Download as pdf or txt
Download as pdf or txt
You are on page 1of 40

mm Principles

40 of Communication
60 80 Systems 100 120
(Subject Code: EC-3501)

40
Lecture 2

60
Dr. Pawan Kumar
Assistant Professor
Department of Electronics and Communication Engineering
National Institute of Technology Rourkela
Rourkela, Odisha
80
Autumn (2021-22)
navigation symbols

1 / 40
Part 1: An overview of Fourier analysis

mm 40 60 80 100 120
Review of the previous lecture

In the last session, we have discussed about

• Communication systems and types


• 40
Analog and digital communications
(Digital communications’ advantages over analog communications: i) data can be stored, ii)
cost effective, iii) more immune to noise )

• Signals and their classifications


60
• Continuous and discrete time signals
• Analog and digital signals
• Periodic and aperiodic signals
• 80
Deterministic and random signals
• Energy and power signals
navigation symbols

2 / 40
Some well
mm known signals/functions
40 60 80 100 120
• Exponential function
• Sinusoidal function
• Rectangular function
40
• Unit step function
• Delta function

Delta function (Unit impulse function)


60
• It was introduced by Paul Direc using the definition

Z ∞
0, for t 6= 0
δ(t) = under the condition δ(t)dt = 1.
not defined, for t = 0, −∞

80

navigation symbols

3 / 40
δ (t) δT (t)
mm 40 60 80 100 120
T

0 t -1/2T 1/2T t
40

• Area under the graph in the second figure is

1
60 Area of δT (t) = base(T) × hight(T) = × T = 1.
T

• Hence, we can write


Z ∞   Z ∞
Area of δ(t) = lim (base(T) × hight(T)) dt = (δ(t)) dt = 1
−∞ T→∞ −∞
80

navigation symbols

4 / 40
mm 40 60 80 100 120

Properties of unit impulse function

1) x(t)δ(t) = x(0)δ(t)
40
2) x(t)δ(t − t0 ) = x(t0 )δ(t − t0 )
R∞  R∞ 
3) −∞
x(t)δ(t)dt = x(0) = x(0) −∞
δ(t)dt
R ∞ 60  R∞ 
4) −∞
x(t)δ(t − t0 )dt = x(t0 ) = x(t0 ) −∞
δ(t − t0 )dt

80

navigation symbols

5 / 40
Systems
• Physical devices which process on the input signals and gives output signals.
mm 40 60 80 100 120
Properties
• Causality
• Memory
• Linearity
40
• Time-invariant
Linearity
Consider inputs x1 (t) and x2 (t) to a system and corresponding outputs y1 (t) and y2 (t),
60
respectively, that is,

x1 (t) → y1 (t)
x2 (t) → y2 (t)
80
Then for linear system

a1 x1 (t) + a2 x2 (t) → a1 y1 (t) + a2 y2 (t). navigation symbols

6 / 40
Time-invariant

If we have input output relation as


mm 40 60 80 100 120
x(t) → y(t),

Then for time-invariant system

40 x(t − t0 ) → y(t − t0 ).

Convolution operation

Let h(t) represents unit impulse response of an linear and time-invariant (LTI) system,
then with input x(t) and output y(t), we have input-output relation as
60
Z ∞ Z ∞
y(t) = x(t) ∗ h(t) = x(τ )h(t − τ )dτ = h(τ )x(t − τ )dτ
−∞ −∞

80 y(n) =
x(t) y(t) = x(n)
h(t) h(n) x(n)*h(n)
x(t)*h(t)
navigation symbols

7 / 40
Fourier Series
mm 40 60 80 100 120
• Introduced by Joseph Fourier to represent periodic signals in terms of sine and cosine
functions. (For the purpose of solving heat equation in a metal plate.)

• An example of periodic signal (rectangular pulse)


40 

A, if |t| < T1
x(t) =

0, if T1 < |t| < T.

