lOMoARcPSD|7564228

Signals and Waves - Part B

MATHEMATICS II: ENGINEERING MATHEMATICS (University of Surrey)

StuDocu is not sponsored or endorsed by any college or university

Downloaded by Cyber City ([email protected])
 lOMoARcPSD|7564228

Mathematics II: Engineering Maths (EEE1032)

Signals and Waves

Part B

Chris Russell
Centre for Vision, Speech and Signal Processing
BA Building, Room 12BA00
[email protected]

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Content
• The importance of signals

• Time and frequency representation Part A (≈ 4 hours)

• The decibel

• Signals and their classification

Part B (≈ 5-6 hours)
• Fourier series; time and frequency domains

• Fourier series – complex form

Part C (≈ 5 hours)
• Introduction to the Fourier transform

B-2

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Signals and their classification

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Periodic and aperiodic signals

• A periodic signal is one that repeats itself after a fixed interval. So for a
signal x(t) there is a value T>0 such that:
x(t+T) = x(t) for all t, -∞<t<∞

• The smallest value of T that satisfies the above is called the period of the
signal T
V T V

t t

• Any signal for which there is no value of T is called aperiodic or non-

periodic
V V

t t
B-4

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Periodic and aperiodic signals

• For example, V = sin ω t is periodic because:

  2 
sin    t    sin t  2  sin t
   

• And the period is:

2 2
1
T  
 2 f f

V T

B-5

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Power and energy

• We have already considered expressions for the instantaneous and
average power, based on the idea of a signal as a voltage dissipating
power in a 1Ω resistor: I
 /2
1
P (t )  x(t ) 2 Pave 
 
 /2
x(t ) 2 dt x(t) R=1Ω

• Since energy = total power × time, the energy dissipated in some

small time interval δt is δE=P(t)·δt and so integrating over a time τ
we get:
 / 2  / 2
E 
 /2
P (t ) dt  
 /2
x (t ) 2 dt

B-6

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Power and energy units

• If we have a signal x(t) representing a voltage, and its instantaneous

power is P(t)=x(t)2 then the units of x(t) are volts and by convention
the units of P(t) are volts squared (V2)

• If the signal is applied to some load R then the power dissipated in

the load is P(t)=x(t)2 / R and the units are watts

• Since energy is the integral of the power, the units of signal energy
are volts squared times seconds (V2s)

B-7

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Power signals and energy signals

• Now extend the time over which the integrals for power and energy
are taken to infinity

• If the average power remains finite and non-zero, the signal x(t) is
called a power signal
 /2
1
Pave  lim
   
 / 2
x(t ) 2 dt

• If the total energy remains finite and non-zero, the signal x(t) is
called an energy signal
 / 2
E  lim
  

 /2
x(t ) 2 dt

B-8

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Example 1
• For the “top hat” waveform shown, calculate the average power and
the total energy. Hence classify the signal as either a power signal
or an energy signal. x(t)
2
0 if   t  3

x(t )  2 if 3t  4 t=3 t=4 t

0 if 4t 


• Approach: for an aperiodic waveform, integrate between –τ/2 and

+τ/2, making sure the interesting part of the waveform is included.
Then increase τ to infinity to get the final result

B-9

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Example 1 (continued)
• Average power:
3 4  /2
1  1 4
  0dt   2 dt 
2
 0dt    0  4  0  
   /2 3 4   
4 x(t)
 Pave  lim  0 2
  
t
• Total energy:  / 2 t=3 t=4  /2
 /2
 3 4

  0dt   2 dt   0dt    0  4  0   4
2

  /2 3 4 
 E  lim  4   4
 

• Because 4/τ tends to zero as τ tends to infinity, the average power is zero, so
it is not a power signal.

