Chapter 2 Continuous-Wave

Modulation

Analog Modulation is the subject concerned

in this chapter.

2.1 Introduction
Analog communication system
The most common carrier is the sinusoidal wave.

(Analog)
Carrier wave

Po-Ning Chen@cm.nctu

Chapter 2-2

2.1 Introduction
Modulation
A process by which some characteristic of a carrier
is varied in accordance with a modulating wave
(baseband signal).

Sinusoidal Continuous-Wave (CW) modulation

Amplitude modulation
Angle modulation

Po-Ning Chen@cm.nctu

Chapter 2-3

Sinusoidal carrier

2.1 Introduction
Baseband signal

Amplitude Modulation

Frequency Modulation

Po-Ning Chen@cm.nctu

Chapter 2-4

2.2 Double-Sideband with Carrier or simply

Amplitude Modulation
Carrier c(t ) = Ac cos(2f c )
Baseband m(t )
Modulated Signal s(t ) = Ac [1 + ka m(t )] cos(2f ct ),
where ka is amplitude sensitivity or modulation index

Two required conditions on amplitude sensitivity

1 + ka m(t) 0, which is ensured by |ka m(t)| 1.

The case of |ka m(t)| > 1 is called overmodulation.

The value of |ka m(t)| is sometimes represented by percentage
(because it is limited by 1), and is named (|ka m(t)|100)%
modulation.

fc >> W, where W is the message bandwidth.

Violation of this condition will cause nonvisualized envelope.

Po-Ning Chen@cm.nctu

Chapter 2-5

2.2 Overmodulation

| ka m(t ) | 1

Po-Ning Chen@cm.nctu

overmodulaion

Chapter 2-6

2.2 Example of Non-Visualized Envelope

Po-Ning Chen@cm.nctu

Chapter 2-7

2.2 Example of Visualized Envelope

Po-Ning Chen@cm.nctu

Chapter 2-8

2.2 Transmission Bandwidth

s(t ) = Ac [1 + ka m(t )] cos(2f c t )
Ac
[ ( f f c ) + ( f + f c )] + ka Ac [M ( f f c ) + M ( f + f c )]
2
2
Transmission bandwidth BT = 2W.

S( f ) =

2W
Po-Ning Chen@cm.nctu

Chapter 2-9

2.2 Transmission Bandwidth

Transmission bandwidth of an AM wave
For positive frequencies, the highest frequency
component of the AM wave equals fc + W, and the
lowest frequency component equals fc W.
The difference between these two frequencies
defines the transmission bandwidth BT for an AM
wave.

Po-Ning Chen@cm.nctu

Chapter 2-10

2.2 Transmission Bandwidth

The condition of fc > W ensures that the sidebands
do not overlap.

Po-Ning Chen@cm.nctu

Chapter 2-11

2.2 Negative Frequency

Operational meaning of negative frequency
in spectrum
If time-domain signal is real-valued, the negative
frequency spectrum is simply a mirror of the
positive frequency spectrum.
We may then define a one-sided spectrum as
Sone-sided ( f ) = 2 S ( f ) for f 0.
Hence, if only real-valued signal is considered, it is
unnecessary to introduce negative frequency.
Po-Ning Chen@cm.nctu

Chapter 2-12

2.2 Negative Frequency

So the introduction of negative frequency part is
due to the need of imaginary signal part.
Signal phase information is embedded in imaginary
signal part of the signal.
mQ(t)
phase
mI(t)

Po-Ning Chen@cm.nctu

Chapter 2-13

2.2 Negative Frequency

As a result, the following two spectrums contain the
same frequency components but different phases
(90 degree shift in complex plane).
2 cos(2f c t )

2 j sin( 2f c t )

( f + fc )

( f fc )

fc

fc

( f + fc )

fc
Po-Ning Chen@cm.nctu

( f fc )
fc
Chapter 2-14

2.2 Negative Frequency

Summary
Complex-valued baseband signal consists of
information of amplitude and phase; while realvalued baseband signal only contains amplitude
information.
One-sided spectrum only bears amplitude
information, while two-sided spectrum (with
negative frequency part) carries also phase
information.
Po-Ning Chen@cm.nctu

Chapter 2-15

2.2 Virtues of Amplitude Modulation

AM receiver can be implemented in terms of simple
circuit with inexpensive electrical components.
E.g., AM receiver

v1 (t ) = s 2 (t ) = Ac2 [1 + ka m(t )]2 cos2 ( 2f c t )

Ac2
=
[1 + k a m(t )]2 [1 + cos(4f c t )]
2
Po-Ning Chen@cm.nctu

Chapter 2-16

2.2 Virtues of Amplitude Modulation

The bandwidth of m2(t) is twice of m(t). (So to
speak, the bandwidth of m(f)*m(f) is twice of m(f).)

Lowpass filter

Po-Ning Chen@cm.nctu

Chapter 2-17

2.2 Virtues of Amplitude Modulation

So if 2fc > 4W,

Ac2
v2 ( t ) =
[1 + ka m(t )]2
2
A
v3 (t ) = c [1 + ka m(t )]
2
block DC

if m(t ) is zero mean

Ac ka
m(t )
2

By means of a squarer, the receiver can recover the

information-bearing signal without the need of a local carrier.
Po-Ning Chen@cm.nctu

Chapter 2-18

2.2 Limitations of Amplitude Modulation

(DSB-C)
Wasteful of power and bandwidth
s(t ) = Ac [1 + ka m(t )] cos(2f c t )
= Ac cos(2f c t ) + ka m(t ) cos(2f c t )
142
4 43
4
with carrier

Waste of power in the information-less with-carrier part.

Po-Ning Chen@cm.nctu

Chapter 2-19

2.2 Limitations of Amplitude Modulation

Wasteful of power and bandwidth
Only requires half of bandwidth after modulation

Po-Ning Chen@cm.nctu

Chapter 2-20

10

2.3 Linear Modulation

Definition
Both sI(t) and sQ(t) in
s(t) = sI(t)cos(2fct) sQ(t) sin(2fct)
are linear function of m(t).

Po-Ning Chen@cm.nctu

Chapter 2-21

2.3 Linear Modulation

For a single real-valued m(t), three types of
modulations can be identified according to
how sQ(t) are linearly related to m(t), at the
case that sI(t) is exactly m(t):
(Some modulation may have mI(t) and mQ(t) that
respectively bear independent information.)

1. Double SideBand-Suppressed Carrier modulation

(DSB-SC)
2. Single SideBand (SSB) modulation
3. Vestigial SideBand (VSB) modulation
Po-Ning Chen@cm.nctu

Chapter 2-22

11

2.3 DSB-SC and SSB

Type of modulation

sI(t)

sQ(t)

DSB-SC

m(t)

SSB

m(t)

m (t )

Upper side band transmission

SSB

m(t)

m (t )

Lower side band transmission

* m (t ) = Hilbert transform of m(t ) , which is used to completely suppress the other sideband.

DSB-SC
SSB

usb

SSB

lsb

Po-Ning Chen@cm.nctu

Chapter 2-23

2.3 DSB-SC
Different from DSB-C, DSB-SC s(t) undergoes
a phase reversal whenever m(t) crosses zero.
s (t ) = m(t ) cos(2f c t )

Require a receiver that can

recognize the phase reversal!
Po-Ning Chen@cm.nctu

Chapter 2-24

12

2.3 Coherent Detection for DSB-SC

For DSB-SC, we can no longer use the envelope
detector (as used for DSB-C), in which no local
carrier is required for the receiver.
The coherent
detection or
synchronous
demodulation
becomes necessary.

Po-Ning Chen@cm.nctu

Chapter 2-25

2.3 Coherent Detection for DSB-SC

v (t ) = Ac' cos(2f c t + ) s(t )
= Ac Ac' cos(2f c t ) cos(2f c t + )m(t )
1
1
Ac Ac' cos(4f c t + )m(t ) + Ac Ac' cos( )m(t )
2
2
LowPass 1

Ac Ac' cos( )m(t ),

2
provided f c > W .
=

Po-Ning Chen@cm.nctu

Chapter 2-26

13

2.3 Coherent Detection for DSB-SC

Quadrature null effect of the coherent detector.
If = /2 or /2, the output of coherent detector for DSBSC is nullified.
If is not equal to either /2 or /2, the output of coherent
detector for DSB-SC is simply attenuated by a factor of
cos(), if is a constant, independent of time.
However, in practice, often varies with time; therefore, it is
necessary to have an additional mechanism to maintain the
local carrier in the receiver in perfect synchronization with
the local carrier in the transmitter.
Such an additional mechanism adds the system complexity of
the receiver.
Po-Ning Chen@cm.nctu

Chapter 2-27

2.3 Costas Receiver for DSB-SC

An exemplified
design of
synchronization
mechanism is the
Costas receiver,
where two
coherent
detectors are
used.

In-phase coherent detector

Quadrature-phase coherent detector

Po-Ning Chen@cm.nctu

Chapter 2-28

14

2.3 Costas Receiver for DSB-SC

Conceptually, the Costas receiver adjusts the
phase so that it is close to 0.
When drifts away from 0, the Q-channel output
will have the same polarity as the I-channel output
for one direction of phase drift, and opposite
polarity for another direction of phase drift.
The phase discriminator then adjusts through the
voltage controlled oscillator.
cos( )
sin( )

Po-Ning Chen@cm.nctu

-pi/2

Chapter 2-29

pi/2

2.3 Single-Sideband Modulation

How to generate SSB signal?
1. Product modulator to generate DSB-SC signal
2. Band-pass filter to pass only one of the sideband
and suppress the other.

The above technique may not be applicable to a

DSB-SC signal like below. Why?
filter

filter

SSB
Po-Ning Chen@cm.nctu

Chapter 2-30

15

2.3 Single-Sideband Modulation

For the generation of an SSB modulated signal
to be possible, the message spectrum must have
an energy gap centered at the origin.

Po-Ning Chen@cm.nctu

Chapter 2-31

2.3 Single-Sideband Modulation

Example of signal with 300 Hz ~ 300 Hz
energy gap
Voice : A band of 300 to 3100 Hz gives good
articulation

Also required for SSB modulation is a highly

selective filter

Po-Ning Chen@cm.nctu

Chapter 2-32

16

2.3 Single-Sideband Modulation

Phase synchronization is also an important
issue for SSB demodulation. This can be
achieved by:
Either a separate low-power pilot carrier
Or a highly stable local oscillator (for voice
transmission)
Phase distortion that gives rise to a Donald Duck voice
effect is relatively insensitive to human ear.

Po-Ning Chen@cm.nctu

Chapter 2-33

2.3 Vestigial Sideband Modulation

Instead of transmitting only one sideband as
SSB, VSB modulation transmits a partially
suppressed sideband and a vestige of the other
sideband.

Po-Ning Chen@cm.nctu

Chapter 2-34

17

SSB

VSB
Still, no information loss in principle.

Po-Ning Chen@cm.nctu

Chapter 2-35

2.3 Requirements for VSB filter

1.

2.

The sum of values of the magnitude response |H(f)| at any

two frequencies equally displaced above and below fc is
unity. I.e., |H(fc f)| + |H(fc + f)| = 1 for fv < f < fv.
H(f fc) + H(f + fc) = 1 for W < f < W.
ffc

fc

f+fc
fc

So the transmission band of VSB filter is BT = W + fv.

Po-Ning Chen@cm.nctu

Chapter 2-36

18

2.3 Generation of VSB Signal

Analysis of VSB
Give a real baseband signal m(t) of bandwidth W.
Then, M ( f ) = M * ( f ) and M ( f ) = 0 for | f |> W .

