15 A white dwarf star has a mass about that of the 20 A block of mass 1.

2 kg is raised a vertical distance

Sun and a radius about that of the Earth. Give an of 5.55 m in 2.450 s. Calculate the power
order-of-magnitude estimate of the density of a mgh
delivered. (P = and g = 9.81 m s−2 )
white dwarf. t
21 Find the kinetic energy (EK = 12mv2 ) of a block of
16 A sports car accelerates from rest to 100 km per
mass 5.00 kg moving at a speed of 12. 5 m s−1.
hour in 4.0 s. What fraction of the acceleration
22 Without using a calculator, estimate the value
due to gravity is the car’s acceleration?
of the following expressions. Then compare
17 Give an order-of-magnitude estimate for the
your estimate with the exact value found using a
number of electrons in your body.
calculator.
18 Give an order-of-magnitude estimate for the
243
ratio of the electric force between two electrons a
43
1 m apart to the gravitational force between the
electrons. b 2.80 × 1.90
19 The frequency f of oscillation (a quantity with 480
c 312 ×
units of inverse seconds) of a mass m attached 160
to a spring of spring constant k (a quantity with 8.99 × 109 × 7 × 10−16 × 7 × 10−6
d
units of force per length) is related to m and k. (8 × 102 )2
By writing f = cmxk y and matching units 6.6 × 10−11 × 6 × 1024
k e
on both sides, show that f = c , where c is a (6.4 × 106)2
m
dimensionless constant.

1.2 Uncertainties and errors Learning objectives

This section introduces the basic methods of dealing with experimental • Distinguish between random
error and uncertainty in measured physical quantities. Physics is an and systematic uncertainties.
experimental science and often the experimenter will perform an • Work with absolute, fractional
experiment to test the prediction of a given theory. No measurement will and percentage uncertainties.
ever be completely accurate, however, and so the result of the experiment • Use error bars in graphs.
will be presented with an experimental error. • Calculate the uncertainty in a
gradient or an intercept.
Types of uncertainty
There are two main types of uncertainty or error in a measurement. They
can be grouped into systematic and random, although in many cases
it is not possible to distinguish clearly between the two. We may say that
random uncertainties are almost always the fault of the observer, whereas
systematic errors are due to both the observer and the instrument being
used. In practice, all uncertainties are a combination of the two.

Systematic errors
A systematic error biases measurements in the same direction; the
measurements are always too large or too small. If you use a metal ruler
to measure length on a very hot day, all your length measurements will be
too small because the metre ruler expanded in the hot weather. If you use
an ammeter that shows a current of 0.1 A even before it is connected to

1 MEASUREMENT AND UNCERTAINTIES 7

 a circuit, every measurement of current made with this ammeter will be
larger than the true value of the current by 0.1 A.
Suppose you are investigating Newton’s second law by measuring the
acceleration of a cart as it is being pulled by a falling weight of mass m
(Figure 1.1). Almost certainly there is a frictional force f between the cart
and the table surface. If you forget to take this force into account, you
would expect the cart’s acceleration a to be:
m
mg
a=
M
Figure 1.1 The falling block accelerates the where M is the constant combined mass of the cart and the falling block.
cart. The graph of the acceleration versus m would be a straight line through
the origin, as shown by the red line in Figure 1.2. If you actually do the
experiment, you will find that you do get a straight line, but not through
the origin (blue line in Figure 1.2). There is a negative intercept on the
vertical axis.
a /m s–2 2.0

1.5

1.0

0.5

0.0
0.1 0.2 0.3 0.4
m / kg
–0.5

–1.0
Figure 1.2 The variation of acceleration with falling mass with (blue) and without
(red) frictional forces.

This is because with the frictional force present, Newton’s second law
predicts that:
mg f
a= −
M M
So a graph of acceleration a versus mass m would give a straight line with
a negative intercept on the vertical axis.
Systematic errors can result from the technique used to make a
measurement. There will be a systematic error in measuring the volume
of a liquid inside a graduated cylinder if the tube is not exactly vertical.
The measured values will always be larger or smaller than the true value,
depending on which side of the cylinder you look at (Figure 1.3a). There
will also be a systematic error if your eyes are not aligned with the liquid
level in the cylinder (Figure 1.3b). Similarly, a systematic error will arise if
you do not look at an analogue meter directly from above (Figure 1.3c).
Systematic errors are hard to detect and take into account.

