Uncertainty Propagation and Numerical Differentiation, ENGR/PHYS 216 Modules 2–3 – Study Notes

Source: Final Exam Practice Problems and Solutions

Tags: uncertainty, error propagation, partial derivatives, kinetic energy uncertainty, Bragg's law, forward finite difference, numerical differentiation, momentum calculation, ENGR 216, PHYS 216


TL;DR

Uncertainty propagation lets you calculate how measurement errors compound when plugged into a formula. You take partial derivatives of the formula with respect to each measured variable, multiply by that variable's uncertainty, and combine in quadrature. Forward finite differences give you a numerical approximation of a derivative (velocity, acceleration) from discrete position-time data.


Key Terms

Uncertainty (error)

The range within which the true value of a measurement is expected to fall, expressed as ± some value.

Propagation of uncertainty

The method for determining how uncertainties in measured quantities carry through to a calculated result. Uses partial derivatives and root-sum-of-squares (RSS).

Partial derivative

The derivative of a multivariable function with respect to one variable, treating all others as constants. Central to the propagation formula.

Root-sum-of-squares (RSS)

The method of combining individual uncertainty contributions: square each, sum them, take the square root. Also called "adding in quadrature."

Forward finite difference

A numerical method for approximating a derivative at a point using the value at that point and the next point: f'(x) ≈ [f(x + h) − f(x)] / h.

Nominal value

The calculated result using the best (central) measured values, before uncertainty is applied.


Core Content

Uncertainty Propagation Formula

For a quantity Q that depends on measured variables x₁, x₂, ..., xₙ:

δQ = √[ (∂Q/∂x₁ · δx₁)² + (∂Q/∂x₂ · δx₂)² + ... + (∂Q/∂xₙ · δxₙ)² ]

The procedure is always the same:

  • Write out the formula for Q

  • Take the partial derivative of Q with respect to each measured variable

  • Multiply each partial derivative by the uncertainty in that variable

  • Square each term, sum them, take the square root

Worked Example: Kinetic Energy Uncertainty

Given: m = 10.0 ± 0.5 kg, v = 3.0 ± 0.2 m/s, KE = ½mv²

  • ∂KE/∂m = ½v² = ½(3.0)² = 4.5

  • ∂KE/∂v = mv = (10.0)(3.0) = 30

δKE = √[(4.5 × 0.5)² + (30 × 0.2)²] = √[5.0625 + 36] = √41.0625 = 6.41 J

Worked Example: Subtraction (Volume Remaining)

Given: V₁ = 125 ± 5 mL, V₂ = 25 ± 2 mL, V_remaining = V₁ − V₂

  • ∂V/∂V₁ = 1, ∂V/∂V₂ = −1

δV = √[(1 × 5)² + (−1 × 2)²] = √[25 + 4] = √29 = 5.39 mL

Note that when you subtract two measurements, the uncertainties still add in quadrature. They never cancel.

Worked Example: Volume of a Sphere

Given: r = 3.2 ± 0.15 cm, V = (4/3)πr³

  • ∂V/∂r = 4πr² = 4π(3.2)² = 128.68

δV = 128.68 × 0.15 = 19.3 cm³

Single-variable case: the partial derivative is just the ordinary derivative.

Worked Example: Bragg's Law

Given: d = 0.025 ± 0.0020 nm, θ = 25 ± 1.0°, nλ = 2d sin(θ), n = 1

Nominal: λ = 2(0.025) sin(25°) = 0.0211 nm

  • ∂λ/∂d = 2 sin(θ) = 2 sin(25°) = 0.8452

  • ∂λ/∂θ = 2d cos(θ) = 2(0.025) cos(25°) = 0.04532

Important: convert the angle uncertainty to radians before plugging in. 1.0° = 0.01745 rad.

δλ = √[(0.8452 × 0.0020)² + (0.04532 × 0.01745)²] = √[2.857 × 10⁻⁶ + 6.257 × 10⁻⁷] = 0.0019 nm

Final answer: λ = 0.021 ± 0.0019 nm

Forward Finite Difference Method

To approximate velocity from position-time data:

v(tᵢ) ≈ [x(tᵢ₊₁) − x(tᵢ)] / [tᵢ₊₁ − tᵢ]

Then momentum p = m × v.

Worked Example: Ball Momentum

Given: m = 2.0 kg, positions in cm at 0.1 s intervals.

Time (s)

Position (cm)

0.1

1.0

0.2

3.9

0.3

9.1

0.4

15.8

0.5

24.9

At t = 0.1 s: v = (3.9 − 1.0) cm / 0.1 s = 29 cm/s = 0.29 m/s, p = 2.0 × 0.29 = 0.58 (kg·m)/s

At t = 0.2 s: v = (9.1 − 3.9) / 0.1 = 52 cm/s = 0.52 m/s, p = 2.0 × 0.52 = 1.04 (kg·m)/s

At t = 0.3 s: v = (15.8 − 9.1) / 0.1 = 67 cm/s = 0.67 m/s, p = 2.0 × 0.67 = 1.34 (kg·m)/s

At t = 0.4 s: v = (24.9 − 15.8) / 0.1 = 91 cm/s = 0.91 m/s, p = 2.0 × 0.91 = 1.82 (kg·m)/s


Formulas / Diagrams

General uncertainty propagation: δQ = √[ Σ (∂Q/∂xᵢ · δxᵢ)² ]

Kinetic energy: KE = ½mv²

Volume of a sphere: V = (4/3)πr³

Bragg's Law: nλ = 2d sin(θ)

Forward finite difference: f'(xᵢ) ≈ [f(xᵢ₊₁) − f(xᵢ)] / h


Why It Matters / Exam Flags

⚠️ Always convert angle uncertainties to radians before using them in propagation with trig functions.

⚠️ Subtraction does not reduce uncertainty. The uncertainties still add in quadrature.

⚠️ For forward finite difference, you cannot compute the derivative at the last data point (there is no "next" value).

⚠️ Watch your unit conversions, especially cm to m when computing momentum.


Practice Q&A

Q: If Q = ½mv² and m = 10.0 ± 0.5 kg, v = 3.0 ± 0.2 m/s, what is δKE?

A: 6.41 J. Take partial derivatives ∂KE/∂m = ½v² and ∂KE/∂v = mv, multiply each by its uncertainty, combine in quadrature.

Q: Why does subtracting two measurements not reduce the uncertainty?

A: Because uncertainties are always squared before summing (RSS). The sign of the partial derivative becomes positive when squared, so addition and subtraction produce the same uncertainty formula.

Q: What is the key limitation of the forward finite difference method?

A: You lose the derivative estimate at the final data point, and the approximation is only first-order accurate (error proportional to the step size h).

Q: When propagating uncertainty through Bragg's Law, what common mistake can double your uncertainty in λ?

A: Forgetting to convert degrees to radians for δθ. The partial derivative ∂λ/∂θ = 2d cos(θ) gives a result in radians, so the uncertainty must also be in radians.


Related Terms / Search Tags

error propagation, uncertainty analysis, RSS, root sum of squares, quadrature, partial derivatives, measurement error, kinetic energy error, Bragg diffraction, X-ray crystallography, forward difference, backward difference, central difference, numerical differentiation, finite difference method, momentum from position data, ENGR 216 Module 2, ENGR 216 Module 3