In the world of metrology and ISO/IEC 17025 accreditation, a measurement is never just a single number. It is a range. The Uncertainty Budget is the financial statement of that measurement. It details every “cost” (source of error) that contributes to the final “balance” (the measurement uncertainty).
Reading an uncertainty budget can feel like deciphering a complex tax return. This guide will break down the anatomy of the budget. It will explain the statistical concepts. It will also help you interpret what the final numbers actually mean for your product quality.
Metrology budget are critical to understand in case of complex test like EMC campaigns and Leakage current test. A clear understanding of the uncertainty can easily change the final PASS or FAIL results.
1. The Core Concept: Error vs. Uncertainty
Before diving into the table, it is vital to understand the distinction between error and uncertainty:
- Error: The difference between the measured value and the “true” value. (We try to correct for known errors).
- Uncertainty: The doubt that remains about the result after correcting for error. It quantifies how much the “true” value might vary from your measured result.
2. The Anatomy of the Budget Table
Most uncertainty budgets are presented as a spreadsheet. Below is a breakdown of the standard columns you will encounter and what they represent.
Column A: The Source of Uncertainty
This lists where the variation is coming from. Common sources include:
- Reference Standard: The uncertainty inherited from the master equipment used for calibration.
- Resolution: The smallest digit the device can read.
- Repeatability: The variation observed when measuring the same thing multiple times.
- Environmental Factors: Temperature, humidity, or vibration effects.
Column B: The Value ±\pm
This is the raw magnitude of the uncertainty source in the unit of measurement (e.g., 0.05 mm, 0.1 °C).
Column C: Probability Distribution
This is often the most confusing part for non-statisticians. It describes how the errors are likely distributed.
- Normal (Gaussian): Used for Type A data (statistics from repeated readings) or calibration certificates where a confidence level is stated.
- Rectangular (Uniform): Used when we know the limits (e.g., resolution or manufacturer specs) but the value is equally likely to be anywhere within those limits. This is divided by 3\sqrt 3.
- Triangular: Used when values are more likely to be near the center of the range than the extremes. This is divided by6\sqrt {6}.
- U-Shaped: Common in temperature control (thermostats), where values hang out at the extremes. This is divided by 2\sqrt {2}.
Column D: The Divisor
Based on the distribution (Column C), we divide the raw value to standardize it to “One Standard Deviation”σ\sigma .
- Normal k=2k=2: Divide by 2.
- Rectangular: Divide by 3\sqrt 3(approx 1.732).
Column E: Sensitivity Coefficient
Sometimes the source of error is in a different unit than the measurement.
- Example: If measuring length (meters) but temperature (°C) causes expansion, the sensitivity coefficient converts °C into meters based on the thermal expansion coefficient.
- If the units are the same, this number is usually 1.
Column F: Standard Uncertainty uiu_i
This is the normalized value.
This column is crucial because this is the only column where the values can be compared apples-to-apples.
3. Calculating the “Combined” Uncertainty Budget
Once all sources are listed and normalized into Standard Uncertainty uiu_i, the lab calculates the Combined Standard Uncertainty ucu_c.
They do not simply add the numbers up. Because it is unlikely that all errors will occur in the worst possible direction simultaneously, labs use the Root Sum Squares (RSS) method:
How to Interpret this: The ucu_c represents one standard deviation. Statistically, this means there is roughly a 68% probability that the true value lies within this range. For most industrial applications, 68% confidence is not high enough.
To understand better the reasoing you can check this uncertanty budget example
Excel is also an important resource to prepare the complete uncertainty budget
4. The Final Result: Expanded Uncertainty U
To provide a result with higher confidence (usually 95%), the lab multiplies the Combined Uncertainty by a Coverage Factor kk.
The Formula:
- The Standard: usually, k=2k=2.
- The Result: If ucu_c is 0.05 and k=2k=2, the Expanded Uncertainty UU is 0.10.
What this tells you:
When you see a calibration certificate that says: 10.00 mm±0.10 mm,k=210.00 \text{ mm} \pm 0.10 \text{ mm}, k=2, it translates to:
“We measured 10.00. We are 95% confident that the true value lies somewhere between 9.90 and 10.10.”
5. Red Flags: How to Spot a Bad Budget
If you are auditing a lab or reviewing a vendor’s budget, look for these common errors:
- The “Dominant” Repeatability: If the repeatability is zero, the budget is suspicious. Nothing measures perfectly every time.
- Missing Resolution: The device’s resolution is a mandatory uncertainty component. It implies you cannot measure better than the screen displays.
- Double Counting: Including “Accuracy” from a spec sheet AND “Linearity” and “Hysteresis” often counts the same error sources twice.
- The “Best Case” Scenario: Using a budget created in a temperature-controlled cleanroom to validate measurements taken on a hot, vibrating factory floor.
6. Application: Pass/Fail Decisions
Reading the budget is only half the battle; applying it is the rest.
ILAC P14: policy for uncertainty in calibration.
If your product specification is 10.0±0.210.0 \pm 0.2, and the lab measures 10.1810.18, is it a Pass?
- If the Uncertainty is 0.01, it is a Pass.
- If the Uncertainty is 0.05, the measurement range extends up to 10.23. This is a conditional pass or indeterminate result because the uncertainty overlaps the limit.
Summary: When interpreting a budget, focus on the Expanded Uncertainty UU. This is your “margin of doubt.” If this number is larger than your tolerance allows (usually a ratio of 4:1 or 3:1 is desired), the measurement process is not capable of verifying your product.
