Error Propagation Calculator

Propagate uncorrelated standard uncertainties through sums, differences, products, quotients and powers.

Sum A + B
15
Uncertainty of sum
0.5
Product A × B
50
Uncertainty of product
4.272001873
Selected result
—
Combined standard uncertainty
—
Relative standard uncertainty (%)
—

First-order propagation for uncorrelated standard uncertainties: u(y) = √Σ(∂f/∂x · u(x))². Inputs are standard uncertainties, not maximum errors. Nonlinear results can be unreliable when uncertainties are large. A zero derivative can give a zero first-order term without implying exactness. Relative uncertainty is undefined when y=0. Negative bases require an integer exponent; noninteger powers require a positive base. 0⁰ and singular powers are rejected. Active values satisfy |x|≤10¹², uncertainties 0–10¹², exponent −100…100; outputs up to 10¹⁰⁰. Inactive fields are ignored. Up to ten significant digits.

NIST · TN 1297

How the calculation works

u(A+B) = √(uA² + uB²); u(A×B) ≈ √((B×uA)² + (A×uB)²)

Inputs are uncorrelated standard uncertainties, not maximum errors. Product uncertainty uses a first-order approximation and can be unreliable for large relative uncertainties.

Values must satisfy the formula’s domain. Nonzero results too small or too large for reliable representation are rejected.

First-order propagation for uncorrelated standard uncertainties: u(y) = √Σ(∂f/∂x · u(x))². Inputs are standard uncertainties, not maximum errors. Nonlinear results can be unreliable when uncertainties are large. A zero derivative can give a zero first-order term without implying exactness. Relative uncertainty is undefined when y=0. Negative bases require an integer exponent; noninteger powers require a positive base. 0⁰ and singular powers are rejected. Active values satisfy |x|≤10¹², uncertainties 0–10¹², exponent −100…100; outputs up to 10¹⁰⁰. Inactive fields are ignored. Up to ten significant digits.

Frequently asked questions

How is the result calculated?

u(A±B)=√(uA²+uB²); u(AB)≈√((B·uA)²+(A·uB)²); u(A/B)≈√((uA/B)²+(A·uB/B²)²); u(Aᵖ)≈|p·Aᵖ⁻¹|uA.

What assumptions apply?

Inputs are uncorrelated standard uncertainties, not maximum errors. Product uncertainty uses a first-order approximation and can be unreliable for large relative uncertainties.

How precise are the displayed results?

Numerical results show up to ten significant digits. Rounding does not improve the accuracy of the input measurements or the model. Values must satisfy the formula’s domain. Nonzero results too small or too large for reliable representation are rejected.

Also try:

Back to category →