Angle modulation

Principles of Communication Prof. V. Venkata Rao
Indian Institute of Technology Madras
5.1
CHAPTER 5
CHAPTER 5
Angle Modulation
5.1 Introduction
Consider a sinusoid, ( )c cA f t 0cos 2π + φ , where cA is the (constant)
amplitude, cf is the (constant) frequency in Hz and 0φ is the initial phase angle.
Let the sinusoid be written as ( )cA tcos⎡ ⎤θ⎣ ⎦ where ( ) ct f t 02θ = π + φ . In chapter
4, we have seen that relaxing the condition that cA be a constant and making it a
function of the message signal ( )m t , gives rise to amplitude modulation. We
shall now examine the case where cA is a constant but ( )tθ , instead of being
equal to cf t 02π + φ , is a function of ( )m t . This leads to what is known as the
angle modulated signal. Two important cases of angle modulation are Frequency
Modulation (FM) and Phase modulation (PM). Our objective in this chapter is to
make a detailed study of FM and PM.
An important feature of FM and PM is that they can provide much better
protection to the message against the channel noise as compared to the linear
(amplitude) modulation schemes. Also, because of their constant amplitude
nature, they can withstand nonlinear distortion and amplitude fading. The price
paid to achieve these benefits is the increased bandwidth requirement; that is,
the transmission bandwidth of the FM or PM signal with constant amplitude and
which can provide noise immunity is much larger than W2 , where W is the
highest frequency component present in the message spectrum.

5.2
Now let us define PM and FM. Consider a signal ( )s t given by
( ) ( )c is t A tcos⎡ ⎤= θ⎣ ⎦ where ( )i tθ , the instantaneous angle quantity, is a
function of ( )m t . We define the instantaneous frequency of the angle modulated
wave ( )s t , as
( )
( )i
i
d t
f t
d t
1
2
θ
=
π
(5.1)
(The subscript i in ( )i tθ or ( )if t is indicative of our interest in the instantaneous
behavior of these quantities). If ( )i ct f t 02θ = π + φ , then ( )if t reduces to the
constant cf , which is in perfect agreement with our established notion of
frequency of a sinusoid. This is illustrated in Fig. 5.1.
Fig. 5.1: Illustration of instantaneous phase and frequency
Curve 1 in Fig. 5.1 depicts the phase behavior of a constant frequency sinusoid
with 0 0φ = . Hence, its phase, as a function of time is a straight line; that is
( )i ct f t2θ = π . Slope of this line is a constant and is equal to the frequency of
the sinusoid. Curve 2 depicts an arbitrary phase behavior; its slope changes
with time. The instantaneous frequency (in radians per second) of this signal at
t t1= is given by the slope of the tangent (green line) at that time.

5.3
a) Phase modulation
For PM, ( )i tθ is given by
( ) ( )i c pt f t k m t2θ = π + (5.2)
The term cf t2π is the angle of the unmodulated carrier and the constant pk is
the phase sensitivity of the modulator with the units, radians per volt. (For
convenience, the initial phase angle of the unmodulated carrier is assumed to be
zero). Using Eq. 5.2, the phase modulated wave ( )s t can be written as
( ) ( )c c pPM
s t A f t k m tcos 2⎡ ⎤⎡ ⎤ = π +⎣ ⎦ ⎣ ⎦ (5.3)
From Eq. 5.2 and 5.3, it is evident that for PM, the phase deviation of ( )s t from
that of the unmodulated carrier phase is a linear function of the base-band
message signal, ( )m t . The instantaneous frequency of a phase modulated
signal depends on
( )
( )
d m t
m t
d t
'= because
( )
( )pi
c
kd t
f m t
d t
1 '
2 2
θ
= +
π π
.
b) Frequency Modulation
Let us now consider the case where ( )if t is a function of ( )m t ; that is,
( ) ( )i c ff t f k m t= + (5.4)
or ( ) ( )
t
i it f d2
− ∞
θ = π τ τ∫ (5.5)
( )
t
c ff t k m d2 2
− ∞
= π + π τ τ∫ (5.6)
fk is a constant, which we will identify shortly. A frequency modulated signal
( )s t is described in the time domain by
( ) ( )
t
c c fFM
s t A f t k m dcos 2 2
− ∞
⎡ ⎤
⎢ ⎥⎡ ⎤ = π + π τ τ⎣ ⎦ ⎢ ⎥
⎣ ⎦
∫ (5.7)
fk is termed as the frequency sensitivity of the modulator with the units Hz/volt.
From Eq. 5.4 we infer that for an FM signal, the instantaneous frequency

5.4
deviation of ( )s t from the (unmodulated) carrier frequency cf is a linear function
of ( )m t . Fig. 5.2 to 5.5 illustrate the experimentally generated AM, FM and PM
waveforms for three different base-band signals. From these illustrations, we
observe the following:
i) Unlike AM, the zero crossings of PM and FM waves are not uniform (zero
crossings refer to the time instants at which a waveform changes from
negative to positive and vice versa).
ii) Unlike AM, the envelope of PM or FM wave is a constant.
iii) From Fig. 5.2(b) and 5.3(b), we find that the minimum instantaneous
frequency of the FM occurs (as expected) at those instants when ( )m t is
most negative (such as t t1= ) and maximum instantaneous frequency
occurs at those time instants when ( )m t attains its positive peak value, pm
(such as t t2= ). When ( )m t is a square wave (Fig. 5.4), it can assume
only two possible values. Correspondingly, the instantaneous frequency
has only two possibilities. This is quite evident in Fig. 5.4(b).

5.5
Fig 5.2: AM and FM with tone modulation

5.6
Fig. 5.3: AM and FM with the triangular wave shown as ( )m t

5.7
Fig. 5.4: AM and FM with square wave shown as ( )m t
iv) A triangular wave has only two possibilities for its slope. In Fig. 5.5(b), it has
a constant positive slope between t1 and t2 , and constant negative slope to
the right of t2 for the remaining part of the cycle shown. Correspondingly,
the PM wave has only two values for its ( )if t , which is evident from the
figure.
v) The modulating waveform of Fig. 5.5(c) is a square wave. Except at time
instants such as t t1= , it has zero slope and
( )
t t
d m t
d t
1=
is an impulse.

5.8
Therefore, the modulated carrier is simply a sinusoid of frequency cf ,
except at the time instants when ( )m t changes its polarity. At t t1= , ( )if t
has to be infinity. This is justified by the fact that at t t1= , ( )i tθ undergoes
sudden phase change (as can be seen in the modulated waveform) which
implies
( )id t
d t
θ
tends to become an impulse.
Eq. 5.3 and 5.7 reveal a close relationship between PM and FM. Let
( ) ( )
t
Im t m d
− ∞
= τ τ∫ . If ( )Im t phase modulates a carrier with modulator
sensitivity p fk k2= π , then the resulting signal is actually an FM signal as given
by Eq. 5.7. Similarly a PM signal can be obtained using frequency modulator by
differentiating ( )m t before applying it to the frequency modulator. (Because of
differentiation, ( )m t should not have any discontinuities.)
As both PM and FM have constant amplitude cA , the average power of a
PM or FM signal is,
c
av
A
P
2
2
= ,
regardless of the value of pk or fk .

5.9
Fig 5.5: PM with ( )m t (a) a sine wave
(b) a triangular wave
(c) a square wave

5.10
Example 5.1
An angle modulated signal is given by
( ) ( ) ( )( )s t t t t6
cos 2 2 10 30sin 150 40cos 150⎡ ⎤= π × + +
⎣ ⎦
Let us find the maximum phase and frequency derivations.
The terms ( ) ( )t t30sin 150 40cos 150⎡ ⎤+⎣ ⎦ can be expressed in the form
( ) ( )t tcos sin 150 sin cos 150⎡ ⎤α + α⎣ ⎦ .
As 2 2
30 40 50+ = , we have
( ) ( ) ( ) ( )t t t t
3 4
30sin 150 40cos 150 50 sin 150 cos 150
5 5
⎛ ⎞
+ = +⎜ ⎟
⎝ ⎠
( )t50sin 150= + ϕ where 1 4
tan
3
−
ϕ =
Hence ( ) ( )s t t t6
cos 4 10 100 sin 150⎡ ⎤= π × + π + ϕ
⎣ ⎦
Evidently, the maximum phase deviation is ( )100π .
Let ( ) ( )t t100 sin 150ψ = π + ϕ
( )
( )
d t
t
d t
1
50cos 150 150
2
ψ
= + ϕ ⋅
π
( )t7500 cos 150= + ϕ
Hence maximum frequency deviation 7500= Hz.
Example 5.2
Let ( )m t be a periodic triangular wave with period 3
10−
sec. with
( ) ( )m t m tmax min
1= − = volt. We shall find the maximum and minimum values
of the instantaneous frequency for
a) FM with fk 4
10= Hz/volt
b) PM with pk = π rad/volt

5.11
Assume the carrier frequency to be 100 kHz.
a) For FM, ( ) ( )i c ff t f k m t= +
( )( )if t 3 4
min
100 10 10 90= × − = kHz
( )( )if t 3 4
max
100 10 10 110= × + = kHz
b) For PM, ( ) ( )p
i c
k
f t f m t'
2
= +
π
Note that ( )m t' is a square wave with maximum and minimum values as
4,000± .
Hence ( )( )if t 3
min
1
100 10 4,000
2
= × − ×
98= kHz
Similarly, ( )( )if t
max
102= kHz
Example 5.3
Let ( )s t be a general angle modulated signal given by
( ) ( ) ( )c i c cs t A t A t tcos cos⎡ ⎤ ⎡ ⎤= θ = ω + ϕ⎣ ⎦ ⎣ ⎦
It is given that when ( ) mm t tcos= ω , ( )s t has the instantaneous frequency
given by ( ) ( )i c m mf t f k f t
2
2 cos= + π ω , where k is a suitable constant. Let us
find the expression for ( )i tθ . If ( )m t is different from m tcos ω , what could be
the expression for ( )i tθ and ( )s t .
( )
( ) ( )i
i c m m
d t
f t f k f t
d t
21
2 cos
2
θ
= = + π ω
π
( )
( )i
c m m
d t
f k t
d t
2
2 cos
θ
= π + ω ω

5.12
or ( ) ( )i c m mt f t k t2 sinθ = π + ω ω
( )c m
d
f t k t
d t
2 cos⎡ ⎤= π − ω⎣ ⎦
Generalizing,
( ) ( )i c
d
t f t k m t
d t
2 ⎡ ⎤θ = π − ⎣ ⎦
and ( )
( )
c c
d m t
s t A f t k
d t
cos 2
⎡ ⎤
= π −⎢ ⎥
⎣ ⎦
Exercise 5.1
A periodic signal ( )m t angle modulates a very high frequency carrier.
Let the modulated signal be given by, ( ) ( )c ps t A t k m t8
cos 2 10⎡ ⎤= π × +
⎣ ⎦
,
where ( )m t is as shown in Fig. 5.6.
Fig. 5.6: Modulating signal for the exercise 5.1
If ( ) ( )i if fmax min
− is to be 100 kHz show that pk 10= π rad/volt.

5.13
5.2 Bandwidth of FM
In this section, we shall make a detailed analysis of the bandwidth
requirements of FM. (PM is considered in section 5.4.) Let f fc k2= π . For
convenience, let ( )s t denote the FM signal. With ( )
−∞
= τ τ∫
t
Im t m d( ) , the pre-
envelope of ( )s t is
( ) ( )( )c c f Ipe
s t A j t c m texp⎡ ⎤⎡ ⎤ = ω +⎣ ⎦ ⎣ ⎦
( ) ( )( ) ( )( ) c
n
n j tnf f
c f I I I
c c
A j c m t m t j m t e
n
2
2
1
2! !
ω⎡ ⎤
= + − + ⋅ ⋅ ⋅ + + ⋅ ⋅ ⋅⎢ ⎥
⎢ ⎥⎣ ⎦
(5.8)
As ( ) ( ){ }pe
s t s tRe ⎡ ⎤= ⎣ ⎦ , we have
( ) ( ) ( ) ( ) ( )( ) ( )f
c c f I c I c
c
s t A t c m t t m t t
2
2
cos sin cos
2!
⎡ ⎤
= ω − ω − ω + ⋅ ⋅ ⋅⎢ ⎥
⎢ ⎥⎣ ⎦
(5.9)
If ( )m t is band-limited to W , then so is ( )Im t , as ( )
( )
( )I
M f
M f
j f2
=
π
. But ( )s t
consists of terms ( )( ) ( ) ( )( ) ( )I c I cm t t m t t
2 3
cos , sinω ω etc. The spectrum of
( )( )Im t
2
is of width W2 , and that of ( )( )Im t
3
is W3 etc. The spectrum of
( )( ) ( )I cm t t
2
cos ω occupies the interval cf W2± and that of ( )( ) ( )
n
I cm t tcos ω
occupies the interval cf nW± . Clearly, the spectrum of the FM signal is not
band-limited. It appears that, at least theoretically, it has infinite bandwidth and
( )S f is not simply related to ( )M f . A similar situation prevails even in the case
of a PM signal. (Recall that the bandwidth of any linear modulated signal is less
than or equal to W2 , where W is the message band-width). Let us take a closer
look at this issue of FM bandwidth.

