EENG22000 Communications EENG22000 Communications

University of Bristol University of Bristol

Part 2: Part 2.1:

‘Basic’ Background
(Analog and Binary)
Bandpass Communications

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 1 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 2

Orthogonality (1) Orthogonality (2)

University of Bristol University of Bristol

• Two signals ai(t) and aj(t) are said to be TB

Area under product zero over
completed bit period
orthogonal if:
TS 
  ai t .a j t .dt  0 sin(t)

 0 i j
cos(t)
• Examples of orthogonal signals are:
t
ai(t) = cos(ωt) and aj(t) = sin(ωt)
ai(t) = cos(ωit) and aj(t) = cos(ωjt),
where i j
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 3 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 4

1
 Orthogonality (3) Bandpass and Baseband Signalling
University of Bristol University of Bristol

• With baseband signalling the channel is assumed to

T
B
extend from 0Hz upwards.
• Transmitting this type of data over conventional
media, such as radio channels, requires a shift in
frequency.
10 kHz t – This process is called ‘carrier modulation’ or just
‘modulation’.
• Often the process of modulation takes a band of
signals based on zero Hz and shifts them to occupy
twice the bandwidth centred on a carrier frequency.
Area under product zero
over completed bit
15 kHz • A corresponding ‘demodulation’ or ‘detection’ process
period is required to recover the data.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 5 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 6

Bandpass Signal Spectra (1) Bandpass Signal Spectra (2)

University of Bristol University of Bristol

• We need to consider what form a bandpass signal will take. • From trig identities, we know:
• One thing we need to consider is its spectrum. cos(ct).cos(mt)
• We will see later that we can often achieve carrier modulation = 0.5cos(c- m)t +0.5cos(c+m)t
through a multiplication process.
A(t)
• We can use simple trig identities to analyse such signals.
0.5 c-m c+m
• We can start with a sinusoidal message for simplicity and
then extend to a more realistic case.
• In block diagrams we show multiplication as a ‘mixer’

Cos(mt) c 
• Non sinusoidal messages will have more complex baseband
Cos(ct)
spectra – we can consider these to consist of lots of
sinusoids of different frequencies…
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 7 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 8

2
 Spectrum of a Carrier Modulated
Spectrum of a Baseband Square Wave (Bandpass) Square Wave
University of Bristol University of Bristol

• It can be shown by fourier • Hence a carrier

analysis that the spectrum of a modulated square
(baseband) square wave signal is wave has the spectrum Baseband Square
described by a sinc or sin(x)/x shown below: Wave Message
function: G(f) spectrum
 sin  . f .T  
G  f   T . A. 
  . f .T  G(f) frequency
• This spectrum extends to infinity. Carrier Modulated
We have already seen that it will be ASK Square Wave
necessary to limit the bandwidth. We frequency Message spectrum
will consider the details of how this is
fc-fm fc fc+fm fc+3fm fc+5fm fc+7fm
done later. frequency
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 9 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 10

Constellation Diagrams Constellation Diagram for Binary OOK

ASK
University of Bristol University of Bristol

• A constellation diagram is a method of representing

• A constellation diagram of binary OOK ASK is
the signal states of the amplitude and phase of the
signal in a diagram. shown below. Sin( t) c

• The horizontal axis displays the components of the

signal that are in phase with the carrier signal
(cos(c.t)). (the real part)
• The vertical axis shows the components that are in Cos(ct)

phase with the quadrature component of the carrier

signal (sin(c.t)). (the imaginary part)

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 11 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 12

3
 Constellation Diagram: 8-ary ASK Carrier Modulation (1)
University of Bristol University of Bristol

• The carrier signal is typically a sine wave. The

• A constellation diagram of 8-ary ASK is transmitted signal can be described by:
shown below. Sin( t) c
s(t) = Acos(t + )
• This signal has three properties which we can
modulate.
– Amplitude, A.
– Frequency, .
– Phase, .
Cos(ct)
• Initially we will consider modulation schemes which
modulate only one of these three properties.
• Also, we will first consider binary digital modulation
schemes and analog modulation schemes.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 13 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 14

Carrier Modulation (2) Carrier Modulation (3)

