4 Analog modulation

4.1 Modulation formats

The message waveform is represented by a low-pass real signal m(t) such that
M(f) = 0 [f[ W
where W is the message bandwidth. m(t) is called the modulating signal.
Carrier modulation: Reversible transformation of m(t) into a bandpass signal x(t) centered around
f
c
W (f
c
is the carrier frequency). Demodulation is the inverse transformation of x(t) into m(t).
x(t) is the modulated signal.
x(t) = x
I
(t) cos 2f
c
t x
Q
(t) sin 2f
c
t
Two types of modulation schemes:
Linear modulation: linear relationship between the modulated signal and the message signal
(ex: AM, DSB-SC, SSB, VSB).
Angle modulation: the angle of the carrier wave is varied according to the message signal
(ex: FM,PM).
4.2 Linear modulation
a) Amplitude modulation AM
x(t) = A
c

1 +k
a
m(t)

cos 2f
c
t
= A
c

1 +m
n
(t)

cos 2f
c
t
x
I
(t) = A
c

1 +m
n
(t)

x
Q
(t) = 0
m
n
(t) =
m(t)
max [m(t)[
k
a
is the amplitude sensitivity, = k
a
max [m(t)[ is modulation index (factor) (0 1).
Exercise: Plot a graph representing x(t) and identify the message signal m(t).
29
The Fourier transform of x(t) is given by
X(f) =
A
c
2

(f f
c
) +M
n
(f f
c
)

+
A
c
2

(f +f
c
) +M
n
(f +f
c
)

where M
n
(f) = T m
n
(t).
Exercise: Plot X(f) and nd the condition on f
c
to avoid distortion.
Let W be the bandwidth of M(f) = T m(t) and M
n
(f), then the bandwidth of the modu-
lated signal x(t) is B = 2W.
Generation of AM:
x(t)
A
c
(1 + m
n
(t))
cos(2f
c
t)
Figure 4: Generation of an AM modulated signal
Demodulation of AM:
Synchronous (coherent) detection
Envelope detection
Coherent detection:
30
W
LPF
x(t)
y(t) u(t)
z(t)
cos(2f
c
t)
Figure 5: Coherent detection of an AM modulated signal
y(t) = A
c

1 +m
n
(t)

cos
2
(2f
c
t)
=
A
c
2

1 +m
n
(t)

+
A
c
2
(1 +m
n
(t)) cos(4f
c
t)
u(t) =
A
c
2

1 +m
n
(t)

(Output of low pass)

z(t) =
A
c

2
m
n
(t) =
A
c
2
k
a
m(t)
Exercise: Assume that the local carrier is not synchronized to the signal, show that the output of
the coherent demodulator is distorted.
Envelope detection:
The output of an envelope detector is the natural envelope [ x(t)[. Envelope detection is feasible
since 1 +m
n
(t) 0.
Draw an example of a RC circuit implementing envelope detection. What conditions on the
resistors and capacitor are necessary to ensure the envelope detector will function properly ?
31
Envelope detection is feasible due to the inclusion of the carrier but the transmission of the
carrier represents a waste of power (contains no information).
b) Double sideband-suppressed carrier: DSB-SC
x(t) = A
c
m(t) cos 2f
c
t
x
I
(t) = A
c
m(t) x
Q
(t) = 0
Exercise: Plot x(t) and identify the message signal m(t). What phenomenon characterizes
DSB-SC ?
The Fourier transform of x(t) is given by
X(f) =
A
c
2
M(f f
c
) +
A
c
2
M(f +f
c
)
where M(f) = T m(t).
Exercise: Plot X(f) and nd the condition on f
c
to avoid distortion.
Let W be the bandwidth of M(f) = T m(t), then the bandwidth of the modulated signal
x(t) is B = 2W.
Generation of DSB-SC:
Draw a block diagram of a DSB-SC modulator.
32
Demodulation of DSB-SC:
Only Synchronous (coherent) detection
Coherent detection: Using a local carrier synchronized to the received signal carrier, draw a block
diagram of a coherent detector for DSB-SC.
Envelope detection:
The output of an envelope detector is the natural envelope [ x(t)[ = A
c
[m(t)[ = A
c
m(t).
AM and DSB-SC are wasting bandwidth, thus ltering of sidebands to reduce bandwidth re-
sults in a more bandwidth efcient scheme.
c) Single sideband modulation: SSB
x(t) =
A
c
2
m(t) cos 2f
c
t
A
c
2
m(t) sin 2f
c
t (Upper sideband SSB)
x(t) =
A
c
2
m(t) cos 2f
c
t +
A
c
2
m(t) sin 2f
c
t (Lower sideband SSB)
x
I
(t) =
A
c
2
m(t) x
Q
(t) =
A
c
2
m(t)
The Fourier transform of x(t) is given by
X(f) =
A
c
2