• The total energy is 4 V2s ; hence this signal is an energy signal

B-10

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Example 2
• Calculate the average power and
total energy of the signal x(t) = cos 2πt V T=1s

• Approach: for a periodic waveform, perform the integral over n

cycles, for example from time t=0 to t=nT. Then increase n to
infinity to get the final result

B-11

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Example 2 (continued)
• Average power:
nT nT
1 1
  cos 2 t 
2

2
x (t ) dt  dt
nT 0
nT 0

• And, using the identity cos2 θ = ½(1+cos 2θ):

nT nT nT
1 1 1  1  nT 1
  cos 2 t   
2

nT 0
dt 
2nT 0 1  cos 4 t dt 
2nT  4 sin 4
t  t 
0

2nT 2
1 1 2
 Pave  lim  V
N  2 2

• Total energy:
nT nT nT
1 1 1  nT
  cos 2 t  0  
2
dt  1  cos 4 t dt   t  sin 4 t 
0
2 2  4 0 2
 nT 
 E  lim  
N 
 2 
• The total energy is infinite, and hence this is a power signal B-12

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.1: Power and energy signals

Consider the signal x(t) defined as follows
 0 if   t  1
x(t )   1t
e if 1 t  

(a) Sketch the function

(b) Calculate the total energy and the average power of the signal

(c) Classify the signal as either a power signal or an energy signal

B-13

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.1: Power and energy signals (solutions)

B-14

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Signal classification
• A power signal has finite non-zero average power and infinite total
energy
• An energy signal has finite non-zero total energy and zero average
power
• A signal with infinite power and infinite energy is neither a power
signal nor an energy signal
• A signal cannot be both a power signal and an energy signal

Power Energy Classification

Finite ∞ Power signal

0 Finite Energy signal
∞ ∞ Neither

B-15

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Signal classification
• Usually a periodic signal is a power signal (like in Example 2). It
lasts "forever" and therefore has infinite energy.
V T=1s

• Usually a non-periodic signal which has a finite duration (like in

Example 1) or an infinite duration but decreases rapidly enough
when (like in Exercise B.1) is an energy signal
x(t) x(t)
2
1

t=3 t=4 t 0 1 t

B-16

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

The importance of power & energy

Examples:
• A radar pulse transmitted to detect aircraft – how much energy is
reflected, and is it detectable?

• The energy transmitted per bit sent in a digital system affects the
bit error rate – for example in a satellite broadcast

• The average power in a signal sent over a communications link

needs to be sufficient for the signal to be received and detected
successfully (link power budget)

• In spread spectrum techniques, the signal power is spread over a

wide frequency range so that the signal seems to disappear among
electromagnetic noise (military applications)
B-17

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Random and deterministic signals

• A random signal is one whose behaviour cannot be predicted and

which cannot be described exactly by any mathematical function
– e.g. radio output when it is not tuned to a station, but is picking
up disturbances / noise

• A deterministic signal is one that can be described exactly by a

mathematical function

B-18

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Random signals: examples

Randomly
generated
signal:

Heartbeat:
Periodic but with
random components

B-19

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Deterministic signals: examples

Exponential function:

0 if   t  0
y (t )    t
e if 0t 

Step function:

0 if   t  0
y (t )  
1 if 0t 

B-20

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Recap: Equivalence of time and frequency domains

Fourier analysis (time to frequency domain)

Fourier synthesis (frequency to time domain)

A-22

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Signal spectra
• We can build up complicated periodic signals as a sum of a series of
sine waves. This is composed of a fundamental frequency and
harmonics:
– Fundamental frequency: 1 / the period of the signal (or, the
frequency of a sine wave that has the same period as the
waveform under consideration)
– Harmonic: a frequency that is an integral multiple of the
fundamental frequency

• Note that the fundamental is also called the first harmonic of the
waveform under consideration

• Let’s look at how we can build up a square wave, looking both at

the time domain and the frequency domain representations
B-23

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Square wave: fundamental

/s

x(t )  cost 

B-24

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Square wave: up to 3rd harmonic

/s
1
x(t )  cost   cos3t 
3

B-25

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Square wave: up to 5th harmonic

/s
1 1
x(t )  cost   cos3t   cos5t 
3 5

B-26

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Square wave: up to 7th harmonic

/s
1 1 1
x(t )  cost   cos3t   cos5t   cos7t 
3 5 7

B-27

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Square wave: up to 9th harmonic

/s
1 1 1 1
x(t )  cost   cos3t   cos5t   cos7t   cos9t 
3 5 7 9

B-28

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Square wave: up to 19th harmonic

/s

B-29

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

The square wave in the frequency domain

The spectrum shows the magnitude of each frequency component

/ Hz

B-30

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Triangle: fundamental

x(t )  cost 

B-31

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Triangle: up to 3rd harmonic