5.14
5.2.1 NarrowBand FM (NBFM)
Assume that ( )f Ic m t
max
1<< . Then ( )s t of Eq. 5.9 can be well
approximated by the first two terms; that is,
( ) ( ) ( ) ( )c c f I cs t A t c m t tcos sin⎡ ⎤ω − ω⎣ ⎦ (5.10)
Eq. 5.10 is similar to that of an AM wave. Now ( )S f is band-limited to W2 as in
the case of an AM signal1
. For this reason, the case ( )f Ic m t
max
1<< is called
narrowband FM (NBFM). Similarly, the narrowband PM (NBPM) signal can be
written as
( ) ( ) ( ) ( )c c p cs t A t k m t tcos sin⎡ ⎤ω − ω⎣ ⎦ (5.11)
Though the expressions for NBFM or NBPM resemble fairly closely that of an AM
signal, there is one important difference; namely, the sideband spectrum of PM
or FM has a phase shift of
2
π
with respect to the carrier. This makes the time
domain signal of PM or FM very much different from that of an AM signal. A
NBFM signal has phase variations with very little amplitude variations whereas
the AM signal has amplitude variations with no phase variations.
Note that from Eq. 5.10, we have the envelope of the NBFM signal given by
( ) ( )c f IA t A c m t2 2
1= +
If ( )f Ic m t 1<< , then ( ) cA t A .
Example 5.4
1
In general, FM is nonlinear as
( ) ( )[ ]{ } ( ){ } ( ){ }c c f I I c c f I c c f IA t c m t m t A t c m t A t c m t1 2 1 2cos cos cosω + + ≠ ω + + ω +

5.15
A NBFM signal, ( )s t , is generated by using ( ) ( )m t c t3 2 3
10 sin 10= . We
shall find ( )S f .
From Eq. 5.10, we have
( ) ( ) ( ) ( )c c f I cs t A t c m t tcos sin⎡ ⎤= ω − ω⎣ ⎦
But ( )
( )
I
M f
m t
j f2
←⎯→
π
and f fc k2= π .
Hence
( ) ( ) ( )
( )
( )
( )
( )
c cc c f
c c
c c
M f f M f fA A K
S f f f f f
f f f f2 2
⎡ ⎤− +
⎡ ⎤= δ − + δ + + −⎢ ⎥⎣ ⎦ − +⎢ ⎥⎣ ⎦
where ( )
f
M f tri 3
10
⎛ ⎞
= ⎜ ⎟
⎝ ⎠
5.2.2 WideBand FM (WBFM)
If the condition ( )f Ic m t
max
1<< is not satisfied, then we have the
wideband FM, which, as mentioned earlier has, at least theoretically, infinite
bandwidth. However, as will be seen a little later, most of the power of the FM
signal resides in a finite bandwidth, called the transmission bandwidth. In order
to estimate this bandwidth, we observe that
( ) ( )i c ff t f k m t= +
Let ( ) ( )pm m t m tmax min
= = . Then the instantaneous frequency varies in the
Exercise 5.2
Let ( )
f
M f ga3 3
1
10 10
⎛ ⎞
= ⎜ ⎟
⎝ ⎠
and cf 6
10= Hz. It is given that fk 250=
Hz/volt, and cA 4= V. Sketch ( )S f for a NBFM signal.

5.16
range ( )c f pf k m− to ( )c f pf k m+ . As the total range of frequency deviation (with
centre at cf ) is f pk m2 , can we assume that, for all practical purposes, the
transmission bandwidth of an FM signal, ( )T FM
B is
( )T f pFM
B k m
?
2=
Let f∆ denote the maximum deviation of the carrier frequency cf . That is,
f pf k m∆ = . Then, can we assume that, to a very good approximation,
( )T FM
B f2= ∆ ?
The above expression for the transmission bandwidth is valid only when
f W∆ >> . In the case f W∆ << , ( )T FM
B f2≠ ∆ but W2 , as will be shown
later. This is indeed the fallacy that gave birth to FM in the first place. In the
1920s radio engineers, thinking that the actual bandwidth of an FM signal is
f2∆ , felt that bandwidth requirement of FM can be made less than that of AM
(that is, less than W2 ) by choosing f∆ appropriately! The fallacy here lies in
equating the instantaneous frequency to the spectral frequency. Although ( )if t is
measured in Hz, it should not be equated with spectral frequency. Spectral
frequency f is the independent variable of the frequency domain where as ( )if t
is a time dependent quantity indicating the time behavior of a signal with angle
modulation. When we refer to the spectrum ( )X f of a signal ( )x t , we imply that
( )x t is composed of the complex exponentials with the magnitude and phase
specified by ( )X f , and each one of these exponentials exists for all t with the
given frequency. ( )if t , on the other hand represents the frequency of a cosine
signal that can be treated to have a constant frequency for a very short duration,
maybe only for a few cycles. We shall now give some justification that bandwidth
of an FM signal is never less than W2 .

5.17
Let the base-band signal ( )m t be approximated by a staircase signal, as
shown in Fig. 5.7(a). From the sampling theorem (to be discussed in chapter 6),
this staircase approximation is justified as long as the width of each rectangle
(separation between two adjacent samples) is less than or equal to
W
1
2
. For
convenience, we shall treat the adjacent sample separation to be equal to
W
1
2
.
The FM wave for the staircase-approximated signal will consist of a
sequence of sinusoidal pulses, each of a constant frequency and duration of
W
1
2
sec. A typical pulse and its spectrum are shown in Fig. 5.7(b) and (c)
respectively.
The frequency of the RF pulse in the interval k kt t
W
1
,
2
⎛ ⎞
+⎜ ⎟
⎝ ⎠
is
( ) ( )i k c f kf t f k m t= + . Hence, its spectral range (between the first nulls) as
shown in Fig. 5.7(c), is from ( )c f kf k m t W2⎡ ⎤+ −⎣ ⎦ to ( )c f kf k m t W2⎡ ⎤+ +⎣ ⎦ .
Clearly the significant part of the spectrum of the FM signal will lie in the interval
( )c f pf k m W2− − to ( )c f pf k m W2+ + . (Note that about 92% of the energy of
a rectangular pulse of duration T sec, lies in the frequency range f
T
1
≤ ).
Hence, we can take as one possible value of the transmission bandwidth of an
FM signal, as
( ) ( )T f pFM
B k m W f W f W2 4 2 4 2 2= + = ∆ + = ∆ + (5.12)
For the wideband FM case, where f W∆ >> , ( )T FM
B can be well approximated
by f2∆ .
In the literature, other rules of thumb for FM bandwidths are to be found.
The most commonly used rule, called the Carson’s rule, is
( ) ( )T FM
B f W2= ∆ + (5.13)

5.18
Fig. 5.7: Estimation of FM spectral width
Carson’s rule gives a better bandwidth estimate than does Eq. 5.12 for the NBFM
( )f W∆ << case. This is in agreement with our result on NBFM, namely, its
bandwidth is approximately W2 . In other cases, where f W∆ << is not satisfied
(wideband and intermediate cases) Eq. 5.12 gives a better estimate than does
Eq. 5.13.

5.19
We now define the deviation ratio D , as
f
D
W
∆
= (5.14)
Then Eq. 5.12 and 5.13 can be combined into
( ) ( )T FM
B W D k2= + (5.15)
where k varies between 1 and 2. As can be seen, Eq. 5.12 and Eq. 5.13 are
special cases of Eq. 5.15.
5.3 Tone Modulation
Let ( ) ( )m mm t A f tcos 2= π
Then ( ) ( )m
I m
m
A
m t tsin= ω
ω
, assuming ( )Im 0− ∞ =
( ) ( )m
pe c c f m
m
A
s t A j t c texp sin
⎡ ⎤⎛ ⎞
= ω + ω⎢ ⎥⎜ ⎟
ω⎝ ⎠⎣ ⎦
( )f m
c c m
m
k A
A j t t
f
exp sin
⎡ ⎤⎛ ⎞
= ω + ω⎢ ⎥⎜ ⎟
⎝ ⎠⎣ ⎦
For tone modulation, f mf k A∆ = and mW f= . In this case, the deviation ratio is
referred to as the modulation index and is usually denoted by the symbol β .
That is, for tone modulation, f m
m
k A
f
β = and
( ) ( )mc j tj t
pe cs t A e e
sinβ ωω ⎡ ⎤= ⎢ ⎥⎣ ⎦
(5.16a)
( ) ( ) ( )( )pe c c ms t s t A t tRe cos sin⎡ ⎤= = ω + β ω⎣ ⎦ (5.16b)
5.3.1 NBFM
( ) ( )c c ms t A t tcos sin⎡ ⎤= ω + β ω⎣ ⎦
( ) ( )( ) ( ) ( )( ){ }c c m c mA t t t tcos cos sin sin sin sin= ω β ω − ω β ω

5.20
NBFM: small β .
Let β be small enough so that we can use the approximations
( )( )m tcos sin 1β ω , and
( )( ) ( )m mt tsin sin sinβ ω β ω .
Then, ( ) ( ) ( ) ( )c c m cs t A t t tcos sin sin⎡ ⎤= ω − β ω ω⎣ ⎦ (5.17a)
( ) ( ) ( )c c c m c mA t t tcos cos cos
2
β⎧ ⎫
⎡ ⎤= ω + ω + ω − ω − ω⎨ ⎬⎣ ⎦
⎩ ⎭
(5.17b)
Corresponding expression for AM (Eq 4.7) is
( ) ( ) ( ) ( )c c c m c mAM
s t A t t tcos cos cos
2
µ⎧ ⎫
⎡ ⎤⎡ ⎤ = ω + ω + ω + ω − ω⎨ ⎬⎣ ⎦ ⎣ ⎦
⎩ ⎭
Eq. 5.17(b) can be written as
( ) ( ) ( )c m c mc j t j tj t c
c
A
s t A e e eRe
2
ω + ω − ω − ωω β⎧ ⎫⎡ ⎤= + −⎨ ⎬⎢ ⎥⎣ ⎦⎩ ⎭
(5.17c)
Fig. 5.8: Phasor diagram for NBFM with tone modulation
Using Eq. 5.17(c), we construct the phasor diagram for NBFM, shown in Fig. 5.8.
(We have taken the carrier phasor as the reference.) Comparing the phasor
diagram for the NBFM with that of the AM signal (Fig. 4.26), we make the
following observation.
In the case of AM, the resultant of the side-band phasors is collinear with
the carrier phasor whereas, it is perpendicular to the carrier phasor in NBFM.

5.21
It is this quadrature relationship between the two phasors of NBFM that
produces angle variations resulting in corresponding changes in ( )if t .
5.3.2 WBFM
The exponential term in the brackets in Eq. 5.16(a) is a periodic signal
with period
mf
1
. Expressing this term in Fourier series, we have
( )m mj t j n t
n
n
e x e
sin
∞
β ω ω
= − ∞
= ∑
where ( )
m
m m
m
f
j t j n t
n m
f
x f e e d t
1
2
sin
1
2
β ω − ω
−
= ∫
Let m tθ = ω . Then,
( )j n
nx e d
sin1
2
π
β θ − θ
− π
= θ
π ∫ (5.18)
The integral on the RHS of Eq. 5.18 is recognized as the nth order Bessel
function of the first kind and argument β and is commonly denoted by the symbol
( )nJ β ; that is,
( ) ( )j n
n nx J e d
sin1
2
π
β θ − θ
− π
= β = θ
π ∫ (5.19a)
That is,
( )
( )m mj t j n t
n
n
e J e
sin
∞
β ω ω
= − ∞
= β∑ (5.19b)
Hence,
( )
( ) ( )mj t
n m
n
e J f nf
sin
∞
β ω
= − ∞
←⎯→ β δ −∑ (5.19c)
Using Eq. 5.19, we can express ( )pes t as

5.22
( ) ( ) ( )c mj f nf t
pe c n
n
s t A J e
2
∞
⎡ ⎤π +⎣ ⎦
= − ∞
= β∑ (5.20)
As ( ) ( )pes t s tRe⎡ ⎤= ⎣ ⎦ , we obtain
( ) ( ) ( )c n c m
n
s t A J f nf tcos 2
∞
= − ∞
⎡ ⎤= β π +⎣ ⎦∑ (5.21)
Taking the Fourier transform of ( )s t , we have
( ) ( ) ( ) ( )c
n c m c m
n
A
S f J f f nf f f nf
2
∞
= − ∞
⎡ ⎤= β δ − − + δ + +⎣ ⎦∑ (5.22)
Properties of ( )nJ β : The following properties of ( )nJ β can be established.
1) ( )nJ β is always real (For all n and β )
2) ( ) ( ) ( )n
n nJ J1 −β = − β
3) For small values of β , ( )
( )n
nJ
n
2
!
β
β
Exercise 5.3
Eq. 5.19(c) gives the Fourier transform of ( )mj t
e
sinβ ω
. Show that
a) ( )
( ) ( ) ( )m nj t
n m
n
e j J f nf
cos
∞
β ω
= − ∞
←⎯→ β δ +∑
b) ( ) ( ) ( )c c m c n c m
n
A t t A J n t ncos cos cos
2
∞
= − ∞
π⎡ ⎤
⎡ ⎤ω + β ω = β ω + ω +⎣ ⎦ ⎢ ⎥
⎣ ⎦
∑
Hint: ( ) m
mm
j t
j t
e e
sin
2cos
⎡ ⎤⎛ ⎞π
β ω −⎢ ⎥⎜ ⎟
ωβ ω ⎢ ⎥⎝ ⎠⎣ ⎦=
If ( ) ( )mj t
x t e
sinβ ω −
= , then ( )mj t
m
e x t
cos
2
β ω ⎛ ⎞π
= −⎜ ⎟
ω⎝ ⎠
.
Now use the Fourier transform properties for time reversal and time shift.

5.23
Hence, for small values of β ,
( )J0 1β
( )J1
2
β
β
( )nJ n0, 1β >
4) ( )n
n
J2
1
∞
= − ∞
β =∑
Fig 5.9 depicts the behavior of ( )nJ β . We now make the following observations
regarding the spectrum ( )f 0> of a tone modulated FM signal.
1) The spectrum of a tone modulated FM wave contains carrier component
and an infinite set of side-frequencies located symmetrically on either side
of the carrier at frequency separations of mnf n, 1, 2,= ⋅ ⋅ ⋅
2) For small β , only ( )J0 β and ( )J1 β are significant. Then the FM spectrum
has only three components: at c c mf f f, ± . This situation corresponds to the
special case of NBFM.
3) The amplitude of the carrier component varies with β according to ( )J0 β .
Thus, in contrast to AM, this amplitude ‘contains’ part of the message
information. Moreover, ( )J0 0β = for some values ( )2.4, 5.5, etc.β β = .
4) From Fig. 5.9(b), we find that ( )nJ β decays monotonically for
n
1>
β
and
that ( )nJ 1β << for
n
1>>
β
.
Fig, 5.10 gives the spectral plots obtained from a spectrum analyzer with
2.4β . As can be seen from the plot, the carrier component is very nearly zero
(approximately 50 dB below the maximum value).

5.24
Fig 5.11(a) gives the theoretical plot of various spectral components for
4β . Fig. 5.11(b) is the spectral plot as observed on a spectrum analyzer. We
see from this plot that the third spectral component has the largest magnitude
which is an agreement with the theory. Let us normalize the dB plot with respect
to the largest magnitude; that is, we treat the largest magnitude as 0 dB and
compare the rest respect to with this value. For n 4= , Fig. 5.11(a), we have
( )J4 0.281β = . Hence 10
0.281
20 log 3.7
0.43
= − and value indicated by the
spectrum analyzer agrees with this. Similarly, for n 5= , we expect
10
0.132
20 log 10.25
0.430
= − ; the value as observed on the plot of Fig. 5.11(b) is in
close agreement with this. For n 1= , theoretical value is
0.066
20 log 17.1
0.43
= −
dB. Spectrum analyzer display is in close agreement with this. The values of the
remaining spectral components can be similarly be verified.
From Eq. 5.21, the average power avP of the FM signal with tone
modulation is ( )c
av n
n
A
P J
2
2
2
∞
= − ∞
= β∑ .