1
2
M(f f
c
) +
1
2
M(f +f
c
)
1
2j

M(f f
c
)

M(f +f
c
)

=
A
c
4
M(f f
c
)

1 + sgn(f f
c
)

+
A
c
4
M(f +f
c
)

1 sgn(f +f
c
)

where M(f) = T m(t).

33
Exercise: Plot X(f).
Let W be the bandwidth of M(f) = T m(t), then the bandwidth of the modulated signal
x(t) is B = W.
Generation of SSB:
Using a Hilbert transformer (but wideband /2 shifter difcult to implement)
Using sideband ltering (but demands a very sharp lter if M(f) contains very low frequen-
cies components, hence Vestigial sideband modulation (VSB) is also used)
Demodulation of SSB:
Only Synchronous (coherent) detection
Envelope detection by adding a strong carrier to the SSB signal but not a regular SSB any-
more (see VSB subsection)
Coherent detection: Using a local carrier synchronized to the received signal carrier, draw a block
diagram of a coherent detector for SSB.
SSB is difcult to implement if the message signal m(t) has a large bandwidth and it is rich in
34
low frequency components. In this case vestigial sideband modulation is used.
d) Vestigial sideband modulation: VSB
x(t) =
A
c
2
Km(t) cos 2f
c
t
A
c
2
m

(t) sin 2f
c
t
x
I
(t) =
A
c
2
Km(t) x
Q
(t) =
A
c
2
m

(t) =
A
c
2
m(t) h

(t)
Rationale (generation of VSB): Assume that a DSB-SC signal is passed through a general bandpass
lter to alter its sidebands (ex. for SSB half of the sidebands ltered out).
The Fourier transforms of the DSB-SC signal y(t) and x(t) are given by
Y (f) =
A
c
2

M(f f
c
) +M(f +f
c
)

DSB-SC
X(f) =
A
c
2

M(f f
c
) +M(f +f
c
)

H(f)
where M(f) = T m(t). The lter H(f) must have spectral characteristics such that the original
message signal m(t) can be recovered from x(t) by coherent detection.
Demodulation of VSB:
Only Synchronous (coherent) detection
Envelope detection by adding a strong carrier to the VSB signal but not a regular VSB
anymore
Coherent detection:
W
LPF
cos(2f
c
t)
v(t) u(t)
x(t)
Figure 6: Coherent detection of VSB
v(t) = x(t) cos 2f
c
t
35
V (f) =
1
2

X(f f
c
) +X(f +f
c
)

=
A
c
4

M(f 2f
c
) +M(f)

H(f f
c
) +

M(f) +M(f + 2f
c
)

H(f +f
c
)

=
A
c
4
M(f)

H(f f
c
) +H(f +f
c
)

+
A
c
4

M(f 2f
c
)H(f f
c
) +M(f + 2f
c
)H(f +f
c
)

The output of the lowpass lter is given by

U(f) =
A
c
4
M(f)

H(f f
c
) +H(f +f
c
)

For distortionless transmission

U(f) =
A
c
4
KM(f) i.e. u(t) =
A
c
4
Km(t)
where K is a constant. Thus the lter H(f) must satisfy the so-called vestigial symmetry condi-
tion:
H(f f
c
) +H(f +f
c
) = K = const. [f[ W
Time domain representation of VSB signals:
x(t) = h(t) y(t) = x(t) =
1
2

h(t) y(t)
Since y(t) = A
c
cos(2f
c
t)m(t), its complex envelope is given by
y(t) = A
c
m(t) =

Y (f) = A
c
M(f)
The Fourier transform of

h(t) is

H(f) = 2H(f +f
c
) f > f
c
hence

X(f) =
1
2

H(f)

