Introduction to

Telecommunication

Frequency Modulation

Asst. Prof. Dr. Ertuğrul SAATÇI

Wideband FM
Spectrum of Sinusoidal Modulation
 To find the spectrum of the single-tone FM wave defined by the
exact formula for an arbitrary value of the modulation index β is
not easy ((t) is inside the cosine).

s (t )  Ac cos  2 f c t    t    Ac cos  2 f c t   sin  2 f m t  

 In order to simplify the spectral analysis of the wideband FM,
we use complex envelope representation of s(t).

 Recall that any bandpass signal can be represented by using one of

the following three forms:
s (t )  sI (t ) cos  2 f c t   sQ (t ) sin  2 f c t  Inphase and Quadrature
s (t )  R (t ) cos  2 f c t   (t )  Polar

s (t )  Re s (t )e j 2 fc t  Complex Envelope

1
Wideband FM
Spectrum of Sinusoidal Modulation – cont.
 For FM signals the complex envelope:
 t 
 
j  2 k f m ( ) d 

s (t )  Ac e j (t )  Ac e  0 

 For single-tone FM wave the complex envelope becomes:

j  t  j  sin  2 f m t 
s (t )  Ac e  Ac e
 Now, recall that the spectrum of a bandpass signal is related to the
spectrum of the complex envelope signal as
1  1
S( f )  S  f  f c   S   ( f  f c ) 
2 2
 Thus, we can examine the spectral properties of the complex
envelope signal in order to determine the spectral properties of
the bandpass signal.

Wideband FM
Spectrum of Sinusoidal Modulation – cont.
 The signal of interest s (t ) is periodic with T=1/fm
j  sin  2 f m t 
s (t )  Ac e
  k 
j  sin  2 f m  t   
j  sin  2 f m t  k 2  j  sin  2 f m t 
s (t  kT )  Ac e   fm  
 Ac e  Ac e
and thus does not have a Fourier Transform in the strict sense.
• However, we can determine the Fourier Transform using the
Fourier Series:
 
s (t )  
n 
cn e j 2 nf0 t  S ( f )   c   f  nf 
n 
n 0

• In this case, the fundamental frequency is the message

frequency: fo=fm

2
Wideband FM
Spectrum of Sinusoidal Modulation – cont.
 The Fourier Series coefficients are determined as
 T0 /2
1 1
s (t )  ce
n 
n
j 2 nf 0 t
, cn 
T0  s (t )e  j 2 nf0 t dt , f0 
T0
 fm
T0 / 2
1/  2 f m 
j  sin  2 f m t   j 2 nf m t
cn  f m 
1/  2 f m 
Ac e e dt

1/  2 f m 
j   sin  2 f m t   2 nf m t 
 Ac f m 
1/  2 f m 
e dt

• If we make a change of variables x=2fmt.


1 1 j   sin  x   nx 
dt 
2 f m
dx  cn  Ac
2  e

dx

Wideband FM
Spectrum of Sinusoidal Modulation – cont.
 Continuing with the coefficient derivation:
 
1 j   sin  x   nx  1 j   sin  x   nx 
cn  Ac
2  e

dx  J n (  ) 
2 
 e dx

• However, this is simply a known integral, known as a Bessel

Function Jn(β) scaled by Ac:

cn  Ac J n   
• Thus, the complex envelope of the single-tone FM wave
 
s (t )  
n 
cn e j 2 nfm t  Ac  J  e
n 
n
j 2 nf m t

and the spectrum of the complex envelope signal can be given as:
 
S  f    c   f  nf   A  J      f  nf 
n 
n m c
n 
n m

Jn(β): n-th order Bessel function of the first kind with argument β.

3
Wideband FM
Spectrum of Sinusoidal Modulation – cont.
 Single-tone FM signal s(t) for an arbitrary value of modulation index β
and its spectrum S(f) can be written as:
 
j 2 f  nf t 
 
s (t )  Re s (t )e j 2 fc t  Re  Ac  J n    e  c m  
 n  

s (t )  Ac  J    cos  2  f
n 
n c  nf m  t 

Taking the Fourier transforms of both sides


S  f   Ac  J    F cos  2  f
n 
n c 
 nf m  t 


 J      f   f  nf m      f   f c  nf m   
Ac
S f  n c
2 n 

Wideband FM
Spectrum of Sinusoidal Modulation – cont.
 Alternatively, using the spectrum of the complex envelope signal, the
spectrum S(f) of the single-tone FM wave s(t) can be written as:

S  f   Ac  J      f  nf 
n 
n m

1  1
S( f )  S  f  f c   S   ( f  f c ) 
2 2

S  f   Ac  J n       f  f c   nf m 
1
2 n 

Ac  J n        f  f c   nf m 
1

2 n 
A 
S  f   c  J n      f   f c  nf m      f   f c  nf m   
2 n 

4
Wideband FM
Spectrum of Sinusoidal Modulation – cont.
 Single-tone FM spectrum contains a carrier component at fc and
infinite numbers of discrete side frequencies located at fc±nfm.
 Theoretically the bandwidth of FM is infinite.


 J      f   f  nf m      f   f c  nf m   
Ac
S f  n c
2 n 

2S  f 

Figure shows the

spectra for positive
frequencies only

Wideband FM
Bessel Function Jn(β)
 The Bessel function is
tabulated in Appendix 3
of your text book.

 This function is also

available in MATLAB
using besselj(n,beta).


1 j   sin  x   nx 
J n ( ) 
2  e

dx

Jn(β): n-th order Bessel function of the first kind with argument β.

5
Wideband FM
Properties of Bessel Function Jn(β)
 For different integer (positive and negative) values of n,
J n     J n    , for n even
J n      J n    , for n odd
 For small values of the modulation index β
J0     1

J1    
2
J n     0, n  1
 The equality holds exactly for any value of β


 J   1
n 
2
n

Wideband FM
Properties of Single-tone FM for Arbitrary Modulation Index β

 Notice the difference between AM and FM:

 Single-tone AM spectrum contains a carrier component at fc and a
single pair of side frequencies located at fc±fm.

 Single-tone FM spectrum contains a carrier component at fc and

infinite numbers of discrete side frequencies located at fc±nfm.
• But only a finite number are significant.
• The number of significant side frequencies will depend on the
modulation index β.

 If β is small compared with unity (corresponds to narrowband FM),

only J0(β) and J1(β) will have significant values and
• the FM wave will be similar to an AM in the frequency domain.
• The FM wave is effectively composed of a carrier at fc and a
single pair of side frequencies at fc±fm.

Ac
S f 
2
 J      f  f
n 
n c  nf m     f  f c  nf m  

6
Wideband FM
Properties of Single-tone FM for Arbitrary Modulation Index β

 The amplitude of the carrier component of an FM wave is dependent

on the modulation index β.

Ac
S f 
2
 J      f  f
n 
n c  nf m     f  f c  nf m  

Ac
n0  S f  J 0      f  f c     f  f c  
2
• The envelope of an FM wave is constant, thus the average power
delivered to a 1 ohm resistor is also constant.
• The average power of an FM wave can be determined as

s  t   Ac  J    cos  2  f
n 
n c  nf m  t 

1 2  2 1
Pav  Ac  J n     Ac2
2 n

 2
1

Narrowband vs. Wideband FM

Single-tone FM
 Recall that if β is small (β << 1)
 the signal can be approximated as a linear modulation
scheme similar to AM.

 If β is bigger than 1
 we call the resulting modulation wideband FM and the
modulation is nonlinear as expected.

 The spectral properties of wideband FM are substantially

different from AM.

 Let’s examine the spectrum for various values of β

 when the message is sinusoidal.

7
 Example 9.1
FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 Consider a LC-AM wave modulated by a sinusoidal message assuming

fc=1kHz and fm=50Hz. Using the same parameters, compare the AM
spectrum with FM when β=0.1, 1.0, 2.0 and 5.0
 Investigate how the variations in the amplitude and frequency of a
sinusoidal modulating wave affect the spectrum of FM wave.

β=0.1, f=5Hz
Figure shows the fm=50Hz
spectra for positive β<<1
frequencies only Narrowband FM

Example 9.1 – cont.

FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 Case 1: The frequency fm of the modulating wave is fixed, but its

amplitude is varied (∆f is varied, since ∆f=kfmax|m(t)|).
• Keep fm constant (i.e., ∆f is increased since β=∆f/fm).

, f=50Hz
fm=50Hz
2f Wideband FM

8
Example 9.1 – cont.
FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 For β=2.0

, f=100Hz
fm=50Hz
2f Wideband FM

Example 9.1 – cont.

FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 For β=5.0

, f=250Hz
fm=50Hz
Wideband FM
2f

9
Example 9.1 – cont.
FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 Case 2: The amplitude of the modulating wave is fixed, but its

frequency fm is varied (∆f is constant, since ∆f=kfmax|m(t)|).
• Compare the AM spectrum with FM when β=0.1, 1.0, 2.0 and 5.0
• Keep ∆f constant (i.e., fm is decreased since β=∆f/fm).

AM Spectrum FM Spectrum
β=0.1, f=5Hz
fm=50Hz
β<<1
Narrowband FM

Example 9.1 – cont.

FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 β=1.0 (f=5Hz, fm=5Hz)

AM Spectrum FM Spectrum
β=1.0
Wideband FM
2f

10
Example 9.1 – cont.
FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 β=2.0 (f=5Hz, fm=2.5Hz)

AM Spectrum FM Spectrum
β=2.0
Wideband FM
2f

Example 9.1 – cont.

FM Spectrum for Varying Amplitude and
Frequency of Sinusoidal Modulating Wave

 β=5.0 (f=5Hz, fm=1Hz)

• We see that when β is increased, we have an increasing number of
spectral lines in the fixed frequency interval fc-∆f<|f|<fc+∆f.
• That is, when β approaches infinity, the bandwidth of the FM wave
approaches the limiting value of 2∆f.

AM Spectrum FM Spectrum
β=5.0
Wideband FM

2f

11
Summary
 Today we have investigated two forms of frequency modulation,
namely narrowband FM and wideband FM.
● Narrowband FM approaches linear modulation and is similar
to AM.
● Wideband FM is decidedly nonlinear and occupies a band
much larger than the bandwidth of the message.

 We restricted ourselves to a sinusoidal message signal

 since the nonlinear nature makes analysis extremely difficult.

Introduction to
Telecommunication

Spectral Characteristics of
Frequency Modulated Signals

Asst. Prof. Dr. Ertuğrul SAATÇI

12
Overview
 Today we continue our discussion of the spectral characteristics
of FM signals.
• For a general measure of bandwidth, we can use Carson’s Rule.
• For a more detailed description of the spectrum, we rely on
approximations.
• Narrowband and wideband FM have very different behavior and
thus we will examine them separately.

 Reading
 4.6

Angle Modulation
 Phase Modulation: sPM (t )  Ac cos  2 f c t  k p m(t ) 

 t

 Frequency Modulation: sFM (t )  Ac cos  2 f c t  2 k f  m( )d 
 0 

where
● m(t) - message signal
● Ac - carrier amplitude
● fc - carrier frequency
● kp - phase sensitivity constant (radians/volt)
● kf - frequency sensitivity constant (hertz/volt)

13
Exact Calculation of FM Bandwidth
 Because FM modulation is nonlinear, calculation of bandwidth is
extremely difficult.

 Book shows an example for m(t)=sin(2fmt).

• This is a standard textbook example.
• Result is somewhat complex and is in terms of the Bessel
function.

 The problem with this type of example:

• Does not tell us what the spectrum is for practical signals.
• Even adding another sinusoid is difficult because of
nonlinear terms.

 We would like a simple approximate formula for bandwidth.

Example
 Recall our previous example for a sinusoidal message with
fc=1kHz and fm=50Hz and let β=1, β=5.
● Case 1: fm is fixed, ∆f is varied.
● Since β=∆f/fm  ∆f=50Hz, 250Hz

14
Example – cont.
 Further for a sinusoidal message with fc=1kHz if we hold ∆f
constant at ∆f=5Hz and let β=1, β=5.
● Case 2: ∆f is fixed, fm is varied.
● Since β=∆f/fm  fm=5Hz, 1Hz.

Transmission Bandwidth of FM Waves

Carson’s Rule
 We can see that while the absolute bandwidth may be infinite,
 the number of significant frequency components (i.e., the
practical bandwidth) is finite and related to ∆f.

 Two limiting cases may be identified

• For large values of the modulation index β, the bandwidth
approaches 2∆f, and is only slightly greater than the total
frequency excursion 2∆f.
• For small values of the modulation index β, the spectrum of the
FM wave is effectively composed of a carrier at fc and one pair of
side frequencies at fc±fm, and the bandwidth approaches 2fm.