University of Bristol University of Bristol

• If we modulate just one property of the signal, we can – By modulating the amplitude of the carrier wave with a continuous
modulate our data signal onto this carrier signal in the message. Thus, A is a continuous function of the modulating signal.
This is known as Amplitude Modulation (AM).
following ways:
– By modulating the frequency of the carrier wave with a continuous
– By modulating the amplitude of the carrier wave with a discrete message. Thus, , is a continuous function of the modulating signal.
message. Thus, A is a discrete function of the modulating signal. This This is known as Frequency Modulation (FM)
is known as Amplitude Shift Keying (ASK).
– By modulating the phase of the carrier wave with a continuous
– By modulating the frequency of the carrier wave with a discrete message. Thus, , is a continuous function of the modulating signal.
message. Thus, , is a discrete function of the modulating signal. This This is known as Phase Shift Keying (PSK)
is known as Frequency Shift Keying (FSK)
– By modulating the phase of the carrier wave with a discrete message. • Note that whilst the message is discrete for ASK, FSK and
Thus, , is a discrete function of the modulating signal. This is known PSK, it will need to be filtered into a non-discrete form in
as Phase Shift Keying (PSK) order to achieve bandlimiting – more on this later.
• FM and PM have been shown to be very similar. The term
FM has become very well known. PM is rarely used.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 15 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 16

4
 EENG22000 Communications Amplitude Modulation (AM)
University of Bristol University of Bristol

• A generic AM modulation process is illustrated below:

Part 2.2: • The message signal is

mixed with a carrier signal.


m(t)
s(t) • A proportion of the carrier
AM signal may be added to the
output of the mixer.
• The result is the
transmitted signal
K
s(t)=[m(t)+K]cos(ct).
•Depending upon the nature
of the message, it might
need to be bandlimited by
application of a suitable filter
cos(ct)

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 17 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 18

Amplitude Modulation with a Sine Wave AM Phasor diagram

University of Bristol University of Bristol

• As a simple example, the case where the message signal,

m(t), is a simple cos wave can be considered.
mt   Am cosmt 
Am Am
s t   cos c   m t   cos c   m t   K cos c t 
2 2 Amplitude
Variation
m m Resultant AM Signal

• The spectrum of the modulated signal is as illustrated:

K
Am/2 Am/2

c -m c c +m Carrier phasor rotates

at c radians/sec
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 19 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 20

5
 Variants of AM FC-AM Signal Envelope
University of Bristol University of Bristol

• Three variants of AM are defined by the values and ratio of • For FC-AM, the power in the carrier is equal to or greater
K and Am. than that in the sidebands.
• These are: • When modulated by a sine wave message (K>Am), the signal
– Full Carrier AM (FC-AM): KAm. envelope is as illustrated below:
– Diminished Carrier AM (DC-AM): K<Am.
Am
– Suppressed Carrier AM (SC-AM): K=0.
K
• It will be shown that these AM variants have different
advantages and disadvantages, particularly in terms of their K
Am
power efficiency and ease of demodulation.

• Note that the envelope never goes to zero. This will only
occur when K=Am.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 21 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 22

FC-AM Signal Spectrum DC-AM Signal Envelope

University of Bristol University of Bristol

• For FC-AM, the power in the carrier is equal to or greater • For DC-AM, the power in the carrier is less than that in the
than that in the sidebands. sidebands.
• When modulated by a sine wave message (K>Am), the signal • When modulated by a sine wave message, the signal
spectrum is as illustrated below: envelope is as illustrated below:
Overmodulation
K
Am
Am/2 Am/2 K
K
Am
c -m c c +m
Phase Phase
Discontinuity Discontinuity
• Note that the amplitude of the carrier is greater than the sum
• Note that the signal envelope does pass through zero and
of the amplitude of the two sidebands. (They will be equal for
that the carrier inverts at the zero crossings. This is termed
the case of K=Am)
over modulation.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 23 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 24

6
 DC-AM Signal Spectrum SC-AM Signal Envelope
University of Bristol University of Bristol

• For DC-AM, the power in the carrier is less than that in the • For SC-AM, there is no power in the carrier.
sidebands. • When modulated by a sine wave message, the signal
• When modulated by a sine wave message, the signal is as envelope is as illustrated below:
illustrated below:
Overmodulation

K Am
Am/2 Am/2

c -m c c +m
Phase Phase
Discontinuity Discontinuity
• Note that the amplitude of the carrier component will be less
than the sum of the sidebands. • Note that, as with DC-AM, the signal envelope exhibits over
modulation.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 25 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 26