1
x(t )  cost   cos3t 
9

B-32

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Triangle: up to 5th harmonic

1 1
x(t )  cost   cos3t   cos5t 
9 25

B-33

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Triangle: up to 7th harmonic

1 1 1
x(t )  cost   cos3t   cos5t   cos7t 
9 25 49

B-34

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Triangle: up to 9th harmonic

1 1 1 1
x(t )  cost   cos3t   cos5t   cos7t   cos9t 
9 25 49 81

B-35

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Sawtooth: fundamental

x(t )   sin t 

B-36

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Sawtooth: up to 2nd harmonic

1
x(t )   sin t   sin 2t 
2

B-37

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Sawtooth: up to 3rd harmonic

1 1
x(t )   sin t   sin 2t   sin 3t 
2 3

B-38

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Sawtooth: up to 4th harmonic

1 1 1
x(t )   sin t   sin 2t   sin 3t   sin 4t 
2 3 4

B-39

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Sawtooth: up to 5th harmonic

1 1 1 1
x(t )   sin t   sin  2t   sin  3t   sin  4t   sin  5t 
2 3 4 5

B-40

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Sawtooth: up to 9th harmonic

1 1 1 1 1
x(t )   sin t   sin 2t   sin 3t   sin 4t   sin 5t   ...  sin 9t 
2 3 4 5 9

B-41

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Effect of harmonics
- The graph obtained by summing the terms progressively
approaches the shape of the function as the number of terms
increases
- Discontinuous functions (e.g. square waves or sawtooth) have
slowly decaying Fourier coefficients, thus the series converges
slowly. This tends to produce ripples which increase in number
and, apart from the one nearest to any discontinuity, decrease in
amplitude
- Near discontinuities, the series tends to overshoot on one side
and undershoot on the other as more terms are added. This
produces some ringing effect and is known as Gibbs'
phenomenon

B-42

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series (trigonometric form)

• Any periodic waveform f(t) of period T can be expressed as a d.c. value (a
constant) plus the sum of an infinite series of sine and cosine terms, where
each sine or cosine is an integral multiple of the frequency 1/T :
a0 
f (t )    an cos n0t  bn sin n0t 
2 n 1
a0
  a1 cos 0t  b1 sin 0t   a2 cos 20t  b2 sin 20t   ...
2
and ω0 = 2 π / T

• Terminology:
– This expression is called the Fourier series expansion of f(t)
– a0, an and bn are called the Fourier coefficients

• When we perform Fourier analysis of a periodic waveform, we are

concerned with calculating all the Fourier coefficients a0, an and bn. Now
let's see how we can calculate these coefficients.
B-43

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Formulae for the coefficients: a0

• Integrate both sides of the expression over one period (here I choose to
integrate from 0 to T, but you could equally well integrate from –T/2 to
+T/2)
a0 
f (t )    an cos n0t  bn sin n0t 
2 n 1
T T  T  T
a0
 f (t )dt  2  dt   an  cos n0t dt   bn  sin n0t dt
0 0 n 1 0 n 1 0

• The second and third terms have the value zero, since sine or cosine
integrated over a period (or n periods) are zero
(T = 2π/ω0 )
T
a a
  f (t )dt  0 t T0  0 T
2 2
0
• Hence: T
2
a0 
T  f (t )dt
0 B-44

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Formulae for the coefficients: an

• Multiply both sides by cos mω0t and integrate over one period (again, here I
choose 0 to T, but could use –T/2 to +T/2):
a0 
f (t )    an cos n0t  bn sin n0t 
2 n 1
T T  T  T
a0
 cos m0t dt   an  cos m0t cos n0t dt   bn  cos m0t sin n0t dt
2 
f (t ) cos m0tdt 
0 0 n 1 0 n 1 0

• Now, as we showed earlier (lecture notes Part A, “orthogonality”) if we

integrate over one or multiple periods:
– cos mω0t integrates to zero always
– cos mω0t . sin nω0t integrates to zero always
– cos mω0t . cos nω0t integrates to zero when m ≠ n
– And when m = n:
T T T T
1 1 1 T
 cos2
n  0 t dt   1  cos 2 n  0 t dt   2
dt  cos 2 n  0 t dt 
0
2 0
2 0 0
2
B-45