Let M VSB ( f ) = M ( f )[1 + H Q ( f ) / j ] / 2, where

f fv
1,
1

H Q ( f ) = (0,1), f v < f < 0, and H Q ( f ) = H Q* ( f ).

j
0,
f =0

The filter is denoted by HQ is because it is used to generate sQ(t) (cf. slide Chapter 2-23)
Po-Ning Chen@cm.nctu

Chapter 2-37

2.3 Generation of VSB Signal

*

LQ ( f ) =

1
1
1
H Q ( f ) is real. LQ ( f ) = H Q ( f ) = H Q* ( f ) = H Q ( f ) = L*Q ( f )
j
j
j
j

= LQ ( f ).
1
LQ ( f ) = H Q ( f )
j

[1 + LQ ( f )] / 2 = 1 + H Q ( f ) / 2
j

1.0

W
Po-Ning Chen@cm.nctu

Chapter 2-38

19

2.3 How to recover from VSB signal?

*
( f )
M VSB ( f ) + M VSB

1
(M ( f )[1 + LQ ( f )] + M ( f )[1 + LQ ( f )]* )
2
1
= (M ( f )[1 + LQ ( f )] + M ( f )[1 + LQ ( f )]),
2
since [1 + LQ ( f )] is real, and M ( f ) = M ( f )
=

1
(M ( f )[2 + LQ ( f ) + LQ ( f )])
2
= M ( f ), because LQ ( f ) = LQ ( f ).
=

Po-Ning Chen@cm.nctu

Chapter 2-39

2.3 VSB upper sideband transmission

sDSB ( f ) = [ M ( f + f c ) + M ( f f c )] / 2 = [ M ( f + f c ) + M * ( f + f c )] / 2
*
sVSB ( f ) = [ MVSB ( f + f c ) + MVSB
( f f c )] / 2

[1 + LQ ( f + f c )]
[1 + LQ ( f + f c )]
1
= M ( f + f c )
+ M * ( f + f c )

2
2
2

[1 + LQ ( f + f c )]
1
1{ f f c + W }
= M ( f + f c )1{ f c W f f c + W }
2
2
[1 + LQ ( f + f c )]

1{ f f c W }
+ M * ( f + f c )1{ f c W f f c + W }
2

sVSB (t )

mVSB (t )
Po-Ning Chen@cm.nctu

Chapter 2-40

20

M L ( f ) FL ( f ) + M R ( f ) FR ( f ) = [ M L ( f ) + M R ( f )][ FL ( f ) + FR ( f )]
if M L ( f ) FR ( f ) = M R ( f ) FL ( f ) = 0.

1
(M ( f + f c )1{ f c W f f c + W } + M * ( f + f c )1{ f c W f f c + W })
2
[1 + LQ ( f + f c )]

[1 + LQ ( f + f c )]
1{ f f c + W } +
1{ f f c W }

2
2

1
= (M ( f + f c ) + M * ( f + f c ) )
2
[1 + LQ ( f + f c )]

[1 + LQ ( f + f c )]

1{ f f c + W } +
1{ f f c W }
2
2

cont.

[1 + LQ ( f + f c )]

[1 + LQ ( f + f c )]
= sDSB ( f )
1{ f f c + W } +
1{ f f c W }
2
2

1
= sDSB ( f ) ([1 + LQ ( f + f c )]1{ f + f c W } + [1 + LQ ( f + f c )]1{ f + f c W })
2
1
= sDSB ( f ) (2 + LQ ( f + f c ) + LQ ( f + f c ) ) (See next slide.)
2
Po-Ning Chen@cm.nctu

Chapter 2-41

[1 + LQ ( f + f c )]1{ f + f c W }
2

[1 + LQ ( f + f c )]1{ f + f c W }

[1 + LQ ( f + f c )]1{ f + f c W }
+ [1 + LQ ( f + f c )]1{ f + f c W }

LQ ( f + f c )
1

LQ ( f + f c )
2

LQ ( f + f c ) + LQ ( f + f c ) + 2
Chapter 2-42

21

Consequently,
1

sVSB ( f ) = sDSB ( f ) (LQ ( f + f c ) + LQ ( f + f c ) + 2 )

2

sVSB ( f ) = sDSB ( f ) H ( f )

H( f ) =

1
(LQ ( f + f c ) + LQ ( f + f c ) + 2)
2

Po-Ning Chen@cm.nctu

Chapter 2-43

H( f )

H ( f fc )

H ( f + fc )
1
1

H ( f fc ) H ( f + fc )

Po-Ning Chen@cm.nctu

Chapter 2-44

22

H ( f fc ) H ( f + fc )

LQ ( f ) = H ( f f c ) H ( f + f c ) for | f | W
H Q ( f ) = j[ H ( f f c ) H ( f + f c )] for | f | W

Po-Ning Chen@cm.nctu

Chapter 2-45

2.3 Mathematical Representation of VSB

signal
M VSB ( f ) = M ( f )[1 jH Q ( f )] / 2
1
1
M ( f ) jM ( f ) H Q ( f )
2
2
1
1
= M ( f ) + jM ' ( f )
2
2
where M ' ( f ) = M ( f ) H Q ( f ) = jM ( f ) LQ ( f ).
=

Notably, m' (t ) is real. This is an extension of Hilbert Transform.

M ' ( f ) = jM ( f ) LQ ( f ) = jM * ( f ) LQ ( f )
= ( j )* M * ( f ) L*Q ( f ) = [ jM ( f ) LQ ( f )]* = [ M ' ( f )]* .
Po-Ning Chen@cm.nctu

Chapter 2-46

23

2.3 Application of VSB Modulation

Television Signals
1. The video signal exhibits a large bandwidth and
significant low-frequency content.
Hence, no energy gap exists (SSB becomes
impractical).
VSB modulation is adopted to save bandwidth.
Notably, since a rigid control of the transmission VSB
filter at the very high-power transmitter is expensive, a
not-quite VSB modulation is used instead (a little
waste of bandwidth to save cost).
Po-Ning Chen@cm.nctu

Chapter 2-47

2.3 Application of VSB Modulation

VSB Filter for Television Signal Transmissions

55.25 MHz

Po-Ning Chen@cm.nctu

59.75 MHz

Chapter 2-48

24

2.3 Application of VSB Modulation

As the transmission signal is not quite VSB modulated,
the receiver needs to re-shape the received signal
before feeding it to a VSB demodulator.

Po-Ning Chen@cm.nctu

Chapter 2-49

2.3 Application of VSB Modulation

2. In order to save the cost of the receiver (in order to
use envelope detector at the receiver), an additional
carrier is added.
Notably, additional carrier does not increase
bandwidth, but just add transmission power.

1
Ac k a (m(t ) cos(2f c t ) m' (t ) sin( 2f c t ) )
2
1
1

= Ac 1 + ka m(t ) cos(2f c t ) k a Ac m' (t ) sin( 2f c t )

2
2

s (t ) = Ac cos(2f c t ) +

Po-Ning Chen@cm.nctu

Chapter 2-50

25

2.3 Application of VSB Modulation

Distortion of envelope detector
2

1
1

s 2 (t ) = Ac2 1 + k a m(t ) cos2 ( 2f c t ) + k a2 Ac2 ( m' (t )) 2 sin 2 ( 2f c t )

4
2

1
1

ka Ac2 m' (t ) 1 + ka m(t ) sin( 2f c t ) cos(2f c t )

2
2

1 2 1
1 2
2
Ac 1 + ka m(t ) + ka ( m' (t ))

2 2
4 4243
1
Distortion

LowPass

The distortion can be compensated by reducing the amplitude

sensitivity ka or increasing the width of the vestigial sideband. Both
methods are used in the design of Television broadcasting system.
Po-Ning Chen@cm.nctu

Chapter 2-51

2.3 Extension Usage of DSB-SC

Quadrature-Carrier Multiplexing or Quadrature
Amplitude Modulation (QAM)

Synchronization is critical in QAM modulation, which is often achieved by a

separate low-power pilot tone outside the passband of the modulated signal.
Po-Ning Chen@cm.nctu

Chapter 2-52

26

2.4 Frequency Translation

The basic operation of SSB modulation is simply a
special case of frequency translation.
So SSB modulation is sometimes referred to as frequency
changing, mixing, or heterodyning.
The mixer is a device that consists of a product modulator
followed by a band-pass filter, which is exactly what SSB
modulation does.

Po-Ning Chen@cm.nctu

Chapter 2-53

2.4 Frequency Translation

The process is named
upconversion, if f1 + f
is the wanted signal,
and f1 f is the
unwanted image
signal.
The process is named
downconversion, if f1
f is the wanted
signal, and f1 + f is
the unwanted image
signal.
Po-Ning Chen@cm.nctu

Chapter 2-54

27

2.5 Frequency-Division Multiplexing

Multiplexing is a technique to combine a
number of independent signals into a composite
signal suitable for transmission.
Two conventional multiplexing techniques
Frequency-Division Multiplexing (FDM)
Time-Division Multiplexing (TDM)
Will be discussed in Chapter 3.

Po-Ning Chen@cm.nctu

Chapter 2-55

2.5 Frequency-Division Multiplexing

Po-Ning Chen@cm.nctu

Chapter 2-56

28

Example 2.1

Po-Ning Chen@cm.nctu

Chapter 2-57

2.6 Angle Modulation

Angle modulation
The angle of the carrier is varied in accordance with
the baseband signal.

Angle modulation provides us with a practical

means of exchanging channel bandwidth for
improved noise performance.
So to speak, angle modulation can provide better
discrimination against noise and interference than
the amplitude modulation, at the expense of
increased transmission bandwidth.
Po-Ning Chen@cm.nctu

Chapter 2-58

29

2.6 Angle Modulation

Commonly used angle modulation
Phase modulation (PM)
s (t ) = Ac cos[2f ct + k p m(t )], where k p is phase sensitivity.
Frequency modulation (FM)

[
]
= A cos[2f t + 2k m( )d ],

s (t ) = Ac cos 2 0( f c + k f m( )) d
t

where k f is frequency sensitivity.

Po-Ning Chen@cm.nctu

Chapter 2-59

2.6 Angle Modulation

Main differences between Amplitude
Modulation and Angle Modulation
1. Zero crossing spacing of angle modulation no
longer has a perfect regularity as amplitude
modulation does.
2. Angle modulated signal has constant envelope; yet,
the envelope of amplitude modulated signal is
dependent on the message signal.

Po-Ning Chen@cm.nctu

Chapter 2-60

30

2.6 Angle Modulation

Similarity between PM and FM
PM is simply an FM with 0m( )d in place of m(t).
t

sPM (t ) = Ac cos[2f c t + k p m(t )]

t
sFM (t ) = Ac cos 2f c t + 2k f m( )d

Hence, the text only discusses FM in this chapter.

Po-Ning Chen@cm.nctu

Chapter 2-61

2.7 Frequency Modulation

s(t) of FM modulation is a non-linear function of m(t).
t
t
s(t ) = Ac cos 2 f i ( )d = Ac cos 2 ( f c + k f m( ))d

0
0
t
= Ac cos 2f c t + 2k f m( )d

So its general analysis is hard.