SC-AM Signal Spectrum Modulation Index of AM

University of Bristol University of Bristol

• For SC-AM, there is no power in the carrier. Thus, all the

• The modulation index m is defined as the ratio of the
power is in the sidebands.
amplitude of the modulating wave to the amplitude of the
• When modulated by a sine wave message the signal is as
illustrated below: carrier wave:
– Thus, m = Am/K

Am/2 Am/2 – For FC-AM, m1

– For DC-AM, m>1
c -m c c +m
– For SC-AM, m=

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 27 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 28

7
 Power Efficiency of AM (1) Power Efficiency of AM (2)
University of Bristol University of Bristol

• Average power in carrier = K2/2R • The carrier bears no information and the power used to
• Average power in sidebands = 2.Am2/8R transmit it is effectively wasted.
• Clearly, the power efficiency of AM depends upon the
= Am2/4R modulation index.
• Ratio of average power in the sidebands to the average • For example, FC-AM is most efficient when m=1.
power in the carrier: (Am2/4)/(K2/2) = m2/2 – The transmit power efficiency in this case is 33%
• Ratio of the average power in the sidebands to the total – Note that this is the BEST efficiency that FC-AM can
average power: 2 achieve.
m • Alternatively, SC-AM is 100% transmit power efficient.
2  m2 • Thus, a high modulation index is highly desirable where it is
desired to achieve good transmit power efficiency.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 29 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 30

Detection of AM Signals The Envelope Detector

University of Bristol University of Bristol

• Analog detectors (demodulators) are either ‘coherent’ or ‘non-

coherent’ depending on whether or not they use a local • The envelope detector consists of a half wave rectifier and a
carrier. low pass filter
– The local carrier is a ‘copy’ of the carrier signal used in the modulator,
available in the demodulator • The filter need not have a sharp roll off since the message
– More realistically, it is an estimate since if we could actually provide the and carrier frequencies are usually well spaced
carrier used in the modulator to the demodulator, we would not need a
communication system!
– Generation of the local carrier will be discussed in part 4 of the course
– Coherent systems use a local carrier, Non-coherent systems do not
• Non-coherent systems have the advantage of lower
complexity but typically offer lower performance. We will see
that further problems are encountered when the AM signal is
over modulated.
• Coherent systems have the advantage of strong performance
but require more complex receiver hardware.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 32 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 33

8
 Envelope Detection of Over Modulated Coherent Detector
Signals University of Bristol University of Bristol

• The envelope detector behaves differently, depending on the • A simple coherent detection system is shown below.
type of AM signal being received. • How this functions as a demodulator is not immediately
obvious but a little mathematical analysis will show that it
FC-AM
does
p(t)
s(t) r(t)

DC-AM

SC-AM

cos(ct+ )
• Only FC-AM is correctly demodulated by an envelope
detector; over modulation causes envelope detection to fail.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 34 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 35

Detection of AM Signals (1) Detection of AM Signals (2)

University of Bristol University of Bristol

• If we assume initially that the local oscillator is perfectly

locked to the carrier (=0) then, for AM modulation: • Thus, provided that the local oscillator is perfectly locked, the
Am Am
s t   cos c   m t   cos c   m t   K cos c t 
output of the coherent detector consists of three terms: The
2 2 message signal (in both positive and negative frequencies)
• The output of the mixer is given by: and a DC offset due to the carrier signal.
Am Am Am
pt   cos  mt   cos2 c   m t   cos mt  • For SC-AM, the DC term is zero (K=0).
4 4 4
• Any error in the local oscillator reduces the performance of
Am K
 cos2 c   m t    K cos2 c t  the receiver:
4 2
• The higher frequency terms are removed by the filter and Am Am K
r t   cos  mt     cos mt     cos 
hence:   Am   Am   K
rt  cos   mt  cos  mt  4 4 2
4 4 2
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 36 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 37

9
 Non-Sinusoidal Message Signals Non-Sinusoidal Message Signals -
University of Bristol Signal Envelopes University of Bristol

• We have analysed the various forms of AM by considering a

sinusoidal message signal. Message
• In practice, transmitting a sinusoid is rather pointless - it
conveys no information.
• More likely message sources are audio or video signals.
• We need to extend the analysis that we have considered FC-AM
here to non-sinusoidal message signals.
– The maths can become extremely tedious!
– The general concepts can be readily extended.
SC-AM

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 38 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 39

Non-Sinusoidal Message Signals - Advantages and Disadvantages of

Signal Spectra Different Types of AM Signals
University of Bristol University of Bristol