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Formulae for the coefficients: an (continued)

• Hence:
T
T
 f (t ) cos n0t dt  0  an
2
0
0

• And finally:

T
2
an 
T  f (t ) cos n0t dt
0

B-46

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Formulae for the coefficients: bn

• Similarly, we can show that (see Self Assessment Question Q.14):

T
2
bn 
T  f (t ) sin n0t dt
0

B-47

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Summary
A periodic waveform can be described as a sum of an infinite series of
sine and cosine waves:
a0 
f (t )    an cos n0t  bn sin n0t 
2 n 1

T
where:
2 2
a0 
T  f (t )dt 0 
T
0
T
2
an 
T  f (t ) cos n0t dt
0
T
2
bn 
T  f (t ) sin n0t dt
0
Note: the integral doesn’t have to be from 0 to T; it can be over any
B-48
whole period, for example –T/2 to +T/2.

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Summary (alternative notation)

An alternative notation, used in some textbooks, is where we substitute
ω0 = 2π / T and write:
a0   2nt 2nt 
f (t )     an cos  bn sin 
2 n 1 T T 

T
2
a0 
T  f (t )dt
0
T
2 2nt
an 
T  f (t ) cos
T
dt
0
T
2 2nt
bn 
T  f (t ) sin
T
dt
0
B-49

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series: Dirichlet conditions

• If the periodic waveform f(t) is such that:
– f(t) is single-valued (i.e. it has only one value v for any time
t)
– f(t) and its derivative df(t)/dt are continuous or have at
most a finite number of finite discontinuities over one
period (i.e. they are piece-wise continuous)
then f(t) can be represented as a Fourier series

• These conditions are known as the Dirichlet conditions (you

do not need to remember these conditions, just be aware that
they exist. In reality any periodic waveform you come across
in this course will have a Fourier series)

B-50

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Amplitude and phase spectra

B-51

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Amplitude and phase spectra: definition

• So far, when working in the frequency domain, we have only considered
the amplitude (i.e. magnitude) of the frequency components.

• However, we also need to consider the phase of each frequency component.

• A graph of the amplitude against angular frequency is called the amplitude

spectrum

• A graph of the phase against angular frequency is called the phase

spectrum

• We need to plot both the amplitude and phase spectra in order to fully
characterise a signal in the frequency domain

• Do not forget to include the d.c. component in these spectra

B-52

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Amplitude and phase spectra: example

V T= 0.1ms
+1
t
-1
• In the case of the square waveform shown above, we will show later that
the Fourier series expansion is:
4 1 1 1 1 
x(t )   cost   cos3t   cos5t   cos7t   cos9t   ...
 3 5 7 9 

• Noting that  cos nt  cosnt    , we can rewrite the expression as

follows in order to show the amplitude and phase of each frequency
component:

4 1 1 1 1 
x(t )   cost   cos3t     cos5t   cos7t     cos9t   ... 
 3 5 7 9 

B-53

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Amplitude and phase spectra: example

4 1 1 1 1 
x(t )   cost   cos3t     cos5t   cos7t     cos9t   ... 
 3 5 7 9 

• This gives the following amplitude and phase spectra:

Relative phase of Relative phase of

3rd harmonic 7th harmonic
ϕ3=π ϕ7=π

Relative phase
of fundamental Relative phase of
frequency ϕ1=0 5th harmonic
ϕ5=0

B-54

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.2: Amplitude and phase spectra

Consider a periodic waveform of period T=1 ms whose Fourier series
expansion is given by:
1 1 1 1
y (t )  1  cos t   sin  2t   cos  3t   cos  4t   cos  5t   ...
2 3 4 5
2
where  
T

(a) Rewrite the Fourier series expansion to express each frequency

component using a cosine function in the following form
cn cos(nt  n )

(b) Draw the magnitude and phase plots of the frequency spectrum up to and
including the 5th harmonic

B-55

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.2: Amplitude and phase spectra (solutions)

B-56

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Amplitude and phase spectra

• Some periodic waveforms have a Fourier series where both cosine and sine
terms are present at some frequencies
a0 
f (t )     an cos n0t  bn sin n0t 
2 n 1