Y (f) = A
c
M(f)H(f +f
c
) f > f
c
(9)
The inphase and quadrature components of x(t) are
x
I
(t) = ' x(t) =
1
2
[ x(t) + x

(t)]
36
x
Q
(t) = x(t) =
1
2j
[ x(t) x

(t)]
their Fourier transform are then given by
X
I
(f) =
1
2

X(f) +

X

(f)

=
A
c
2
[M(f)H(f +f
c
) +M

(f)H

(f +f
c
)] from (9)
=
A
c
2
M(f) [H(f +f
c
) +H(f f
c
)] (since m(t) and h(t) are real.)
=
A
c
2
M(f)K from the vestigial symmetry (10)
X
Q
(f) =
1
2j

X(f)

X

(f)

=
A
c
2j
M(f) [H(f +f
c
) H(f f
c
)]
=
A
c
2
M(f)H

(f) (11)
where the lter H

(f) is dened as
H

(f) =
1
j
[H(f +f
c
) H(f f
c
)]
From (10) and (11), the inphase and quadrature components of a VSB signal are given by
x
I
(t) =
A
c
2
Km(t)
x
Q
(t) =
A
c
2
m(t) h

(t) =
A
c
2
m

(t)
Bandwidth of m(t): W
Bandwidth of x(t): W < B < 2W (typically B
VSB
= 1.25B
SSB
)
If H

(f) = j sgn(f), then we obtain the upper sideband SSB.

Envelope detection of SSB and VSB:
We add a strong carrier A

c
cos(2f
c
t) (with A

c

Ac
2
K[m(t)[,
Ac
2
[m

(t)[) such that the signal

to be demodulated is given by
x(t) =

c
+
A
c
2
Km(t)

cos(2f
c
t)
A
c
2
m

(t) sin(2f
c
t)
37
= A

1 +
A
c
K
2A

c
m(t)

cos(2f
c
t)
A

c
2
A
c
A

c
m

(t) sin(2f
c
t)
For simplicity assume K = 1, and let
Ac
A

c
= k
a
x(t) = A

1 +
k
a
2
m(t)

cos(2f
c
t)
A

c
2
k
a
m

(t) sin(2f
c
t) (12)
x(t) = A

1 +
k
a
2
m(t)

+jA

c
k
a
2
m

(t) (complex envelope of x(t))

The output of an envelope detector is the natural envelope of the input, hence the output of the
envelope detector is given by
[ x(t)[ =

2
c

1 +
k
a
2
m(t)

2
+A

2
c
k
2
a
4
m
2

(t)

1/2
= A

1 +
k
a
2
m(t)

1 +
k
2
a
4
m
2

(t)

1 +
ka
2
m(t)

1/2
A

1 +
k
a
2
m(t)

since

ka
2
m(t)

<1 and

ka
2
m

(t)

<1. This method is used in TV systems. Distortion due to the

envelope detection of VSB is reduced by reducing k
a
ensuring the conditions

ka
2
m(t)

< 1 and

ka
2
m

(t)

<1.
4.3 Multiplexing
The purpose of multiplexing is to transmit several signals m
1
(t), . . . , m
N
(t) at the same time
by the use of a single communication system. This can be achieved by combining the signals into
one signal s(t) such that each of the signals m
k
(t) can be extracted from s(t). In this section, we
present two types of multiplexing; quadrature carrier multiplexing and frequency division multi-
plexing (FDM). A third multiplexing technique called Time Division Multiplexing (TDM) will be
considered in the context of signal sampling.
a) Quadrature carrier multiplexing
Since cos(2f
c
t) and sin(2f
c
(t) are orthogonal functions, the principle of quadrature multiplex-
ing of two signals is to transmit one signal using a carrier of the form cos(2f
c
t) and to transmit
the other signal using a carrier of the form sin(2f
c
t). Let m
1
(t) and m
2
(t) be two lowpass signals
with bandwidth W. The schemes of multiplexing and demultiplexing follows:
38
m
2
(t)
sin(2f
c
t)
m
1
(t)
cos(2f
c
t)
s(t) =
m
1
(t) cos(2f
c
t) + m
2
(t) sin(2f
c
t)
Figure 7: Quadrature carrier multiplexing
W
LPF
LPF
2 cos(2f
c
t)
2 sin(2f
c
t)
m
1
(t)
m
2
(t)
v
1
(t)
v
2
(t)
W
s(t)
Complete the proof that the scheme of Fig. 8
is a quadrature carrier demultiplexer:
v
1
(t) = 2s(t) cos(2f
c
t)
= +
....
lowpass term
. .. .
bandpass term
v
2
(t) = 2s(t) sin(2f
c
t)
= +
....
lowpass term
. .. .
bandpass term
Since the bandpass terms are removed by the
lowpass lter, the result follows.
Figure 8: Quadrature carrier demultiplexing
39
b) Frequency division multiplexing (FDM)
The principle of frequency division multiplexing is to modulate each signal m
0
i
(t) using a different
carrier frequency f
c
i
such that the spectrum of the modulated signals x
i
(t) do not overlap. Then a
FDM signal is obtained by adding the modulated signal yielding a signal with a higher bandwidth.
The multiplexed signal can be further modulated before transmission. Hence the modulation of
the signals m
i
(t) (bandlimited version of m
0
i
(t)) to be multiplexed is called sub-modulation and
the carrier f
c
i

