CMC Uncertainty Budgets

An Uncertainty Budget is the fundamental document that proves a calibration laboratory understands and controls its measurement processes. For laboratories seeking or maintaining ISO/IEC 17025 accreditation, establishing a Calibration Measurement Capability (CMC) uncertainty budget for every parameter on their scope of accreditation is mandatory.

The CMC represents the smallest uncertainty that a laboratory can achieve when calibrating a "nearly ideal" device under test (DUT) or Unit Under Test (UUT). It is the baseline capability of the laboratory. Constructing a CMC budget requires identifying all significant sources of error, quantifying them, and combining them using the GUM framework.

Step 1: Define the Measurement Process

Before any math begins, the measurement process must be explicitly defined. What is the parameter? What is the range? What reference standard is being used? What is the method (direct measurement, substitution, ratio)? This information dictates the mathematical model of the measurement.

Step 2: Identify Uncertainty Components

For a typical electrical or physical calibration CMC budget, you must consider at minimum the following universal contributors:

• Reference Standard Uncertainty (Type B): The uncertainty of the reference standard itself. This is always taken directly from the calibration certificate provided by the higher-echelon laboratory (NMI or accredited primary lab) that calibrated your standard. It is typically a Normal distribution ($k=2$).
• Reference Standard Long-Term Drift (Type B): Standards drift over time. You must estimate the maximum possible drift that could occur between calibration intervals. This is usually derived from historical data or manufacturer specifications and treated as a Rectangular distribution.
• Process Repeatability (Type A): The random scatter of the measurement process. This is determined by taking a series of independent measurements (e.g., $n=10$ or $n=30$) of a stable artifact and calculating the standard deviation of the mean.
• Resolution of the Reference Standard (Type B): The smallest increment the reference standard can display. Evaluated as a Rectangular distribution.
• Resolution of the "Best Existing" UUT (Type B): Because a CMC budget represents a laboratory's best capability, you must include the resolution of a hypothetical "nearly ideal" UUT. Evaluated as a Rectangular distribution.
• Environmental Influences (Type B): Changes in temperature, humidity, or barometric pressure can affect the standard or the measurement process. The magnitude of these effects requires calculating the temperature coefficient of the standard multiplied by the laboratory's environmental control limits. Evaluated as a Rectangular or U-shaped distribution.

Step 3: Convert to Standard Uncertainties

Every identified component must be converted to a Standard Uncertainty ($u_i$) operating at a 1-sigma (approx. 68%) confidence level, in the exact same units of measure.

• Normal ($k=2$) → Divide by 2
• Rectangular Limits ($\pm a$) → Divide by $\sqrt{3}$
• Triangular Limits ($\pm a$) → Divide by $\sqrt{6}$
• U-Shaped Limits ($\pm a$) → Divide by $\sqrt{2}$
• Type A → Use standard deviation of the mean directly

Step 4: Apply Sensitivity Coefficients ($c_i$)

Ensure all standard uncertainties are in the same units as the measurand. If a component is evaluated in degrees Celsius, but the measurand is Ohms, you must apply a sensitivity coefficient (e.g., Ohms per degree Celsius) to convert the variance into the domain of the final measurement result. For direct measurements where all components are evaluated in the base unit, $c_i = 1$.

Step 5: Calculate Combined and Expanded Uncertainty

Root-Sum-Square (RSS) the individual standard uncertainties to find the Combined Standard Uncertainty ($u_c$). Then, determine the Effective Degrees of Freedom ($\nu_{eff}$) using the Welch-Satterthwaite equation to find the correct coverage factor $k$ for a 95.45% confidence interval. Finally, multiply $u_c$ by $k$ to yield the Expanded Uncertainty ($U$), which becomes the CMC value.