5.25
Fig. 5.9: Bessel Functions:
(a) ( )nJ β vs. β for fixed n
(b) ( )nJ β vs.
n
β
for fixed β

5.26
Fig. 5.10: Spectrum analyzer output with 2.4β
But from property 4 of ( )nJ β , we have ( )n
n
J2
1
∞
= − ∞
β =∑ , which implies that the
avP of the tone modulated FM signal is cA2
2
. This result is true in general
(whether ( )m t is a tone or not) because an FM signal is essentially a constant
amplitude cosine signal with time varying frequency. As the RMS value of a
sinusoid is independent of the frequency, we state that
( ) c
av FM
A
P
2
2
= (5.23)
For large β , if we assume that we can neglect ( )nJ β for n 2> β + , then
( ) ( )T m mFM
B nf f2 2 2= = β + (5.24a)
This in agreement with Eq. 5.15, with k 2= . If we neglect the values of ( )nJ β
for n 1> β + , then,
( ) ( )T mFM
B f2 1= β + (5.24b)
which corresponds to Eq. 5.15 with k 1= .

5.27
Fig. 5.11(a): ( )nJ β for 4β =
(b) Spectrum analyzer output for 4β
Consider 5β = . From Appendix A5.1, we have ( )J7 5 0.053= and
( )J8 5 0.018= . P1, the average power of these two is,

5.28
( ) ( )c cA A
P
2 2
1 0.0028 0.0003 0.0031
2 2
= + = .
Hence,
( )av FM
P
P
1 0.0031 0.31%= = . As ( )nJ 5 for n 9≥ are much smaller than
( )J8 5 , we see that practically the entire power of the FM signal is confined to the
frequency range [ ]±c mf f6 . As such, we find that even Eq. 5.24(b) is a fairly
good measure of ( )T FM
B for tone modulation.
With tone modulation, it is possible to estimate the value of TB to the
accuracy desired. In this case, the magnitude of a spectral component depends
only on ( )nJ β and these Bessel coefficients have been tabulated extensively.
For a given β , let us define transmission bandwidth such that it includes all those
spectral components whose magnitude is greater than one percent of the
unmodulated carrier amplitude. If we take cA 1= , then
T sig mB n f2= (5.24c)
where sign is such that ( )nJ 0.01β ≥ for sign n≤ . For example, if 20β = ,then
sign would be 25. (For 10, 5,β = etc., sign can be found from the table in
Appendix A5.1.) Let mf f0= when 20β = . Let us tabulate TB for a few other
values of β . We will keep f f020∆ = , but reduce the value of β by increasing
mf . These are listed in Table 5.1. Also listed in the table are TB ,1 and TB ,2 where
( )T mB f,1 2 1= β + (Carson’s rule), and
( )T mB f,2 2 2= β +

5.29
mf β 2 sign TB ,1TB ,2TB
f0 20 50 50 f0 42 f0 44 f0
2 f0 10 28 56 f0 44 f0 48 f0
4 f0 5 16 64 f0 48 f0 56 f0
10 f0 2 8 80 f0 60 f0 80 f0
20 f0 1 6 120 f0 80 f0 120 f0
40 f0 0.5 4 160 f0 120 f0 200 f0
60 f0 0.33 4 240 f0 160 f0 280 f0
200 f0 0.1 2 400 f0 440 f0 840 f0
Table 5.1: TB of Eq. 5.24c for various values of β with ∆ =f f020
From the above table, we see that
i) For small values of β (less than or equal to 0.5), f∆ becomes less and less
significant in the calculation of TB . For 0.1β = , we find that
T
ff
B f
0
0
402 1
400 10
∆
= =
ii) For small values of β , bandwidth is essentially decided by the highest
frequency component in the input spectrum. As such, as 0β → , TB does
not go zero. In fact, in absolute terms, it increases. (From the table, with
20β = , we require a TB of f050 where as with 0.1β = , TB required is
f0400 .)
iii) TB ,1, which is based on Carson’s rule is in close agreement with TB only
for very small values of β . Otherwise, TB ,2 is better.
It is interesting to note that for β = 5, 10 and 20, TB ,2 (which is generally
considered to overestimate the bandwidth requirement) is less than TB as given

5.30
by Eq. 5.24(c). This is because TB of Eq. 5.24(c) neglects only those frequency
components whose magnitude is less than or equal to cA0.01 . In other words,
the ratio
T
av
B
P
Power of any spectral component not included in
is less than
( )c
c
A
A
2
4
2
0.01 2
10
2
−
=
With 5β = , TB ,2 takes into account only upto ( )J7 5 and
( )( )
( )
−
= ×
c
c
J A
A
2
8 4
2
5 2
3.38 10
2
. That is, TB as given by Eq. 5.24(c) includes few
more spectral components in the bandwidth than that taken into account by TB ,2
resulting in a larger value for the bandwidth.
It is instructive to analyze the FM signal when ( )m t is a sum of sinusoids.
This has been done in the Appendix A5.3. Another measure of bandwidth that is
useful in the study of frequency modulation is the rms bandwidth. This has been
dealt with in Appendix A5.4.
5.4 Phase Modulation
All the results derived for FM can be directly applied to PM. We know that,
for phase modulation,
( )
( )
( )pi
i c
kd t
f t f m t
d t
1 '
2 2
θ
= = +
π π
(5.25a)
where ( )m t' is the derivative of ( )m t . Therefore,
p
p
k
f m'
2
∆ =
π
, where ( )pm m t
max
' '= (5.25b)

5.31
( ) [ ]T PM
B f kW k2 , 1 2= ∆ + < <
p pk m
kW
'
2
2
⎡ ⎤
⎢ ⎥= +
⎢ ⎥π
⎣ ⎦
(5.26)
With respect to f∆ , there is one important difference between PM and
FM. For FM, as f pf k m∆ = , it depends only on the peak value of the modulating
signal. Hence the bandwidth of WBFM is essentially independent of the spectrum
of ( )m t (It has a weak dependence on W ). On the other hand, for PM, f∆ is a
function of pm' and this value is strongly dependent on ( )M f . Strong high
frequency components in ( )M f would result in large values of pm' . Similarly,
predominance of lower-frequency components will result in a lower value of pm' .
As a result, for signals with strong high frequency components, ( )T PM
B will be
much larger than for the case of the signals with strong spectral components at
lower frequencies.
Example 5.5
To determine the frequency sensitivity of an FM source, the following
method has been used. The FM generator is fed a tone modulating signal
( )mA t3
cos 4 10⎡ ⎤π ×
⎣ ⎦
. Starting near about zero, mA is gradually increased and
when mA 2= V, it has been found that the carrier component goes to zero for
the first time. What is the frequency sensitivity of the source? Keeping mA at 2 V,
frequency mf is decreased until the carrier component goes zero for the second
time. What is the value of mf for this to happen?

5.32
With tone modulation, we know that carrier goes to zero for the first time
when 2.4β = . (That is, the smallest value of β for which ( )J0 0β = is
2.4β = .)
But
m
f
f
2.4
∆
β = = . That is,
m f pf f k m2.4 4.8 kHz∆ = = =
As pm is 2 V, we have fk
4.8
2.4
2
= = kHz/V
The carrier component is zero for the second time when 5.5β = . Hence,
mf
3
4.8 10
5.5
×
=
or mf
3
4.8 10
872
5.5
×
= = Hz
Example 5.6
A 1.0 kHz tone is used to generate both an AM and an FM signal.
Unmodulated carrier amplitude is the same for both AM and FM. The modulation
index β of FM is 8. If the frequency components at ( )cf 1000± Hz have the
same magnitude in AM as well as FM, find the modulation index of AM.
For AM, the magnitude of the spectral components at ( )cf 1000± Hz is
cA
2
µ
. For FM, the magnitude of the spectral components at ( )cf 1000± Hz is
( )cA J1 8 .
( ) ( )c
c
A
A J J1 18 2 8
2
µ
= ⇒ µ =
0.47=

5.33
Example 5.7
In this example we will show that the bandwidth of the PM signal is
strongly dependent on ( )M f whereas this dependence is weak in the case of an
FM signal.
Let ( ) ( ) ( )m t f t f t1 2cos 2 0.85 cos 2= π + π be used to generate an FM as
well as a PM signal. Modulator constants are fk 5= kHz/V and pk 10= rad/V.
Let us compute (a) ( )T FM
B and (b) ( )T PM
B for the following combinations
of f1 and f2 .
i) f1 500= Hz and f2 700= Hz
ii) f1 1000= Hz and f2 1400= Hz
We will assume that ( )nJ β can be neglected if n 2> β + .
a) ( )T FM
B
Case i) f1 500= Hz and f2 700= Hz
From Eq. A5.3.1 (Appendix 5.3), we have
( ) ( ) ( ) ( )c m n c
m n
s t A J J m n t1 2 1 2cos⎡ ⎤= β β ω + ω + ω⎣ ⎦∑∑
3
1
5 10
10
500
×
β = =
3
2
5 10 0.85
6
700
× ×
β =
We shall take into account ( )mJ 1β upto ( )m 112 2= = β + and
( )nJ 2β upto ( )n 28 2= = β + . Then,
( ) ( )T FM
B 2 12 500 8 700= × + ×
3
23.2 10= × Hz

5.34
In the equation for ( )s t above, we see that the magnitude of
any spectral component depends on the product. ( ) ( )m nJ J1 2β β . Let
us calculate the ratio
( ) ( )
( ) ( )m n
J J
J J
13 8
max
10 6
10 6⎡ ⎤⎣ ⎦
so as to get an idea of the
magnitude of the spectral components that have been neglected in
calculating ( )T FM
B .
Let ( ) ( )A J J13 910 6 0.03 0.02= = ×
0.0006=
From the tables (Appendix A5.1), we find that the maximum value of
the product ( ) ( )m nJ J10 6 occurs for m 8= and n 5= . Let,
( ) ( )B J J8 510 6 0.318 0.362= = ×
0.115= ,
Then,
A
B
0.0052= .
Case ii) f1 1000= Hz and f2 1400= Hz
Now, we have 1 5β = and 2 3β .
If we account for ( )mJ 1β and ( )nJ 2β upto m 7= and n 5= ,
then
( ) ( )T FM
B 3
2 7 10 5 1400= × + ×
= × 3
28 10 Hz
Of course the maximum value of ( )mJ 1β and ( )nJ 2β occurs for
m 4= and n 2= , and this product is
( ) ( )C J J4 25 3 0.391 0.486 0.19= = × =
However, ( ) ( )D J J8 65 3 0.018 0.0114= = ×
0.0002=

5.35
Hence
D
C
0.0002
0.001
0.19
=
This is much less than ( )A B which implies that we have taken into
account more number of spectral components than case (i) above. If
we restrict ourselves to the number of spectral components upto
m 1= β + , then,
( ) ( )
( ) ( )
J J
J J
7 6
4 2
5 3
0.003
5 3
= which is closer to ( )A B than 0.001.
Then,
( ) ( )T FM
B 3
2 6000 5 1400 26 10= + × = × Hz
which is fairly close to the previous value of 3
23.2 10× Hz.
b) ( )T PM
B
( ) ( ) ( ) ( )m t f f t f f t1 1 2 2
' 2 sin 2 0.85 2 sin 2⎡ ⎤= − π π + π π⎣ ⎦
Now ( )m t' frequency modulates the carrier
p
p
k f
k
f
1
1
1
2
10
2
π
β = ⋅ = =
π
Similarly,
pk2 0.85 8.5β = =
Case i) f1 500= Hz and f2 700= Hz
( ) ( )T PM
B 3
2 12 500 10 700 26 10= × + × = × Hz
Case ii) f1 1000= Hz and f2 1400= Hz
As 1β and 2β remains unchanged, we have
( ) ( )T PM
B 3
2 12 10 10 1400= × + ×
3
52 10= × Hz

5.36
This is twice ( )T PM
B of case (i). This example shows that the
bandwidth of the PM signals is strongly dependent on ( )M f .
5.5 Generation of FM
We had earlier identified two different categories of FM, namely, NBFM
and WBFM. We shall now present the schemes for their generation.
Exercise 5.4
Consider the scheme shown in Fig. 5.12.
Fig. 5.12: Scheme for the Exercise 5.4
a) Let cf be the frequency of the modulator when ( )m t 0= . If ( )s t can be
expressed in the form
( ) ( )⎡ ⎤= ω − β ω
⎢ ⎥⎣ ⎦
cos ' sin 2c c ms t A t t ,
what are the values of cf ' and β ?
b) What are the frequency components in the spectrum of ( )s t ?
Ans: f m
c c
k A
f f
2
'
2
⎛ ⎞
= +⎜ ⎟⎜ ⎟
⎝ ⎠
f m
m
k A
f
2
2
β =

5.37
5.5.1 Narrowband FM
One of the principal applications of NBFM is in the (indirect) generation of
WBFM as explained later on in this section. From the approximation (5.10) we
have,
( ) ( ) ( ) ( )c c f I cs t A t c m t tcos sin⎡ ⎤ω − ω⎣ ⎦
The system shown in Fig. 5.13 can be used to generate the NBFM signal.
Applying ( )m t directly to the balanced modulator, results in NBPM with a
suitable value for pk .
Fig. 5.13: Generation of NBFM signal
5.5.2 WBFM: Indirect and direct methods
There are two distinct methods of generating WBFM signals: a) Direct FM
b) Indirect FM. Details on their generation are as follows.
a) Indirect FM (Armstrong’s method)
In this method - attributed to Armstrong - first a narrowband FM signal is
generated. This is then converted to WBFM by using frequency multiplication.
This is shown schematically in Fig. 5.14.