To simplify the analysis, we may assume a single-tone
transmission, where
m(t ) = Am cos(2f m t )
Po-Ning Chen@cm.nctu

Chapter 2-62

31

 From the formula in the previous slide,

f i (t ) = f c + k f m(t )
= f c + k f Am cos(2f m t )
= f c + f cos(2f m t )
where f = k f Am is the frequency deviation.
t
s(t ) = Ac cos 2 f i ( )d

0
t
= Ac cos 2 [ f c + f cos(2f m )]d

f
= Ac cos 2f c t +
sin( 2f m t )
fm

where = f / f m is often called the modulation index

of FM signal.
Po-Ning Chen@cm.nctu

Chapter 2-63

Modulation index is the largest deviation from 2fct in

FM system.
s(t ) = Ac cos[2f c t + sin(2f m t )]
As a result,
f c f m = f c f f i (t ) = f c + f cos(2f m t ) f c + f = f c + f m

1. A small corresponds to a narrowband FM.

2. A large corresponds to a wideband FM.

Po-Ning Chen@cm.nctu

Chapter 2-64

32

2.7 Narrowband Frequency Modulation

s(t ) = Ac cos[2f c t + sin( 2f m t )]

= Ac cos(2f c t ) cos[ sin( 2f m t )] Ac sin( 2f c t ) sin[ sin( 2f m t )]
Ac cos(2f c t ) Ac sin(2f c t ) sin(2f m t )

(Often, < 0.3.)

Po-Ning Chen@cm.nctu

Chapter 2-65

2.7 Narrowband Frequency Modulation

Comparison between approximate narrowband
FM modulation and AM (DSB-C) modulation
sFM (t ) Ac cos(2f c t ) Ac sin( 2f c t ) sin(2f m t )
= Ac cos(2f c t ) +

Ac
2

cos(2 ( f c + f m )t )

Ac
2

cos(2 ( f c f m )t )

s AM (t ) = Ac [1 + k a m(t )] cos(2f c t )
= Ac [1 + k a Am cos(2f m t )] cos(2f c t )
= Ac cos(2f c t ) +
Po-Ning Chen@cm.nctu

k a Am
k A
cos(2 ( f + f m )t ) + a m cos(2 ( f f m )t )
2
2
Chapter 2-66

33

2.7 Narrowband Frequency Modulation

Represent them in terms of their low-pass
isomorphism.
A
~
sFM (t ) = ( Ac + j 0) + c [cos( 2f m t ) + j sin(2f m t )]
2
Ac

[cos(2f m t ) j sin( 2f m t )]
2
k A
~
s AM (t ) = ( Ac + j 0) + a m [cos(2f m t ) + j sin(2f m t )]
2
k A
+ a m [cos(2f m t ) j sin(2f m t )]
2
Po-Ning Chen@cm.nctu

Chapter 2-67

2.7 Narrowband Frequency Modulation

Phaser diagram

~
sFM (t )

Ac
2

[cos(2f m t ) j sin( 2f m t )]

Ac

( Ac + j 0)

Let Ac = ka Am .
Po-Ning Chen@cm.nctu

(cos( 2f m t )
2
+ j sin( 2f m t ))

Ac
2

[cos(2f m t ) j sin(2f m t )]

~
s AM (t )
Chapter 2-68

34

2.7 Spectrum of Single-Tone FM

Modulation
s(t ) = Ac cos[2f c t + sin( 2f m t )]

= Re{Ac exp( j[2f c t + sin(2f m t )])}

= Re{~
s (t ) exp( j 2f t )}
c

~
s (t ) = Ac exp( j[ sin( 2f m t )])

~
s (t ) = Ac

J n ( )e j 2fmnt

n =

(Slide Chapter 1-247)

n =

( x )e jn = e jx sin( )

where J n () is the nth order Bessel function of the first kind.

Po-Ning Chen@cm.nctu

Chapter 2-69

2.7 Spectrum of Single-Tone FM

Modulation

j 2ft
S~ ( f ) = ~
s (t )e dt

= Ac J n ( )e j 2f mnt e j 2ft dt

n =

= Ac

n =

= Ac

j 2 ( f nf m ) t
dt
n ( ) e

( ) ( f nf m )

n =

Po-Ning Chen@cm.nctu

Chapter 2-70

35

2.7 Spectrum of Single-Tone FM

Modulation
Consequently,

S( f ) =

Ac
2

1 ~
~
S ( f fc ) + S * ( f fc )
2
A
= c J n ( )[ ( f f c nf m ) + ( f f c nf m )]
2 n =

( )[ ( f f c nf m ) + ( f + f c + nf m )]

n =

Po-Ning Chen@cm.nctu

Chapter 2-71

2.7 Spectrum of Single-Tone FM

Modulation
The power of s(t)
By definition, the time-average autocorrelation function
is given by:

1
T
2T

Rs ( ) = lim

1
T
2T

T E[ s(t ) s(t + )]dt = lim

T

s(t ) s(t + )dt

Hence, the power of s(t) is equal to:

1
T
2T
1
= lim
T
2T

Rs (0) = lim

1 T 2
Ac cos2 [2f ct + sin(2f m t )]dt

T
T
2T
T
Ac2
2 1 + cos[4f c t + 2 sin( 2f m t ) ]

A
dt
T c
2
2
2
T s (t )dt = lim
T

Po-Ning Chen@cm.nctu

Chapter 2-72

36

2.7 Spectrum of Single-Tone FM

Modulation
The time-average power spectral density of a
deterministic signal s(t) is given by
PSD ( f ) = lim
T

1
S ( f ) S2*T ( f )
2T

where S2T ( f ) is the Fourier transform of s(t ) 1{| t | T }.

From

s2T (t ) = Ac J n ( ) cos(2 ( f c + nf m )t ) 1{| t | T }

we obtain:

n =

S2T ( f ) = AcT J n ( )[sinc( 2T ( f f c nf m )) + sinc(2T ( f + f c + nf m ))]

n =

Po-Ning Chen@cm.nctu

Chapter 2-73

For simplicity, assume that 2T increases along the multiple of

1/fm., i.e., 2T = p/fm, where p is an integer. Also assume that fc
is a multiple of fm, i.e., fc = qfm., where q is an integer. Then

PSD ( f )
1
= lim
S ( f ) S2*T ( f )
T
2T
Ac2
= lim
J k ( )[ ( f f c kf m ) + ( f + f c + kf m )]
p
4 k =

J n ( )[sinc( p( f f c nf m ) / f m ) + sinc( p ( f + f c + nf m ) / f m )]
n =

f c + (n 1) f m

f c + nf m
f c + nf m +

Po-Ning Chen@cm.nctu

f c + (n + 1) f m
fm
p
Chapter 2-74

37

PSD ( f ) =
+

Ac2

lim
J
(

)
sinc
(
p
(
f
f
nf
)
/
f
)
J k ( ) ( f f c kf m )

n
c
m
m
4 p n =
k =

n =

+
+

( )sinc( p ( f f c nf m ) / f m ) J k ( ) ( f + f c + kf m )

n =

k =

J n ( )sinc( p( f + f c + nf m ) / f m ) J k ( ) ( f f c kf m )

)
sinc(
p
(
f
f
nf
)
/
f
)
J k ( ) ( f + f c + kf m )
+
+

n
c
m
m
k =

Ac2 2
J
(

(
f
f
nf
)

+
n
J n ( )J n2 q ( ) ( f f c nf m )
c
m
4 n =
n =

J n ( )J n2 q ( ) ( f + f c + nf m ) +

n =

k =

n =

n =

2
n

( ) ( f + f c + nf m )

2
J n ( )[ ( f f c nf m ) + ( f + f c + nf m )]
4 n =

2
c

Po-Ning Chen@cm.nctu

Chapter 2-75

2.7 Average Power of Single-Tone FM

Signal
Hence, the power of a single-tone FM signal is
given by:

Ac2 2
(
)
(
)
PSD
f
df
=
J
+
J n ( ) J n 2 q ( )

2 n =
n =

A
= c 1 + ( 1) n J n ( ) J n + 2 q ( )
2 n =

2
A
c
2
Question: Can we use 2f to be the bandwidth of a
single-tone FM signal?
Po-Ning Chen@cm.nctu

Chapter 2-76

38

Example 2.2
Fix fm and kf,
but vary = f/fm =
kf Am/fm.

Po-Ning Chen@cm.nctu

Chapter 2-77

Example 2.2
Fix Am and kf,
but vary = f/fm
= kf Am/fm.

Po-Ning Chen@cm.nctu

Chapter 2-78

39

2.7 Spectrum of Narrowband Single-Tone

FM Modulation
J 0 ( ) 1

When is small, J 1 ( )
.
2

J n ( ) 0 for n > 2

Po-Ning Chen@cm.nctu

Chapter 2-79

2.7 Spectrum of Narrowband Single-Tone

FM Modulation
This results in an approximate spectrum for
narrowband single-tone FM signal spectrum as
Ac2 2
J n ( )[ ( f f c nf m ) + ( f + f c + nf m )]
4 n =
A2
c J 21 ( )[ ( f f c + f m ) + ( f + f c f m )]
4
A2
+ c J 02 ( )[ ( f f c ) + ( f + f c )]
4
Ac2 2
+
J 1 ( )[ ( f f c f m ) + ( f + f c + f m )]
4

PSD ( f )

Po-Ning Chen@cm.nctu

Chapter 2-80

40

J n ( ) = ( 1)n J n ( ) J n2 ( ) = J 2n ( )
Ac2 2
PSD ( f ) =
J 1 ( )[ ( f f c + f m ) + ( f + f c f m )]
4
Ac2 2
+
J 0 ( )[ ( f f c ) + ( f + f c )]
4
A2
+ c J 12 ( )[ ( f f c f m ) + ( f + f c + f m )]
4
2 Ac2
[ ( f f c + f m ) + ( f + f c f m )]
=
16
A2
+ c [ ( f f c ) + ( f + f c )]
4
2 Ac2
[ ( f f c f m ) + ( f + f c + f m )]
+
16
Po-Ning Chen@cm.nctu

Chapter 2-81

Ac2
4

Ac2
4

2 Ac2

2 Ac2

2 Ac2

2 Ac2

16

16

16

16

fc

Po-Ning Chen@cm.nctu

fm
fc

Chapter 2-82

41

2.7 Transmission Bandwidth of FM signals

Carsons rule An empirical bandwidth
An empirical rule for Transmission Bandwidth of
FM signals
For large , the bandwidth is essentially 2f.
For small , the bandwidth is effectively 2fm.
So Carson proposes that:

1
BT 2f + 2 f m = 2 f 1 +

Po-Ning Chen@cm.nctu

Chapter 2-83

2.7 Transmission Bandwidth of FM signals

Universal-curve transmission bandwidth
The transmission bandwidth of an FM wave is the
minimum separation between two frequencies
beyond which none of the side frequencies is
greater than 1% of the carrier amplitude obtained
when the modulation is removed.
S( f ) =

Ac
2

Ac cos(2f c t )
Po-Ning Chen@cm.nctu

( )[ ( f f c nf m ) + ( f + f c + nf m )]

n =

Ac
[ ( f f c ) + ( f + f c )]
2
Chapter 2-84

42

A
A
BT = 2nmax f m , where nmax = n : c | J n ( ) |> 0.01 c .
2
2

2nmax BT / f
0.1 2 20.0
BT 2nmax f m 2nmax
=
=
0.3 4 13.3
f m

f
8.0
0.5 4
6.0
For fixed f , a smaller causes a larger BT . 1.0 6
4.0
2.0 8
3.2
5.0 16
2.8
10.0 28
2.5
20.0 50
2.3
30.0 70
Po-Ning Chen@cm.nctu

Chapter 2-85

2.7 Bandwidth of a General FM wave

Now suppose m(t) is no longer a single tone but a
general message signal of bandwidth W.
Hence, the worst-case tone is fm = W, and D = f / W.
For nonsinusoidal modulation, the deviation ratio D is
used instead of the modulation index .
The derivation ratio D plays the same role for
nonsinusoidal modulation that the modulation index
plays for the case of sinusoidal modulation.