i=1,...,N
are sub-carriers. In a FDM system, the sub-carriers are selected such that
the spectrum of the sub-modulated signals do not overlap. Therefore the original message signals
m
0
i
(t) have to be passed rst through a lowpass lter that limit them to a predetermined bandwidth
W. If the original message signals are already bandlimited to W, no lowpass ltering is required.
Let B be the bandwidth of each of the sub-modulated signals x
i
(t). To avoid overlapping of the
spectrum of the sub-modulated signals (and hence to ensure distortionless demultiplexing), we
must have
[f
c
i
f
c
k
[ B
FDM multipler and demultiplexer follows:
x
1
(t)
LPF
Modulator
m
N
(t)
f
c
N
x
N
(t) m
0
N
(t)
LPF
m
1
(t)
Modulator
f
c
1
m
0
1
(t)
s(t) =

N
i=1
x
i
(t)
Figure 9: Frequency Division multiplexing
example of FDM:
s(t) =
N

i=1

1 +k
a
i
m
i
(t)

cos(2f
c
i
t) AM sub-modulation (BW of s(t): 2NW)
s(t) =
N

i=1
A
c
2

m
i
(t) cos(2f
c
i
t) m
i
(t) sin(2f
c
i
t)

SSB sub-modulation (BW of s(t): NW)

40
s(t)
BPF
B
Demodulator
N
fc
N
f
c
N
x
N
(t) m
N
(t)
BPF
fc
1
1
B
Demodulator
f
c
1
m
1
(t) x
1
(t)
Figure 10: Frequency Division demultiplexing
Complete the proof that the scheme of Fig. 10 is a FDM demultiplexer:
s(t) = +
. .. .
bandpass term around f
c
i
. .. .
sum of bandpass terms around f
c
j
(j = i)
The bandpass lter around f
c
i
keeps the bandpass term around f
c
i
(namely x
i
(t)) and removes all
the other bandpass terms. It is seen that non-overlapping of the spectrums of x
i
(t) is needed to
avoid distortion. Then each x
i
(t) can be demodulated.
Note that for SSB sub-modulation, since the spectrum of x
i
(t) contains only one sideband, the
required bandpass lter should pass only one sideband as seen in Fig. 11.
BPF
n
f
c
n
f
c
n
+ W
f
Figure 11: Bandpass lter required for upper sideband SSB
41
4.4 Angle modulation
With m(t) the message signal, an angle modulated signal is dened as
x(t) = A
c
cos

(t)

= A
c
cos

2f
c
t +(t)

(13)
where (t) is given by
(t) = K

m()h(t )d = 2K
f

m()h(t )d
where K = 2K
f
is the phase sensitivity (of the modulator) expressed in rd/Volt, K
f
=
K
2
is the frequency sensitivity (of the modulator) expressed in Hz/Volt and h(t) is assumed to be
causal. (t) is the phase of x(t) and (t) = 2f
c
t +(t) is the angle of x(t). The instantaneous
frequency of x(t) is dened as
f(t) =
1
2
d(t)
dt
= f
c
+
1
2
d(t)
dt
The maximum phase deviation of x(t) is
= max [(t)[
The maximum frequency deviation of x(t) is
f = max [f(t) f
c
[ = max