5.38
Fig. 5.14: Generation of WBFM (Armstrong method)
The generation of NBFM has already been described. A frequency
multiplier is a nonlinear device followed by a BPF. A nonlinearity of order n can
give rise to frequency multiplication by a factor of n . For simplicity, consider a
square law device with output ( ) ( )y t x t2
= where ( )x t is the input. Let ( )x t be
the FM signal given by,
( ) ( )x t tcos⎡ ⎤= θ⎣ ⎦ , where ( ) ( )
t
c ft t k m d2
− ∞
θ = ω + π α α∫
Note that we have dropped the subscript i from ( )i tθ . Then,
( ) ( )y t t2
cos ⎡ ⎤= θ⎣ ⎦
( ){ }t
1
1 cos 2
2
⎡ ⎤= + θ⎣ ⎦
( )
t
c ft k m d
1 1
cos 2 4
2 2 − ∞
⎡ ⎤
⎢ ⎥= + ω + π α α
⎢ ⎥
⎣ ⎦
∫ (5.27)
The DC term in Eq. 5.27 can be filtered out to give an FM output with the
carrier frequency cf2 and frequency deviation twice that of the input FM signal.
An input-output relation of the type ( ) ( ) ( ) ( )n
ny t a x t a x t a x t2
1 2 ....= + + will
give rise to FM output components at the frequencies c c cf f nf, 2 , ...., with the
corresponding frequency deviations f f n f, 2 , ....,∆ ∆ ∆ , where f∆ is the
frequency deviation of the input NBFM signal. The required WBFM signal can be
obtained by a suitable BPF. If necessary, frequency multiplication can be
resorted to in more than one stage. The multiplier scheme used in a commercial
FM transmitter is indicated in Fig. 5.15.

5.39
Fig. 5.15: Multiplier chain used in typical commercial FM transmitter
The carrier frequency of the NBFM signal cf 1, is 200 kHz with the corresponding
f1 25∆ = Hz. Desired FM output is to have the frequency deviation f4 75∆
kHz and a carrier ( )cf 4 of 91.2 MHz.
To obtain f4 75∆ = kHz starting from f1 25∆ = Hz, we require a total
frequency multiplication of
( )3
75 10
3000
25
×
= . In the scheme of Fig. 5.11, this
has been accomplished in two stages, namely, multiplication by 64 followed by
multiplication by 48, giving a total multiplication by the factor 64 48 3072× = .
(Actually each stage of multiplication is implemented by a cascade of frequency
doublers or triplers. Thus multiplication by 64 is obtained by 6 doublers in
cascade where as multiplication by 48 is implemented by a cascade of a
frequency tripler and 4 doublers.) Multiplication of cf 1 200= kHz by 3072 gives a
carrier frequency cf 4 614.4= MHz. As the final required carrier frequency is
91.2 MHz, a frequency conversion stage is used to down convert cf 2 (12.8 MHz)
to cf 3 (1.9 MHz). In this process of down conversion, frequency deviation is
unaffected ( )f f2 3 1.6 kHz∆ = ∆ = . The possible drawbacks of this scheme are
the introduction of noise in the process of multiplication and distortion in the

5.40
generation of NBFM signal especially for low modulating frequencies as
m
f
f
∆
could become excessive.
Example 5.8
Armstrong’s method is to be used to generate a WBFM signal. The NBFM
signal has the carrier frequency cf 1 20= kHz. The WBFM signal that is required
must have the parameters cf 6= MHz and f 10∆ = kHz. Only frequency
triplers are available. However, they have a limitation: they cannot produce
frequency components beyond 8 MHz at their output. Is frequency conversion
stage required? If so, when does it become essential? Draw the schematic block
diagram of this example.
Total frequency multiplication required
6
3
6 10
300
20 10
×
= =
×
. Using only
frequency triplers, we have 5
3 243= and 6
3 729= . Hence a set of six
multipliers is required. But these six cannot be used as a single cascade
because, that would result in a carrier frequency equal to 3 6
20 10 3 14.58× × =
MHz and the last multiplier cannot produce this output. However, cascade of 5
triplers can be used. After this, a frequency conversion stage is needed.
Fig 5.16: Generation of WBFM from NBFM of example 5.8
Block diagram of this generation scheme is shown in Fig. 5.16. As the final
frequency deviation required is 10 kHz, the NBFM must have a

5.41
( )f
3
1
10 10
13.71
729
×
∆ = = Hz. After the frequency conversion stage, we have
one more stage multiplication by 3. If cf 3 is the carrier frequency at the mixer
output, then cf 3 3 6× = MHz cf 3 2⇒ = MHz. Assuming that Lof is greater than
incoming carrier frequency of 4.86 MHz, we require Lof 6.86= MHz so that the
difference frequency component is 2 MHz.
Exercise 5.5
In the indirect FM scheme shown in Fig. 5.17, find the values of c if , and
if∆ for i 1, 2 and 3= . What should be the centre frequency, f0 , of the BPF.
Assume that LO cf f ,2> .
Exercise 5.6
In the indirect FM scheme shown in Fig. 5.18, find the values of the
quantities with a question mark. Assume that only frequency doublers are
available. It is required that LO cf f ,2< .

5.42
b) Direct FM
Conceptually, the direct generation of FM is quite simple and any such system
can be called as a Voltage Controlled Oscillator (VCO). In a VCO, the oscillation
frequency varies linearly with the control voltage. We can generate the required
FM signal from a VCO by using the message signal ( )m t in the control voltage.
We shall explain below two methods of constructing a VCO.
One can construct a VCO by using an integrator and hysteric comparator
(such as Schmitt trigger circuit). Method 1 below gives the details of this scheme.
Another way of realizing a VCO is to vary the reactive component of a tuned
circuit oscillator such as Colpitt’s oscillator or Hartley oscillator. Method 2 below
gives the details of this scheme.
Method 1: Consider the scheme shown in Fig. 5.19(a). The Schmitt trigger
output ( )y t is V0+ when the integrator output ( )x t is increasing, and V0−
when ( )x t is decreasing. Further, ( )y t changes from V0+ to V0− when ( )x t
reaches E0+ and from V0− to V0+ when ( )x t reaches E0− . The relationship
between ( )x t and ( )y t is shown in Fig. 5.19(b). The electronic switch is
designed such that it is in position 1 when ( )y t V0= and goes to position 2
when ( )y t V0= − . Let us first examine the case when the input ( )mv t is a
positive constant v0 . Consider the situation at t 0= when ( )y t has just
switched to V0 and the switch going from position 2 to position 1. At this point,
( )x t has attained the value E0− . Then, for t t10 < ≤ , we have
( )
t
x t E v d
RC0 0
0
1
= − + τ∫
When t t1= , let ( )x t become E0 . Then ( )y t goes to V0− and the electronic
switch assumes position 2. The value of t1 can be obtained from

5.43
t
E E v d
RC
1
0 0 0
0
1
= − + τ∫
or
RCE
t
v
0
1
0
2
=
Fig. 5.19: Direct FM generation (method 1)
The output ( )x t now keeps decreasing until t t2= when ( )x t E0= − . It is easy
to see that ( )
RCE
t t
v
0
2 1
0
2
− = . That is, ( )x t and ( )y t are periodic with period
RCE
v
0
0
4
or the fundamental frequency of these waveforms is
RCE
f
v
1
0
0
0
4
−
⎡ ⎤
= ⎢ ⎥
⎣ ⎦
.
Note that f0 depends on the input signal v0 . If v E0 0= then f
RC
0
1
4
= . By
properly choosing the values of R and C , we can have cf f
RC
0
1
4
= = , where

5.44
cf is the required carrier frequency. If the triangular wave ( )x t is the input to a
narrowband BPF tuned to cf , the output ( )s t is ( )c cA f tcos 2π . Then, by making
( ) ( )mv t v k m t0= + where k is a constant such that ( )k m t v0< , we can have
the instantaneous frequency of ( )s t as ( )c ff k m t+ , which is the desired result.
Method 2: The oscillation frequency f0 of a parallel tuned circuit with inductance
L and capacitance C is given by
f
LC
0
1
2
=
π
or
LC
0
1
ω =
Let C be varied by the modulating signal ( )m t , as given by
( ) ( )C t C k m t0= −
where k is an appropriate constant.
Then ( )
( )
i t
k m t
LC
C0
0
1
1
ω =
⎡ ⎤
−⎢ ⎥
⎣ ⎦
( )k m t
CLC 00
1
1
2
⎡ ⎤
+⎢ ⎥
⎣ ⎦
when
( )k m t
C0
1<<
Let c
LC0
1
ω = . Then,
( ) ( )i c ft c m tω = ω + , where c
f
k
c
C02
ω
= .
One of the more recent devices for obtaining electronically variable capacitance
is the varactor (also called varicap, or voltacap). In very simple terms, the
varactor is a junction diode. Though all junction diodes have inherent junction
capacitance, varactor diodes are designed and fabricated such that the value of
the junction capacitance is significant (varactors are available with nominal
ratings from 0.1 to 2000 pF). Varactor diodes, when used as voltage-variable
capacitors are reverse biased and the capacitance of the junction varies
inversely with the applied (reverse) voltage.

5.45
Fig. 5.20: Direct FM generation (method 2)
Consider the scheme shown in Fig. 5.20. The frequency of the oscillator
output depends on L1 and eqC where eq dC C C1 ||= , dC being the capacitance
of the varactor D . BV reverse biases D such that when ( )m t 0= , eqC is of the
correct value to result in output frequency cf . When the message signal ( )m t is
on, dC can be taken as, ( )dC C k m t0
'= − where C0
' is the value of dC , with
( )m t 0= . Hence ( )C t of the oscillator circuit is,
( ) ( ) ( ) ( )C t C C k m t C k m t1 0 0
'= + − = − , where C C C0 1 0
'= +
The other components in Fig. 5.14 have the following functions. C2 is a
blocking capacitor to isolate the DC level of the varactor from the rest of the
modulator circuit. Note that dC C2 >> or C1 and is essentially a short at the
required FM frequencies. RFC is an RF choke that prevents the RF energy of the
oscillator circuit from feeding into the audio transformer. R1 limits the current in
the circuit in the event that the peaks of the audio signal voltage exceed the
voltage of the DC source and momentarily forward bias the diode D .
Direct FM generation can produce sufficient frequency deviation and
requires little frequency multiplication. However, it suffers from the carrier

5.46
frequency instability as the carrier frequency is not obtained from a highly stable
oscillator. Some additional circuitry is required to achieve frequency stability.
Voltage controlled oscillators are available in the form of IC chips. For
example XR-2206 of the EXAR Corporation, USA has an operating frequency
range of 0.01 Hz to 1 MHz. Another chip with an operating frequency range of
about 1 MHz is LM 566. Texas Instruments CD 4046 is an inexpensive VCO chip
with a normal operating frequency range upto 1.4 MHz. (Actually CD 4046 is a
PLL and VCO is a part of it. PLLs are discussed in sec. 5.5.4.) To make the VCO
chip functional, what is required is an external capacitor and one or two external
resistors. CD 74HC7046 is another PLL chip with VCO. Free running frequency
of the VCO is 18 MHz. Koster, Waldow and Ingo Wolf describe a VCO operating
in the frequency range 100 MHz to 4 GHz [1]. Donald Tillman describes a new
VCO design (called a quadrature trapezoid VCO) especially suited for electronic
music applications. Details are available at:
http://www.till.com/articles/QuadTrapVCO/discussion.html.

5.47
Exercise 5.7
Consider the circuit of Fig. 5.20 for the direct generation of FM. The
diode capacitance dC , is related to the reverse bias as,
d
d
C
v
100
1 2
=
+
pF
where dv is the voltage across the varactor. Let BV 4= V and
( ) ( )m t t3
0.054 sin 10 10⎡ ⎤= π ×
⎣ ⎦
. It is given that C1 250= pF and the circuit
resonates at 2 MHz when ( )m t 0= .
a) Show that, by using the binomial approximation, dC can be put in the
form
( )dC t
10
12 310
0.2 10 sin 10 10
3
−
− ⎡ ⎤= − × π ×
⎣ ⎦
b) Show that
( ) ( ) ( )if t t6 3
2 10 705.86 sin 10 10⎡ ⎤= × + π ×
⎣ ⎦
(Note that C2 is a short circuit at the frequencies being generated.)
c) Let cA be the amplitude of oscillations of the VCO. Write the expression
for the generated FM signal.

5.48
Exercise 5.8
A simple variation of the circuit of Fig. 5.20 is shown in Fig. 5.21.
Fig. 5.21: A variation of the VCO circuit of Fig. 5.14
The varactor diodes D1 and D2 are connected back-to-back. This
arrangement helps to mitigate the effect of the RF signal of the tuned circuit
(also called tank circuit) driving a single diode into conduction on its peaks
which will change the bias voltage (and thereby the frequency that is
generated). Of course, as far as the tuned circuit is concerned, the two diodes
are in series; this means the capacitance of the combination is one-half that of
a single diode.
Let BV 5= V, ( ) ( )m t t3
0.5cos 2 10= π × . Calculate the carrier
frequency cf , ( )if max
and ( )if min
from the above scheme, given that C1 50=
pF, L1 50= µ H and the capacitance of the varactor diode dC , follows the
relation
d
d
C
v
50
= pF
where dv is the voltage across the diode. Do not use any approximations.
What are the values of ( )if max
and ( )if min
, if you use binomial approximation.
Ans: cf 91 kHz, ( )if max
91.4 kHz and ( )if min
90.6 kHz, without any
approximation.

5.49
5.6 Demodulation of FM
A variety of techniques and circuits have been developed for
demodulating FM signals. We shall consider a few of these techniques falling
under the following categories:
1) FM-to-AM conversion
2) Phase shift discrimination
3) Zero crossing detection
4) Phase Locked Loop (PLL)
5.6.1 FM-to-AM conversion
The instantaneous frequency of an FM signal is given by ( )i c ff f k m t= + .
Hence a frequency selective network with a transfer function of the from
( )H f f= α + β , (f 0> , and α and β are constants) over the FM band would
yield an output proportional to the instantaneous frequency. That is, the circuit
converts the frequency deviation into a corresponding amplitude change, which
in this case is proportional to ( )m t , the message signal. It is assumed that the
time constant of the network is small enough in comparison with the variations in
the instantaneous frequency of the FM signal. We shall now indicate three ways
of implementing this demodulation scheme.
Consider the scheme shown in Fig. 5.22 where
d
d t
represents a band-
pass differentiator with the magnitude characteristic ( )H f f= α + β, (for f 0> ),
over the required bandwidth. BPL1
is a Band-Pass Limiter which eliminates
amplitude fluctuations from the received FM signal.
1
Band-pass limiters have been analyzed in section 5.7.