• We can combine the sine and cosine terms for each harmonic to obtain a
compact form of the trigonometric series:
a0 
f (t )    cn cos(n0t  n )
2 n 1

where cn and n are the amplitude and phase of that harmonic

• Let's see how the amplitude and phase can be expressed as a function of the
an and bn coefficients

B-57

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Amplitude and phase spectra

• The coefficients must satisfy
an cos n0t  bn sin not  cn cos( n0t  n )

• Using the trigonometric identity cos  A  B   cos A cos B  sin A sin B

we have:
an cos n0t  bn sin not  cn cos(n ) cos( n0t )  cn sin(n ) sin(n0t )

• Therefore  an  cn cos(n )
 an2  bn2
bn  cn sin(n ) -bn
ϕn
2 2 an
• It follows that: cn  an  bn
 bn 
and if an ≠ 0: n  tan 1
 B-58
a
 n 
Downloaded by Cyber City ([email protected])
 lOMoARcPSD|7564228

Examples of Fourier series

calculations

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a square wave (1)

• To find the Fourier components of a waveform we have to evaluate
all the coefficients a0, an and bn.

• In practice, for many waveforms we find that not all coefficients

exist (i.e. they are zero).

• Find the Fourier series of the square wave with period T where:
 T T
 1 2
t
4

 T T
f (t )    1 t
 4 4
 T T
 1 4
t
2

f T
+1
t B-60
-1
Downloaded by Cyber City ([email protected])
 lOMoARcPSD|7564228

Fourier series of a square wave (2)

T / 2 (The integral is performed over any one
2
To find an: an  
T T /2
f (t ) cos n0t dt period of the waveform. Here I choose
to integrate from –T/2 to +T/2)

And, using the mathematical description of the square wave:

T T T
2 4
2 4
2 2
an   (1)  cos n0t dt   (1)  cos n0t dt   ( 1)  cos n0t dt
T T T T T T
2 4 4

2 T 2 T 2 T
an    sin n0t T 2 
4
sin n0t T 4 
4
  sin n0t T 24
Tn0 Tn0 Tn0

Noting that ω0 = 2 π / T so ω0 T = 2 π
and 2 2 1
 
Tn0 2n n
n0T n2
sin  sin  sin n  0
2 2
n0T n2 n B-61
sin  sin  sin
4 4 2
Downloaded by Cyber City ([email protected])
 lOMoARcPSD|7564228

Fourier series of a square wave (3)

We therefore get:

1  n   n n   n 
an    sin  0    sin  sin    0  sin 
n  2   2 2   2 

4 n
an  sin
n 2
y = sin θ
y
 4
 n n  1, 5, 9, 
 θ
an   0 n even
 4 π/2 π 3π/2 2π
 n  3, 7, 11,  1π/2 2π/2 3π/2 4π/2
 n

B-62

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a square wave (4)

4 n
To find a0 we can’t use a n  sin when n= 0, so use the formula for a0:
n 2
 T /2
2
a0  
T T / 2
f (t )dt

T T T
2 4
2 4
2 2
a0   (1)  dt   (1)  dt   (1)  dt
T T T T T T
2 4 4

2 T 2 T 4 2 T
a0   t T  t T   t T 2
4
T 2 T 4 T 4

2  T T   T T   T T  
a0           0
T  4 2   4 4   2 4  

The waveform has no dc component and hence a0  0

B-63

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a square wave (5)

T / 2
2
For bn: bn  
T T / 2
f (t ) sin n0t dt

T T T
2 4
2 4
2 2
bn   (1)  sin n0t dt   (1)  sin n0t dt   ( 1)  sin n0t dt
T T T T T T
2 4 4

2 T 2 T 2 T
bn  cos n0t T 42    cos n0t T 44   cos n0t T 24
Tn0 Tn0 Tn0

1  n   n n   n 
bn   cos  cos n     cos  cos 
  cos n  cos   0
n  2   2 2   2 

All the bn terms are zero.

B-64

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a square wave (6)

• Finally, we obtain the square wave’s Fourier series as:

4 1 1 1 
f (t )   cos 0t  cos 30t  cos 50t  cos 70t.... where ω0 = 2π/T
 3 5 7 

• Note that we can also write the Fourier series as:

4 
(1) n 1
f (t ) 


n 1  2n  1
cos(2n  1)0t

• Make sure you can see that these two expressions are equivalent!