8
a b c
Figure 1.3 Parallax errors in measurements.

Random uncertainties
The presence of random uncertainty is revealed when repeated
measurements of the same quantity show a spread of values, some too large
some too small. Unlike systematic errors, which are always biased to be in
the same direction, random uncertainties are unbiased. Suppose you ask ten
people to use stopwatches to measure the time it takes an athlete to run a
distance of 100 m. They stand by the finish line and start their stopwatches
when the starting pistol fires.You will most likely get ten different values
for the time. This is because some people will start/stop the stopwatches
too early and some too late.You would expect that if you took an average
of the ten times you would get a better estimate for the time than any
of the individual measurements: the measurements fluctuate about some 21 22 23 24 25 26 27
value. Averaging a large number of measurements gives a more accurate
estimate of the result. (See the section on accuracy and precision, overleaf.)
We include within random uncertainties, reading uncertainties (which
really is a different type of error altogether). These have to do with the
precision with which we can read an instrument. Suppose we use a ruler
to record the position of the right end of an object, Figure 1.4. 30 31 32 33 34 35 36

The first ruler has graduations separated by 0.2 cm. We are confident
that the position of the right end is greater than 23.2 cm and smaller
than 23.4 cm. The true value is somewhere between these bounds. The Figure 1.4 Two rulers with different
average of the lower and upper bounds is 23.3 cm and so we quote the graduations. The top has a width between
measurement as (23.3 ± 0.1) cm. Notice that the uncertainty of ± 0.1 cm graduations of 0.2 cm and the other 0.1 cm.
is half the smallest width on the ruler. This is the conservative way
of doing things and not everyone agrees with this. What if you scanned
the diagram in Figure 1.4 on your computer, enlarged it and used your
computer to draw further lines in between the graduations of the ruler.
Then you could certainly read the position to better precision than
the ± 0.1 cm. Others might claim that they can do this even without a
computer or a scanner! They might say that the right end is definitely
short of the 23.3 cm point. We will not discuss this any further – it is an
endless discussion and, at this level, pointless.
Now let us use a ruler with a finer scale. We are again confident that the
position of the right end is greater than 32.3 cm and smaller than 32.4 cm.
The true value is somewhere between these bounds. The average of the
bounds is 32.35 cm so we quote a measurement of (32.35 ± 0.05) cm. Notice

1 MEASUREMENT AND UNCERTAINTIES 9

 again that the uncertainty of ± 0.05 cm is half the smallest width on the
ruler. This gives the general rule for analogue instruments:

The uncertainty in reading an instrument is ± half of the smallest

width of the graduations on the instrument.

For digital instruments, we may take the reading error to be the smallest
Instrument Reading error
division that the instrument can read. So a stopwatch that reads time to
ruler ± 0.5 mm two decimal places, e.g. 25.38 s, will have a reading error of ± 0.01 s, and a
vernier calipers ± 0.05 mm weighing scale that records a mass as 184.5 g will have a reading error of
micrometer ± 0.005 mm ± 0.1 g. Typical reading errors for some common instruments are listed in
electronic weighing ± 0.1 g
Table 1.6.
scale
stopwatch ± 0.01 s
Accuracy and precision
In physics, a measurement is said to be accurate if the systematic error
Table 1.6 Reading errors for some common
instruments. in the measurement is small. This means in practice that the measured
value is very close to the accepted value for that quantity (assuming that
this is known – it is not always). A measurement is said to be precise
if the random uncertainty is small. This means in practice that when
the measurement was repeated many times, the individual values were
close to each other. We normally illustrate the concepts of accuracy and
precision with the diagrams in Figure 1.5: the red stars indicate individual
measurements. The ‘true’ value is represented by the common centre
of the three circles, the ‘bull’s-eye’. Measurements are precise if they are
clustered together. They are accurate if they are close to the centre. The
descriptions of three of the diagrams are obvious; the bottom right clearly
shows results that are not precise because they are not clustered together.
But they are accurate because their average value is roughly in the centre.

not accurate and not precise accurate and precise

not accurate and not precise accurate and precise

not accurate but precise accurate but not precise

not accurate but precise accurate but not precise
Figure 1.5 The meaning of accurate and precise measurements. Four different sets of
four measurements each are shown.

