Uncertainties and errors
In a minute
Every measurement has an . You record it, carry it through calculations and quote results to sensible . and affect and differently, and shows whether a result agrees with an accepted value.
You'll be able to
- Tell random from systematic errors, and say how each affects precision and accuracy.
- Round values to a sensible number of significant figures or decimal places.
- Record a reading with its , such as A.
- Find and and percentage error, and judge whether a result agrees with an accepted value.
- Work out and interpret and .
- through sums, differences, products, quotients and powers.
Before you start3 quick questions on what this builds on
Random and systematic errors
Time the same event five times and you get five slightly different readings. Errors come in two kinds.
Random errors come from things you can't control exactly, like reaction time. Systematic errors come from the instrument or method, like a zero error or heat lost in every trial.
Try it first, then reveal one step at a time.
- Marking point 1: the readings agree within : precise.
- Marking point 2: every reading is high: systematic, so not accurate.
- Marking point 3: subtract the zero reading: .
- The readings agree within 0.1 N, so they are precise. The zero error adds 0.3 N to every reading: a systematic error, so they are not accurate. Repeating does not reduce it. Subtract 0.3 N from each reading, or reset the meter to zero, giving about 4.3 N.
Try it first, then reveal one step at a time.
- Marking point 1: varying reaction time is a random error.
- Marking point 2: averaging repeats lets the errors partly cancel.
- Marking point 3: light gates remove reaction time.
- Reaction time in starting and stopping the watch varies between trials. This random error makes the readings scatter above and below the true time. More repeats and a mean help, because random errors partly cancel. Light gates would remove reaction time altogether.
Significant figures
says more than : the final zero shows the length was measured to the nearest centimetre. Digits that carry this information are .
Non-zero digits always count, and so do zeros between them: has three. Leading zeros never count: has two.
Trailing zeros count after a decimal point: has three. In a whole number they are unclear, so write or , not .
Round only the final answer, never along the way.
Try it first, then reveal one step at a time.
- .
- A division: has three significant figures and has two, so give two.
Try it first, then reveal one step at a time.
- In series, resistances add: .
- An addition: has one decimal place and has two, so give one.
Recording a measurement with its uncertainty
No instrument gives the exact true value, so every reading comes with an .
Write value absolute uncertainty, with the unit: A, or . The current lies between and .
1. Analogue scale: half the smallest division, unless the reading is harder to judge.
2. Digital display: one unit in the last digit. A balance showing has an uncertainty of .
3. Repeated readings: half the range, , if that is bigger than the instrument's uncertainty.
Try it first, then reveal one step at a time.
- Analogue scale, so take half the smallest division: .
- The uncertainty is in the second decimal place, and so is the reading, so no rounding is needed.
Try it first, then reveal one step at a time.
- Mean: .
- Half the range: . It is bigger than the stopwatch's , so use it.
- Round the mean to the second decimal place, like the uncertainty.
Percentage uncertainty and percentage error
is tiny on a rod but huge on a bead. To compare uncertainties, divide by the value.
is the absolute uncertainty in . Fractional and percentage uncertainties have no unit; give them to one or two significant figures.
Try it first, then reveal one step at a time.
- Fractional: , with no unit.
- Percentage: , to two significant figures.
For the same absolute uncertainty, small readings have large percentage uncertainties. To shrink one, measure something bigger: time ten swings instead of one, or measure a stack of 50 sheets and divide by 50.
Try it first, then reveal one step at a time.
- Marking point 1: the range is to , and lies outside it.
- Marking point 2: percentage error ; percentage uncertainty .
- Marking point 3: so a systematic error is likely; air resistance would make every value too small.
- The range is 9.1 to 9.7 m s⁻² and the accepted value lies outside it, so the result does not agree. Random errors alone cannot explain the gap, so a systematic error, such as air resistance, is likely. (The percentage error, 4.2%, is also clearly larger than the percentage uncertainty, 3.2%.)
Percentage change and percentage difference
Two more percentages compare values, not uncertainties. Which one you use depends on whether one value came first.
A rise gives a positive percentage change and a fall a negative one. Keep the sign.
Try it first, then reveal one step at a time.
- Change: . It is a fall, so it is negative.
- Divide by the original: . The change has two significant figures, so give two.
- The resistance fell by just under a third.
Use percentage difference when neither value is the right one or the starting one, such as two groups' results. With an accepted value, use percentage error.
Try it first, then reveal one step at a time.
- Difference ; mean .
- . The difference has one significant figure, so give one.
- Whether 4% matters depends on each group's percentage uncertainty.
Propagating uncertainties
Most results are calculated from several measurements, each with its own uncertainty. combines them into the uncertainty of the result.
Three rules cover sums, differences, products, quotients and powers; combine them for longer formulas. They are printed in the data booklet under Uncertainties, so the skill is choosing the right one.
1. Adding or subtracting: add the absolute uncertainties, even for a subtraction. They never cancel.
2. Multiplying or dividing: add the fractional (or percentage) uncertainties, then multiply by for the absolute uncertainty.
3. Raising to a power: multiply the fractional uncertainty by the size of the power. A square root, power , halves it.
Exact numbers, such as the in or the and in , have no uncertainty.
Try it first, then reveal one step at a time.
- Value: .
- Dividing, so add fractional uncertainties:
- , to one significant figure.
- That is the first decimal place, so give as . With an uncertainty, it sets the decimal places, not the significant-figure rule.
Try it first, then reveal one step at a time.
- Value: .
- Fractional uncertainties: mass ; speed , doubled because is squared: . The adds nothing.
- Total , so .
- Round the value to the first decimal place: .
Common mistakes
- Writing 'human error' or 'to reduce error'. Name the error as random or systematic, give its cause, and say what your change does to it.
- Quoting too many digits, such as . Round the uncertainty to one significant figure, then match the value: .
- Subtracting uncertainties when values are subtracted. In a sum or difference, absolute uncertainties always add.
- Forgetting the power. In the fractional uncertainty in counts twice; an exact or adds nothing.
- Dividing by your own result in a percentage error. Divide by the accepted value.
- Calling a result correct because its percentage error is small. It agrees only if the accepted value lies inside its range.
Exam tips
- 'Determine the absolute uncertainty' needs working: the fractional uncertainties, their sum, then multiplying by the value.
- To 'state' a result to the appropriate precision, give the uncertainty to one significant figure, match the value's decimal places, and add the unit.
- To 'comment on' agreement with an accepted value, give the range, say whether the accepted value is inside it, and if not name a likely systematic error. Paper 1B data-based questions test this.
Key takeaways
Practise this
37 past-paper questions on Uncertainty, error and significant figures in the Physics HL bank, each with a mark scheme and a worked explanation.