60
x(t)

80 -T -T1 0 T1 T t
-T/2 T/2

navigation symbols

8 / 40
Definition
mm 40 60 80 100 120
In terms of sine and cosine functions:
A T-periodic signal x(t) can be represnted in terms of sine and cosine functions as
∞     
a0 X 2πnt 2πnt
40 x(t) = + an cos + bn sin ,
2 n=1
T T

where Z
2
a0 = x(t)dt,
T T
60 Z  
2 2πnt
an = x(t) cos dt,
T T T
and Z  
2 2πnt
bn = x(t) sin dt.
80 T T T

navigation symbols

9 / 40
In terms of complex exponentials:
mm 40 60 80 100 120
Alternatively, in a more generalized form, any T-periodic signal x(t) can be represented
in terms of complex exponentials as

X
x(t) = cn exp(2πnft),
40 n=−∞

where f = 1/T is the signal’s fundamental frequency and


Z
1
cn = x(t) exp(−2πnft)dt.
60 T T

Practice 1: One can verify that c0 = a0 , cn = (an − bn ) and c−n = (an + bn ), using the
exp(t)+exp(−t) exp(t)−exp(−t)
relations exp(t) = (cos(t) +  sin(t)), cos(t) = 2
, and sin(t) = 2
.
80

navigation symbols

10 / 40
Importance of Fourier Series
mm
Approximation theory
40 60 80 100 120
Solving differential equations
Fourier analysis in signal processing and communication engineering, and many more...

40
Examples
Example 1: Fourier series representation of rectangular pulses:

A, if |t| < T1
x(t) =
60 0, if − T1 < |t| < T/2

Solution: Using complex exponential form:


X
80 x(t) = cn exp(2πnt/T)
n=−∞

navigation symbols

11 / 40
Z T/2
1
cn = x(t) exp(−2πnt/T)dt
T −T/2
mm 40 Z T1
60 80 100 120
1
= A · exp(−2πnt/T)dt
T −T1
 
1 A
= (exp(−2πnT1 /T) − exp(2πnT1 /T))
T −2πnf
40 
1 −2A sin(2πnT1 //T)

=
T −2πn/T
= A sin(2πnT1 /T)/(πn) = A(2T1 /T) sinc(2nT1 /T),
where sinc(πz) = sin(πz)/(πz). This gives
60 X∞
x(t) = A(2T1 /T) sinc(2nT1 /T) exp(2πnt/T)
n=−∞

Practice 2: Verify that



80 2AT1 X 2AT1
x(t) = + sinc(2nT1 /T) cos(2πnt/T).
T n=1
T
navigation symbols

12 / 40
Approximation of the square wave for A = 1, T1 = 1/4 and T = 1 in terms of Fourier series
harmonics. (Even hormonic coefficients are zero!)
mm 40 60 80 100 120
n=1 harmonic n=2 harmonics n=3 harmonics
1.5 1.5 1.5

1 1 1

f(t)
f(t)

f(t)
0.5 0.5 0.5

40 0 0 0

-0.5 -0.5 -0.5


-1 0 1 -1 0 1 -1 0 1
t t t
n=5 harmonics n=10 harmonics n=50 harmonics
1.5 1.5 1.5
60
1 1 1

f(t)
f(t)

f(t)

0.5 0.5 0.5

0 0 0

-0.5 -0.5 -0.5


80 -1 0 1 -1 0 1 -1 0 1
t t t
navigation symbols

13 / 40
Fourier Transform

It can
mmbe viewed as an40extension of the
60Fourier series80
to aperiodic signals.
100 120
It enables trasnforming a signal from one (time) domain to another (frequency).

Applicable to aperiodic and as well as periodic signals.