B-65

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Some useful observations: sin nπ and cos nπ

By observation we see that:

y = sin θ
y

sin n  0 θ

π/2 π 3π/2 2π

y = cos θ
y

cos n  (1) n θ

π/2 π 3π/2 2π

B-66

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Some useful observations: sin nπ/2 and cos nπ/2

By observation we see that:

y = sin θ
y
n  ( 1) ( n 1)/ 2
(n odd)
sin 
2  0 (n even) θ

π/2 π 3π/2 2π
1π/2 2π/2 3π/2 4π/2

y = cos θ
y
n (1) n / 2 (n even)
cos 
2  0 (n odd) θ

π/2 π 3π/2 2π
1π/2 2π/2 3π/2 4π/2

B-67

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series shortcuts

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Even and odd functions (revision!)

• A function f ( t ) is said to be even if: f ( −t ) = f ( t )

An even function is symmetrical about the vertical axis t = 0

V T

• A function f ( t ) is said to be odd if: f ( −t ) = − f ( t )

An odd function is anti-symmetrical about the vertical axis t = 0

V T

t
B-69

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Even and odd functions: examples

• x(t)=t2 is even because x(−t)=(−t)2=t2=x(t)
x
t02

t 0 t0 t

• y(t)=t3 is odd because y(− t)=(−t)3= −t3= − y(t)

y

t03

t 0
t0 t

t03 B-70

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Even and odd functions: examples

• cos is even
y y = cos θ

-2π -π π 2π θ

• sin is odd

y y = sin θ

-2π -π π 2π θ

B-71

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series: useful short cuts

1. If the waveform is symmetrical about the axis t = 0 then it is

composed only of cosine terms (since cosine is symmetrical, or an
even function). So the an terms have to be calculated, but we can
see immediately that all bn terms are zero.

2. If the waveform is anti-symmetric about the axis t = 0 then it is

composed only of sine terms (sine is an odd function), and all the
an terms are zero – this includes the a0 term, which is also zero.

3. If the area of the waveform above the time axis ( y = 0 ) equals

the area below the axis, then the dc term is zero, i.e. a0 = 0.

B-72

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Example (1)
• Consider the signal shown. Without performing any detailed
calculations, what general comments can you make about the
frequency spectrum and Fourier series of the signal?

V (t)

–2m s –1m s 0 +1m s +2m s t

B-73

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Example (2): things to consider

• Is the waveform periodic or non-periodic? Hence does it have a

fundamental and harmonics?

• What is the period? Hence fundamental frequency?

• Consider the Fourier components:

– dc component a0 (compare waveform areas over and below the
time axis)
– cosine terms an (even or odd function?)
– sine terms bn (even or odd function?)

B-74

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Example (3): answer

V (t)

–2m s –1m s 0 +1m s +2m s t

• It’s a periodic waveform, so the spectrum consists of components at a fundamental

frequency and integral multiples of that frequency (the harmonics)

• The period is 1 ms, and frequency f = 1 / T so the fundamental frequency is 1 kHz

• The dc component a0 is non-zero, because the waveform areas above and below y =
0 are different (“useful short cut 3”)

• The waveform is symmetric about t = 0 and therefore an even function, so (“useful

short cut 1”):
– The an terms are in general non-zero
– The bn terms are all zero B-75

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Square waveform revisited using shortcuts

 T T
1 t f T

2 4 +1
 T T
f (t )    1 t
 4 4 t
 T T -1
1 4
t
2


• We calculated earlier that the Fourier series of the square waveform is

4 
( 1) n 1
f (t ) 


n 1  2n  1
cos(2n  1)0t

• Using the shortcuts we could have predicted that:

– The d.c. component a0 is zero, because the waveform areas above and
below y = 0 are equal
– The an terms are in general non-zero and the bn terms are all zero,
because the waveform is symmetric about t = 0 and therefore an even
function
B-76

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.3: Fourier series shortcuts

• For each waveform shown below, identify the coefficients a0, an and bn as
zero or non-zero using the previous shortcuts

(a) y(t)
(b)
y(t)

-T/2 T/2
0 t t
0
-T/2 T/2

(c) y(t) (d) y(t)

t -T 0 T t
-T/2 0 T/2

B-77

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.3: Fourier series shortcuts (solutions)