• Choice of modulation type between FC-AM, DC-AM and SC-

Message AM is a function of receiver cost and transmit power
0 frequency efficiency.
– FC-AM achieves poor power efficiency but facilitates the use of
cheap, simple (low performance) non-coherent receivers. Coherent
FC-AM receivers will achieve better performance.
0 fc frequency – DC-AM offers improved power efficiency but mandates the use of
more costly and complex coherent receivers.
– SC-AM offers maximal power efficiency.
SC-AM
0 fc frequency

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 40 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 41

10
 EENG22000 Communications Amplitude Shift Keying (ASK)
University of Bristol University of Bristol

• The simplest form of bandpass modulation is

Amplitude Shift Keying (ASK). With this modulation
Part 2.3: the data is represented as various (discrete)
amplitude levels of a fixed frequency and phase
ASK offset carrier.
• The simplest form of ASK is to switch a fixed
frequency oscillator ON for a ‘one’, and OFF for a
‘zero’. This is known as ON-OFF KEYING (OOK).
• If more than two symbol states are used, then M-ary
ASK is generated.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 42 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 43

Constellation Diagram for Binary OOK

Amplitude Shift Keying (ASK) ASK
University of Bristol University of Bristol

0 1 1 0 1 • A constellation diagram of binary OOK ASK is

shown below. Sin( t) c
Binary ASK
or OOK

M-ary Cos(ct)

ASK

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 44 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 45

11
 Generation of ASK (1) Generation of ASK (2)
University of Bristol University of Bristol

• This diagram illustrates one way in which an ASK signal may • Another method is to multiply a carrier signal by the
be generated. However, good bandpass filtering at RF is very baseband signal stream (similar to the AM modulator).
difficult to achieve in practice. • The multiplier is known as a mixer.
• This modulation process enables bandlimiting at baseband
Data input ASK output

cos( c .t )
Pulse Shaping
Filter

switch

Bandpass filter cos( c .t )

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 46 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 47

ASK spectrum (1) ASK spectrum (2)

University of Bristol University of Bristol

• If we can achieve ASK

• Note that the spectrum is two sided, and therefore the
modulation my Baseband Square
bandwidth required to transmit it is 2m. That is, twice the
multiplying the Wave Message baseband bandwidth. Spectral efficiency is limited to
message by the carrier spectrum 1bit/s/Hz
G(f)
then we can use our • This spectrum cannot be transmitted in practice.
earlier analysis to • It must be bandlimited somehow – a filter must be applied
determine the spectrum frequency during the modulation process.
of the modulated signal: • This turns out to be the case for almost all modulation
Carrier Modulated schemes (and all digital modulation schemes).
ASK Square Wave • We will consider later what the implications of bandlimiting
Message spectrum are and what form a good bandlimiting filter should take.

fc-fm fc fc+fm fc+3fm fc+5fm fc+7fm

frequency
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 48 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 49

12
 Non Coherent Detection of ASK (1) Non-Coherent Detection of ASK (2)
University of Bristol University of Bristol

• Non coherent detection of ASK is straight forward. A circuit to • Just like the envelope detector for AM but will a
do this is shown below. hardlimiting process applied to the output.
• There is no overmodulation issue.
• The simplicity of this approach is often outweighed
by the inability of a non-coherent detector to extract
signal from noise.
Comparator • A coherent detector can do this and can achieve a
lower Bit Error Rate (BER) over the same channel,
with the same S/N ratio.

LPF
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 50 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 51

Coherent Detection (1) Coherent Detection (2)

University of Bristol University of Bristol

• Coherent Detection is based upon multiplying the • Coherent detection requires the local generation of a replica
received signal by a local replica of the carrier signal. of the carrier signal.
• If a local carrier is available, then the reverse process to
modulation can be performed at the receiver.
• Essentially the locally generated carrier is multiplied by the
incoming signal. Thus:
A(t)cos(ct)  cos(ct+) = 0.5A(t)cos() + 0.5A(t)cos(2 ct+)
• The process results in two output components. One is the
baseband message signal (multiplied by a cos() term) and
the other is the baseband message signal shifted to twice the
cos( c .t ) carrier frequency.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 52 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 53