5.50
Fig. 5.22: Schematic of an FM demodulator based on FM - to - AM conversion
( )s t is the constant amplitude FM signal to be demodulated. ( )s t' , the
output of the differentiator, is given by
( ) ( ) ( )
t
c c f c fs t A k m t t k m d' 2 sin 2
− ∞
⎡ ⎤
⎢ ⎥⎡ ⎤= − ω + π ω + π α α⎣ ⎦ ⎢ ⎥
⎣ ⎦
∫ (5.28)
Eq. 5.28 represents a signal that is both amplitude and frequency modulated.
The envelope of ( ) ( )c c fs t A k m t' 2⎡ ⎤= ω + π⎣ ⎦ (we assume that f p cf k m f∆ = ≤ ;
hence, ( )c fk m t2 0⎡ ⎤ω + π ≥⎣ ⎦ ). As c cA ω represents a DC term, signal ( )m t can
be obtained from ( )s t' , after the DC-block.
The need for a BPL is as follows. Assume that the received FM signal
(with amplitude fluctuations) is applied directly as the input to the differentiator.
Let ( )cA t denote the envelope of the FM signal. Then, there would be an
additional term,
( )cd A t
d t
on the RHS of Eq. 5.28. Even if this term were to be
neglected, the envelope of ( )s t' would be ( ) ( )c c fA t k m t2⎡ ⎤ω + π⎣ ⎦ , which implies
that the envelope of ( )s t' does not contain a term proportional to ( )m t .
Therefore, it is essential to maintain the FM envelope at a constant level.
(Several factors such as channel noise, fading etc. cause variations in cA ).
Band-pass limiter eliminates the amplitude fluctuations, giving rise to an FM
signal with constant envelope.
We shall now indicate some schemes to implement this method of demodulation.

5.51
Scheme 1: Circuit implementation of the scheme of Fig. 5.22, can be carried out
by constructing an op-amp differentiator circuit followed by an envelope detector.
Scheme 2 (Slope detection): Another way of implementing the FM-to-AM
conversion scheme is through a simple tuned circuit followed by an envelope
detector. The transfer characteristic of a tuned circuit in a small region off
resonance is approximately linear.
Fig. 5.23: Magnitude characteristics of a tuned circuit
In Fig. 5.23, we have shown the resonance characteristic of a tuned
circuit. The parts of the characteristic shown in red have been drawn as straight
lines. (This is a good approximation to the actual characteristic.) Assuming that
( )m t 0> produces an increase in the instantaneous frequency, we can use the
straight line segment between A and B (for f 0> ; for f 0< , we have the
corresponding segment, A' to B' ) for demodulation purposes. Let us assume
that the FM signal to be demodulated has the carrier frequency c
f f
f 1 2
2
+
= and
T
c
B
f f1
2
≤ − where as T
c
B
f f2
2
≥ + . This is shown in Fig. 5.24.

5.52
Fig. 5.24: Segment A to B of Fig. 5.23
As can be seen from the figure, changes in the instantaneous frequency
will give rise to corresponding changes in the output amplitude. Envelope
detection of this output will produce the required message signal.
Fig. 5.25: Tuned circuit demodulator
Consider the demodulator circuit shown in Fig. 5.25. The primary of the
coupled circuit is tuned to cf 1
whereas the secondary is tuned to f0 , where
cf f0 > . If, over the frequency range T
c
B
f f
2
± ≤ , the output of the primary is
fairly constant, then, we can expect the ( )m t to resemble ( )m t fairly closely.
This method of demodulating an FM signal is also known as slope detection.
Though the method is fairly simple, it suffers from the following
disadvantages: the resonant circuits on the primary and secondary side are
1
In a superheterodyne receiver, the detector stage follows the IF stage. As such, cf gets
converted to IFf .

5.53
tuned to two different frequencies; the frequency range of linear amplitude
response of a tuned circuit is somewhat limited, making the demodulation of a
WBFM unsatisfactory.
Scheme 3 (Balanced slope detection): This latter problem is partially overcome
by using a balanced configuration (balanced slope detection). The scheme,
shown in Fig. 5.26(a) has three tuned circuits: two on the secondary side of the
input transformer and one on the primary. The resonant circuit on the primary is
tuned to cf whereas the two resonant circuits on the secondary side are tuned to
two different frequencies, one above cf and the other, below cf . The outputs of
the tuned circuits on the secondary are envelope detected separately; the
difference of the two envelope detected outputs would be proportional to ( )m t .
Though the balanced configuration has linearity over a wider range (as
can be seen from Fig. 5.26(b), the width of linear frequency response is about
B3 , where B2 is the width of the 3-dB bandwidth of the individual tuned circuits)
and does not require any DC bock (The two resonant frequencies of the
secondary are appropriately selected so that output of the discriminator is zero
for cf f= ), it suffers from the disadvantage that the three tuned circuits are to be
maintained at three different frequencies.

5.54
Fig. 5.26: Balanced slope detection (a) circuit schematic (b) response curve
Example 5.9
Let the FM waveform of Fig. 5.4(b) be the input to a differentiator followed
by an envelope detector. Let us find an expression for the output of the
differentiator and sketch the output of the envelope detector. We shall assume
that the differentiator will produce the appropriate step change in the output for
sudden changes in the input frequency.

5.55
Let the FM waveform between the time instants ( )t t1 2, be taken as
( )f t1cos 2π and that between time instants ( )t t2 3, be taken as ( )f t2cos 2π ,
where f f2 1> .
Then the output of the differentiator is,
t t t1 2< < : ( ) ( )f f t A f t1 1 1 12 sin 2 sin 2π π = π
t t t2 3< < : ( ) ( )f f t A f t2 2 2 22 sin 2 sin 2π π = π
with A A2 1> and A A1 2, 0> .
(Imagine Figure 5.4(a), with two different frequencies.) The output of the ED is
proportional to A1 during t t t1 2< < and proportional to A2 during t t t2 3< < .
Taking the constant of proportionality as unity, we have the output of ED as
shown in Fig 5.27
Fig. 5.27: Output of the ED
After DC block, we obtain the modulating square wave signal. Hence,
d
d t
followed by ED with a DC block, would act as a demodulator for the FM.
Example 5.10
a) Consider the RC network shown in Fig. 5.28. For the values of R and C
given, we will show that for frequencies around 1.0 MHz, this can act as a
differentiator.

5.56
Fig. 5.28: The RC network of Example 5.10
b) Let us find the condition on R and C such that this network can act as a
differentiator for frequencies around some frequency cf .
c) For the above network, let ( ) ( ) ( )in c f Iv t s t A t c m t6
cos 2 10⎡ ⎤= = π × +
⎣ ⎦
where ( )fk m t 6
10<< Hz. If ( )outv t is envelope detected, we will show
that, we can recover ( )m t from the ED output.
a) ( )
π
=
+ π
j f RC
H f
j f RC
2
1 2
RC 12 9
100 50 10 5 10− −
= × × = × and for frequencies around 1.00 MHz,
f RC 6 9 3
2 2 10 5 10 10 10
100
− − π
π = π × × × = π × =
As 1
100
π
<< , we can take ( )H f as
( )H f j f RC2π , which is a differentiator.
( )j f9
5 10 2−
= × π
b) for frequencies around some cf , we require, π <<f RC2 1.
c) With ( )s t as the input, ( )outv t is
( ) ( )out
d
v t s t
d t
9
5 10−
⎡ ⎤= × ⎣ ⎦

5.57
( ) ( ){ }c f I fA t c m t k m t9 6 6
5 10 sin 2 10 2 10 2− ⎡ ⎤= × π × + π × + π
⎣ ⎦
Output of the ED ( )c fA k m t9 6
5 10 2 10 2− ⎡ ⎤= × π × + π
⎣ ⎦
( )f
c
k
A m t6 9
6
2 10 5 10 1
10
− ⎡ ⎤
= π × × × +⎢ ⎥
⎣ ⎦
.
( )c fA k
m t6
1
100 10
π ⎡ ⎤
= +⎢ ⎥
⎣ ⎦
5.6.2 Phase shift discriminator
This method of FM demodulation involves converting frequency variations
into phase variations and detecting the phase changes. In other words, this
method makes use of linear phase networks instead of the linear amplitude
characteristic of the circuits used in the previous method. Under this category, we
have the Foster-Seely discriminator (and its variant the ratio detector) and the
quadrature detector. Foster-Seely discriminator and the ratio detector have
been discussed in appendix A5.2. We shall now explain the operation of the
quadrature detector.
Quadrature Detector (QD): Consider the FM signal
( ) ( )c cs t A t tcos⎡ ⎤= ω + φ⎣ ⎦ where
( ) ( )
t
ft k m d2
− ∞
φ = π λ λ∫
Then ( )
( )
( )f
d t
t k m t
d t
' 2
φ
φ = = π
( ) ( )t t t
t
1
⎡ ⎤φ − φ − ∆⎣ ⎦∆
, provided t∆ is small.
( )t tφ − ∆ can be obtained from ( )tφ with a delay line or a network with
linear phase.

5.58
Consider the scheme shown in Fig. 5.29. ( )H f has the transfer
characteristic given by
( ) ( )j A f
H f e=
where ( )A f , the phase function, can be well approximated by a linear phase
function, namely,
( )
( )
( )
T
c c
T
c c
B
f f t f f f
A f
B
f f t f f f
2 , 0 ,
2 2
2 , 0 ,
2 2
π⎧
− − π − ∆ > − <⎪⎪
⎨
π⎪ − π + ∆ < + <
⎪⎩
(5.29)
Fig. 5.29: Block diagram of a quadrature detector
With ( )H f specified as above, let us calculate the output of the network with the
input ( )
( ) ( )c cj t t j t t
c c c
e e
A t t Acos
2
⎡ ⎤ ⎡ ⎤ω + ϕ − ω + ϕ⎣ ⎦ ⎣ ⎦+
⎡ ⎤ω + φ =⎣ ⎦ . We can take
( )cj t t
cA e
1
2
⎡ ⎤ω + ϕ⎣ ⎦ to represent the positive part of the spectrum. As ( )H f has the
linear phase term j f t
e 2− π ∆
, (which will contribute a delay of t∆ ), the term at the
output of the filter is
( ) ( ) ( )c c cj t t t t t j t t t
c cA e A e2 21 1
2 2
π π⎡ ⎤ ⎡ ⎤
ω − ∆ + ϕ − ∆ − + ω ∆ ω + ϕ − ∆ −⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦= .

5.59
Similarly, the output corresponding to the negative part of the input spectrum
would be
( )cj t t t
cA e 21
2
π⎡ ⎤
− ω + ϕ − ∆ −⎢ ⎥
⎣ ⎦ . Combining these two terms, we have at the
output of the filter, the quantity
( ) ( )c c c cA t t t A t t tcos sin
2
π⎡ ⎤
⎡ ⎤ω +ϕ − ∆ − = ω + ϕ − ∆⎣ ⎦⎢ ⎥
⎣ ⎦
.
The
2
π
phase shift provided by ( )H f at cf f= ± gives rise to the term
quadrature detector. Multiplication of this output by ( )c cA t tcos⎡ ⎤ω + ϕ⎣ ⎦ followed
by low pass filtering yields the output ( )y t proportional to ( ) ( )t t tsin⎡ ⎤φ − φ − ∆⎣ ⎦ .
Assuming t∆ to be very small, ( )y t can be approximated as,
( ) ( ) ( ){ }y t k t t t1 φ − φ − ∆
( )k t t1
'= ∆ ϕ
( )k m t2=
where k1 and k2 are constants with = ∆fk c k t2 1 .
Several tuned circuits, when properly designed, can provide the band-
pass response with the phase characteristic given by Eq. 5.29. Consider the
series RLC circuit shown in Fig. 5.30.
Fig. 5.30: A network to provide the phase of Eq. 5.29

5.60
( )
( )
( )i
V f j f C
H f
V f R j f L
j f C
0
1
2
1
2
2
π
= =
+ π +
π
( )j f RC f LC
2
1
2 2 1
=
π − π +
(5.30)
Let f
LC
0
1
2
=
π
and b
R
f
L2
=
π
. Then, Eq. 5.30 reduces to
( )
( ) ( )b
f
H f
f f j f f
2
0
2 2
0
=
− +
Consider f 0>
( ) bf f
H f
f f
1
2 2
0
arg tan−
⎛ ⎞
⎡ ⎤ = − ⎜ ⎟⎣ ⎦ ⎜ ⎟−⎝ ⎠
b
f f
f f
2 2
1 0tan
2
−
⎛ ⎞−π
= − − ⎜ ⎟⎜ ⎟
⎝ ⎠
( ) ( )
b
f f f f
f f
1 0 0
tan
2
− ⎡ ⎤+ −π
= − − ⎢ ⎥
⎣ ⎦
Let the circuit be operated in a small frequency interval, around f0 so that f f0 .
Then, f f f0 02+ ,
( )
b b
f f
f f f
0 2+
and
( )
( )
b
f
H f
f
1 2
arg tan
2
− δπ
⎡ ⎤ − −⎣ ⎦
( )f Q
f
1
0
2
tan
2
− δπ
− −
where
b
f
Q
f
0= and ( )f f f0δ = −
If
( )f Q
f0
2
1
δ
<< , then,
( )
( )Q f
H f
f0
2
arg
2
δπ
⎡ ⎤ − −⎣ ⎦ (5.31)

5.61
As ( )H f is the frequency response of a network with real impulse response, we
will have
( ) ( )⎡ ⎤ ⎡ ⎤− = −⎣ ⎦ ⎣ ⎦H f H farg arg
By choosing cf f0 = , we have ( )cf f fδ = − and
c
Q
t
f
1
∆ = ⋅
π
There are other circuit configurations (other than the one given in Fig.
5.30) that can provide the required phase shift for the quadrature detector. The
circuit given in Fig. 5.31 is another possibility.
Fig. 5.31: Another phase shift circuit for the QD
Here C provides a very high reactance at the carrier frequency and the parallel
tuned circuit resonates at cf f= .
Quadrature detector is well suited to IC construction. Companies such as
Signetics, have built high quality FM receivers using QD as the FM demodulator.
Some details on these FM receivers can be found in Roddy and Coolen [2].
5.6.3 Zero-crossing detection
Consider the scheme shown in Fig. 5.32(a). When the input to the hard
limiter is a sine wave of period T0 , it produces at its output a square wave of the
same period, with the transitions in the square wave occurring at the zero

5.62
crossings of the input sine wave. When the input to the hard limiter is an FM
signal, the hard limiter output appears as a square wave of varying frequency.
The hard limiter output ( )Hv t , triggers a monostable pulse generator, which
produces a short pulse of amplitude A and duration τ at each upward (or
downward) transition of ( )Hv t .
Consider a time interval T such that
c
T
W f
1 1
>> >> , where W is the
highest frequency present in the input signal. We shall assume that during the T
sec. interval, the message signal ( )m t is essentially constant which implies that
instantaneous frequency ( )if t is also a near constant (Fig. 5.32(b)). Then the
monostable output, ( )pv t , looks like a pulse train of nearly constant period. The
number of such pulses in this interval ( )T in T f t with an average value,
( ) ( )
t
I p
t T
v t v d
T
1
−
= λ λ∫
( )T in A A f t
T
1
= τ τ
After the DC block, we will have ( )y t being proportional to ( )m t , which is the
desired result.