40
Definition

For an aperiodic signal x(t), Fourier transform is given by


Z ∞
60 FT[x(t)] = X(f ) = x(t) exp(−2πft)dt
−∞

and the inverse transform is


Z ∞
x(t) = X(f ) exp(2πft)df
80 −∞

navigation symbols

14 / 40
Importance
Tellsmm
about what frequencies
40 are present
60 in a given signal
80 and in what
100proportions. 120
Derivatives of signals can be easily calculated using the relation
FT[dk x/dk t] = (2πf )k X(f )
In system analysis.
40
Examples
Example 1: x(t) = x̃(t) for |t| < T/2.
Solution: Its Fourier transform is obtained as follows.
Z ∞ Z T1
60 X(f ) = x(t) exp(−2πft)dt = A exp(−2πft)dt
−∞ −T1
!
A 
= exp(−2πfT1 ) − exp(2πfT1 )
−2π

80 A sin(2πfT1 )
= = 2AT1 sinc(2fT1 )
πf

navigation symbols

15 / 40
Consider
mm 40 60
x(t) A = 1, -T
801 = ¼, T → ∞ 100 120
A

-T1 0 T1 t
-T/2 T/2
40
0.6

2AT1
0.4
Amplitude, |X(f)|

60 0.2

-1/T1 1/T1
-0.2
80 -10 -5 0 5 10
frequency

navigation symbols

16 / 40
mm
Example 2: x(t) = 1, t >40
0. 60 80 100 120

Solution: The Fourier transform is


Z ∞
X(f ) = x(t) exp(−2πft)dt
40 −∞
Z ∞
= 1 · exp(−2πft)dt
0
!
1  
= exp(−(2πf )∞) − exp(−(2πf ) × 0)
60 −2πf | {z } | {z }
0 1
−1
=
2πf

80

navigation symbols

17 / 40
Another example for Fourier transform x(t)
x(t)
mm A
40 60 80 A 100 120
-T1 0 T1 t
0 2T1 t -T/2 T/2

• Consider signal x(t) = A for 0 ≤ t ≤ 2T1 . Its Fourier transform is obtained as follows:
Z ∞ Z 2T1
X(f40
) = x(t) exp(−2πft)dt = A exp(−2πft)dt
−∞ 0
A   
= exp − 2πf (2T1 ) − 1
−2πf
A   
= exp − 2πf (T1 ) exp − 2πf (T1 ) − exp 2πf (T1 )
60 −2πf
A  
= exp − 2πf (T1 ) − 2 sin(2πfT1 )
−2πf
A 
= sin(2πfT1 ) exp − 2πfT1
πf • Even and odd signals
80 = 2AT1 sinc(2fT1 ) exp −2πfT1
 A signal x(t) is said to be even if x(t) = x(−t),
| {z } and odd if x(t) = −x(−t).
| {z }
There it may happen that a signal is neither even nor
Amplitude |X(f )| phase ∠X(f )
odd.
navigation symbols

18 / 40
Another example for Fourier series expansion
x(t)
mm 40 60A 80 100 120

-T 0 2T1 T t

• Consider a rectangular pulse given by


40

A, for 0 ≤ t < 2T1
xT (t) =
0, for 2T1 ≤ t < T.

60
The Fourier series coefficient cn of the expansions can be obtained as
Z
1 T A  
cn = xT (t) exp(−2πnt/T)dt = exp(−2πn(2T1 )/T) − 1
T 0 T(−2πn/T)
 
80 A sin(2πnT1/T) 2AT1 2nT1 
= exp(−2πfnT1 ) = sinc exp −πnT1 /T
πn T T | {z }
| {z } phase ∠cn
Amplitude |cn | navigation symbols

19 / 40
Spectra/Responses
• Three kinds
mm of spectra of
40a signal 60 80 100 120
 Power spectrum, |cn |2 versus frequency f (Power spectral density (PSD))
 Amplitude spectrum, |cn | versus frequency f
 Angle spectrum, ∠|cn | versus frequency f . In the following figure, ∠cn = θn .
40
Amplitude and phase spectra (f = n/T)
θ-3 θn (=-2πnT1/T)
|cn|
T1 = 1/4, T =1
|c0|
θ-2
|c-1| |c1| θ-1
60
|c-1| |c5| 0 f
|c-4| |c4|
θ1
|c-2| 0 |c2| f
θ2
80 θ3
|c-3| |c3|