13
 Coherent Detection (3) Bit Error Rate (BER) Performance of
OOK ASK
University of Bristol University of Bristol

• By passing these two products through a low pass

filter, the higher frequency component can be Non Coherent ASK

removed. Thus only A(t)cos() is left. P 

1 
.erfc 
Eb  1  
  exp   E b 
 2  4 .N 
•  defines the phase error between the locally 2  2.N 0   0 
generated carrier and the received carrier. Any non-
zero value of  (i.e. any phase error) causes a loss Coherent ASK
in performance since cos()<1 if 0.
1  E 
• Thus, it is crucial for coherent detection to be able to P  .erfc b 
accurately generate the local carrier signal. 2  2N0 

• One way to do this is to use a Phase Lock Loop

(PLL). We will consider this in more detail later in the
course.
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 54 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 55

EENG22000 Communications Frequency Shift Keying (FSK)

University of Bristol University of Bristol

• Binary FSK uses 2 distinct frequencies to represent the

different data bits. Thus, s(t) = Acos((t)t+)
Part 2.4: • The amplitude of the signal is thus to some degree
unimportant, and amplitude limiting can be used. This is a
good, since FSK is thus robust to fading in the channel.
FSK • An FSK signal can be viewed as two ASK signals with one or
the other switched on or off depending upon the data signal.
ASK-1

0 1 0 1

ASK-2
Switch

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 57 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 58

14
 FSK Spectrum FSK Generation (1)
University of Bristol University of Bristol

• It is not easy to determine • FSK can be generated by switching between distinct

the spectrum of an FSK frequency sources as shown. It is likely that the phase
signal analytically. However, transition in this switching process won’t be smooth, and
it can be determined there will be phase jumps.
graphically as the composite
of two appropriately spaced • Also it is not possible to apply pulse shaping before the
ASK spectra. switch. As we have discussed, filtering at carrier frequencies
• Clearly, the FSK spectrum is impractical.
and bandwidth depend upon cos(  1t ) Data Input
the spacing of the two centre
frequencies
• Whatever, its exact form,
bandlimiting will again be FSK Output
Switch
required in the modulation
process cos(  2 t )
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 59 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 60

FSK Generation
FSK Generation via VCO
University of Bristol University of Bristol

• The alternative method of FSK generation is to use the data • Has the advantage of giving smooth phase transitions.
input to control a Voltage Controlled Oscillator (VCO). • Data input can be low pass filtered before the VCO.
• This will control the shape of the FSK spectrum.

Data Input Filtered Data

Data Input
FSK Output FSK Output
Voltage Controlled
Oscillator (VCO)
Voltage Controlled
Oscillator (VCO)
• This achieves Continuous Phase FSK (CPFSK)
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 61 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 62

15
 Minimum Shift Keying FSK Band Limiting
University of Bristol University of Bristol

• If the two frequencies are spaced at a multiple of 1/2 of the • A common pulse shaping filter used for FSK (or MSK) is the
symbol rate, they will be orthogonal. Gaussian shape.
• Thus the minimum spacing between the two frequencies • This filter ensures that the spectrum of the FSK signal rolls
required to ensure that they are orthogonal is half the symbol off rapidly outside of the main lobe.
rate. • This is termed Gaussian FSK (GFSK) or Gaussian MSK
• When this minimum spacing is used, the modulation is (GMSK)
termed Minimum Shift Keying (MSK). Spectral efficiency can • GMSK is used as modulation for the GSM cellular radio
approach 2bits/s/Hz. system.
• GFSK is used as the modulation scheme for Bluetooth

f1 f2
frequency

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 63 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 64

FSK Detection Coherent FSK Detection

University of Bristol University of Bristol

• As with ASK, FSK can be detected coherently or non coherently.

• A typical coherent FSK detector is shown below.
• This is based around having filters tuned to the signal frequencies.
• The output of the filters is measured with a peak detector.
• The filters with the greatest output is assumed to correspond to
correspond to the signal frequency.
• This is measured in a comparator circuit. LPF
Data
FSK Input output

cos(1.t ) Comparator
FSK Data
Input Bandpass filter output
f1
Peak detectors
Comparator
LPF

Bandpass filter
f2
cos( 2 .t )
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 65 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 66

16
 Coherent FSK Detection
BER Performance of FSK
University of Bristol University of Bristol

• The two carrier frequencies required for binary FSK are • It can be seen that only about
generated at the receiver. a 1dB advantage accrues
from using coherent detectors
• Each detector only responds to the signal frequency it is at low BER
set to demodulate. • For this reason it is very often
• With binary FSK, one signal frequency should cause a acceptable to use non
detector to respond. The other signal frequency should coherent FSK detectors.
provoke no response from the detector. This occurs
when the frequencies of the signals are orthogonal. The Non-Coherent FSK
product of the carrier with the incorrect signal frequency, Ps=exp(-Eb/2No)
over the symbol period, will integrate to zero.
Coherent FSK
Ps=erfc(Eb/2No)1/2
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 67 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 68

