Calculation Method of System Uncertainty for Energy Meter Test Benches

The half-width is a/2. The standard uncertainty is:

u_res = (a / 2) / √3

For a = 0.001%:

u_res = (0.001 / 2) / √3 = 0.000288675%

The rectangular distribution assumes that the true value is equally likely to lie anywhere within half of one displayed digit.

Step 5: Convert the Reference Certificate Uncertainty

The calibration certificate of the reference standard meter states an expanded uncertainty U_ref = 0.0075% with coverage factor k_ref = 2. The standard uncertainty is:

u_ref = U_ref / k_ref

u_ref = 0.0075 / 2 = 0.00375%

The certificate also provides the reference meter error at the relevant test point. In the example, e_ref = −0.0002%.

Step 6: Correct the Meter Error

The measured mean error is corrected by the reference-standard error. Using the sign convention in the workbook:

E_MTE = e_ref − x̄

E_MTE = −0.0002 − 0.0016 = −0.0018%

This corrected value is the error attributed to the meter under test at the selected point.

Step 7: Combine the Standard Uncertainties

Because all components are expressed in percent error and the sensitivity coefficients are 1, the combined standard uncertainty is the root-sum-square:

u_c = √(u_rep² + u_res² + u_ref²)

Substituting the example values:

u_c = √(0.000266667² + 0.000288675² + 0.00375²)

u_c = 0.003770536%

Step 8: Calculate the Expanded Uncertainty

The final result is reported as an expanded uncertainty using coverage factor k = 2:

U = k × u_c

U = 2 × 0.003770536 = 0.007541073%

After rounding, the corrected meter error and its expanded uncertainty at this point are:

E_MTE = −0.0018%
U = 0.00754% (k = 2)

The complete uncertainty statement should always include the coverage factor. A value of U without k is incomplete.

Worked Uncertainty Budget

Quantity	Estimate	Standard uncertainty	Distribution	Sensitivity	Contribution
Mean measurement result	0.0016%	0.000266667%	Normal	1	0.000266667%
Display resolution	0	0.000288675%	Rectangular	1	0.000288675%
Reference standard meter	−0.0002%	0.00375%	Normal	1	0.00375%
Combined standard uncertainty	—	0.003770536%	Normal	—	—
Expanded uncertainty, k = 2	—	—	Normal	—	0.007541073%
Corrected error of the meter	−0.0018%	—	—	—	—

The reference-standard component is about 0.00375%, while repeatability and resolution are each below 0.0003%. This difference controls the shape of the final result.

Summary of the 24 Test Points

The workbook repeats the same uncertainty model at 24 active-energy points covering low and high voltage, low and high current, and several power factors. The resulting values vary only slightly because the reference certificate uncertainty remains constant and dominates the budget.

Parameter	Result from the workbook
Number of test points	24
Nominal voltages	58 V and 230 V
Current range	0.05 A to 120 A
Power factors	1.0, 0.8, and 0.5
Individual measurement mean	−0.0040% to 0.0016%
Sample standard deviation	0.000316% to 0.000949%
Corrected meter error	−0.00241% to 0.00470%
Combined standard uncertainty	Approximately 0.00376% to 0.00377%
Expanded uncertainty, k = 2	0.007525% to 0.007546%
Reference-standard contribution	Approximately 98.8% to 99.4% of the combined variance

The near-constant expanded uncertainty is not a coincidence. The reference meter has a fixed certificate uncertainty of 0.00375% standard uncertainty, while the repeatability of the mean is generally between 0.0001% and 0.0003%. The resolution contribution is also fixed at 0.000288675%.

How to Interpret the Result

The final statement should combine the corrected error and its uncertainty:

E_MTE = −0.0018%
U = 0.00754% (k = 2)

This means that, under the assumptions of the model, the expanded uncertainty interval associated with the corrected error is approximately −0.00934% to 0.00574%. The interval is calculated by subtracting and adding U to the corrected error.

If the acceptance limit is very close to this interval, the measurement result may be inconclusive. A formal decision rule and guard band may be required. This is especially important when the test result is used for conformity assessment or customer acceptance.

Why the Reference Standard Dominates

The squared uncertainty contributions are:

• Reference standard: 0.00375² = 1.40625 × 10⁻⁵ %²
• Resolution: 0.000288675² = 8.333 × 10⁻⁸ %²
• Repeatability of the mean: approximately 7.1 × 10⁻⁸ %² in the example

The reference contribution is more than 98% of the total variance. Improving repeatability or increasing the number of measurements will therefore produce only a small reduction in the final expanded uncertainty unless the reference standard is upgraded or its calibration uncertainty is reduced.

Limitations of the Example Model

The workbook is a practical uncertainty calculation, not a complete uncertainty budget for every possible influence quantity. Its main terms are repeatability, display resolution, and the calibration uncertainty of the reference standard meter. These are the components explicitly represented in the spreadsheet.