Amplitude spectrum Phase spectrum


navigation symbols

20 / 40
Parseval’s Theorem
(For periodic signals)
• Relatesmm 40in time domain
power of a signal 60with that of its80
components in100 120
frequency domain.
∗ For a T-periodic signal x(t), its power is given by
Z T
1
P= |x(t)|2 dt, where |x(t)|2 = x(t)x∗ (t)
T 0
40
∗ Fourier series of x(t)

X ∞
X
x(t) = cn exp(−2πnt/T), x∗ (t) = c∗n exp(2πnt/T)
n=−∞ n=−∞
60
∗ Hence, we can represent
Z Z ∞ ∞
1 T 1 T X X
P = |x(t)|2 dt = ck c∗l exp(−2π(k − l)t/T)dt
T 0 T 0 k=−∞ l=−∞
80 ∞
X X∞ 1Z T  X∞
= ck c∗l exp(−2π(k − l)t/T)dt = |ck |2
k=−∞ l=−∞
T 0 k=−∞
| {z } navigation symbols
=1 if k=l and =0 if k6=l
21 / 40
mm
PSD of the previous Fourier
40 series example
60 80 100 120
2
|cn|

|c-1|2 |c1|2
40
2
|c-3| |c3|2
|c-5|2 |c5|2

60

|c-4|2 |c-2|2 0 |c2|2 |c4|2 f

Power spectral density


80

navigation symbols

22 / 40
Dirichlet Conditions
mm for the evaluation
• Required 40 of Fourier60series and Fourier
80 transform of
100 120
real and periodic
signals.
• The three conditions are
1) The signal of consideration, x(t) is absolutely integrable on [−T/2, T/2]
40 Z
x(t)dt < ∞
T

2) x(t) has finite number of discontinuities in period T.


3) It has finite number of minima and maxima.
60
• If these conditions are satisfied, Fourier series of x(t) converges to its actual form.
• At a discontinuity t0 , the Fourier series converges to

1h i
lim x(t0 − ǫ) − x(t0 + ǫ)
80 ǫ→0 2

• Types of convergence: point-wise, uniform, and L − 2 norm.


navigation symbols

23 / 40
Properties of Fourier transform
1) Conjugate symmetry: For real signal x(t),
mm 40 60 80 100 120
X(f ) = X ∗ (−f ) ⇒ |X(f )| = |X(−f )| and θ(f ) = −θ(−f )

2) Linearity: If x1 (t) ←→ X1 (f ) and x2 (t) ←→ X2 (f ), then for constants a1 and a2

40 a1 x1 (t) + a2 x2 (t) ←→ a1 X1 (f ) + a2 X2 (f )

3) Time scaling: For a signal x(t)


1 f 
x(at) ←→ X
|a| a
60
4) Time shifting: For a signal x(t)

x(t − t0 ) ←→ X(f ) exp(−2πft0 )

80 shifting: For a signal x(t)


5) Frequency

x(t) exp(2πf0 t) ←→ X(f − f0 ) navigation symbols

24 / 40
5) Convolution
mm in time 40
domain is multiplication
60 in frequency
80 domain:100 120
For signals x(t) and y(t)
x(t) ∗ y(t) ←→ X(f )Y(f )

6) Multiplication in time domain is convolution in frequency domain:


40 x(t) and y(t)
For signals
x(t)y(t) ←→ X(f ) ∗ Y(f )

(Frequency modulation property or frequency translation property.)


6) Duality property: If a signal x(t) has Fourier transform X(f ), that is,
60
x(t) ←→ X(f )

then
80 X(−t) ←→ x(f ).

navigation symbols

25 / 40
Fourier transform of some well known signals

impulse function40x(t) = δ(t)


1) Unitmm 60 80 100 120
Z ∞
X(f ) = δ(t) exp(−2πft)dt = 1 ⇒ X(f ) = 1
−∞

x(t) X(f)
40
δ(t) 1

0 t 0 f

60
2) x(t) = 1 for all t
X(f ) = δ(f )

3) Fourier transform of unit step function, x(t) = u(t) = 1 for t ≥ 0


Z ∞ Z ∞ Z ∞
80
X(f ) = exp(−2πft)dt = cos(2πft)dt −  sin(2πft)dt.
0 0 0
navigation symbols