5.63
Fig 5.32: Zero-crossing detector
5.6.4 FM demodulation using PLL
PLL is a versatile building block of the present day communication
systems. Besides FM demodulation, it has a large number of other applications
such as carrier tracking (in schemes with a pilot carrier and even suppressed
carrier; recall the squaring loop, Section 4.2.3), timing recovery, frequency
synthesis etc.
The basic aim of a PLL is to lock (or synchronize) the instantaneous angle
of a VCO output to the instantaneous angle of a signal that is given as input to
the PLL. In the case of demodulation of FM, the input signal to PLL is the
received FM signal.
In its simplest form, PLL consists of a phase detector and a VCO
connected as shown in Fig. 5.33(a).

5.64
Fig. 5.33: Phase lock loop (a) Basic configuration
(b) PD of (a) in functional form
(c) PLL as a negative feedback loop
PD makes the comparison of the instantaneous phase of ( )x t and ( )v t ,
and is designed such that ( )v tθ , the instantaneous phase of ( )v t locks on to
( )x tθ , the instantaneous phase of ( )x t , if necessary with some fixed phase
difference. (This will become evident later.)
A number of circuits are available which have been used as phase
detectors. In the context of FM demodulation, the most common PD is that of an
analog multiplier followed by a LPF (Fig. 5.33(b)). The scheme resembles closely
that of a negative feedback amplifier configuration, shown in Fig. 5.33(c). In this
figure, s is the variable of the Laplace transform, ( )G s is system function in the

5.65
forward path whereas ( )H s is the network in the feedback path. A properly
designed negative feedback system ensures that the error quantity ( )e t2 is fairly
close to zero so that ( ) ( )v t x t1 1 . This is ensured by providing sufficiently high
loop gain. Similarly, by making the amplifier gain ag sufficiently large, it is
possible to make ( )v tθ follow the changes in ( )x tθ .
Let ( ) ( )cx t t tcos⎡ ⎤= ω + ϕ⎣ ⎦ and let the VCO output be,
( ) ( )cv t t tcos⎡ ⎤= ω + ψ⎣ ⎦ . Then, from Fig. 5.33(b),
( ) ( ) ( ) ( ) ( )c cw t x t v t t t t tcos cos⎡ ⎤ ⎡ ⎤= = ω + ϕ ω + ψ⎣ ⎦ ⎣ ⎦
( ) ( ) ( ) ( ){ }c t t t t t
1
cos 2 cos
2
⎡ ⎤ ⎡ ⎤= ω + ϕ + ψ + ϕ − ψ⎣ ⎦ ⎣ ⎦
Only the term ( ) ( )t t
1
cos
2
⎡ ⎤ϕ − ψ⎣ ⎦ will appear at the output of the LPF; that is
( ) ( ) ( )e t t t1
1
cos
2
⎡ ⎤= ϕ − ψ⎣ ⎦ (5.32)
(We are assuming that the LPF has unit gain)
As the phase detector, we want ( )e t to be zero where ( ) ( )t tϕ = ψ ; but from Eq.
5.32, ( )e t1 is maximum when ( ) ( )t tϕ = ψ . This is not the characteristic of a
proper phase detector. This anomaly can be corrected, if the loop provides a
2
π
phase shift so that the output of the VCO is ( )c t tsin⎡ ⎤ω + ψ⎣ ⎦ . That is, the loop
locks in phase quadrature. Here after, we shall assume this
2
π
phase shift in the
VCO output.
Now let us look at the demodulation of FM. Let
( ) ( ) ( )c cx t s t A t tcos⎡ ⎤= = ω + ϕ⎣ ⎦

5.66
where ( ) ( )
t
ft k m d2
− ∞
ϕ = π τ τ∫ . The VCO is designed such that when ( )y t , the
control voltage is zero, its frequency is cf . (As mentioned earlier, in a superhet,
the demodulator follows the IF stage and cf is actually IFf , which is a known
quantity). This is called the free running frequency of the VCO. Hence, the
VCO output can be written as
( ) ( )v cv t A t tsin⎡ ⎤= ω + ψ⎣ ⎦ (5.33a)
where ( ) ( )
t
vt k y d2ψ = π τ τ∫ (5.33b)
and vk is the voltage sensitivity of the VCO, in units of Hz/volt.
Or ( ) ( )vt k y t' 2ψ = π (5.33c)
( )K y t1= , where vK k1 2= π
( )e t1 of Fig. 5.33(b) is,
( ) ( ) ( ) ( )c vA A
e t t t h t1 sin
2
⎡ ⎤= ϕ − ψ ∗⎣ ⎦ (5.34a)
where ( )h t is the impulse response of the LPF,
and ( ) ( )ay t g e t1= (5.34b)
Let ( ) ( ) ( )e t t tθ = ϕ − ψ (5.35)
Then ( ) ( ) ( )c v a
e
A A g
y t t h tsin
2
⎡ ⎤= θ ∗⎣ ⎦
( ) ( )eK t h t2 sin⎡ ⎤= θ ∗⎣ ⎦ , where c v aA A g
K2
2
= (5.36)
Using Eq. 5.35, 5.36 and 5.33(c), we can draw following block diagram (Fig.
5.34) for the PLL.

5.67
Fig. 5.34: Equivalent circuit of a PLL
Fig. 5.34 brings out much more clearly the negative feedback nature of the PLL,
where in the quantities involved are instantaneous phase deviations of the input
and the VCO output.
Let the loop be in lock so that ( )θ <<e t 1 for all t . Then
( ) ( )e et tsin θ θ , ( ) ( )t tψ ϕ and ( ) ( )t t' 'ψ ϕ . As ( ) ( )
t
ft k m d2
− ∞
ϕ = π τ τ∫ ,
we have
( ) ( )ft k m t' 2ϕ = π
Hence ( )
( )
( )ft k
y t m t
K K1 1
'
2ψ π
= . That is, ( ) ( )y t m tα .
The above has been a very elementary analysis of the operation of PLL.
Literature on PLL is very widespread. For a more detailed and rigorous analysis,
refer Taub and Shilling [3].
Summary of the Detectors:
Slope detection using a single tuned circuit has been presented to show
how simple an FM demodulator could be; it is not used in practice because of its
limited range of linearity. Though balanced slope detection offers linearity over a
wider frequency range, it suffers from the problem of tuning. The Foster-Seely
discriminator and the ratio detector have been the work horses of the FM industry

5.68
until recently; they are now becoming less and less important as better circuit
configurations have been developed especially from the point of view of IC
design. The quadrature detector offers very high linearity and is commonly used
in high quality receivers. Except the phase-shifting network, the rest of the
detector, including the IF amplifier is available in a single chip (e.g. RCA CA
3089E). Commercial zero crossing detectors have better than 0.1 % linearity and
can operate from 1Hz to 10 MHz. A divide by ten counter inserted after the hard
limiter extends the range up to 100 MHz. This type of detector is best suited
when exceptional linearity over a very large frequency deviation is required. It is
less useful when the frequency deviation is a small fraction of the carrier
frequency. PLL performs better than other demodulators when the signal-to-
noise ratio at the input to the detector is somewhat low.
5.7 BandPass Limiter (BPL)
We know that information in an FM wave resides in the instantaneous
frequency (or in the zero-crossings) of the signal. As such, the amplitude
changes of the carrier are irrelevant. In fact, as was pointed out in section 5.6.1,
if the envelope ( )A t is not a constant, it will give rise to distortion in the
demodulated output. In other words, from the point of view of proper
demodulation, we want ( )A t A= , a constant.
Though the FM signal that is generated at the transmitter has a constant
envelope, the received signal may not possess this property. This is due to
various impairments during propagation on the channel, namely, channel noise,
distortion introduced by the channel, etc. BPL helps us to recover the constant
envelope FM signal from the one that has envelope fluctuations.
A band-pass limiter consists of a hard limiter followed by a band-pass
filter. The input-output relationship of a hard-limiter is given by

5.69
( )
( )
( )
A x t
y t
A x t
, 0
, 0
⎧ >⎪
= ⎨
− <⎪⎩
(5.37)
where ( )x t is the input, ( )y t is the output and A is a constant. Let ( )x t be as
shown in Fig. 5.35(a).
Fig. 5.35: (a) Input to the hard limiter
(b) Output of the hard limiter
Then the output with A taken as 1, would be as shown in Fig. 5.35(b). If the input
to the hard limiter is the FM signal (with or without envelope fluctuations), then
the output would be a sequence of alternate positive and negative rectangular
pulses with durations that are not uniform. Such a wave is difficult to analyze.
However, cosθ as a function of θ is always periodic with period 2π . Hence the
hard limiter output, when considered as a function of θ , will be a periodic square
wave (with period 2π) when the input to the limiter is a cosine signal. Hence, if
( )x t tcos cos= ω = θ , then

5.70
( )y
1, cos 0
1, cos 0
θ >⎧
θ = ⎨
− θ <⎩
Fig. 5.36: Hard limiter output with cosθ as input
( )y θ being a periodic signal (Fig. 5.36), we can expand it in terms of Fourier
series, namely
( )y
4 1 1
cos cos3 cos5
3 5
⎡ ⎤
θ = θ − θ + θ + ⋅ ⋅ ⋅⎢ ⎥π ⎣ ⎦
(5.38)
(Note that ( )y θ is real, and has half-wave symmetry. Hence Fourier series
consists of only cosine terms with the even harmonics missing. See Sec. 1.2.2.)
Let θ be the instantaneous angle of the FM signal; that
( ) ( )θ = θ = ω + τ τ = ω + ϕ∫
t
i c f ct t c m d t t( )
Then, from Eq. 5.38, we have
( ) [ ] ( )⎡ ⎤θ = ω + ϕ − ω +ϕ + ⋅ ⋅ ⋅⎣ ⎦π π
c cy t t t t
4 4
cos ( ) cos 3 ( )
3
At the output of BPL, we have the constant envelope FM waves with carrier
frequencies cnf and frequency n f∆ respectively, where n 1, 3, 5= etc. With an
appropriate BPF, it is possible for us to obtain constant envelope FM signal with
carrier frequency cf and deviation f∆ . We will assume that BPF will pass the
required FM signal, suppressing the components with spectra centered at the

5.71
harmonics of cf . Incidentally, BPL can be used as a frequency multiplier,
multiplication factors being 3, 5, 7, etc.
A hard limiter, as defined by Eq. 5.37 can be easily be realized in practice.
One such circuit is shown in Fig. 5.37.
Fig. 5.37: Circuit realization of a hard limiter
As shown in the figure, the circuit consists of a high gain amplifier, a current
limiting resistor R and two zener diodes arranged in a back-to-back
configuration. Zeners can be chosen to have the appropriate break down
voltages.
5.8 Broadcast FM
5.8.1 Monophonic FM Reception
FM stations operate in the frequency range 88.1 to 107.9 MHz with
stations being separated by 200 kHz; that is, the transmission bandwidth
allocation for each station is about 200 kHz. The receiver for the broadcast FM is
of the superheterodyne variety with the intermediate frequency of 10.7 MHz. As
the audio bandwidth is 15 kHz, these stations can broadcast high quality music.
Let us now look at the receiver for the single channel (or monophonic)
broadcast FM. Like the superhet for AM, the FM receiver (Fig. 5.38) also has the

5.72
front end tuning, RF stage, mixer stage and IF stage, operating at the allocated
frequencies. This will be followed by the limiter-discriminator combination, which
is different from the AM radio.
Fig. 5.38: FM broadcast superhet receiver
The need for a band-pass limiter has already been explained. As far as the
frequency discriminator is concerned, we have a lot of choices: Foster-Seely,
ratio detector, quadrature detector, PLL etc. This is unlike AM, where the
envelope detector is the invariable choice. Also, the output of the discriminator is
used in a feedback mode to control the frequency stability of the local oscillator.
This is very important as a change in the frequency of the LO can result in
improper demodulation. (Recall that, in the case AM, the envelope detector
output is used for AVC.) However, most of the present day receivers might be
using frequency synthesizers in place of a LO. These being fairly stable, AFC is
not a requirement.
Basic operation of the AFC block is as follows. Assume that the receiver is
properly tuned so that LO c IFf f f− = . Then the discriminator input will have equal
frequency variations with respect to IFf ; hence the discriminator output will vary
symmetrically with respect to zero output. As such, the net DC voltage is zero
and LO frequency will not be changed. Assume, however, that LO frequency is
not correct. Let LO c IFf f f '− = , where IF IFf f' < . Then the input to the S-curve of

5.73
the discriminator will be operating mostly in the negative output region. This
means that the discriminator output will have a net negative voltage. If this
voltage is applied to the varicap in the LO circuit, LOf will increase which implies
IFf ' will be increased. In other words, IFf ' tends towards IFf . Similarly, if
LO c IFf f f '− = where IF IFf f' > , then the discriminator output is positive most of
the time which implies a net positive DC value. This voltage will increase the
capacitance of the varicap which implies LOf decreases and this makes IFf ' tend
to IFf . Discriminator output goes through a base-band amplifier1
(with a
bandwidth of 15 kHz), whose output drives the speaker.
5.8.2 Two-channel (stereo) FM
Two-channel (stereo) FM is fairly common these days and stereo
transmission has been made compatible with mono-aural reception. As the name
indicates, in the two-channel case, audio signal is derived as the output of two
separate microphones. These are generally called as left microphone and the
right microphone. Let the corresponding output signals be denoted by ( )Lm t
and ( )Rm t . The two- channel transmitter is shown in Fig. 5.39.
Fig. 5.39: FM stereo transmission scheme
1
Output of the discriminator goes through a de-emphasis network before being applied to a base-
band amplifier. Pre-emphasis at the transmitter and de-emphasis at the receiver are used to
improve the signal-to-noise ratio performance. Pre-emphasis and de-emphasis will be discussed
in Chapter 7.