We can then use Carsons rule or universal curve to

determine the transmission bandwidth BT.
Po-Ning Chen@cm.nctu

Chapter 2-86

43

2.7 Bandwidth of a General FM wave

Final notes
Carsons rule usually underestimates the
transmission bandwidth.
Universal curve is too conservative in bandwidth
estimation.
So, a choice of a transmission bandwidth inbetween is acceptable for most practical purposes.

Po-Ning Chen@cm.nctu

Chapter 2-87

Example 2.3
FM radio in North America requires the
maximum frequency derivation f = 75 kHz.
If some message signal has bandwidth W = 15 kHz,
then the deviation ratio D = 75/15 = 5.
Then
1
1

BT ,Carson = 2f 1 + = 2 751 + = 180 kHz

D
D
2n
16
BT ,Universal Curve = max f = 75 = 240 kHz
D
5
Po-Ning Chen@cm.nctu

Chapter 2-88

44

Example 2.3
In practice, a bandwidth of 200 kHz is allocated to
each FM transmitter.
So Carsons rule underestimates BT, while
Universal Curve overestimates BT.

Po-Ning Chen@cm.nctu

Chapter 2-89

2.7 Generation of FM Signals

Direct FM
Carrier frequency is directly varied in accordance with the
message signal as accomplished using a voltage-controlled
oscillator.
Indirect FM
The message is first integrated and sent to a phase modulator.
So, the carrier frequency is not directly varied in accordance to
the message signal.

Po-Ning Chen@cm.nctu

Chapter 2-90

45

2.7 Generation of FM Signals

s(t ) = Ac cos 2f c t + 2k f m( )d

m( )d

(See next slide.)

Po-Ning Chen@cm.nctu

Chapter 2-91

2.7 Generation of FM Signals

Frequency multiplier

v (t ) = a1s(t ) + a2 s 2 (t ) + L + an s n (t )
n

t
= Aj cos2if c t + 2ik f m( )d (See next slide.)

0
i =1
t
An cos2nf c t + 2nk f m( )d

BandPass

t
= Ac' cos2f c't + 2k 'f m( )d

0
Po-Ning Chen@cm.nctu

Chapter 2-92

46

1
2
cos ( x ) = 2 (cos(2 x ) + 1)

cos3 ( x ) = 1 (cos(3x ) + 3 cos( x ))

1
4
cos ( x ) = (cos(4 x ) + 4 cos(2 x ) + 3 cos( x ))
8

1
5
cos ( x ) = 16 (cos(5 x ) + 5 cos(3x ) + 10 cos( x ))

Po-Ning Chen@cm.nctu

Chapter 2-93

2.7 Demodulation of FM Signals

Indirect Demodulation Phase-locked loop
(Will be introduced in Section 2.14)

Direct Demodulation
Balanced frequency discriminator
s1 (t )

|~
s1 (t ) |
~
so (t )

s(t )

differentiation filters
s2 ( t )
Po-Ning Chen@cm.nctu

|~
s2 (t ) |
Chapter 2-94

47

j 2a f f c +

H 1 ( f ) = j 2a f + f c

0,

BT
2
BT
2

, | f f c | T
2

, | f + f c | T
2

elsewhere

BT
BT

j 2a f + f c + 2 , | f + f c | 2

B
B

H 2 ( f ) = j 2a f f c T , | f f c | T
2
2

0,
elsewhere

Po-Ning Chen@cm.nctu

Chapter 2-95

Po-Ning Chen@cm.nctu

Chapter 2-96

48

2.7 Analysis of Direct Demodulation in

terms of Low-Pass Equivalences
B

~
2 H ( f + f c ), | f | T
H1 ( f ) = 1
2

0,
otherwise

B
B

j 4a f + T , | f | T
=
2
2

0,
otherwise

B ~
B

1 ~
~
~
j 2a f + T S ( f ), | f | T
S1 ( f ) = H 1 ( f ) S ( f ) =
2
2

0,
elsewhere
s (t )

d~
s1 (t ) = a
s ( t )
~
+ jBT ~

dt
Po-Ning Chen@cm.nctu

Chapter 2-97

t
s(t ) = Ac cos2f c t + 2k f m( )d

0
t
s (t ) = Ac exp j 2k f m( )d
~

s (t )
d~

+ jBT ~
s1 (t ) = a
s ( t )
~
dt

= a jAc 2k f m(t ) exp j 2k f m( )d

+ jBT Ac exp j 2k f m( )d

t
2k

= jBT aAc 1 + f m(t ) exp j 2k f m( )d

0
BT

Po-Ning Chen@cm.nctu

Chapter 2-98

49

 s1 (t ) = Re{~
s1 (t ) exp( j 2f c t )}
t

2k
= Re jBT aAc 1 + f m(t ) exp j 2k f m( )d exp( j 2f c t )

0
BT

t
2k

= BT aAc 1 + f m(t ) sin 2f c t + 2k f m( )d

0

BT

t

2k

= BT aAc 1 + f m(t ) cos 2f c t + 2k f m( )d +

0
2
BT

If

2k f
BT

m(t ) < 1 and f c >> W , then envelope detector can be used

to obtain the amplitude of the lowpass equivalent message.

2k

s1 (t ) |= BT aAc 1 + f m(t )
|~
BT

Po-Ning Chen@cm.nctu

Similarly,

Chapter 2-99

B
B

~
j 4a f T , | f | T
H2 ( f ) =
2
2

0,
otherwise

B ~
B

1 ~
~
~
j 2a f T S ( f ), | f | T
S2 ( f ) = H 2 ( f ) S ( f ) =
2
2

0,
elsewhere
s (t )
d~

~
s2 (t ) = a
s ( t )
jBT ~
dt

t
2k

= jBT aAc 1 f m(t ) exp j 2k f m( )d

0
BT

Po-Ning Chen@cm.nctu

Chapter 2-100

50

s2 (t ) exp( j 2f c t )}
s2 (t ) = Re{~
t
2k

= BT aAc 1 f m(t ) cos 2f c t + 2k f m( )d +

0
2
BT

2k

s2 (t ) |= BT aAc 1 f m(t )
| ~
BT

~
s (t ) =| ~
s (t ) | | ~
s (t ) |= 4k aA m(t )
o

Final Note: a is a filter parameter of the filters, which can be

used to adjust the amplitude of the resultant output.

Po-Ning Chen@cm.nctu

Chapter 2-101

2.7 FM Stereo Multiplexing

How to do Stereo Transmission in FM radio?
Two requirements:
Backward compatible with monophonic radio receivers
Operate within the allocated FM broadcast channels

To fulfill these requirements, the baseband message

signal has to be re-made.
m(t ) = [ml (t ) + mr (t )] + [ml (t ) mr (t )] cos(4f m t ) + K cos(2f m t )
where ml (t ) and mr (t ) are message signals picked up by left
and right microphones, respectively.
Po-Ning Chen@cm.nctu

Chapter 2-102

51

fm = 19 kHz

m(t ) = [ml (t ) + mr (t )] + [ml (t ) mr (t )] cos(4f m t ) + K cos(2f m t )

144244
3
142
4 43
4
For monophonic reception

Po-Ning Chen@cm.nctu

For coherent detection

Chapter 2-103

Demultiplexer in receiver of FM stereo.

Po-Ning Chen@cm.nctu

Chapter 2-104

52

2.8 Nonlinear Effects in FM Systems

The channel (including background noise, interference and
circuit imperfection) may introduce non-linear effects on
the transmission signals.
For example, non-linearity due to amplifiers.
0.8

1 3

vi (t ) 3 vi (t ), | vi (t ) | 1

2
vo (t ) =
vi (t ) > 1
,
3

vi (t ) < 1
,
3

0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1.5

-1

-0.5

Po-Ning Chen@cm.nctu

0.5

1.5

Chapter 2-105

2.8 Nonlinear Effects in FM Systems

Suppose

2
3
vo (t ) = a1vi (t ) + a2 vi (t ) + a3vi (t )

vi (t ) = Ac cos[2f c t + (t )]

t
(t ) = 2k f 0m( )d
2
3
Then vo (t ) = a1vi (t ) + a2 vi (t ) + a3vi (t )

= a1 Ac cos[2f c t + (t )] + a1 Ac2 cos2 [2f c t + (t )]

+ a1 Ac3 cos3 [2f c t + (t )]
Po-Ning Chen@cm.nctu

Chapter 2-106

53

1
vo (t ) = a1 Ac cos[2f c t + (t )] + a2 Ac2 (1 + cos[4f c t + 2 (t )])
2
1
+ a3 Ac3 (3 cos[2f c t + (t )] + cos[6f c t + 3 (t )])
4
1
3

= a1 Ac2 + a1 Ac + a3 Ac3 cos[2f c t + (t )]

2
4 44
1444
42444444
3
BT ,Carson = 2 f + 2W

1
1
+ a2 Ac2 cos[4f c t + 2 (t )] + a1 Ac3 cos[6f c t + 3 (t )]
2 444424444
4 444424444
1
3 1
3
BT ,Carson = 4 f + 2W

BT ,Carson = 6 f + 2W

Thus, in order to recover s(t) from vo(t) using band-pass filter,

it requires:

2 f c ( 4 f + 2W ) / 2 > f c + ( 2 f + 2W ) / 2

Po-Ning Chen@cm.nctu

Chapter 2-107

or equivalently, f c > 3f + 2W .
The filtered output is therefore:

vo ,filtered (t ) = a1 Ac + a3 Ac3 cos[2f c t + (t )]

4

Observations
Unlike AM modulation, FM modulation is not affected
by distortion produced by transmission through a channel
with amplitude nonlinearities.
So the FM modulation allows the usage of highly nonlinear amplifiers and power transmitters.

Po-Ning Chen@cm.nctu

Chapter 2-108

54

2.8 AM-to-PM Conversion

Although FM modulation is insensitive to amplitude
nonlinearity, it is indeed very sensitive to phase
nonlinearity.
A common type of phase nonlinearity encountered in
microwave radio transmission is the AM-to-PM conversion.
The AM-to-PM conversion is owing to that the phase
characteristic of amplifiers (or repeaters) also depends on the
instantaneous amplitude of the input signal.
Notably, the nonlinear amplifiers discussed previously will
leave the phase of the input unchanged.
Often, it requires that the peak phase change for a 1-dB
change in input envelope is less than 2%.
Po-Ning Chen@cm.nctu

Chapter 2-109

2.9 Superheterodyne Receiver

A commercial radio communication system
contains not only the transmission but also
some other functions, such as:
Carrier-frequency tuning, to select the desired
signals
Filtering, to separate the desired signal from other
unwanted signals
Amplifying, to compensate for the loss of signal
power incurred in the course of transmission
Po-Ning Chen@cm.nctu

Chapter 2-110

55

2.9 Superheterodyne Receiver

A superheterodyne receiver or superhet is designed to
facilitate the fulfillment of these functions, especially
the first two.
It overcomes the difficulty of having to build a tunable highly
selective and variable filter ( rather a fixed filter is applied on
IF section).

heterodyning function
Po-Ning Chen@cm.nctu

Chapter 2-111

2.9 Superheterodyne Receiver

Example

AM Radio

FM Radio

RF carrier range

0.535-1.605 MHz

88-108 MHz

Midband frequency of IF section

0.455 MHz

10.7 MHz

IF bandwidth

10 kHz

200 kHz

First detector
Po-Ning Chen@cm.nctu

Second detector
Chapter 2-112

56

2.9 Image Interference

Fix fIF and fLO at the receiver end. What is the
fRF that will survive at the IF section output?