 This leads to an approximate rule (known as Carson’s rule) for the

transmission bandwidth of a sinusoidally modulated FM signal:
 1
B  2f  2 f m  2    1 f m  2f  1  
 

15
Transmission Bandwidth of FM Waves
Universal Curve for FM Transmission Bandwidth

 Carson’s rule does not always provide a good estimate of the

bandwidth requirements of wideband FM.

 We may use a definition based on retaining the maximum number of

significant side frequencies whose amplitudes are all greater than
some selected value.
 A convenient choice for this value is one percent of the
unmodulated carrier amplitude.

 Thus, the transmission bandwidth of an FM wave is defined as

● The separation between the two frequencies beyond which none
of the side frequencies is greater than one percent of the
unmodulated carrier amplitude (all |Jn()|<0.01).
● B=2nmaxfm, where fm is the modulation frequency and nmax is the
largest value of the integer n beyond which all |Jn()|<0.01.

Transmission Bandwidth of FM Waves

Universal Curve for FM Transmission Bandwidth

 Transmission bandwidth of single tone FM: B=2nmaxfm

● Separation between two frequencies beyond which all |Jn()|<0.01.
● The Bessel function table can be used to find the value of nmax that
satisfies the threshold requirement.


 J      f   f  nf m      f   f c  nf m   
Ac
S f  n c
2 n 
J n     0.01 B=2nmaxfm J n     0.01

2S  f 

16
Transmission Bandwidth of FM Waves
Universal Curve for FM Transmission Bandwidth

 The transmission bandwidth calculated on the one percent basis,

 can be presented in the form of a universal curve by normalizing it
with respect to frequency deviation f and then plotting it versus .
B 2nmax f m 2nmax
B  2nmax f m   
f f 
From Carson ' s Rule :
B  1
 2 1  
f   

Universal curve for evaluating one percent bandwidth.

This curve shows that the small values of modulation index  are relatively
more extravagant in transmission bandwidth than the larger values of .

Transmission Bandwidth of FM Waves

Universal Curve for FM Transmission Bandwidth

 It can be seen that

B 2nmax B/f is decreasing
B  2nmax f m   and approaches to 2
f  as the increase of .

 Therefore the
bandwidth B
approaches 2f
when  is large.

 Recall that the range

Universal curve for evaluating one percent bandwidth. of the instantaneous
frequency is 2f:
 fc  f , f c  f 

17
Transmission Bandwidth of FM Waves
Carson’s Rule – Arbitrary Modulating Wave
 Carson’s rule as described, applies to sinusoidal signals.

 However, we can define a ratio similar to the modulation index 

for general message signals.
f k f max  m(t ) 
 Deviation Ratio D: D  where
W W
● ∆f is the highest instantaneous frequency deviation and
● W is the bandwidth of the message signal.
 The deviation ratio D plays the same role for nonsinusoidal
modulation that the modulation index  plays for sinusoidal
modulation.

 Hence, replacing  by D and replacing fm with W, the Carson rule

can be generalized.
 The generalized Carson Rule for the transmission bandwidth of
an arbitrary FM signal can be given as

B  2  f  f m   2    1 f m  B  2  f  W   2  D  1 W

Transmission Bandwidth of FM Waves

Carson’s Rule – Arbitrary Modulating Wave
 Approximately 98% of the power in an FM signal lies within the
bandwidth:
B  2  f  W   2  D  1 W
 This formula provides a very good approximation to bandwidth.
 No information concerning the shape of spectrum.

 Interpretation:

B  2f  2W  2  D  1 W
• Bandwidth corresponds to frequency deviation around
carrier frequency plus bandwidth of message signal.
• Deviation ratio tells us the extra bandwidth required by the
FM signal (compared to 2W required by DSB-SC).

18
Transmission Bandwidth of FM Waves
Example 9.2 – Commercial FM Broadcasting
 In North America, the maximum value of frequency deviation f is
fixed at 75kHz for commercial FM broadcasting by radio. If we take
the modulation frequency W=15kHz (typically the maximum audio
frequency of interest in FM transmission), find the corresponding
transmission bandwidth of the FM wave.