B-78

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Examples of Fourier series

calculations using shortcuts

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a sawtooth wave (1)

• Find the Fourier series of the sawtooth waveform with period T such
that:
a T T
f (t )  t,  t
T 2 2

f(t)
a/2
t
0
-a/2
-T/2 T/2

B-80

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a sawtooth wave (2)

f(t)
a/2
a T T
f (t )  t ,  t t
T 2 2 0
-a/2
-T/2 T/2

• Before doing any calculation, let us check whether any of the shortcuts can
be used to identify any zero coefficients (and save ourselves some
unnecessary calculations!)

• The function is odd because the waveform is anti-symmetric about t = 0.

Using the shortcuts we can conclude that all the an terms and a0 are zero.
We can also say that in general the bn terms are non-zero.

a0  0

an  0
B-81

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a sawtooth wave (3)

Let us now calculate bn , we have:
T / 2
2
T T/ 2
bn  f (t ) sin n0t dt with ω0 = 2 π / T

Using the mathematical description of the sawtooth waveform, we have

T / 2  T /2
2 a 2a
bn  
T T / 2 T
t sin n0t dt  2  t sin n0t dt
T T / 2
௕ ௗ௩ ௕ ௕ ௗ௨
Reminder on integration by parts: ௔ ௔ ௔
ௗ௧ ௗ௧

ௗ௩
Integrating by parts with and ଴ , we obtain:
ௗ௧
T / 2
2a 1 2a 1
t cos n0t T / 2  2
T / 2
bn   2
T n0 T n0 
T / 2
cos n0t dt

Using elementary integration for the remaining integral we obtain:

2a 1 2a 1
t cos n0t T / 2  2 2 2 sin n0t T /2
T / 2  T /2
bn   2
T n0 T n 0 B-82

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a sawtooth wave (4)

Noting that 0 T  2 T  
2 T 2
2a 1 T T  2a 1
bn    cos n  cos(  n )   2 2 2  sin n  sin( n ) 
T2 n0  2 2  T n 0
2a 1 2a 1
bn   2 T cos n  2 2 2 2sin n
T n0 T n 0
2a 1
bn   cos n
T n0

1 T
Noting that  and cos n  (1) n
0 2

2a T a
bn   (1) n   (1) n
T n 2 n

a
bn  (1) n 1
n
B-83

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Fourier series of a sawtooth wave (5)

• Finally, we obtain the sawtooth wave’s Fourier series as:
a 1 1 1 1 
f (t )  sin  t  sin 2 t  sin 3 t  sin 4 t  sin 5 t ....
  0
2
0
3
0
4
0
5
0 

where ω0 = 2π/T

• Or, equivalently, using the sigma notation:

a (1) n 1

f (t )   sin n0t where ω0 = 2π/T
 n 1 n

B-84

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.4: Fourier series calculation

Consider the function y(t) defined as follows:
y (t )  t 2 ,  1  t  1
y (t  2)  y (t ), for any t
(a) Sketch three cycles centred on the origin of the function y(t)

(b) What is its period T? What is the angular frequency ω0 of its fundamental
frequency?

(c) Calculate its Fourier series, i.e.:

– Calculate its a0 term
– Calculate its an terms
– Calculate its bn terms
– Write down the Fourier series.
Hint: Make sure you use the shortcuts to avoid calculating the zero terms B-85

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.4: Fourier series calculation (solutions)

B-86

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.4: Fourier series calculation (solutions)

B-87

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.4: Fourier series calculation (solutions)

B-88

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Application to communications:
signal and channel bandwidth

B-89

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Signal bandwidth
• The bandwidth of a signal that contains lots of different frequencies
is the difference between its highest and lowest frequency
components

• The units of bandwidth are hertz (Hz)

– Because the bandwidth is simply a difference between two
frequencies

• Example: if a telephone can transmit speech signals in the frequency

range 300 Hz to 3.4kHz, then the bandwidth of this signal is
( 3400 – 300 ) = 3100 Hz. (Note: we sometimes say loosely that the
upper frequency limit is 3.4kHz; we actually filter out the signal
above 3400 Hz, but filters can’t in practice stop a signal perfectly at
a single given frequency)
B-90