Answer: f RF =| f LO f IF |
Example. Suppose the receiver use 1.105MHz local
oscillator, and receives two RF signals respectively
centered at 0.65MHz and 1.56MHz.
f IF

f IF

0.455 MHz 0.455 MHz

f RF

f Interference

0.65 MHz

1.56 MHz

f LO
1.105 MHz
Po-Ning Chen@cm.nctu

Chapter 2-113

2.9 Image Interference

Then at the output of the IF section, we respectively obtain:

f IF = f LO f RF

f Im age = f LO + f RF = 1.755 MHz

0.455 MHz

f Image Interference = f Interference f LO = 0.455 MHz

Po-Ning Chen@cm.nctu

f Interference + f LO = 2.665 MHz

Chapter 2-114

57

2.9 Image Interference

A cure of image interference is to employ a highly
selective stages in the RF session in order to favor the
desired signal (at fRF) and discriminate the undesired
signal (at fRF + 2fIF or fRF 2fIF).

Po-Ning Chen@cm.nctu

Chapter 2-115

2.9 Advantage of Constant Envelope for FM

modulation
Observations
For FM modulation, any variation in amplitude is
caused by noise or interference.
For FM modulation, the information is resided on
the variations of the instantaneous frequency.
So we can use an amplitude limiter to remove the
amplitude variation, but to retain the frequency
variation, after the IF section.

Po-Ning Chen@cm.nctu

Chapter 2-116

58

2.9 Advantage of Constant Envelope for FM

modulation
Amplitude limiter
Clipping the modulated wave at the IF section
output (almost to the zero axis) to result in a nearrectangular wave.
Pass the rectangular wave through a bandpass filter
centered at fIF to suppress harmonics (due to
clipping).
Then the filter output retains the frequency
variation with constant amplitude.
Po-Ning Chen@cm.nctu

Chapter 2-117

2.10 Noise in CW Modulation Systems

To simplify the system analysis, we assume:

ideal band-pass filter (that is just wide enough to pass the
modulated signal s(t) without distortion),
ideal demodulator,
Gaussian distributed white Noise process.

So the only source of imperfection is from the noise.

Po-Ning Chen@cm.nctu

Chapter 2-118

59

2.10 Noise in CW Modulation Systems

BT
So after passing through the ideal bandpass filter, s(t) is
unchanged but w(t) becomes a narrowband noise n(t).
Hence,
x (t ) = s(t ) + n(t ),

where n(t ) = n I (t ) cos(2f c t ) nQ (t ) sin(2f c t ).

Po-Ning Chen@cm.nctu

Chapter 2-119

2.10 Noise in CW Modulation Systems

Input signal-to-noise ratio (SNRI)
The ratio of the average power of the modulated
signal s(t) to the average power of the filtered noise
n(t).

Output signal-to-noise ratio (SNRO)

The ratio of the average power of the demodulated
message signal to the average power of the noise,
measured at the receiver output.

Po-Ning Chen@cm.nctu

Chapter 2-120

60

2.10 Noise in CW Modulation Systems

It is sometimes advantageous to look at the lowpass
equivalent model.
Channel signal-to-noise ratio (SNRC)
The ratio of the average power of the modulated
signal s(t) to the average power of the channel noise
in the message bandwidth, measured at the receiver
input (as illustrated below).

Po-Ning Chen@cm.nctu

Chapter 2-121

2.10 Noise in CW Modulation Systems

Notes
SNRC is nothing to do with the receiver structure, but
depends on the channel characteristic and modulation
approach.
SNRO is however receiver-structure dependent.
Finally, define the figure of merit for the receiver as:

figure of merit =

Po-Ning Chen@cm.nctu

SNRO
SNRC

Chapter 2-122

61

2.11 Noise in Linear Receivers Using

Coherent Detection
Recall that for AM demodulation
When the carrier is suppressed, linear coherent
detection is used. (Section 2.11)
When the carrier is additionally transmitted,
nonlinear envelope detection is used. (Section 2.12)

The noise analysis of the above two cases are

respectively addressed in Sections 2.11 and
2.12.
Po-Ning Chen@cm.nctu

Chapter 2-123

2.11 Noise in Linear Receivers Using

Coherent Detection

cos(2f c t )

m(t ) : stationary with zero meam and PSD S M ( f ) bandlimited to W .

s (t ) = CAc m(t ) cos( 2f c t )

Po-Ning Chen@cm.nctu

Chapter 2-124

62

2.11 Noise in Linear Receivers Using

Coherent Detection
Average signal power
1 T
1
2
lim
[
(
)]
lim
=
E
s
t
dt

T
T
2T T
2T
1
= lim
T
2T

E [C
T

Ac2 cos2 ( 2f ct )m 2 (t )]dt

C 2 Ac2 cos2 ( 2f ct ) E [m 2 (t )]dt

1
T
2T

= C 2 Ac2 P lim
=

cos2 ( 2f c t )dt

C 2 Ac2 P
2

where P = E [m 2 (t )] = S M ( f )df is the message power.

W

Po-Ning Chen@cm.nctu

Chapter 2-125

2.11 Noise in Linear Receivers Using

Coherent Detection
Noise power in the message bandwidth

S w ( f )df = W
W
W

Po-Ning Chen@cm.nctu

N0
df = WN 0
2

Chapter 2-126

63

2.11 Noise in Linear Receivers Using

Coherent Detection

cos(2f c t )

Channel SNR for DSC-SC and coherent detection (observed at x(t))

C 2 Ac2 P 2 C 2 Ac2 P
=
WN o
2WN o
Next, we calculate output SNR (observed at y(t)) under the condition
that the transmitter and the receiver are perfectly synchronized.
SNRC ,DSB-SC =

Po-Ning Chen@cm.nctu

Chapter 2-127

x (t ) = s (t ) + n (t )
= CAc m(t ) cos(2f ct ) + nI (t ) cos(2f c t ) nQ sin(2f c t )
v (t ) = x (t ) cos(2f ct )
= [CAc cos(2f ct )m(t ) + nI (t ) cos(2f ct ) nQ (t ) sin(2f ct )] cos(2f ct )
= CAc cos2 ( 2f ct )m(t ) + nI (t ) cos2 ( 2f ct ) nQ (t ) sin(2f c t ) cos(2f ct )
1
1
CAc m(t ) + nI (t )
2
2
1
1
(Recall E[n 2 (t )] = E [n I2 (t )] = E [nQ2 (t )].)
y (t ) = CAc m(t ) + nI (t )
2
2
LowPass

SNRO , DSB-SC =

E [C 2 Ac2 m 2 (t ) / 4] C 2 Ac2 P C 2 Ac2 P

=
=
E [nI2 (t ) / 4]
E[n 2 (t )] 2WN 0

Po-Ning Chen@cm.nctu

Chapter 2-128

64

 Figure of merit for DSB - SC and coherent detection = 1.

Similar derivation on SSB and coherent detection yields the

same figure of merit.
Conclusions
Coherent detection for SSB performs the same as
coherent detection for DSB-SC.
There is no SNR degradation for SSB and DSB-SC
coherent receivers. The only effect of these modulation
and demodulation processes is to translate the message
signal to a different frequency band to facilitate its
transmission over a band-pass channel.
No trade-off between noise performance and bandwidth.
This may become a problem when high quality
transceiving is required.
Po-Ning Chen@cm.nctu

Chapter 2-129

2.12 Noise in AM Receivers Using Envelope

Detection

s(t ) = Ac [1 + ka m(t )] cos( 2f c t )

lim
T

1
2T

E [ s 2 (t )]dt = Ac2 E [(1 + ka m(t )) 2 ]lim

T
=

Po-Ning Chen@cm.nctu

Ac2
(1 + ka2 P )
2

1
2T

cos2 ( 2f c t )dt

(Assume m(t ) zero mean.)

Chapter 2-130

65

Also, W S w ( f )df = W
W

N0
df = WN 0
2

Hence, channel SNR for DSB-C is equal to:

SNRC ,AM =

Ac2 (1 + ka2 P )
2WN o

Next, we calculate output SNR (observed at y(t)) under the condition that
the transmitter and the receiver are perfectly synchronized.

x (t ) = s (t ) + n(t )
= Ac [1 + ka m(t )] cos( 2f c t ) + nI (t ) cos(2f c t ) nQ (t ) sin( 2f ct )
= [ Ac (1 + ka m(t )) + nI (t )] cos(2f ct ) nQ (t ) sin( 2f c t )
Po-Ning Chen@cm.nctu

Chapter 2-131

envelop detector
x
x

y (t ) =

(x ( t ) )
2

LowPass

1
[ Ac (1 + ka m(t )) + nI (t )]2 + nQ2 (t )
2
1
[ Ac (1 + ka m(t )) + nI (t )] if Ac [1 + ka m(t )] >> n~(t )

2
(Refer to slide 136.)
block DC
1
=
[ Ac ka m(t ) + nI (t )]
2

Po-Ning Chen@cm.nctu

Chapter 2-132

66

 SNRO ,AM =

E [Ac2 ka2 m 2 (t ) / 2]
Ac2 ka2 P
Ac2 ka2 P
=
=
E [n I2 (t ) / 2]
E[n 2 (t )] 2WN 0

SNRO ,AM
A2 k 2 P ( 2WN 0 )
k 2P
2 c a 2
= a 2 <1
SNRC ,AM Ac (1 + ka P ) ( 2WN o ) 1 + ka P

Conclusion
Even if the noise power is small compared to the
average carrier power at the envelope detector output,
the noise performance of a full AM receiver is
inferior to that of a DSB-SC receiver due to the
wastage of transmitter power.
Po-Ning Chen@cm.nctu

Chapter 2-133

Example 2.4 Single-Tone Modulation

Assume m(t ) = Am cos(2f m t )
s(t ) = Ac [1 + ka Am cos(2f m t )] cos(2f c t )
Hence,
T
1
1 T
2
2
[1 + ka Am cos(2f mt )]2 cos2 (2f ct )dt
lim
E
[
s
(
t
)]
dt
A
lim
=
c

T
T
T
T
2T
2T
1 T
(cos2 (2f ct ) + 2ka Am cos(2f mt ) cos2 (2f ct )
= Ac2 lim
T
2T T
+ ka2 Am2 cos2 ( 2f m t ) cos2 ( 2f c t ) )dt
k 2 A2 A2
1
= Ac2 + 0 + a m = c (1 + ka2 P )
4 2
2
Po-Ning Chen@cm.nctu

Chapter 2-134

67

Am2
1 T 2
2
A
f
t
dt
=

cos
(
2
)
.
m
m
T 0
2
Following similar procedure as previous discussion,
where P = lim
T

SNRO ,AM
k 2P
k 2 A2 / 2
a 2 = a 2m 2 .
SNRC ,AM 1 + ka P 1 + ka Am / 2
So even if for 100% percent modulation (kaAm = 1), the figure of merit
= 1/3. This means that an AM system with envelope detection must
transmit three times as much average power as DSB-SC with coherent
detector to achieve the same quality of noise performance.