Advantages of FSK Disadvantages of FSK

University of Bristol University of Bristol

• Detection can be based on the relative frequency changes

between the states and thus does not require absolute • Can be slightly less bandwidth efficient than either ASK or
frequency accuracy at the receiver. FSK is relatively tolerant PSK. (Not true for MSK)
to local oscillator drift and Doppler shift.
• Constant modulation envelope means this modulation • Bit/symbol error rate worse than PSK
method is insensitive to amplitude variations of the channel.
• This factor also has an advantage in mobile radio.
• The power drain of the radio is largely determined by the
power drain of the final power stage.
• In general, the more non linear power stage, the more
efficient that stage is.
• Using a modulation that doesn’t rely on amplitude being
correct, is thus an advantage in this situation.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 69 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 70

17
 EENG22000 Communications Phase Shift Keying (PSK)
University of Bristol University of Bristol

• PSK modulates the phase of the signal, s(t)=Acos(t+(t))

• The phase can be measured relative to a fixed carrier.
Part 2.5: • This is known as coherent PSK.
• Alternatively the information can be sent as a phase change
between consecutive symbols.
PSK • This is known as differentially coherent PSK .
• There is no non-coherent PSK

1 0 1 0

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 74 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 75

Binary PSK Constellation Spectral Occupancy of PSK

University of Bristol University of Bristol

Q Axis
• The spectrum of binary PSK is identical to that of binary ASK.
sin(ct) • Bandlimiting is required, as always.
• A binary PSK signal can be regarded as an ASK signal that
has been modulated with a signal of +/- 1 (rather than 0 and
1). This is Polar ASK!

I Axis
cos(ct) G( f )

c 
4
c 
3
c 
2
c 
1 c c 
1
c 
2
c 
3
c 
4
T T T T T T T T
f
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 76 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 77

18
 Generation of Binary PSK (1) Generation of Binary PSK (2)
University of Bristol University of Bristol

• One method of generating PSK is shown below • A more practical alternative method uses a linear multiplier to
• It uses a switch to switch between a signal of 0 and change the phase of the signals.
180 degree phase shift.
• But this requires bandlimiting at the carrier
frequency… Pulse shaping
cos( c t )
Data Input
filter
Data PSK
Input output

Mixer

Switch
cos( c t )
Inverter
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 78 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 79

Detection of PSK Carrier recovery for PSK.

University of Bristol University of Bristol

• We will see later how a local carrier replica can be generated

• There is no non-coherent method of detecting PSK. but there is a danger that the local carrier will be generated in
• Coherent detection can be used as for ASK. anti-phase to the carrier used in the modulator.
• In this event the ones will be detected as zeros and the zeros
as ones.
Comparator • To overcome this problem, it is possible to send a training
sequence in which the position at which the ones and the
PSK signal in
zeros appearing in the training sequence are known. A
maximum likelihood decision can then be made at the
receiver as to whether the local carrier is inverted or not.
• The need for a training sequence can be obviated via the use
cos(ct) of Differential PSK.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 80 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 81

19
 Differentially Encoded PSK (DPSK) DPSK (2)
University of Bristol University of Bristol

• Differential encoder and decoder are shown below. • Note that the output of the receiver (decoder) is the same as
the input data stream to the encoder.
At the transmitter
Input data stream 1 1 1 0 0 1 0 1 0 1
In Ex -Or Ex -Or
Delayed Data stream 1 0 1 0 0 0 1 1 0 0
Out Exclusive Or 0 1 0 0 0 1 1 0 0 1

Arbitary starting bit

Output from transmitter

Ts Ts At the receiver
Input to receiver 0 1 0 0 0 1 1 0 0 1
Encoder Decoder Delayed Data 1 0 1 0 0 0 1 1 0 0
Output 1 1 1 0 0 1 0 1 0 1

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 82 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 83

DPSK Modulation DEPSK Demodulation

University of Bristol University of Bristol

• The modulator of a DEPSK system is shown below

• The circuit diagram for a DEPSK demodulator is shown below.

