Professional Documents
Culture Documents
Any measurement performed in a laboratory is affected by experimental errors. These errors do not necessarily imply a
human mistake in taking the measurements. They likely result from limitations in instrumental and human measurement
capabilities.
Types of errors
1. Instrumental errors
Any measuring device can only be used to measure with a certain degree of fineness. For example you cannot measure
length to micrometer precision if you are using a ruler.
2. Systematic errors
These errors are caused by a mistake that does not change during the measurements, affecting all experimental data the
same. Examples: a poorly calibrated instrument or a systematically incorrect reading of an instrument scale.
3. Random errors
Oftentimes when you perform the same measurement multiple trials, you can get different results each time. These errors
are due to variations in the experimental technique/procedure, unpredictable fluctuations in the output of the measuring
instrument, etc. A typical example of random errors is the time measurement performed using a stop watch. For example,
you may repeat the same time measurement 3 times and get values of 1.634s, 1.711s and 1.552s. These values are not
unreasonable, given that a rough estimation of human reaction time is about 0.1s. In this case, the instrumental error is
irrelevant as compared to the errors induced by the reaction time.
Uncertainties
Errors are quantified by associating an uncertainty value (denoted Δ) with each measurement. For example, the best
estimate of a time is 2.4s, but due to uncertainty, the time might be as short as 2.3s or as large as 2.5s. In terms of
uncertainty, this can be expressed in two different ways:
1. Absolute uncertainty (expressed in the units of the quantity): 𝑡 ± ∆𝑡 = 2.4 ± 0.1𝑠
∑𝑁
1 (𝑥𝑖 − 𝑥𝑎𝑣𝑒 )
2
Standard Deviation (σ) A measure of the variability of the data 𝜎=√
𝑁−1
𝝈
Standard Error (∆𝒙𝒂𝒗𝒆 ) Uncertainty of the mean value ∆𝒙𝒂𝒗𝒆 =
√𝑵
1
Experimental Errors and Uncertainties
Propagation of Uncertainties
Often, the quantity of interest cannot be determined in a one-step measurement (for example, area of a rectangle using a
ruler). We combine several measurements, each with their associated uncertainties, into an equation. These uncertainties
propagate (are carried) into the uncertainty of the final answer.
𝝏𝒇 𝟐 𝝏𝒇 𝟐
In general, for any function of two variables, 𝒛 = 𝒇(𝒙, 𝒚), the uncertainly is ∆𝒛 = √( ) (∆𝒙)𝟐 + ( ) (∆𝒚)𝟐 .
𝝏𝒙 𝝏𝒚
Some of the most common cases, all derived from the general formula above, are shown in the table below.
∆𝑥 2 ∆𝑦 2
Multiplication 𝑧 = 𝑥𝑦 ∆𝑧 = |𝑥𝑦|√( ) + ( )
𝑥 𝑦
𝑥 𝑥 ∆𝑥 2 ∆𝑦 2
Division 𝑧= √
𝑦 ∆𝑧 = | | ( ) + ( )
𝑦 𝑥 𝑦
Power 𝑧 = 𝑥𝑛 ∆𝑧 = |𝑛|𝑥 𝑛−1 ∆𝑥
Multiplication by a constant 𝑧 = 𝑎𝑥 ∆𝑧 = |𝑎|∆𝑥
∆𝑥
𝑧 = 𝑙𝑛(𝑥) ∆𝑧 =
𝑥
Logarithms 1 ∆𝑥
𝑧 = 𝑙𝑜𝑔(𝑥)
∆𝑧 =
ln(10) 𝑥
2. Division
An object travels a certain distance, d, with constant speed and you want to determine this speed, and its associated
uncertainty, from time and distance measurements.
You measure the time to be 𝑡 = 5.3 ± 0.2𝑠 and the distance 𝑑 = 2.355 ± 0.001𝑚.
𝑑 2.3550𝑚
𝑣= = ≈ 0.444 𝑚⁄𝑠
𝑡 5.3𝑠
𝑑 ∆𝑑 2 ∆𝑡 2 0.001 2 0.2 2
∆𝑣 = | | √( ) + ( ) = 0.44 𝑚⁄𝑠 √( ) + ( ) ≈ 0.016 𝑚⁄𝑠
𝑡 𝑑 𝑡 2.355 5.3
Considering the rules for rounding numbers, the final answer will be: 𝑣 = 0.44 ± 0.02 𝑚⁄𝑠