10
Averages
In an experiment a measurement must be repeated many times, if at all
possible. If it is repeated N times and the results of the measurements are
x1, x2, …, xN, we calculate the mean or the average of these values (x– )
using:
x + x + … + xN
x– = 1 2
N
This average is the best estimate for the quantity x based on the N
measurements. What about the uncertainty? The best way is to get the
standard deviation of the N numbers using your calculator. Standard
deviation will not be examined but you may need to use it for your
Internal Assessment, so it is good idea to learn it – you will learn it
in your mathematics class anyway. The standard deviation σ of the N
measurements is given by the formula (the calculator finds this very
easily):

(x1 – x– )2 + (x2 – x– )2 + … + (xN – x– )2

σ=
N–1

A very simple rule (not entirely satisfactory but acceptable for this course)
is to use as an estimate of the uncertainty the quantity:
xmax − xmin
∆x =
2

i.e. half of the difference between the largest and the smallest value.
For example, suppose we measure the period of a pendulum (in
seconds) ten times:

1.20, 1.25, 1.30, 1.13, 1.25, 1.17, 1.41, 1.32, 1.29, 1.30

We calculate the mean:

t + t + … + t10
t– = 1 2 = 1.2620 s
10
and the uncertainty:
tmax − tmin 1.41 − 1.13
∆t = = = 0.140 s
2 2 Exam tip
How many significant figures do we use for uncertainties? The general There is some case to be made
rule is just one figure. So here we have ∆t = 0.1 s. The uncertainty is in the for using two significant figures
first decimal place. The value of the average period must also be in the uncertainty when the
expressed to the same precision as the uncertainty, i.e. here to one first digit in the uncertainty
decimal place, t– = 1.3 s. We then state that: is 1. So in this example,
since ∆t = 0.140 s does begin
period = (1.3 ± 0.1) s with the digit 1, we should
state ∆t = 0.14 s and quote
(Notice that each of the ten measurements of the period is subject to a
the result for the period as
reading error. Since these values were given to two decimal places, it is
‘period = (1.26 ± 0.14) s’.
implied that the reading error is in the second decimal place, say ± 0.01 s.

1 MEASUREMENT AND UNCERTAINTIES 11

 This is much smaller than the uncertainty found above so we ignore the
reading error here. If instead the reading error were greater than the error
due to the spread of values, we would have to include it instead. We will
not deal with cases when the two errors are comparable.)
You will often see uncertainties with 2 s.f. in the scientific
literature. For example, the charge of the electron is quoted as
e = (1.602 176 565 ± 0.000 000 035) × 10−19 C and the mass of the electron
as me = (9.109 382 91 ± 0.000 000 40) × 10−31 kg. This is perfectly all right
and reflects the experimenter’s level of confidence in his/her results.
Expressing the uncertainty to 2 s.f. implies a more sophisticated statistical
analysis of the data than is normally done in a high school physics course.
With a lot of data, the measured values of e form a normal distribution
with a given mean (1.602 176 565 × 10−19 C) and standard deviation
(0.000 000 035 × 10−19 C). The experimenter is then 68% confident that
the measured value of e lies within the interval [1.602 176 530 × 10−19 C,
1.602 176 600 × 10−19 C].

Worked example
1.6 The diameter of a steel ball is to be measured using a micrometer caliper. The following are sources of error:
1 The ball is not centred between the jaws of the caliper.
2 The jaws of the caliper are tightened too much.
3 The temperature of the ball may change during the measurement.
4 The ball may not be perfectly round.
Determine which of these are random and which are systematic sources of error.