1
2
d(t)
dt

The complex envelope of x(t) is given by

x(t) = A
c
e
j(t)
Phase modulation (PM):
h(t) = (t) =(t) = Km(t)
H(f) = 1
The phase of a PM signal is proportional to the message signal.
Frequency modulation (FM):
h(t) = u(t) =

1, t > 0
1
2
, t = 0
0, t < 0
=(t) = K

m()d
42
H(f) =
1
j2f
+
1
2
(f)
The instantaneous frequency of an FM signal is proportional to the message signal up to a carrier
frequency shift
f(t) =
K
2
m(t) +f
c
= K
f
m(t) +f
c
Frequency modulation with pre-emphasis/de-emphasis(FM):
When noise analysis is done for FM, it can be shown that the power spectral density of noise
at the FM receiver output is proportional to f
2
in the frequency-band of the message, thus the
noise power is higher at high frequencies. To increase the overall signal-to-noise ratio, practical
systems use a pre-emphasis lter h
pe
(t) before frequency modulation. The purpose of h
pe
(t) is
to articially increase the high-frequency components of the message signal to compensate for
the high noise level. After FM detection, the recovered emphasized message signal is passed
through a de-emphasis lter h
de
(t) which must be ideally the inverse lter of the pre-emphasis
lter (i.e. H
de
(f) =
1
Hpe(f)
). If pre-emphasis is used, the emphasized message signal is given by
m
pe
(t) = m(t) h
pe
(t)
and the FM modulated signal is
x(t) =

m
pe
()d = m
pe
(t) u(t) = m(t) h
pe
(t) u(t)
Therefore (13) corresponds to an FM signal with pre-emphasis when h(t) = h
pe
(t) u(t). Equiv-
alently in the frequency domain
H(f) = H
pe
(f)

1
2
(f) +
1
j2f

=
1
2
H
pe
(0)(f) +
H
pe
(f)
j2f
An example is H
pe
(f) = 1 +
jf
c
.
a) Narrow-band angle modulation
Narrow-band angle modulation: (t) <1
If (t) <1, then
x(t) = A
c
e
j(t)
A
c
(1 +j(t))
Hence
x(t) = '

x(t)e
j2fct

(complete this line)

A
c
cos(2f
c
t) A
c
(t) sin(2f
c
t)
43
Thus narrow-band angle modulation has similar features to AM.
Bandwidth of narrow-band angle modulation: 2 bandwidth of (t).
Generation of narrow-band angle modulation:
Exercise: Draw a block diagram that generates a narrow-band angle modulated signal.
Detection of narrow-band angle modulation:
Exercise: Draw a block diagram of a system with input x(t) and output
Ac
2
(t):
b) Wide-band angle modulation
Wide-band angle modulation: [(t)[ 1
x(t) = A
c
e
j(t)
Generation of wide-band angle modulation using indirect method (Armstrongs method) :
Generates a narrow-band angle modulated signal
x
0
(t) = A
c
cos

2f
c
t +K

m()h(t )d

with

m()h(t )d

<1.
Pass x
0
(t) through a Frequency multiplier by N whose block diagramis illustrated in Fig. 12.
The output to the frequency multiplier by N is then
x(t) = A
c
cos

2Nf
c
t +NK

h(t )m()d

44
x
0
(t) x
N
0
(t)
Nf
c
BPF
x(t)
( )
N
Figure 12: frequency multiplier by N
If N 1 we have a wide-band angle-modulated signal around a carrier at Nf
c
.
Analysis of a frequency multiplier by N:
y(t) = (x
0
(t))
N
=
1
2
N1
cos

2Nf
c
t +NK

m()h(t )d

+ other terms like cos(2nf

c
t +. . .) with n < N
since
(cos )
2n
=
1
2
2n

n1

k=0
2

n
k

cos

2(n k)

2n
n

(cos )
2n1
=
1
2
2n2

n1

k=0

2n 1
k

cos

(2n 2k 1)