Po-Ning Chen@cm.nctu

Chapter 2-135

2.12 Threshold Effect

What if Ac [1 + ka m(t )] >> n~(t )
with envelope detection)?

is violated (in AM modulation

x (t ) = s(t ) + n(t )
= Ac [1 + ka m(t )] cos( 2f c t ) + nI (t ) cos(2f c t ) nQ (t ) sin( 2f ct )
= [ Ac (1 + ka m(t )) + nI (t )] cos(2f ct ) nQ (t ) sin( 2f c t )
B >>| n~ | ( B + nI ) 2 + nQ2 = B 2 + 2nI B + nI2 + nQ2 B 2 + 2nI B + nI2 B + nI
B <<| n~ | ( B + nI ) 2 + nQ2 = B 2 + 2nI B + | n~ |2 B 2 + 2 | n~ | B + | n~ |2 B + | n~ |
Po-Ning Chen@cm.nctu

Chapter 2-136

68

y (t ) =

(x ( t ) )
2

LowPass

Assume Ac [1 + ka m(t )] << n~(t ) .

1
[ Ac (1 + ka m(t )) + nI (t )]2 + nQ2 (t )
2
1
=
Ac2 (1 + ka m(t )) 2 + 2nI (t ) Ac (1 + ka m(t )) + | n~ (t ) |2
2
1

Ac2 (1 + ka m(t )) 2 + 2 | n~ (t ) | Ac (1 + ka m(t )) + | n~(t ) |2

2
1
=
( Ac (1 + ka m(t )) + | n~(t ) |)2
2
1

( Ac (1 + ka m(t )) + | n~(t ) |)
2
block DC
1
=
[ Ac ka m(t ) + | n~(t ) |]
2
=

Po-Ning Chen@cm.nctu

Chapter 2-137

Ac ka m(t ) + nI (t ), Ac [1 + ka m(t )] >>| n~(t ) |

2 y (t ) =
~
~
Ac ka m(t ) + | n (t ) |, Ac [1 + ka m(t )] <<| n (t ) |

SNRO ,AM =

E [Ac2 ka2 m 2 (t )]
Ac2 ka2 P
Ac2 ka2 P
=
=
E [| n~(t ) |2 ]
E[nI2 (t )] + E [nQ2 (t )] 4WN 0

SNRO ,AM
1
A2 k 2 P ( 4WN 0 )
ka2 P
2 c a 2
=
<
2
SNRC ,AM Ac (1 + ka P ) ( 2WN o ) 2(1 + ka P ) 2

Po-Ning Chen@cm.nctu

Chapter 2-138

69

2.12 Threshold Effect

Threshold effect

Ac2 / 2
Ac2
=
=
2WN 0 4WN 0

For AM with envelope detection, there exists a

carrier-to-noise ratio (namely, the power ratio
between unmodulated carrier Ac cos(2fct) and the
passband noise n(t) ) below which the noise
performance of a detector deteriorates rapidly.
Ac2 ka2 P
2
~
2WN = 2ka P , if Ac [1 + ka m(t )] >>| n (t ) |
SNRO ,AM = 2 20
Ac ka P = ka2 P , if Ac [1 + ka m(t )] <<| n~(t ) |
4WN 0
Po-Ning Chen@cm.nctu

Chapter 2-139

2.12 General Formula for SNRO in Envelope

Detection
For envelope detector, the noise is no longer additive; so the
original definition of SNRO (which is based on additive
noise) may not be well-applied.
A new definition should be given:
Definition. The (general) output signal-to-noise ratio
for an output y(t) due to a carrier input is defined as
SNRO =

so2
Var[ y (t )]

where so = E [ y (t )] E [ yo (t )], and yo (t ) is equal to y (t )

in the presense of noise alone.
Po-Ning Chen@cm.nctu

y (t ) = so + yo (t )
Chapter 2-140

70

2.12 General Formula for SNRO in Envelope

Detection
so is named the mean output signal.
Var[y(t)] is named the mean output noise power.
Example. y(t) = A + nI(t), where nI(t) is zero mean..
so = E[ A + nI (t )] E [nI (t )] = A

2
Var[ y (t )] = Var[n I (t )] = E[nI (t )]

SNRO =

A2
E[nI2 (t )]

This somehow shows the backward compatibility of the new

definition.
Po-Ning Chen@cm.nctu

Chapter 2-141

2.12 General Formula for SNRO in Envelope

Detection
Now for an envelope detector, the output due to a carrier
input and additive Gaussian noise channel is given by:
y (t ) = ( A + nI (t )) 2 + nQ2 (t )
I ( ) = modified Bessel function

y (t ) is Rician distributed with pdf of0 the first kind of zero order.
y 2 + A2 Ay
y
I 0 2 for y 0, where N2 = E [n 2 (t )].
f y ( t ) ( y ) = 2 exp
2
N
2

yo (t ) = n I2 (t ) + nQ2 (t ) is Rayleigh distributed with pdf

y2
y
for y 0.
f y ( t ) ( y ) = 2 exp
2
N
2 N
o

Po-Ning Chen@cm.nctu

Chapter 2-142

71

E[ y (t )] = 0 yf y ( t ) ( y )dy

= 0

y 2 + A2 Ay
dy
I0

exp
N2
2 N2 N2

y2

u2
N
2
=
exp( ) 0 u exp
I 0 (u )du,
(2 )3/ 2
4

by taking u = Ay / N2 and = A2 /(2 N2 ) = A2 /(4WN 0 ).

E [ yo (t )] = yf y
0

( t ) ( y ) dy =

y2

exp
2
N2
2 N

y2

dy

2 N2 z exp( z )dz, by taking z = y 2 /( 2 N2 )

= 2 N z ( 3 / 2 ) 1 exp( z )dz
0

3
.
= 2 N = N
2
2
Po-Ning Chen@cm.nctu

Chapter 2-143

Appendix 4 Confluent Hypergeometric

Functions
The Kummer confluent hypergeometric function is a
solution of Kummers equation
dy
d2y
+ (b x ) ay = 0 for a, b complex
2
dx
dx
with boundary conditions y (0) = 1 and y ' (0) = a / b.
x

For b 0, 1, 2, ., the Kummer confluent

hypergeometric function is equal to 1F1(a;b;x).
Generalize d hypergeometric function

( a1 ) k ( a2 ) k L ( a p ) k x k
( a ) k = a ( a + 1) L ( a + k 1)
r r
(
;
;
)
.
F
a
b
x
=
, where

p q
k!
k =0 ( b1 ) k ( b2 ) k L ( bq ) k
( a )0 = 1
Po-Ning Chen@cm.nctu

Chapter 2-144

72

A4.2 Properties of the Confluent

Hypergeometric Function
a
x as
b
2. 1 F1 ( 1;1; x ) = 1 x.

1. 1 F1 ( a; b; x ) 1 +

x 0.

x
x
x
3. 1 F1 ( 1 / 2;1; x ) = exp (1 + x ) I 0 + xI 2
2
2
2
2
4.

as x .

u m 1 exp( b 2u 2 ) I 0 (u )du =

( m / 2) m
1
;1; 2
1 F1
m

2b
2 4b

5. exp( u ) 1 F1 ( ; ; u ) =1 F1 ( ; ;u )
Po-Ning Chen@cm.nctu

Chapter 2-145

2.12 General Formula for SNRO in Envelope

Detection
Hence,
E [ y (t )] =
=
=
=

u2
N
2

exp(
)
exp

u

I 0 (u )du
0
(2 )3 / 2
4

( 3 / 2) 3
N

exp( )
F ;1; By Property 4
3/ 2
3 / 2 1 1
(2 )
2( 4 ) 2

Po-Ning Chen@cm.nctu

3
2

N exp( ) 1 F1 ;1;

1
2

By Property 5

N 1 F1 ;1;

Chapter 2-146

73

2.12 General Formula for SNRO in Envelope

Detection
As a result,
so = E [ y (t )] E [ yo (t )] =

2
Similarly, we can obtain:

1
2

N 1 F1 ;1;

A2
Var[ y (t )] = 2 1 F1 1;1; 2
2 N

2
N

A2
2 N2

1.

1
A2
1 F1 ;1; 2
2 N
4 2

A2 1
A2

F
= 2 1 +

;
1
;

1
1
2
2 N2
2 N 4 2

2
N

Po-Ning Chen@cm.nctu

By Property 2

Chapter 2-147

2.12 General Formula for SNRO in Envelope

Detection
This concludes to:
( SNR )O =

[1 F1 ( 1 / 2;1; ) 1]2 , where = A2

4
2
2 N2
(1 + ) [1 F1 ( 1 / 2;1; )]

2 / 1

, as
4 (1 + ) 2 / 2

2
[(1 + / 2) 1]
, as 0
4
2

+
(
1
)
(
1
/
2
)

Po-Ning Chen@cm.nctu

Chapter 2-148

74

(Continue from the previous slide.)

+ ,

4
( SNR )O =
2

16(1 + ) ( 2 + ) 2
as
,

2

16 4 , as 0
as
,
=
2
0.91 , as 0

as
as 0

Po-Ning Chen@cm.nctu

Chapter 2-149

2.12 General Formula for SNRO in Envelope

Detection
Curve of SNRO =

[1 F1 ( 1 / 2;1; ) 1]2
4
2
(1 + ) [1 F1 ( 1 / 2;1; )]

and the two limiting approximates.

000

100

10

SNRO ,AM = 2ka2 P

SNRO ,AM = ka2 P

0.1
0.1
Po-Ning Chen@cm.nctu

10

100

Chapter 2-150

75

2.12 General Formula for SNRO in Envelope

Detection
Remarks
For large carrier-to-noise ratio , the envelope detector
behaves like a coherent detector in the sense that the output
SNR is proportional to .
For small carrier-to-noise ratio , the (newly defined) output
signal-to-noise ratio of the envelope detector degrades faster
than a linear function of (decrease at a rate of 2).
From threshold effect and the general formula for SNRO,
we may see that the envelope detector favors a strong signal.
This is sometimes called weak signal suppression.

Po-Ning Chen@cm.nctu

Chapter 2-151

2.13 Noise in FM Receivers

To simplify the system analysis, we assume:

ideal band-pass filter (that is just wide enough to pass the
modulated signal s(t) without distortion),
ideal demodulator,
Gaussian distributed white Noise process.

So the only source of imperfection is from the noise.

Po-Ning Chen@cm.nctu

Chapter 2-152

76

nI (t ) = r (t ) cos( (t ))

nQ (t ) = r (t ) sin( (t ))

s (t ) = Ac cos[2f c t + (t )], where (t ) = 2k f m( )d .

t

x (t ) = Ac cos[2f ct + (t )] + r (t ) cos[2f ct + (t )]
= Ac cos[2f ct + (t )] + r (t ) cos[2f ct + (t ) + (t ) (t )]
= ( Ac + r (t ) cos[ (t ) (t )]) cos[2f ct + (t )]
r (t ) sin[ (t ) (t )] sin[2f ct + (t )]

( Ac + r (t ) cos[ (t ) (t )])2 + r 2 (t ) sin 2 [ (t ) (t )] cos[2f ct + (t )]

r (t ) sin[ (t ) (t )]
where (t ) = (t ) + tan 1

Ac + r (t ) cos[ (t ) (t )]
Po-Ning Chen@cm.nctu

x (t ) =

(A

Chapter 2-153

+ r (t ) cos[ (t ) (t )]) + r 2 (t ) sin 2 [ (t ) (t )] cos[2f ct + (t )]

2

Limiter

A cos[2f ct + (t )]

Next, the signal will be passed through a Discriminator.