5.74
As shown in the figure, ( ) ( ) ( )L Rv t m t m t1 = + and ( ) ( ) ( )L Rv t m t m t2 = − . The
signals ( )v t1 and ( )v t2 go through a pre-emphasis stage. The signal ( )y t2 is
applied to a balanced modulator, the other input to modulator being the carrier
with a frequency of 38 kHz. Hence ( )y t3 is a DSB-SC signal. The carrier 38 kHz
is derived from the primary source at 19 kHz and a frequency doubler. The final
base-band signal ( )Bm t consists of the sum of ( )y t1 , ( )y t3 and 19 kHz primary
carrier. Typical spectrum (for f 0> ) of the base-band signal as shown in Fig.
5.40.
( ) ( ) ( )L RV f m f m f1 = +
( ) ( ) ( )Y f K Y f f Y f f3 2 0 2 0⎡ ⎤= − + +⎣ ⎦
f0 38= kHz and K is a constant
Fig. 5.40: Spectrum of the final base-band signal
The signal ( )Bm t is used to frequency modulate the carrier allotted to the station.
The resulting signal ( )s t is transmitted on to the channel.
The block schematic of the stereo receiver upto the discriminator is the
same as shown in Fig. 5.38 (that is, monophonic case). Hence, let us look at the
operations performed by the stereo receiver after recovering ( )Bm t from ( )s t .

5.75
These operations are indicated in Fig. 5.41. The DSB-SC signal is coherently
demodulated by generating the 38 kHz carrier from the pilot carrier of 19 kHz.
(Note that if a pilot carrier of 38 kHz had been sent, it would have been difficult to
extract it at the receiver.) ( )r t1 and ( )r t2 , after de-emphasis will yield ( )v t1 and
( )v t2 . (Note that constants of proportionality are ignored and are taken as 1.)
Fig. 5.41: Scheme to recover ( )Lm t and ( )Rm t
From ( )v t1 and ( )v t2 , the individual channels signals, namely ( )Lm t and ( )Rm t
are obtained. These signals, after suitable power amplification, will drive the two
speakers, arranged such that one is on the left and the other is on the right. If the
receiver is not stereophonic, it would respond only to ( )v t1 thereby making it
stereo transmission and monophonic reception compatible.

5.76
Appendix A5.1
Table of Bessel Functions
( )nJ β
β
n
0.5 1 2 3 4 5 6 8 10
0 0.9385 0.7652 0.2239 - 0.2601 - 0.3971 - 0.1776 0.1506 0.1717 - 0.2459
1 0.2423 0.4401 0.5767 0.3391 - 0.0660 - 0.3276 - 0.2767 0.2346 0.0435
2 0.0306 0.1149 0.3528 0.4861 0.3641 0.0465 -0.2429 - 0.1130 0.2546
3 0.0026 0.0196 0.1289 0.3091 0.4302 0.3648 0.1148 - 0.2911 0.0584
4 0.0002 0.0025 0.0340 0.1320 0.2811 0.3912 0.3576 - 0.1054 - 0.2196
5 - 0.0002 0.0070 0.0430 0.1321 0.2611 0.3621 0.1858 - 0.2341
6 - 0.0012 0.0114 0.0491 0.1310 0.2458 0.3376 - 0.0145
7 0.0002 0.0025 0.0152 0.0533 0.1296 0.3206 0.2167
8 - 0.0005 0.0040 0.0184 0.0565 0.2235 0.3179
9 0.0001 0.0009 0.0055 0.0212 0.1263 0.2919
10 - 0.0002 0.0014 0.0070 0.0608 0.2075
11 - - 0.0020 0.0256 0.1231
12 0.0005 0.0096 0.0634
13 0.0001 0.0033 0.0290
14 - 0.0010 0.0120

5.77
Appendix A5.2
Phase Shift Discriminator
(i) Foster-Seely Discriminator
(ii) The Ratio Detector
A5.2.1 Foster-Seely discriminator
Fig. A5.2.1 illustrates the circuit diagram of this discriminator where all the
resonant circuits involved are tuned to the same frequency. Note the similarity
between this circuit and the circuit of Fig. 5.26. Major differences are a by-pass
capacitor C between the primary and secondary, an additional inductance L and
only a single tuned circuit on the secondary ( )L C2 2|| .
Fig. A5.2.1: Circuit schematic of Foster-Seely discriminator
In the frequency band of operation, C , C3 and C4 are essentially short circuits,
which implies that the entire input voltage inV would appear across L . The
mutually coupled double tuned circuit has high primary and secondary Q and
low mutual inductance. When evaluating the primary current, we may neglect the
primary resistance and any impedance coupled from the secondary. Hence, the
primary current

5.78
in
p
V
I
j L1
=
ω
(A5.2.1)
(Note that all the voltages and currents are phasor quantities)
The voltage induced in series with the secondary as a result of the primary
current is given by
s pV j M I= ± ω (A5.2.2)
where the sign depends on the direction of the winding. Taking the negative sign
and using Eq. A5.2.1 in Eq. A5.2.2, we have
s in
M
V V
L1
⎛ ⎞
= − ⎜ ⎟
⎝ ⎠
Assuming the diode circuit will draw very little current, we calculate the
current in the secondary because of sV as,
( )
s
s
L C
V
I
R j X X2 22
=
+ −
where R2 is the resistance associated with the inductance L2 , LX L2 2= ω and
CX
C2
2
1
=
ω
. Hence, the voltage across the terminals 2, 3 is given by
( )s CV I j X 223 = −
( )
( )
s C
L C
V j X
R j X X
2
2 22
−
=
+ −
in CV Xj M
L R j X
2
1 2 2
=
+
where ( )L CX X X2 22 = −
The voltage applied to diode D1, V62 , is
LV V V62 23
1
2
= +
inV V23
1
2
= + (A5.2.3)

5.79
Similarly, the voltage applied to diode D2 , V63 , is
= −inV V V63 23
1
2
(A5.2.4)
The final output voltage V54 is, V V V54 64 65= − which is proportional to
{ }V V62 63− . We will consider three different cases: cf f= 1
, cf f> and cf f< .
i) When the input frequency cf f= we have
in C c
in
V X M Xj M
V j V
L R L R
2 2
23
1 2 1 2
⎛ ⎞
= = ⎜ ⎟⎜ ⎟
⎝ ⎠
(A5.2.5a)
That is, the secondary voltage V23 leads the primary voltage by 90 .
Thus V23
1
2
will lead inV by 90 and V23
1
2
− will lag inV by 90 . Let us
construct a phasor diagram by taking inV as reference. This is shown in Fig.
A5.2.2(a). As the magnitude of the voltage vectors applied to the diodes D1
and D2 , V62 and V63 respectively are equal, the voltages V64 and V65 are
equal and hence the final output V54 is zero.
1
Note that in a superheterodyne receiver, the demodulator follows the IF stage. Hence, cf is
actually IFf , and the discriminator circuit is always tuned to IFf , irrespective of the incoming
carrier frequency.

5.80
Fig. A5.2.2: Phasor diagram illustrative of the operation of Foster-Seely
discriminator
ii) when the input frequency exceeds cf , L CX X X2 22 = − is positive. Let
j
R j X Z e2 2 2
θ
+ = . Then,
( )in C in C jV X V X Mj M
V e
L R j X L Z
2 2 90
23
1 2 2 1 2
− θ
= =
+
(A5.2.5b)
That is, V23 leads in inV by less than 90 and V23− lags in inV by
more than 90 . This is shown in Fig. A5.2.2(b). As the magnitude of the
vector V62 is greater than that of V63 , V V64 65> which implies the final
output V V V54 64 65= − is positive.

5.81
Fig. A5.2.3: Response curve of the Foster-Seely discriminator
iii) Similarly, based on the phasor diagram of Fig. A5.2.2(c), we can easily
argue that the final output would be negative when cf f< . The actual value
of the final output depends on how far away the input frequency is from cf .
Fig. A5.2.3 gives the plot of the frequency response of the Foster-Seely
discriminator, which is usually termed as the S-curve of the discriminator.
Useful range of the discriminator (frequency range of linear response,
shown in red in Fig. A5.2.3) normally lies between the 3 dB points of the
tuned circuit which forms part of the discriminator circuit.
Foster-Seely discriminator responds also to input amplitude variations. Let
the input to the discriminator be ( )i cf t f= . Then, the voltages across R3 and R4
are equal and let this value be 3 V. Now, let ( )if t be such that voltage across R3
increases while that across R4 decreases. Let the voltage increase on R3 be 2
Volts. We can take the voltage decrease on R4 also as 2 V. In other words, for
the given frequency deviation, say f1∆ , we have the voltage at point 4 equal to 5
volts where as the voltage at point 5 equal to 1 V. This implies

5.82
( )outV 5 1 4= − = . Let V64 denote voltage across R3 and V65 , the voltage
across R4 . Then
V
V
64
65
5= .
Let the input signal strength be increased such that, when ( )i cf t f= ,
V V64 65 6= = . Now let ( )if t change such that we have the deviation f1∆ as in
the previous case. Then V64 will become 10 Volts whereas V65 becomes 2 V,
with their difference being equal to 8 V. Though the ratio
V
V
64
65
remains at 5, outV
changes from the previous value. That is, the circuit responds not only to
frequency changes but also to changes in the incoming carrier strength. Hence,
Foster-Seely discriminator has to be the preceded by a BPL.
A5.2.2 Ratio Detector
By making a few changes in the Foster-Seely discriminator, it is possible
to have a demodulator circuit which has built in capability to handle the amplitude
changes of the input FM signal, thereby obviating the need for an amplitude
limiter. The resulting circuit is called the ratio detector which has been shown in
Fig. A5.2.4.
Comparing the ratio detector circuit with that of the Foster-Seely
discriminator, we find the following differences: direction of D2 is reversed, a
parallel RC combination consisting of ( )R R5 6+ and C5 has been added and
the output outV is taken across a different pair of points. We shall now briefly
explain the operation of the circuit.

5.83
Fig. A5.2.4: Circuit schematic of a ratio detector
Reexamining Fig.A5.2.1 and the corresponding phasor diagrams, we find that by
and large, the sum V V62 63+ remains constant. Hence, any variation in the
magnitude of this sum voltage can be considered to be spurious. Suppression of
these spurious variations would result in a detector that is unaffected by input
voltage fluctuations which implies that circuit does not require a separate limiter
stage. How the sum voltage is kept constant would be explained a little later.
With the diode D2 being reversed, we find that the voltages V64 and V65
are series aiding rather than series opposing and as such, the voltage V54
represents the sum voltage. Taking R R5 6= , we find
outV V V V V64 47 64 74= + = −
V V64 54
1
2
= −
V V
V 56 64
64
2
+
= −
[ ]V V64 56
1
2
= −
k V V62 63⎡ ⎤= −⎣ ⎦

5.84
Usually, C C3 4= and R R3 4= . Hence at resonance, V V64 56= which implies
that outV is zero. Above resonance, as V V64 56> , the output is positive whereas
below resonance V V56 64> , and the output is negative.
In the circuit Fig. A5.2.4, C5 is a capacitor of a rather large value. For
example, C5 is of the order of 5 Fµ whereas C3 and C4 are of the order 300
pF. If inV is constant, C5 charges to the full potential existing between the points
5 and 4, which, as indicated earlier is essentially a constant. If inV tries to
increase, C5 will tend to oppose any rise in outV . This is because as the input
voltage tries to rise, extra diode current flows trying to charge C5 . But V54
remains constant at first because C5 is a fairly large capacitance and it is not
possible for the voltage across it to change instantaneously. The situation now is
that the current in the diodes' load has risen but the voltage across the load has
not changed. This being so, the secondary of the ratio detector transformer is
more heavily damped, Q falls and so does the gain of the amplifier driving the
detector. This nearly counteracts the rise in the input voltage. Similarly, when inV
increases, the damping is reduced. The gain of the driving amplifier increases
thereby counteracting the fall in the input voltage. Thus the ratio detector
provides variable damping.
For a large number of years, the Foster-Seely discriminator and the ratio
detector have been the work horses of the FM industry. As these circuit
configurations are not very convenient from the point of view of IC fabrication, of
late, their utility has come down. Companies such as Motorola have built high
quality FM receivers using the Foster-Seely discriminator and the ratio detector.
Some details can be found in Roddy and Coolen [2].

5.85
Appendix A5.3
Multi-tone FM
Let ( )m t A t A t1 1 2 2cos cos= ω + ω where f1 and f2 are arbitrary. Then,
( ) ( ) ( ) cj t j t j t
pe cs t A e e e1 1 2 2sin sinβ ω β ω ω⎡ ⎤ ⎡ ⎤= ⎢ ⎥ ⎢ ⎥⎣ ⎦ ⎣ ⎦
where fA k
f
1
1
1
β = and fA k
f
2
2
2
β =
( ) ( ) ( ) cj tj m t j n t
pe c m n
m n
s t A J e J e e1 2
1 2
ωω ω
⎧ ⎫⎡ ⎤ ⎡ ⎤⎪ ⎪
= β β⎨ ⎬⎢ ⎥ ⎢ ⎥
⎪ ⎪⎣ ⎦ ⎣ ⎦⎩ ⎭
∑ ∑
Hence, ( ) ( ) ( ) ( )c m n c
m n
s t A J J m n t1 2 1 2cos⎡ ⎤= β β ω + ω + ω⎣ ⎦∑∑ (A5.3.1)
The (discrete) spectrum of ( )s t can be divided into 4 categories:
1) Carrier component: amplitude ( ) ( )J J0 1 0 2= β β when cf f=
2) A set of side frequency components due to f1: These components have
amplitudes ( ) ( )mJ J1 0 2β β at frequencies ( )cf mf m1 , 1, 2, 3,± = ⋅ ⋅ ⋅
3) A set of side frequency components due to f2 : These components have
amplitudes ( ) ( )nJ J0 1 2β β at frequencies ( )cf nf n2 , 1, 2, 3,± = ⋅ ⋅ ⋅
4) A set of cross modulation (or beat frequency) terms with amplitudes
( ) ( )m nJ J1 2β β at frequencies ( )cf mf nf1 2± ∓ where m 1, 2, 3,= ⋅ ⋅ ⋅;
n 1, 2, 3,= ⋅ ⋅ ⋅
Cross spectral terms of the type given at (4) above clearly indicate the
non-linear nature of FM. These terms are not present in linear modulation. Even
with respect to the terms of the type (2) and (3), linear modulation generates only
those components with m 1= and n 1= .