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Channel bandwidth
• In communications we want to transmit signals from one place to
another along a channel. The channel can itself only support a
range of frequencies, the exact details depending on the transmission
medium. The bandwidth that can be used also depends on how good
the transmitter and receiver are, but with current technologies you
can typically get:
– Twisted pair cable: 0 to 109 Hz (depends on cable length)
– Coaxial cable: 0 to 1010 Hz
– Fibre optic cable: 1014 to 1016 Hz (bandwidth = 1016−1014 = 9.9 x
1015 Hz)

Information Information
Transmitter Channel Receiver
source destination

B-91

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Channel bandwidth (2)

• Optical fibres have one of the highest bandwidths available (and can
support these high bandwidths over very long distances), and are
consequently used for high data volumes, for example transmitting
data through the “core” of the Internet

• The bandwidth of the channel also affects the information carrying

capacity of the channel – in essence, the greater the bandwidth the
more information we can carry
– The information carrying capacity also depends on the amount of
noise that is present – but that’s a subject for another lecture
course!

B-92

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Question
What happens if I have a square wave of frequency f0=10kHz and I try
to transmit it along a cable with a channel bandwidth of 55 kHz (dc to
55kHz) ?
V T0= 0.1ms
+π/4
Signal:
t
-π/4

Spectrum:

B-93
Bandwidth of channel

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Half-range series

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Half range series

• Sometimes a given function is not periodic but is only defined over a finite interval
[0, T/2]. For example: V

t
T/2
• We can define the function arbitrarily outside this interval. In particular, we can
choose to define it such that it is periodic with period T. For example:
V

-T -T/2 T/2 T t

• This is called a periodic extension; depending on how we chose to define the

function outside of the interval, we obtain different types of period extensions

• Now that the extended function is periodic, we can calculate its Fourier series. The
series obtained will represent the original function only over the interval [0, T/2]
since we do not know the original function outside this interval. For this reason, it is
called a half-range Fourier series B-95

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Half-range cosine series

• Take a function f(t) defined over a half period, i.e. from t = 0 to t = T / 2
V

t
T/2

• If we extend the waveform by reflecting it in the vertical axis and repeating

it periodically, we obtain a periodic even function
V

t
-T -T/2 T/2 T

• We know from the useful shortcuts that the function will contain only
cosine terms (and possibly a d.c. component). This is called a half-range
cosine series.
T /2 T /2
4 4
a0 
T  f (t )dt an 
T  f (t ) cos n0t dt bn  0
B-96
0 0

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Half-range sine series

• Take a function f(t) defined over a half period, i.e. from t = 0 to t = T / 2
V

t
T/2

• If we extend the waveform by reflecting it in both the vertical axis and

horizontal axis, and repeating it periodically, we obtain a periodic odd
function V

t
-T -T/2 T/2 T

• We know from the useful shortcuts that the function will contain only sine
terms (and no d.c. component). This is called a half-range sine series.
T /2
4
a0  a n  0 bn 
T  f (t ) sin n0t dt B-97
0

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Half-range series
• Note how in the case of the cosine half-range series and the sine half-range
series, integration is performed over the interval [0, T/2] and how the
coefficient 2/T has been replaced by 4/T

• These expressions can also be useful to calculate the Fourier coefficients of

periodic functions which are odd or even. More specifically:
– If a periodic waveform is even, we can calculate the a0 and an
coefficients using the half-range series expressions:
T /2 T /2
4 4
a0 
T  f (t )dt an 
T  f (t ) cos n0t dt
0 0

– If a periodic waveform is odd, we can calculate the bn coefficients using

the half-range series expression:
T /2
4
bn 
T  f (t ) sin n0t dt
0 B-98

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.5: Half range series

Given the function defined over the half range 0 ≤ t < 2 by
f (t )  2  t 0t 2

(a) Draw the function over the interval [0,2] over which it is defined

(b) Draw an appropriate periodic extension of the waveform suitable

for representation using a half-range sine series

(c) Calculate the waveform's half range sine series

B-99

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.5: Half range series (solutions)

B-100

Downloaded by Cyber City ([email protected])

 lOMoARcPSD|7564228

Exercise B.5: Half range series (solutions)

B-101

Downloaded by Cyber City ([email protected])