Recall on slide 94, we have talked about the Balanced
Frequency Discriminator, whose input and output satisfy:
Input s(t ) = A cos(2f c t + (t )) Output ~
so (t ) = 2aA ' (t )

Po-Ning Chen@cm.nctu

Chapter 2-154

77

s (t ) = A exp( j (t )), we have :

Specifically, with ~
s (t )
d~

~
s1 (t ) = a
+ jBT ~
s (t ) = aAj[ ' (t ) + BT ]exp[ j (t )]
dt

s (t )
d~

~
s2 (t ) = a
jBT ~
s (t ) = aAj[ ' (t ) BT ]exp[ j (t )]
dt

~
so (t ) =| ~
s1 (t ) | | ~
s2 (t ) |= 2aA ' (t )
Thus, after passing through the discriminator

r (t ) sin[ (t ) (t )]
d (t ) + tan 1

+

A
r
t
t
t
(
)
cos[

(
)

(
)]
c

v (t ) = 2aA ' (t ) = 2aA

dt
Po-Ning Chen@cm.nctu

Chapter 2-155

(r (t ) sin[ (t ) (t )])'
1

v (t ) = 2aA ' (t ) +
2
r (t ) sin[ (t ) (t )] Ac + r (t ) cos[ (t ) (t )]

1 +

Ac + r (t ) cos[ (t ) (t )]

r (t ) sin[ (t ) (t )](r (t ) cos[ (t ) (t )])'

( Ac + r (t ) cos[ (t ) (t )])2

Po-Ning Chen@cm.nctu

Chapter 2-156

78

We then make the following assumption (to simplify the derivation).

Assumption 1: Ac >> r(t) with high probability.
Assumption 2: 2aA = 1/(2).
(r (t ) sin[ (t ) (t )])'
2v (t ) ' (t ) + 1
Ac

' (t ) +

r (t ) sin[ (t ) (t )](r (t ) cos[ (t ) (t )])'

Ac2

(r (t ) sin[ (t ) (t )])'

= 2k f m(t ) +

Ac

Omit the 2nd term by Assumption 1.

1
(r (t ) sin[ (t ) (t )])'
Ac

= 2k f m(t ) + 2nd (t )

We then obtain the desired additive form.

Po-Ning Chen@cm.nctu

Chapter 2-157

There is however one problem remaining, i.e., nd(t) depends on

the message signal (t)!
Assumption 3: (t) (t) is uniformly distributed, and is
independent of m(t) and r(t).
The above assumption is true for any deterministic (t)!
It has been shown that Assumption 3 is justifiable for high
carrier-to-noise ratio (or equivalently, under Assumption 1).
v (t ) k f m(t ) + nd (t ), where nd (t ) =

Table 6.2 : 8.

d
g (t ) j 2fG ( f )
dt

Po-Ning Chen@cm.nctu

1 dr (t ) sin[ (t )]
1 dnQ (t )
=
2Ac
2Ac dt
dt
nQ (t )

H( f ) =

j 2f
2Ac

nd (t )

Chapter 2-158

79

 S N ( f ) = H ( f )H ( f )S N ( f ) =

f2
jf j ( f )
SN ( f ) = 2 SN ( f )
Ac Ac
Ac

f2
B
N , | f | T
S N ( f ) = Ac2 0
2
0,
otherwise

S N ( f f c ) + S N ( f + f c ),

SN ( f ) =
for | f |< BT / 2
0, otherwise

vo (t ) k f m(t ) + no (t )

f2
N , | f | W
S N ( f ) = Ac2 0
0,
otherwise

Bandwidth W < BT/2

that is just enough
to pass m(t).

Po-Ning Chen@cm.nctu

N
E[n (t )] = 20
Ac
2
o

Chapter 2-159

2 N 0W 3
f df =
3 Ac2
2

Observation from the above formula:

In an FM system, increasing carrier power Ac2 / 2
Decreasing noise power. This is named the noise quieting effect.
Po-Ning Chen@cm.nctu

Chapter 2-160

80

As vo (t ) k f m(t ) + no (t ),

SNRO , FM =

k 2f E [m 2 (t )] 3 Ac2 k 2f P
=
, provided Ac >> r (t ).
2 N 0W 3
2 N 0W 3
3 Ac2

We next turn to SNRC,FM.

s (t ) = Ac cos[2f c t + (t )], where (t ) = 2k f m( )d .
t

average power in the modulated signal s (t ) is Ac2 / 2.

Average noise power in the message bandwidth is

SNRC , FM =

2
c

N0
df = WN 0 .
2

2
c

A /2
A
=
.
N 0W 2 N 0W

Po-Ning Chen@cm.nctu

Chapter 2-161

SNRO ,FM
SNRC ,FM

3 Ac2 k 2f P
2
2 N 0W 3 3k f P
=
=
.
Ac2
W2
2WN 0

Remarks : For fixed W , increasing BT increasing

1. Deviation ratio D =
Hence,

k P1 / 2
f
f
.
W
W

SNRO ,FM
.
SNRC ,FM

f = k f max | m(t ) |

SNRO ,FM
D2.
SNRC ,FM

1
1

2. BT ,Carson = 2f 1 + = 2 DW 1 + = 2W ( D + 1)
D
D
Po-Ning Chen@cm.nctu

Chapter 2-162

81

2.13 Summary
Specifically,
for high carrier-to-noise ratio (equivalently to the
assumption made in Assumption 1), an increase in
transmission bandwidth BT provides a corresponding
quadratic increase in figure of merit of a FM system.
So, there is a tradeoff between BT and figure of merit.

Notably, figure of merit for an AM system is

nothing to do with BT.
Po-Ning Chen@cm.nctu

Chapter 2-163

Example 2.5 Single-Tone Modulation

m(t) = Am cos(2fmt)
Then we can represent the figure of merit in terms of
modulation index (or deviation ratio) as (cf. slide 2-63):
SNRO ,FM 3k 2f P 3k 2f ( Am2 / 2) 3 f 2 3 2

=
=
=
= .
SNRC ,FM
W2
W2
2 W2 2
In order to make the figure of metric for an FM system to
be superior to that for an AM system with 100%
modulation, it requires:
3 2 1
2
>
= 0.471
2
3
3
Po-Ning Chen@cm.nctu

Chapter 2-164

82

2.13 Capture Effect

Recall that in Assumption 1, we assume Ac >> r(t).
This somehow hints that the noise suppression of an FM
modulation works well when the noise (or other unwanted
modulated signal that cannot be filtered out by the bandpass or
lowpass filters) is weaker than the desired FM signal.
What if the unwanted FM signal is stronger than the desired FM
signal.
The FM receiver will capture the unwanted FM signal!
What if the unwanted FM signal has nearly equal strength as the
desired FM signal.
The FM receiver will fluctuates back and forth between them!
Po-Ning Chen@cm.nctu

Chapter 2-165

2.13 FM Threshold Effect

Recall that in Assumption 1, we assume Ac >> r(t)
(equivalently, a high carrier-to-noise ratio) to simplify (t)
so that the next formula holds.
3 A2 k 2 P
SNRO ,FM = c f 3 .
2 N 0W
However, a further decrease of carrier-to-noise ratio will
break the FM receiver (from a clicking sound down to a
crackling sound).
As the same as the AM modulation, this is also named the
threshold effect.
Po-Ning Chen@cm.nctu

Chapter 2-166

83

2.13 FM Threshold Effect

Consider a simplified case with m(t) = 0 (no message
signal).

r (t ) sin[ (t )]
d tan 1

A
+
r
(
t
)
cos[

(
t
)]
c

2v (t ) = ' (t ) =
dt
A r (t ) ' (t ) cos[ (t )] + Ac r ' (t ) sin[ (t )] + r 2 (t ) ' (t )
= c
Ac2 + 2 Ac r (t ) cos[ (t )] + r 2 (t )
To facilitate the understanding of clicking sound effect,
we let r(t) = Ac, a constant ratio of Ac.
Po-Ning Chen@cm.nctu

Chapter 2-167

2.13 FM Threshold Effect

2v (t ) = ' (t ) = ' (t )

cos[ (t )] +
1 + 2 cos[ (t )] + 2

Then at the time, say, (t ) , and > 1 but 1

2v (t ) = ' (t ) = ' (t )

cos[ (t )] +

' (t )
2
1 + 2 cos[ (t )] +
1

Thus a sign change in (t) will cause a spike!

Notably, when = 0 (no noise), the output equals m(t)
= 0 as desired.
Po-Ning Chen@cm.nctu

Chapter 2-168

84

2.13 FM Threshold Effect

(t ) = sin(t ) 2v(t ) = ' (t ) = cos(t )

cos[ sin(t )] +
1 + 2 cos[ sin(t )] + 2

= 1.05
=5
= 0.05

Po-Ning Chen@cm.nctu

Chapter 2-169

2.13 Numerical Experiment

Take BT/(2W) = 5.
The average output signal power is calculated in the
absence of noise.
The average output noise power is calculated when
there is no signal present.
SNRO ,FM =
=

3 Ac2 k 2f P
2 N 0W 3

, provided Ac >> r (t )

Am2
3 Ac2 f 2
=
P
,
and f = k f Am
4 N 0W 3
2
3

B
A2 / 2
B
= 3 T , f = T and = c
is the carrier to noise ratio.
BT N 0
2
2W
Po-Ning Chen@cm.nctu

Chapter 2-170

85

2.13 Numerical of the

formula in slide 2-156
Curve I: The average noise
power is calculated assuming an
unmodulated carrier.
Curve II: The average noise
power is calculated assuming a
sinusoidally modulated carrier.
As text mentioned, for <
11 dB, the output signal-tonoise ratio deviates
appreciably from the linear
curve.
A true experiment found that
occasional clicks are heard at
around 13 dB, only
slightly larger than what
theory indicates.
Po-Ning Chen@cm.nctu

The deviation of
curve II from
curve I shows that
the average noise
power is indeed
dependent on the
modulating signal.

Chapter 2-171

2.13 How to avoid clicking sound?

For given modulation index (or deviation ratio) and message
signal bandwidth W,
1. Determine BT by Carsons rule or Universal curve.
2. For a specified noise level N0, select Ac to satisfy:
Ac2
Ac2
13 dB or equvalently,
10 log10
20.
2 BT N 0
2 BT N 0

Po-Ning Chen@cm.nctu

Chapter 2-172

86

2.13 Threshold Reduction

After our learning
that FM
modulation has
threshold effect,
the next question
is naturally on
how to reduce
the threshold?

Po-Ning Chen@cm.nctu

Chapter 2-173

2.13 Threshold Reduction

Threshold reduction in FM receivers may be
achieved by
1. negative feedback (commonly referred to as an
FMFB demodulator), or
2. phase-locked loop demodulator.
Why PLL can reduce threshold effect is not covered in
this course. Notably, the PLL system analysis in
Section 2.14 assumes noise-free transmission.

Po-Ning Chen@cm.nctu

Chapter 2-174

87

v (t )

x (t )

s (t ) = Ac cos[2f c t + (t )], where (t ) = 2k f m( )d .

t

svco (t ) = 2 cos[2f vco t + vco (t )], where vco (t ) = 2k f m( )d .

t

s (t ) svco (t ) = 2 Ac cos[2f c t + (t )] cos[2f vco t + vco (t )]

Bandpass

Ac cos[2 ( f c f vco )t + (1 ) (t )]

The new frequency deviation f new = (1 ) f original .