 Using the generalized Carson rule for the transmission bandwidth

B  2  f  W   2  75  15   180kHz

 The deviation ratio D

f 75kHz
D  5
W 15kHz

Transmission Bandwidth of FM Waves

Example 9.2 – cont.
 On the other hand, the transmission bandwidth of FM wave can be
obtained using the universal curve or the table for D=5 (=5).
B 2nmax 16
   3.2  B  3.2f  3.2  75  240kHz
f  5
 Generalized Carson rule underestimates the transmission bandwidth
by 25 percent compared with the result obtained from the curve.
240  180 60
  0.25
240 240

3.2

19
Wideband and Narrowband FM
General Message Signals
 Depending on value of D FM performs quite differently.
 Identical to changing β for sinusoidal message signals.

 Wideband FM
• D>1
• We will usually consider this case.

 Narrowband FM
• D < 0.2
• Behaves similarly to AM.

 There is an in between region: 0.2 < D < 1

 For simplicity, we will consider D < 1 to be narrowband.

Spectrum of Narrowband FM
General Message Signals
 t

sFM (t )  Ac cos  2 f c t  2 k f  m( )d   Ac cos  2 f c t   (t ) 
 0 
 For (t)<<1
cos c t   (t )   cos c t    (t ) sin c t 
 Small D (D=kf max{|m(t)|}/W) implies small kf.
 Therefore for D << 1

 t

sNBFM (t )  Ac cos  2 f c t   Ac  2 k f  m( )d  sin  2 f c t 
 0 
 This resembles large carrier AM except that the message is in
quadrature.

20
 Spectrum of Narrowband FM – cont.
General Message Signals
 Spectrum:
 t

sNBFM (t )  Ac cos  2 f c t    Ac 2 k f  m( )d  sin  2 f c t 
 0 
A
S NBFM  f   c   f  f c     f  f c  
2
  
F  Ac cos 2 f c t 

Ac k f  1 1 
  M  f  fc   M  f  fc 
2   f  fc   f  fc 
  
 t  
 
 
F  Ac 2 k f m ( ) d  sin  2 f c t  

 0  
t
1 1 1
 x   d 

j 2 f
X  f  , If X (0)  0, sin  2 f 0 t  
2j
  f  f0     f  f0 
2j

Spectrum of Narrowband FM
Example
 M(f)=F{m(t)}

 BW of NBFM: 2W (close to Carson’s Rule)

Carson ' s Rule : B  2  D  1 W  2W

Since D  1

21
Spectrum of Wideband FM
General Message Signals
 The instantaneous frequency of the signal is directly
proportional to the value of m(t).
 If we know the probability density function (pdf) of m(t) we can
determine the average spectral properties of s(t) (i.e., power
spectral density).

Ac2   1   1 
SWBFM  f    pm  f  f c    pm    f  f c   
4k f   k f  k 
  f 
where pm(x) is the pdf of the message signal.
 Careful!
● PDFs and PSDs are two very different things.
● For this case only, the PSD is proportional to the PDF.

Spectrum of Wideband FM
Example of PSD for WBFM
 Probability Density Function (PDF) for Message Signal

 Power Spectral Density PWBFM(f)

Carson ' s Rule : B  2f  2W  2f

Since W  f  D  1

22
Example A
 Consider the following message signal

Example A – cont.
 Message Spectrum
 W≈6Hz

23
Example A – cont.
Narrowband FM Signal
 For D=0.16 and ∆f=1
 Carson’s Rule: B=2∆f+2W=14Hz
 Spectrum resembles message spectrum as expected.

Example A – cont.
Wideband FM Signal
 For D=1.6 and ∆f=10
 Carson’s Rule: B=2∆f+2W=32Hz
 Spectrum more closely resembles the pdf.

24
Example B
Digital Message
 Message Signal

Example B – cont.
Digital Message
 Message Spectrum
 First Null bandwidth=1Hz

25
Example B – cont.
Narrowband FM
 ∆f=1Hz, W=1Hz
 Carson’s Rule: B=2∆f+2W=4Hz
 Spectrum resembles message spectrum as expected.

Example B – cont.
Wideband FM
 ∆f=10Hz, W=1Hz
 Carson’s Rule: B=2∆f+2W=22Hz
 Spectrum more closely resembles the pdf.

26
Summary
 If all you need is decent measure of bandwidth, use Carson’s
Rule.

 If you need to determine a better approximation of the

spectrum, determine from D whether you have WBFM or NBFM.
● If the signal is narrowband, use spectrum of the message
signal to approximate the spectrum of the FM signal, similar
to AM.
● If the signal is wideband, use the probability density
function to approximate the spectrum (PSD).

27