2n
n

The other terms are removed by the bandpass lter yielding x(t) as output.
Generation of any FMmodulation using direct method; use of Voltage-Controlled-Oscillator (VCO):
x(t)
VCO
V
c
(t)
Figure 13: Voltage-Controlled Oscillator (VCO)
If the input of a VCO is a voltage V
c
(t) then the output of a VCO with unmodulated frequency
of oscillation f
c
has an instantaneous frequency given by
f(t) = f
c
+K
f
V
c
(t)
45
Thus its output is
x(t) = A
c
cos

2f
c
t +K

V
c
()d

see for example Hartley oscillator .

d) Tone modulation
Tone modulation corresponds to a sinusoidal message signal. Let m(t) = A
m
cos(2f
c
t) (A
m

0) applied at some time t
0
such that t
0
< t. Let us calculate the steady state expression of (t)
corresponding to a general angle modulated signal.
(t) = K

t
t
0
A
m
cos(2f
m
)h(t )d
K

A
m
cos(2f
m
)h(t )d (steady state)
= K

A
m
cos(2f
m
)h(t )d (h(t) is causal)
= K

A
m
cos(2f
m
) T

h(t )

df Generalization of Parseval theorem (t xed)

=
KA
m
2

[(f f
m
) +(f +f
m
)]

H(f)e
j2ft

df
=
KA
m
2

(f
m
)e
j2fmt
+H

(f
m
)e
j2fmt

= KA
m
'

H(f
m
)e
j2fmt

(h(t) is real, thus H

(f) = H(f))
= KA
m
[H(f
m
)[ cos

2f
m
t + arg [H(f
m
)]

(steady state) (14)

Show that for PM and FM (t) is given by
(t) =

KA
m
cos(2f
m
t) PM
KAm
2fm
sin(2f
m
t) FM
The modulation index is dened as the maximum phase deviation, or equivalently the maxi-
mum deviation of the angle (t) from 2f
c
t .
= max [(t)[ = max [(t) 2f
c
t[
46
= KA
m
[H(f
m
)[ assuming K, A
m
0
Hence from (14) the steady state expression of (t) for tone modulation is also given by
(t) = cos

2f
m
t + arg [H(f
m
)]

= sin (2f
m
t +)
where = arg [H(f
m
)] +

2
.
Phase modulation (PM):

PM
= KA
m
= : maximum phase deviation
Frequency modulation (FM):

FM
=
KA
m
2f
m
=
K
f
A
m
f
m
=
f
f
m
f = K
f
A
m
: maximum frequency deviation
Transmission bandwidth of an angle modulated signal with tone modulation:
For convenience, we dene = arg [H(f
m
)] such that
FM
= 0 and
PM
=

2
and we use
(t) = sin (2f
m
t +)
The complex envelope of the angle-modulated signal is
x(t) = A
c
e
j(t)
= A
c
exp j sin(2f
m
t +)
Unlike the original signal x(t), the complex envelope x(t) is periodic with period T
m
=
1
fm
,
therefore x(t) admits a Fourier series representation
x(t) =

n=
c
n
e
j2nfmt
where the Fourier coefcients are given by
c
n

=
1
T
m
Tm
2

Tm
2
x(t)e
j2nfmt
dt
= A
c
f
m
Tm
2

Tm
2
x(t)e
j[ sin(2fmt+)2nfmt]
dt
=
A
c
f
m
2f
m

e
j[ sin(u)nu+n]
du (u = 2f
m
t +)
= A
c
e
jn
1
2

e
j[ sin(u)nu]
du
47
= A
c
e
jn
J
n
()
where J
n
() is the n
th
Bessel function of the rst kind dened as
J
n
() =
1
2

e
j[ sin(u)nu]
du (integral independent of
1
)
=

k=0
(1)
k

n+2k
k!(k +n)!
Therefore the modulated signal x(t) with tone modulation is given by
x(t) = '

x(t)e
j2fct

= '

n=
A
c
e
jn
J
n
()e
j[2fct+2nfmt]

= A
c

n=
J
n
() cos

2(f
c
+nf
m
)t +n

and has a Fourier transform given by

X(f) =
A
c
2

n=
J
n
()

(f f
c
nf
m
) +(f +f
c
+nf
m
)

cos(n)
+j

(f f
c
nf
m
) (f +f
c
+nf
m
)

sin(n)

Thus it is seen that angle-modulated signals have an innite bandwidth.

