drift rates, and signal noise. The proper corrections can be applied as needed once the influence of the systematic error is identified.

The influences that have the highest effect on the measurement will have a greater value in the uncertainty calculations, while minor influences will not considerably contribute to the overall measurement uncertainty value.

Expanded Measurement Uncertainty

Most of the uncertainty calculations are based on a confidence level of one standard deviation, which states that approximately 68% of the measurement results will lie within the calculated uncertainty range. For calibration laboratories, a much higher level is required in their reported measurement results.

Expanded Uncertainty refers to uncertainty calculations that are calculated to a confidence level of two standard deviations, which implies a 95% chance that the measured value lies within the uncertainty range. For most situations, the normal uncertainty calculations are multiplied by a value of 2 to achieve the expanded uncertainty value.

Measurement Uncertainty values evolve constantly, and laboratories must routinely review their measurement uncertainty values at an established interval. The uncertainty values for the reference standards can change if calibrated by different vendors so those calibration certificates need to be monitored to detect any significant changes. Turnover in laboratory personnel may also require the performance of new repeatability and reproducibility measurements for certain laboratory functions, such as dimensional measurements.

Is the calibration lab you’re using giving you all the information you need?
Find out here, and whether you’re actually saving any money.