Sources 3 and 4 lead to unpredictable results, so they are random errors. Source 2 means that the measurement of
diameter is always smaller since the calipers are tightened too much, so this is a systematic source of error. Source 1
certainly leads to unpredictable results depending on how the ball is centred, so it is a random source of error. But
since the ball is not centred the ‘diameter’ measured is always smaller than the true diameter, so this is also a source
of systematic error.

Propagation of uncertainties
A measurement of a length may be quoted as L = (28.3 ± 0.4) cm. The value
28.3 is called the best estimate or the mean value of the measurement
and the 0.4 cm is called the absolute uncertainty in the measurement.
The ratio of absolute uncertainty to mean value is called the fractional
uncertainty. Multiplying the fractional uncertainty by 100% gives the
percentage uncertainty. So, for L = (28.3 ± 0.4) cm we have that:
• absolute uncertainty = 0.4 cm
0.4
• fractional uncertainty = 28.3 = 0.0141
• percentage uncertainty = 0.0141 × 100% = 1.41%

12
In general, if a = a0 ± ∆a, we have:
The subscript 0 indicates the mean
• absolute uncertainty = ∆a
∆a value, so a0 is the mean value of a.
• fractional uncertainty = a0
∆a
• percentage uncertainty = a0 × 100%
Suppose that three quantities are measured in an experiment: a = a0 ± ∆a,
b = b0 ± ∆b, c = c0 ± ∆c. We now wish to calculate a quantity Q in terms of
a, b, c. For example, if a, b, c are the sides of a rectangular block we may
want to find Q = ab, which is the area of the base, or Q = 2a + 2b, which
is the perimeter of the base, or Q = abc, which is the volume of the block.
Because of the uncertainties in a, b, c there will be an uncertainty in the
calculated quantities as well. How do we calculate this uncertainty?
There are three cases to consider. We will give the results without proof.

Addition and subtraction

The first case involves the operations of addition and/or subtraction. For
example, we might have Q = a + b or Q = a − b or Q = a + b − c. Then,
in all cases the absolute uncertainty in Q is the sum of the absolute
uncertainties in a, b and c.
Exam tip
Q=a+b ⇒ ∆Q = ∆a + ∆b In addition and subtraction,
Q=a−b ⇒ ∆Q = ∆a + ∆b we always add the absolute
Q=a+b−c ⇒ ∆Q = ∆a + ∆b + ∆c uncertainties, never subtract.

Worked examples
1.7 The side a of a square, is measured to be (12.4 ± 0.1) cm. Find the perimeter P of the square including the
uncertainty.

Because P = a + a + a + a, the perimeter is 49.6 cm. The absolute uncertainty in P is:

∆P = ∆a + ∆a + ∆a + ∆a

∆P = 4∆a

∆P = 0.4 cm

Thus, P = (49.6 ± 0.4) cm.

1.8 Find the percentage uncertainty in the quantity Q = a − b, where a = 538.7 ± 0.3 and b = 537.3 ± 0.5. Comment
on the answer.

The calculated value is 1.7 and the absolute uncertainty is 0.3 + 0.5 = 0.8. So Q = 1.4 ± 0.8.
0.8
The fractional uncertainty is = 0.57, so the percentage uncertainty is 57%.
1.4
The fractional uncertainty in the quantities a and b is quite small. But the numbers are close to each other so their
difference is very small. This makes the fractional uncertainty in the difference unacceptably large.

1 MEASUREMENT AND UNCERTAINTIES 13

 Multiplication and division
The second case involves the operations of multiplication and division.
Here the fractional uncertainty of the result is the sum of the
fractional uncertainties of the quantities involved:
∆Q ∆a ∆b
Q = ab ⇒ = +
Q0 a0 b0
a ∆Q ∆a ∆b
Q= ⇒ = +
b Q0 a0 b0
ab ∆Q ∆a ∆b ∆c
Q= ⇒ = + +
c Q0 a0 b0 c0

Powers and roots

The third case involves calculations where quantities are raised to powers
or roots. Here the fractional uncertainty of the result is the fractional
uncertainty of the quantity multiplied by the absolute value of the
power:
∆Q ∆a
Q = an ⇒ = |n|
Q0 a0
n ∆Q 1 ∆a
Q = √a ⇒ =
Q0 n a0

Worked examples
1.9 The sides of a rectangle are measured to be a = 2.5 cm ± 0.1 cm and b = 5.0 cm ± 0.1 cm. Find the area A of the
rectangle.