X(f) =

Ac
2

n=
J
n
()

(f f
c
nf
m
) +(f +f
c
+nf
m
)

FM
Ac
2

p=

J
2p
()

(f f
c
2pf
m
) +(f +f
c
+ 2pf
m
)

(1)
p
+ jJ
2p+1
()

(f f
c
(2p + 1)f
m
) (f +f
c
+ (2p + 1)f
m
)

PM
Using J
n
() = (1)
n
J
n
() and mathematical tables, the Fourier transform X(f) for FM is
illustrated for example in Fig. 14 and Fig. 15. Note that usually only the magnitude of X(f) is
drawn. It is obtained by reversing the negative peaks to become positive peaks.
1
Expand the integral as the sum of the three integrals

and makes the change of variable v =

u + 2 in the rst one.
48
f + 3f
m
f + f
c m
f
c
f - 2f
c
c
c m
f - f f - 3f
m c

m
( = 2)
X(f)
f
(positive frequencies only)
Figure 14: Fourier transform of X(f) ( = 2) with FM modulation
c
f + f
c m
c
f + 3f
m
f
f - f
m
f - 3f
m c
( = 8)
c
X(f)
f
(positive frequencies only)
Figure 15: Fourier transform of X(f) ( = 8) with FM modulation
Some properties of Bessel functions:
J
n
() = (1)
n
J
n
()
When <1,
J
n
()

n
1
n!
=

J
0
() 1
J
1
()

2
J
n
() 0, n 2
lim
0
J
n
() =

1 n = 0,
0 else.
49
To obtain a denition of an effective (or essential) bandwidth, that is a bandwidth that contains
most of the total power (usually 98% or 99%), let us consider the average power of x(t). The
average power of x(t) is given by
P
x

= lim
T
1
2T

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

n=

p=
J
n
()J
p
()

T
T
cos

2(f
c
+nf
m
)t +n

cos

2(f
c
+pf
m
)t +p

dt
=
A
2
c
2

n=

p=
J
n
()J
p
() lim
T
1
2T

T
T
cos

2(2f
c
+ (n +p)f
m
)t + (n +p)

dt
+

T
T
cos

2(n p)f
m
t + (n p)

dt

=
A
2
c
2

n=

p=
J
n
()J
p
()

1
2T
n,p

Tn,p
Tn,p
cos

2t
T
n,p
+ (n +p)

dt
+
1
2T

n,p

n,p
T

n,p
cos

2t
T

n,p
+ (n p)

dt

where T
n,p
= [2f
c
+ (n +p)f
m
]
1
and T

n,p
= (np)
1
f
1
m
. Since the two trigonometric integrals
are zero unless n = p and are equal to 1 when n = p, the average power of x(t) is given by
P
x
=
A
2
c
2

n=
J
2
n
() = P
c
+

n=
P
n
P
c
=
A
2
c
2
J
2
0
()
P
n
=
A
2
c
2
J
2
n
() n
th
side-band power
P
n
=
A
2
c
2
J
2
n
() =
A
2
c
2
J
2
n
() = P
n
Note that from x(t) = A
c
cos (2f
c
t +(t)), the average power of x(t) is also given by
P
x
=
A
2
c
2
therefore we can deduce that

n=
J
2
n
() = 1
50
An effective (essential) bandwidth can be dened as
B
T
= 2n
max
f
m
where n
max
is the largest n such that [J
n
()[ 0.01 and k > n[J
k
()[ < 0.01. It turns out
that n
max
depends on . In terms of power, it is equivalent to neglect side-bands that contribute
to less than 0.01% of the total power. Another rule for the effective bandwidth is obtained by
approximating n
max
by a linear curve,
B
T
= 2f
m
( +c) 1 c 2
When c = 1, the classical Carsons rule is obtained
B
Cars
T
= 2f
m
( + 1)
example: Show that Carsons rule for FM is
B
FM
T
= 2 (f
m
+ f)
d) General modulating signal with bandwidth W
If m(t) is periodic, m(t) should be expressed in terms of its Fourier series representation. If m(t)
is a general non-periodic deterministic signal, we can use e
j(t)
=