Data Input
Exclusive OR
Comparator
PSK output

PSK signal in Ex -Or

Time Delay
(T )
s Out
cos(ct)
Differential PSK generator Ts
cos(ct)
Decoder

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 84 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 85

20
 DPSK Demodulation (1) DPSK Demodulation (2)
University of Bristol University of Bristol

• Alternatively, we can combine the detection and differential

decoding to achieve Differentially Coherent Detection
• Data must be differentially encoded first. At the transmitter
Input data stream 1 1 1 0 0 1 0 1 0 1
Delayed Data stream 1 0 1 0 0 0 1 1 0 0
Exclusive Or 0 1 0 0 0 1 1 0 0 1
DPSK Input
Transmitted Phase 0 180 0 0 0 180 180 0 0 180
Data output At the receiver
Comparator Phase of input signal 0 180 0 0 0 180 180 0 0 180
Ts
Differential PSK demodulation Delayed phase 180 0 180 0 0 0 180 180 0 0
Output phase 180 180 180 0 0 180 0 180 0 180
Output bit 1 1 1 0 0 1 0 1 0 1
• This is a Hybrid of Coherent Detection and Differential
Decoding
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 86 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 87

Disadvantages of Differential PSK PSK Constellation Diagram (1)

University of Bristol University of Bristol

• Note that the noise immunity of DPSK is poorer than • The constellation diagram of a PSK signal is
conventional PSK because the phase reference for PSK
comes directly from the line and is not as stable as a phase shown below.
reference generated by other means.
• Also, there is potential for error propagation. Since each bit is Q Axis Q Axis
‘decoded’ relative to the previous one, if one symbol is in sin( c .t ) sin( c .t )

error then the next is also likely to be decoded in error.

I Axis I Axis
cos( c .t ) cos( c .t )

PSK Constellation OOK ASK Constellation

Diagram Diagram

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 88 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 89

21
 PSK Constellation Diagram (2) BER versus Eb/N0 for Binary PSK (1)
University of Bristol University of Bristol

• This type of constellation diagram is referred to as “antipodal”

(as in antipodes). • As anticipated
• This means that the symbols are equal and opposite to each Differential PSK has
other in constellation space. slightly worse
• Antipodal signals are seen as giving good noise immunity. performance
• Noise immunity is a function of the separation of the signal compared with
states. Coherent PSK.
Differential PSK
Ps = 0.5exp(-Eb/No)
Coherent PSK
Ps = 0.5erfc(Eb/No)1/2
EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 90 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 91

BER versus Eb/N0 for Binary PSK (2) EENG22000 Communications

University of Bristol University of Bristol

• Note that the expression for the probability of error for the
coherent PSK is similar to that for OOK ASK.
• The only difference is lack of a “2” in the denominator of the
“erfc” function.
Part 2.6:
• This is a reflection of the separation of the signal states, and
indicates that Coherent PSK is 3dB better than Coherent
Comparisons of Binary
ASK.
• This is to be expected. As we showed earlier Binary PSK is
(Digital) Modulation
equivalent to Binary Polar ASK. We already know that for
baseband signals Polar signals offer 3dB advantage over
Schemes
On-Off signals.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 92 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 94

22
 BER versus Eb/N0 for Binary Bandpass BER versus Eb/N0 for Binary Bandpass
Signalling (1) Signalling (2)
University of Bristol University of Bristol

• Of all the binary modulation schemes, Coherent PSK is the

 Eb  most robust to additive noise.
Coherent ASK P  0.5erfc 

and FSK  2.N 0  • Non-coherent FSK performs less well but places lower
 E 
Differential P  0.5 exp   b  demands on hardware. It is thus typically easier and cheaper
 N0 
PSK to implement. When implemented as MSK it is also the most
 Eb 
P  0.5erfc  

spectrally efficient.
Coherent PSK  N0 

Non coherent
 E 
P  0.5 exp  b 
• ASK offers a poor compromise between SNR requirements
FSK  2. N 0  and spectral efficiency and is thus rarely used in practical
Non coherent
applications.
ASK
 E   Eb  • Thus, for binary modulation, the choice of modulation scheme
P  0.5 exp  b   0.5erfc 

 4.N 0   2.N 0  is often between the cheaper simpler non-coherent FSK and
the more robust coherent PSK.

EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 95 EENG22000 Communications, Part 2: Basic Bandpass Communications. Simon Armour 17 February 2009 96

23