Thus, the bandpass filter can conceptually have a smaller passband
as wide as (1 ) BT , centered at ( f c f vco ).
Po-Ning Chen@cm.nctu

Chapter 2-175

w(t ) svco (t ) = 2 w(t ) cos[2f vco t + vco (t )]

The noise at the Mixer output can be treated white with the same
noise level as the input white noise.

x (t ) = Ac cos[2f c't + (1 ) (t )] + r (t ) cos[2f c t + (t )]

Limiter

cos[2f c't + (t )],

where E [n I2 (t )] = E [nQ2 (t )] = E [n 2 (t )] = (1 ) BT N 0 ,
r (t ) sin[ (t ) (1 ) (t )]
and (t ) = (1 ) (t ) + tan 1

Ac + r (t ) cos[ (t ) (1 ) (t )]
Since E[nI2 (t )] = E[nQ2 (t )] is smaller, and Ac remains the same,
the condition of Ac >> r (t ) = nI2 (t ) + nQ2 (t ) holds with higer probability.
Experiments show that an FMFB receiver is capable of realizing a
threshold extension on the order of 5~7 dB.
Po-Ning Chen@cm.nctu

Chapter 2-176

88

2.13 Threshold Reduction of an FMFB

receiver
To sum up:
An FMFB demodulator is essentially a tracking
filter that can track only the slowly varying
frequency of a wideband FM signal (i.e., the
message signal part), and consequently it responds
only to a narrowband of noise centered about the
instantaneous carrier frequency.

Po-Ning Chen@cm.nctu

Chapter 2-177

2.13 Pre-Emphasis and De-emphasis in FM

Recall that the noise PSD at the output shapes like a bowel.
So if we can place more signal power at low frequency, a
better noise performance should result.

Po-Ning Chen@cm.nctu

Chapter 2-178

89

2.13 Pre-Emphasis and De-emphasis in FM

In order to produce an undistorted version of the original

message at the receiver output, we must have:

H pe ( f ) H de ( f ) = 1 for W f W .
This relation guarantees the intactness of the message power.
Next, we need to find Hde(f) such that the noise power is
optimally suppressed.
Po-Ning Chen@cm.nctu

Chapter 2-179

2.13 Pre-Emphasis and De-emphasis in FM

Under the assumption of high carrier-to-noise ratio, the

noise PSD at the de-emphasis filter output is given by:
N0 f 2
| H de ( f ) |2 , | f | W

| H de ( f ) |2 S N ( f ) = Ac2

0,
otherwise
2
W N f
0
Average noise power =
| H de ( f ) |2 df
2
W
Ac
o

Po-Ning Chen@cm.nctu

Chapter 2-180

90

2.13 Pre-Emphasis and De-emphasis in FM

Since the message power remains the same, the
improvement factor of the output signal-to-noise ratio
before and after pre/de-emphasis is:

I=

N0 f 2
df
Ac2

N0 f 2
| H de ( f ) |2 df
Ac2

f 2 df

f 2 | H de ( f ) |2 df

Po-Ning Chen@cm.nctu

2W 3
3

f 2 | H de ( f ) |2 df

Chapter 2-181

Example 2.6
Assume R << r and 2fCr << 1 for | f | W .

Pre-emphasis filter

H pe ( f ) = 1 + j ( f / f 0 ) and H de ( f ) =

Po-Ning Chen@cm.nctu

De-emphasis filter

1
, where f 0 = 1 /( 2Cr ).
1 + j( f / f0 )
Chapter 2-182

91

Example 2.6
I =

2W 3
3

f 2 | H de ( f ) |2 df

2W 3
(W / f 0 )3
=
2
W
f
3[(W / f 0 ) tan 1 (W / f 0 )]
3
df
2
W 1 + ( f / f )
0

With f0 = 21 KHz and W = 15 KHz, we obtain I = 22 = 13 dB.

A significant noise performance improvement is therefore obtained.

Po-Ning Chen@cm.nctu

Chapter 2-183

2.13 Pre-Emphasis and De-emphasis in FM

Final remarks:
The previous example uses a linear pre-emphasis
and de-emphasis filter to improve the noise.
A non-linear pre-emphasis and de-emphasis filters
have been applied successfully to application like
tape recording. These techniques, known as DolbyA, Dolby-B, and DBX systems, use a combination
of filtering and dynamic range compression to
reduce the effects of noise.
Po-Ning Chen@cm.nctu

Chapter 2-184

92

2.14 Computer Experiments: Phase-Locked

Loop
Phase-locked loop
s(t ) = Ac sin[2f c t + 1 (t )], where 1 (t ) = 2k f m( )d .
t

r (t ) = Av cos[2f c t + 2 (t )], where 2 (t ) = 2kv v ( )d .

t

km

Po-Ning Chen@cm.nctu

Chapter 2-185

Loop filter = low pass filter + filter h().

e( t ) = k m s ( t ) r ( t )
= km Ac sin[2f c t + 1 (t )] Av cos[2f c t + 2 (t )]
k m Ac Av
(sin[4f ct + 1 (t ) + 2 (t )] + sin[1 (t ) 2 (t )])
2
Low Pass k A A
m c v sin[e (t )], where e (t ) = 1 (t ) 2 (t ).
2

Also, v (t ) = {e( )}LowPass h(t )d .

km

Po-Ning Chen@cm.nctu

Chapter 2-186

93

e (t ) = 1 (t ) 2 (t ) = 1 (t ) 2kv v ( )d
t

de (t ) d1 (t )
=
2kv v (t )
dt
dt

d (t )
= 1 2kv {e( )}LowPass h (t )d
dt

d (t )
= 1 2k0 sin[e ( )]h(t )d
dt
where k0 = km kv Ac Av / 2.

de (u )
du
du
t d ( u )
t
= 0 1 du 2k0 0 sin[e ( s )]h (u s )ds du
du

e (t ) = 0

= 1 (t ) 0(2k0 sin[e (u )] * h(u ) )du

t

Po-Ning Chen@cm.nctu

Chapter 2-187

The previous formula suggests an equivalent analytical model for PLL.

When e(t) = 0, the system is said to be in phase-lock.

In this case, 1(t) = 2(t) or equivalently, kvv(t) = kfm(t).
When e(t) is small (< 0.5 radians), the system is said to be in near
phase-locked.
In this case, we can approximate sin[e(t)] by e(t); hence, a linear
approximate model is resulted.
Po-Ning Chen@cm.nctu

Chapter 2-188

94

Linearization approximation model for PLL.

We can transform the above time-domain system to its

equivalent frequency domain to facilitate its analysis.

Po-Ning Chen@cm.nctu

Chapter 2-189

Linearization approximation model for PLL.

e ( f )

1 ( f )

H( f )

2 ( f )
G( f )

G( f ) =

e ( f )
e ( f )
=
1 ( f ) [1 ( f ) 2 ( f )] + 2 ( f )
=

e ( f )
e ( f ) + 2k0 e ( f ) H ( f )G ( f )

1
jf
=
1 + 2k0 H ( f )G ( f ) jf + k0 H ( f )

Po-Ning Chen@cm.nctu

1
j 2f

Chapter 2-190

95

2.14 Experiment 0: First-Order PLL

H ( f ) = 1.
e ( f )
j ( f / k0 )
=
1 ( f ) 1 + j ( f / k0 )

A parameter k0 controls both the loop gain and bandwidth of the

filter. In other words, there is no why of adjusting loop gain
without changing the filter bandwidth.

Po-Ning Chen@cm.nctu

Chapter 2-191

2.14 Experiment 1: Second-Order

Acquisition PLL
H ( f ) = 1 + a /( jf ) and using linear PLLmodel.

e ( f )
jf
jf
( jf ) 2
=
=
=
1 ( f ) jf + k0 H ( f ) jf + k0 (1 + a /( jf )) ( jf ) 2 + k0 ( jf ) + k0 a
=

( jf / f n ) 2
1 + 2 ( jf / f n ) + ( jf / f n ) 2

where natural frequency f n = ak0 and damping factor = k0 /( 4a ) .

Po-Ning Chen@cm.nctu

Chapter 2-192

96

2.14 Experiment 1: Second-Order

Acquisition PLL
Take f n = 1 /( 2 ) Herz.

Fast response but require

longer time to stablelize

Slow response but quick

stablelizatoin.

Po-Ning Chen@cm.nctu

Chapter 2-193

2.14 Experiment 2: Phase-Plain Portrait

H ( f ) = 1 + a /( jf ) and using nonlinear PLL model.
50

f n = 2 2 Herz
Take
= 1 (critical damping)
2

Let m(t ) = Am cos(2f m t ), where f m =

50
Hz.
2 2

1 de (t )
1 d1 (t )
=
sin[e ( )]h(t )d

2k0 dt
2k0 dt
=
Po-Ning Chen@cm.nctu

k f Am
k0

cos(2f m t ) sin[e ( )]h(t )d

Chapter 2-194

97

Phase-Plane Portrait

1 d
e (t )
2k0 dt

Initial value

e (t )

Po-Ning Chen@cm.nctu

Chapter 2-195

Remarks on the previous figure

Each curve corresponds to different initial frequency error
(i.e., boundary condition for the differential equation).
The phase-plane portrait is periodic in e(t) with period 2,
but the phase-plane portrait is not periodic in de(t)/dt.
There exists an initial frequency error (such as 2 in the
previous figure) at which a saddle line appears (i.e., the
solid line in the previous figure).
In certain cases, the PLL will ultimately reach an
equilibrium (stable) points at (0,0) or (2,0).
A slightest perturbation to the saddle line causes it to shift
to the equilibrium points.

Po-Ning Chen@cm.nctu

Chapter 2-196

98

2.15 Summary and Discussion

Four types of AM modulations are introduced
(expensive) DSB-C transmitter + (inexpensive) envelope
detector, which is good for applications like radio
broadcasting.
(less expensive) DSB-SC transmitter + (more complex)
coherent detector, which is good for applications like limitedtransmitter-power point-to-point communication.
(less bandwidth) VSB transmitter + coherent detector, which
is good for applications like television signals and high speed
data.
(minimum transmission power/bandwidth) SSB transmitter +
coherent detector, which is perhaps only good for
applications whose message signals have an energy gap on
zero frequency.
Po-Ning Chen@cm.nctu

Chapter 2-197

2.15 Summary and Discussion

FM modulation, a representative of Angle
Modulation
A nonlinear modulation process
Carsons rule and universal curve on transmission
bandwidth

Po-Ning Chen@cm.nctu

Chapter 2-198

99

2.15 Summary and

Discussion

Threshold effect

Noise performance
I.
II.
III.

IV.

Full AM (DSB-C) with

100 percent modulation
Coherent DSB-SC &
SSB
FM with = 2 and 13
dB pre/de-emphasis
improvement
FM with = 5 and 13
dB pre/de-emphasis
improvement

Po-Ning Chen@cm.nctu

Chapter 2-199

2.15 Summary and Discussion

Normalized transmission bandwidth in the previous figure

Bn

AM, DSB-SC
2

Bn ,UniversalCurve =

SSB
1

BT BT / f
B
=
= T
W W / f
f

Po-Ning Chen@cm.nctu

FM(=2)
8

FM(=5)
16
= 5 3.2
= 24
(Refer to slide 85.)

Chapter 2-200

100

2.15 Summary and Discussion

Observations from the figure
1. SSB modulation is optimum in noise performance
and bandwidth conservation in AM family
2. FM modulation improves the noise performance of
AM family at the expense of an excessive
transmission bandwidth.
3. Curves III an IV indicates FM modulation offers
the tradeoff between transmission bandwidth and
noise performance.
Po-Ning Chen@cm.nctu

Chapter 2-201

101