k=0
[(t)]
k
k!
.
Generalization of Carsons rule to a general modulating signal
Worst-case tone approach. Let m(t) be a message signal with bandwidth W and maximum
amplitude max [m(t)[. Assume we model m(t) as an innite number of tones of frequency f
m
k
and maximum amplitude A
m
k
, then its effective bandwidth would be
B
T
= max
k
2f
m
k
[
k
+ 1]
= max
k
2f
m
k
+ max
k
KA
m
k
f
m
k
[H(f
m
k
)[
max
k
2f
m
k
+K max
k
A
m
k
max
k
[f
m
k
[H(f
m
k
)[] (15)
= W +K max [m(t)[ max
k
[f
m
k
[H(f
m
k
)[]
where to get (15) we have applied the worse case tone approach. The worse case tone approach
consists of evaluating the bandwidth obtained by considering a tone at the highest possible fre-
quency and the highest possible amplitude, thus maximizing the product of the second term in
B
T
by maximizing its two terms A
m
k
and f
m
k
[H(f
m
k
)[ separately. Let us consider now the term
max
k
[f
m
k
[H(f
m
k
)[] for the special case of FM and PM modulation. For FM modulation
max
k
[f
m
k
[H(f
m
k
)[] = max
k

f
m
k
2f
m
k

= max
k
1 = 1 = W[H(W)[
51
consistent with the worse tone at W. Note that for FM maximizing A
m
k
is equivalent to maximize
the frequency deviation f
k
. For PM modulation
max
k
[f
m
k
[H(f
m
k
)[] = max
k
f
m
k
= W = W[H(W)[
Hence we obtain
B
T
= 2W +K max([m(t)[)W[H(W)[ = 2W( + 1)
where called for general modulating signal the deviation ratio is dened as
= K max([m(t)[)[H(W)[ =

f
W
=
K
f
max |m(t)|
W
FM
= K max [m(t)[ PM
Exercise:
Calculate using Carsons rule the bandwidth of commercial FM broadcasting characterized
by a maximum allowed maximal frequency deviation of 75kHz and a maximum audio signal of
15kHz.
e) Detection of FM signals
Basic structure: using a differentiator and an envelope detector
Show that for large f
c
, if x(t) is applied to a differentiator followed by an envelope detector,
the resulting output is proportional to the message signal:
x(t) = A
c
cos

2f
c
t +K

m()d

dx(t)
dt
= complete
z(t) = Output of envelope detector (complete)
Frequency domain differentiation (slope demodulator)
Draw the transfer function of a real lter that can implement the differentiation.
52
Note that the differentiation operation has to be implemented only in the bandwidth of the
modulated signal.
Time domain differentiation
Based on the following approximation
dx
dt

x(t) x(t t)
t
draw the block diagram of a time domain differentiator.
Quadrature detector for an angle-modulated signal
Using a differentiator and a /2 phase shifter, draw the block diagram of a demodulator for an
angle modulated signal, that is based on similar principles as coherent detection for AM.
53
Show that the output of the quadrature detector is given by
A
2
c
2
d(t)
dt
.
f) Form of H(f) for FM
The phase of an FM modulated signal is given by
(t) = K

h(t )m()d = K

m()d
Its Fourier transform is given by
T (t) = KH(f)M(f) = K

1
j2f
M(f) +
M(0)
2
(f)

Hence in theory
H(f) =
1
j2f
+
(f)
2
Let H

(f) =
1
j2f
, then the corresponding phase

(t) is given by

(t) = T
1

KH

(f)M(f)

= T
1

K
1
j2f
M(f)

and the original phase is

(t) = T
1
[KH(f)M(f)] = T
1

K
1
j2f
M(f)

+T
1

K
2
(f)M(0)

= T
1

K
1
j2f
M(f)

+
K
2
M(0)
Hence the difference between the two phases is only a constant phase shift which is equivalent to
a change of the time of origin. Furthermore the instantaneous frequency given by
f(t) = f
c
+
1
2
d(t)
dt
54
is independent of the constant
KM(0)
2
. Finally the two possible expressions of H(f)
1. H(f) =
1
j2f
+
1
2
(f) 2. H(f) =
1
j2f
yield the same result for the deviation ratio . Hence in the derivation of results
H(f) =
1
j2f
can be considered instead of H(f) =
1
j2f
+
1
2
(f).
55