The fractional uncertainty in a is:

∆a 0.1
= = 0.04 or 4%
a 2.5

The fractional uncertainty in b is:

∆b 0.1
= = 0.02 or 2%
b 5.0

Thus, the fractional uncertainty in the area is 0.04 + 0.02 = 0.06 or 6%.
The area A0 is:
A0 = 2.5 × 5.0 = 12.5 cm2
∆A
and = 0.06
A0

⇒ ∆A = 0.06 ×12.5 = 0.75 cm2

Hence A = 12.5 cm2 ± 0.8 cm2 (the final absolute uncertainty is quoted to 1 s.f.).

14
1.10 A mass is measured to be m = 4.4 ± 0.2 kg and its speed v is measured to be 18 ± 2 m s−1. Find the kinetic
energy of the mass.

The kinetic energy is E = 12mv2, so the mean value of the kinetic energy, E0, is:
E0 = 12 × 4.4 × 182 = 712.8 J
Using:

∆E ∆m 2× ∆v
= +
E0 m0 v0
because of
the square
we find:
∆E0.2 2
= 2 × = 0.267
=
712.8 4.4 18

So:
∆E = 712.8 × 0.2677 = 190.8 J
Exam tip
To one significant figure, the uncertainty The final absolute uncertainty must be expressed to one
is ∆E = 200 = 2 × 102 J; that is E = (7 ± 2) × 102 J. significant figure. This limits the precision of the quoted
value for energy.

1.11 The length of a simple pendulum is increased by 4%. What is the fractional increase in the pendulum’s
period?

L
The period T is related to the length L through T = 2π .
g
Because this relationship has a square root, the fractional uncertainties are related by:

∆T 1 × ∆L
=
T0 2 L0
because of the
square root

∆L
We are told that = 4%. This means we have :
L0
∆T 1
= × 4% = 2%
T0 2

1 MEASUREMENT AND UNCERTAINTIES 15

 1
1.12 A quantity Q is measured to be Q = 3.4 ± 0.5. Calculate the uncertainty in a and b Q2.
Q

1 1
a = = 0.294 118
Q 3.4

∆(1/Q) ∆Q
=
1/Q Q

∆Q 0.5
⇒ ∆(1/Q) = 2 = = 0.043 25
Q 3.42
1
Hence: = 0.29 ± 0.04
Q

b Q2 = 3.42 = 11.5600

∆(Q2 ) ∆Q
=2×
Q2 Q

⇒ ∆(Q2 ) = 2Q × ∆Q = 2 × 3.4 × 0.5 = 3.4

Hence: Q2 = 12 ± 3

1.13 The volume of a cylinder of base radius r and height h is given by V = πr2h. The volume is measured with an
uncertainty of 4% and the height with with an uncertainty of 2%. Determine the uncertainty in the radius.

V
We must first solve for the radius to get r = . The uncertainty is then:
πh
∆r
r
× 100% =
2 V (
1 ∆V ∆h
+
h
1
)
× 100% = (4 + 2) × 100% = 3%
2

Best-fit lines
In mathematics, plotting a point on a set of axes is straightforward. In
physics, it is slightly more involved because the point consists of measured
or calculated values and so is subject to uncertainty. So the point
(x0 ± ∆x, y0 ± ∆y) is plotted as shown in Figure 1.6. The uncertainties are

2∆ x
y0 + ∆ y

y0 2∆ y

y0 – ∆ y

0 x0 – ∆ x x0 x0 + ∆ x x
Figure 1.6 A point plotted along with its error bars.

16