5.86
Appendix A5.4
RMS Bandwidth of WBFM
We have already used a number of measures for the bandwidths of
signals and systems, such as 3-dB bandwidth, null-to-null bandwidth, noise
equivalent bandwidth, etc. In the context of WBFM, another meaningful and
useful bandwidth quantity is the rms bandwidth. The basic idea behind the rms
bandwidth is as follows.
Let ( )XS f denote the power spectral density of a random process ( )X t .
Then, the normalized PSD, ( )
( )
( )
X
X N
X
S f
S f
S f d f
, ∞
− ∞
=
∫
, has the properties of a
PDF. Let fσ denote the standard deviation of ( )X NS f, . Then, f2 σ can be used
as a measure of the spectral width of the process.
The WBFM process, with cf f>> ∆ , can be treated as a band-pass
random process. We shall define the rms bandwidth of a band-pass process as
( )rmsB f f
1
2 2
02⎛ ⎞= −⎜ ⎟
⎝ ⎠
(A5.4.1a)
or ( )rmsB f f
22
04= − (A5.4.1b)
where ( )
( ) ( )
( )
X
X
f f S f d f
f f
S f d f
2
0
2 0
0
0
∞
∞
−
− =
∫
∫
( ) ( )X
T
f f S f d f
P
2
0
0
2
∞
−
=
∫
(A5.4.2a)
where TP is the total power of the process and

5.87
( )X
T
f f S f d f
P0
0
2
∞
= ∫ (A5.4.2b)
Let ( )M t be a strict sense stationary message process with ( )m t as a
sample function. Let ( )Mp m denote the PDF of the process. (Note that we are
using the symbol ( )p to denote the PDF, instead of the earlier symbol ( )f ; this
has been done to avoid confusion. In this derivation, f denotes only the
frequency variable.) For the FM case, we have
i c ff f k m= +
where m is a specific value of ( )m t for some t . Then
( )i c
f
f f
m
k
−
= (A5.4.3)
Using quasi-static approximation, it has been shown by Peebles [4] that,
( ) c cT
X M MWBFM
f f f
f f f fP
S f p p
k k k2
⎡ ⎤⎛ ⎞ ⎛ ⎞− +
⎡ ⎤ = +⎢ ⎥⎜ ⎟ ⎜ ⎟⎣ ⎦
⎝ ⎠ ⎝ ⎠⎣ ⎦
(A5.4.4)
(Note that in quasi-static approximation, it is assumed that ( )if t remains constant
for a sufficiently long period; as such, FM wave appears to be a regular sinusoid
and if can be replaced by f .) For the WBFM process, c
T T
A
P S
2
2
= = .
In Eq. A5.4.4, c
M
f
f f
p
k
⎛ ⎞−
⎜ ⎟
⎝ ⎠
is the positive part of the spectrum. We will now
show that ( ) ( )rms f MWBFM
B k R
1
22 0⎡ ⎤= ⎣ ⎦ where ( ) ( )MR M t2
0 = , the mean
square value of the process. (Note that ( )MR τ is the ACF of the process.)

5.88
For f 0> , ( )XS f is symmetrical with respect to cf f= (we assume that
the input PDF is symmetric about zero) and hence cf f0 = . Using Eq. A5.4.4 in
A5.4.2(a), we have
( ) ( ) cT
rms c M
WBFM T f f
f fP
B f f p d f
P k k
22
0
8
2
∞
⎛ ⎞−
= − ⋅ ⎜ ⎟
⎝ ⎠
∫ .
Let c
f
f f
k
−
= λ . Then,
( ) ( ) ( )
c
f
rms f M
WBFM
f
k
B k p d
22
4
∞
−
= λ λ λ∫
( ) ( )f M fk p d k M t2 2 2 2
4 4
∞
− ∞
λ λ λ =∫
( )f Mk R2
4 0=
That is,
( ) ( )rms f MWBFM
B k R2 0= (A5.4.5)
Example A5.4.1
Let ( )m t be a sample function of a strict sense stationary process ( )M t .
A WBFM signal is generated using ( )m t as the message signal. It is given that
( )M
m
p m
otherwise
1
, 1
2
0 ,
⎧
<⎪
= ⎨
⎪⎩
a) Find ( )X WBFM
S f⎡ ⎤⎣ ⎦ .
b) What is the value of rmsB ?
As ( )Mp m is uniform, we expect the PSD of the resulting WBFM also to
be uniform over the appropriate frequency range. When m 0= , we have cf f= ,

5.89
whereas m 1= ± results in the instantaneous frequency to be c ff k±
respectively. In other words, the frequency range over which ( )MS f exists is
c f c ff k f f k− < < + . Let c c
M M
f f
f f f f
p p
k k
⎛ ⎞ ⎛ ⎞− +
+⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ ⎠
be as shown in Fig. A5.4.1.
Fig. A5.4.1: Shape of the PSD of Example A5.4.1
To find the value of α , we require that the RHS of Eq. A5.4.4, when integrated
over the entire range should be equal to cA2
2
. That is,
c f
c f
f k
c c
f f k
A A
d f
k
2 2
2
4 2
+
−
⎡ ⎤
⎢ ⎥α =
⎢ ⎥
⎣ ⎦
∫
or ( )c c
f
f
A A
k
k
2 2
2
2 2
⋅ α =
That is,
1
2
α =
Hence, ( )
c
c f c f
X fWBFM
A
f k f f k
S f k
otherwise
2
,
8
0 ,
⎧
− < < +⎪
⎡ ⎤ = ⎨⎣ ⎦
⎪
⎩
In a few situations ( )rmsB might be quite meaningful. For example, if ( )m t
is the sample function of a Guassian process, there is a small but finite

5.90
probability that ( )m t will assume very large values. This implies that f∆
becomes excessive and as such TB given by Carson’s rule or its variants would
be extremely high. As ( )rmsB weights the large frequency derivations with small
probabilities, ( )rmsB will not be unduly excessive. For example if ( )m t belongs
to a Guassian process and say ( )Mp m is ( )N 0, 9 , then
( )rms f fWBFM
B k k2 3 6= × × =
(Note that ( )MR 0 9 3= = )
Exercise A5.4.1
We define the RMS bandwidth of any lowpass process ( )M t as,
( )
( )
( )
M
rms M
M
f S f d f
B
R
2
2
0
∞
− ∞⎡ ⎤ =
⎣ ⎦
∫
(A5.4.6)
Let ( )rms PM
B denote the RMS bandwidth of the PM signal. Show that
( ) ( ) ( )⎡ ⎤=
⎣ ⎦rms p M rmsPM M
B k R B2 0 (A5.4.7)

5.91
Appendix A5.5
Modulation Techniques in TV
Color television is an engineering marvel. At the flick of a button on a
remote, we have access to so much information and entertainment in such a fine
detail (audio as well as video with all its hues and colors) that the angels could
envy the humans on this count. Distances are no longer a barrier; we have real
time reception almost at any point on this globe. Maybe, not too far into the
future, we may be able to watch any program of our choice in the language we
prefer to listen either in real time or near real time. We shall now take a closer
look at the modulation techniques used in (commercial) TV, based on the NTSC
standard used in North America and Japan. Our purpose here is to illustrate how
various modulation techniques have been used in a practical scheme. Hence
PAL and SECAM systems have not been discussed.
As the TV transmission and reception started with monochrome (black
and white) signals we shall begin our discussion with this scheme. (Note that
color transmission can be viewed on a monochrome receiver. Similarly, black-
and-white transmission can be viewed on a color receiver.)
The bandwidth of a monochrome video signal is about 4.2 MHz. Let
( )vm t denote this signal. The bandwidth allocated to each TV station by the
regulatory body is about 6 MHz. Hence the use of DSB is ruled out. It is very
difficult to generate the SSB of the video signal, ( )vm t . (Filtering method is ruled
out because of appreciable low frequency content in ( )vM f and because of the
fairly wide bandwidth, designing the HT with required specifications in extremely
difficult.) As such, VSB becomes the automatic choice. With a suitable carrier
component, we have seen that VSB can be envelope detected. Actually at the
transmitter, it is only partially VSB; because of the high power levels at the
transmitter, it is difficult to design a filter with an exact vestigial sideband. It is at

5.92
the IF stage of the receiver that perfect VSB shaping is achieved; subsequently it
is demodulated.
Fig. A5.5.1: Modulation stages in a monochrome TV transmitter
Fig. A5.5.1 shows the block diagram of the modulation scheme of a
monochrome TV transmitter. Details of the magnitude spectrum of the
transmitted signal is given in Fig. A5.5.2.
Fig. A5.5.2: Audio and video spectrum of a monochrome TV signal

5.93
In this figure, cvf is the video carrier and caf is the carrier for audio, which is
higher than cvf by 4.5 MHz. (The video carrier cvf is 1.25 MHz above the lower
frequency limit allotted to the station)
The USB of the video occupies the full bandwidth of 4.2 MHz. That part of
LSB spectrum between ( )cvf 0.75− to cvf is transmitted without any attenuation.
Below the frequency ( )cvf 0.75− MHz, LSB is gradually attenuated reaching
almost zero level at ( )cvf 1.25 MHz− . As this is not exactly a VSB characteristic,
it is termed as partial VSB.
The audio signal, band-limited to 10 kHz frequency modulates the audio
carrier. The maximum frequency deviation is 25 kHz. Hence, the audio
bandwidth can be taken as ( ) ( )f W2 2 25 10 70∆ + = + = kHz.
TV receiver is of the superheterodyne variety. A part of the reciever structure is
shown in Fig. A5.5.3.
Fig. A5.5.3: (Partial) Block diagram of monochrome TV receiver
The mixer output, ( )v t1 , is applied to an IF amplifier and VSB shaping
network. The IF amplifier has a pass-band of 41 to 47 MHz. The characteristic of
the VSB shaping filter is shown in Fig. A5.5.4.

5.94
Fig. A5.5.4: VSB shaping in TV receiver
The frequency modulated audio signal is also passed by the IF stage but with
much less gain than the video part. That is, ( )v t2 consists of the video signal of
the (VSB+C) type and the FM audio signal, with a carrier at 4.5 MHz and with
ca cvA A<< where caA is the audio carrier amplitude and cvA is the amplitude of
the video carrier. Under these conditions, it can be shown that ( )v t3 , the output
of the envelope detector, has the video as well as required audio. The video
amplifier removes the audio from ( )v t3 . The output of the video amplifier is
processed further and is displayed on the picture tube. ( )v t3 is also applied as
input to an IF stage, with the IF of 4.5 MHz. The audio part of the ( )v t3 is passed
by this IF stage; the FM demodulator that follows produces the audio signal.
Color TV: The three primary colors, whose linear combination can give rise to
other colors are: Red, Blue and Green. These color signals, pertaining to the
scene that is being shot, are available at the outputs of three color cameras. Let
us denote these signals as ( )Rm t , ( )Gm t and ( )Bm t respectively. These basic
color components are linearly combined to produce (i) the video signal of the
monochrome variety (this is called the luminance signal and is denoted by
( )Lm t )) and (ii) two other independent color signals (called the in-phase
component of the color signal and the quadrature component of the color signal).

5.95
Let us denote these components as ( )Im t and ( )Qm t respectively. The
equations of the linear transformation are given below in the form of a matrix
equation.
( )
( )
( )
( )
( )
( )
L R
I G
BQ
m t m t
m t m t
m tm t
0.3 0.59 0.11
0.6 0.28 0.32
0.21 0.52 0.31
⎡ ⎤ ⎡ ⎤⎡ ⎤
⎢ ⎥ ⎢ ⎥⎢ ⎥= − −⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥−⎣ ⎦ ⎣ ⎦⎣ ⎦
M
(A5.5.1)
From Eq. A5.5.1, we obtain
( ) ( ) ( ) ( )L R G Bm t m t m t m t0.3 0.59 0.11= + + (A5.5.2)
It has been found that ( )Lm t as given by Eq. A5.5.2 closely resembles ( )vm t of
the monochrome system. ( )Lm t has the same bandwidth as ( )vm t namely, 4.2
MHz. This is required to preserve the sharp transitions in the intensity of light at
the edges in a scene. However, the eye is not as sensitive to color transitions in
a scene and is possible to reduce the bandwidth occupancy of ( )Im t and ( )Qm t .
In the NTSC system, bandwidth allocation for ( )Im t is 1.5 MHz and that of
( )Qm t is 0.5 MHz. These chrominance signals are quadrature multiplexed
(QAM) on the color subcarrier as shown in Fig. A5.5.5.
Fig. A5.5.5: Generation of composite baseband video

5.96
The composite video signal ( )bvm t is modulated on to the carrier cvf .
This implies that color subcarrier, ccf , is 3.58 MHz above cvf and caf is 4.5 MHz
above cvf . As such, the color information gets interleaved in between the spectral
lines of the luminance signal and no additional bandwidth is required for the color
TV system. In order to facilitate coherent demodulation of the QAM signal, a few
cycles of the color subcarrier (called color burst) is sent along with the
transmitted signal. This reference carrier is tracked by a PLL in the receiver. The
VCO output of the PLL is used in the demodulation of the chrominance signal.
For more details on TV transmission and reception, refer to Carlson, Crilly and
Rutledge [5] or Leon Couch [6].
TV modulator circuit is available in an IC chip. For example, Motorola MC
1374 includes an FM audio modulator, sound carrier oscillator, RF dual input
modulator. It is designed to generate a TV signal from audio and video inputs. It
is also suited for applications such as video tape recorders, video disc players,
TV games etc. The FM system can also be used in the base station of a cordless
telephone. Circuit details and other parameters can be obtained from the manual.
Exercise A5.5.1
Indirect method can used to generate the FM signal for the audio in TV.
Required carrier is 4.5 MHz and f 25∆ = kHz. Using cf 1 200= kHz and
f1 20∆ < Hz, design the modulator such that frequency at any point in the
modulator does not exceed 100 MHz. Use the shortest possible chain of
frequency doublers and triplers.

5.97
References
1) Bettina Koster, Peter Waldow and Ingo Wolff, A unique low voltage,
source-coupled J-FET VCO, RF signal processing, April 2001, PP58-66
(website: http://www.rfdesign.com)
2) Dennis Roddy and John Coolen, Electronic communications, (4th
ed.), PHI,
1995
3) Herbert Taub and Donald L. Schilling, Principles of communication
systems, (2nd
ed.), McGraw Hill International ed., 1986
4) Peebles P. Z., Communication system principles, Addison-Wesley, 1976
5) A. Bruce Carlson, Paul B. Crilly and Janet C. Rutledge, Communication
systems (4th
ed.), McGraw Hill International ed., 2002
6) Leon W. Couch II, Digital and Analog Communication systems (6th
ed.),
Pearson Education-Asia, 2001

Angle modulation

More Related Content

What's hot (20)

Similar to Angle modulation (20)

More from Punk Pankaj (20)

Recently uploaded (20)

Angle modulation