A formal uncertainty budget should also evaluate any significant source that is not already covered by the reference certificate or the test-system validation. Depending on the bench and test point, additional components may include:

• Source stability and regulation during the test interval
• Voltage and current measurement uncertainty
• Waveform distortion, harmonics, and frequency variation
• Phase-angle error and power-factor influence
• Lead, clamp, contact, and burden effects
• Temperature, humidity, and magnetic-field influence
• Pulse measurement, timing, and synchronization
• Operator technique and meter-position variation
• Drift of the reference standard between calibrations

If a component is not covered by the reference certificate and its effect is not negligible relative to the combined uncertainty, it should be included in the budget. Otherwise, the reported uncertainty may be too optimistic.

Why the Sensitivity Coefficients Are 1

All uncertainty components are expressed directly as errors in percent. For example, repeatability is in percent, resolution is in percent, and the reference certificate uncertainty is in percent. When the output is also in percent error, the derivative of error with respect to each component is 1.

This simplifies the combined uncertainty equation. If a component were expressed in voltage, current, time, or phase angle, a sensitivity coefficient would be needed to convert its effect into percent error.

Coverage Factor and Probability Assumptions

The final expanded uncertainty uses k = 2, which corresponds approximately to a 95% coverage interval when the combined distribution is approximately normal. The workbook assumes a normal distribution for the final result.

The repeatability component is based on a sample of 10 measurements. In a more advanced analysis, the effective degrees of freedom and Welch–Satterthwaite approximation could be used to determine a more appropriate coverage factor. For most production and calibration applications with a stable process and a well-characterized reference standard, k = 2 is a practical and widely accepted choice.

Decision Rules and Guard Bands

When the measured error is close to a specification limit, uncertainty must be considered in the decision rule. One common approach is to establish an acceptance zone that is reduced by the expanded uncertainty, often called a guard band.

For example, if a meter must remain within ±0.2% and the test-system expanded uncertainty is 0.0075%, a simple guard band would require the measured error to remain within approximately ±0.1925% for an acceptance decision. The exact rule depends on the applicable standard, contract, and quality procedure.

This approach follows the principle described in documents such as ILAC-G8, which addresses decision rules and statements of conformity. It prevents a meter that is very close to the limit from being accepted without considering measurement uncertainty.

Practical Recommendations

1. Record at least 10 repeated measurements at each critical test point.
2. Calculate the mean and sample standard deviation from the raw data.
3. Use the standard deviation of the mean, not the standard deviation of individual readings, for repeatability.
4. Confirm the resolution and probability model used for rounding.
5. Use the reference certificate’s stated expanded uncertainty and coverage factor exactly as issued.
6. Apply the reference error correction with a consistent sign convention.
7. Combine all standard uncertainties by root-sum-square when they are independent and expressed in the same unit.
8. Report both the corrected error and the expanded uncertainty with its coverage factor.
9. Review additional influence quantities that may not be covered by the reference certificate.
10. Use a guard band when the result is close to the acceptance limit.

Frequently Asked Questions

Can I divide the reference certificate uncertainty by 2?

Only when the certificate states that the expanded uncertainty is based on k = 2. Always use the coverage factor stated by the calibration laboratory.

Should I use the standard deviation or the standard deviation of the mean?

Report the sample standard deviation as a measure of scatter, but use s/√n as the standard uncertainty of the mean in the uncertainty budget.

Why is display resolution included?

Every displayed value is rounded to the last digit. This rounding contributes uncertainty even when the underlying measurement is stable.

Why does the uncertainty change so little between test points?

The reference standard contributes most of the combined uncertainty. Its certificate uncertainty is approximately constant, so changes in repeatability have only a small effect on the final expanded uncertainty.

Can the test bench be used if the uncertainty is larger than the meter specification?

The bench may still be useful for relative or production checks, but it is not suitable for unambiguous conformity decisions at that specification level. A more accurate reference standard or a different test method may be required.

Conclusion

The system uncertainty of an energy meter test bench can be calculated from repeated measurements, display resolution, and the calibration certificate of the reference standard meter. The method is transparent and practical:

u_c = √(u_rep² + u_res² + u_ref²)
U = k × u_c

In the analyzed workbook, the combined standard uncertainty is approximately 0.00376%, dominated by the reference standard contribution. The expanded uncertainty at k = 2 is approximately 0.00753% across the 24 test points.

The most important improvement is not more repeated measurements alone. It is a better reference standard, better traceability, and a documented treatment of every significant influence quantity. That is how the test bench moves from a simple pass/fail tool to a metrologically defensible measurement system.

To review a suitable test bench configuration, see our three-phase smart energy meter test bench, three-phase reference standard meter, and smart meter compliance and accuracy testing guide. You can also contact our engineering team for application support.

Related Resources

• Energy Meter Testing Equipment: Types & Applications
• Smart Energy Meter Test Bench: Compliance & Accuracy Testing
• How to Choose an Energy Meter Test Bench
• IEC 62053 vs IEC 62057: Energy Meter Testing Guide
• Three-Phase Reference Standard Meter Selection Guide
• Onsite Energy Meter Calibration: Equipment, Workflow, and Accuracy