26 / 40
• Consider

mm 40 60
exp(−at), if 80t ≥ 0 100 120
x(t) =
0, if t<0

for a → 0.
• The corresponding Fourier transform is
40 1 a − 2πf a 2πf
X(f ) = = 2 = 2 − 2
a + 2πf a + (2πf )2 a + (2πf )2 a + (2πf )2
| {z } | {z }
I1 I2

where I2 → −1/(2πf ) and


60 a 1 2a 1
I1 = lim = lim 2 = δ(f )
a→0 a2 + (2πf )2 2 a→0 a + (2πf )2 2 |{z}
| {z } (Area = 1)
(Area = 1)

• Hence
1 1
80 X(f ) = δ(f ) +
2 2πf

navigation symbols

27 / 40
4) Fourier transform of x(t) = cos(2πfm t)
Z ∞ Z ∞  
1
X(f ) = mm cos(2πfm t) exp(−2πft)dt
40 = 60 exp(2πf
80 m t)+exp(−2πfm t) exp(−2πft)dt
100 120
−∞ −∞ 2

1 
⇒ X(f ) = δ(f − fm ) + δ(f + fm )
2
5) Fourier transform of x(t) = sin(2πfm t)
Z 40
∞ Z ∞
1 
X(f ) = sin(2πfm t) exp(−2πft)dt = exp(2πfm t)−exp(−2πfm t) exp(−2πft)dt
−∞ −∞ 2

 
⇒ X(f ) = − δ(f − fm ) + δ(f + fm )
2
60
FT[cos(2πfmt)] FT[sin(2πfmt)]

1/2 1/2 j/2

f
δ(f+fm) 0 δ(f-fm) f δ(f+fm) 0 δ(f-fm)
80
-j/2

navigation symbols

28 / 40
mm 40 60 80 100 120
t

5) Fourier transform of rectangular pulse x(t) = A Π T

X(f ) = AT sinc(πfT)

40

x(t)

A
60

-T/2 0 T/2 t

80

navigation symbols

29 / 40
An example for duality property
• Fourier transform of x(t) = AB sinc(πtB)

∗ mm 60 = A Π Tt is 80
40 transform of y(t)
We know the Fourier Y(f ) = AT sinc(πfT)
100 120
∗ Therefore, using duality property, Fourier transform of x(t) is
t
X(f ) = y(−f ) = y(f ) = B Π .
T
• Bandlimited signals
Fourier transform or spectral density
has finite support.
40
• Baseband signals
Most of the energy/power is centered
around zero-frequency. X(f)

60 -B/2 B/2 f
0

x(t)
80
A

navigation symbols

-T/2 0 T/2 t
30 / 40
Orthogonality

• Twomm 40 x(t) and y(t)60


complex functions 80 an interval100
are orthogonal over t1 ≤ t ≤ t2 , if 120
Z t2 Z t2
x(t)y∗ (t)dt = 0 or x∗ (t)y(t)dt = 0
t1 t1

where40(·)∗ represents complex conjugate.

• An extension for two orthogonal signals:


Z t2 Z t2 Z t2
|x(t) + y(t)|2 dt = |x(t)|2 dt + |y(t)|2 dt
t1 t1 t1
60 Z t2 Z t2

+ x(t)y (t)dt + x∗ (t)y(t)dt
t1 t1
Z t2 Z t2
2
= |x(t)| dt + |y(t)|2 dt
t1 t1
80

The sinusoidal components in Fourier series are orthogonal.


navigation symbols

31 / 40
• Recall the complex exponential form Fourier series representation
mm 40 ∞60
X 80 100 120
x(t) = cn exp(2πnt/T),
n=−∞
| {z }

The complex
40 sinusoidal components
. . . exp(−2π(−2)t/T), exp(−2π(−1)t/T), 1, exp(−2π(1)t/T), exp(−2π(2)t/T), · · ·
are orthogonal.

Z T, if m=k
60 exp(−2π(m − k)t/T)dt =
T 0, if m 6= k.

• Similarly, it is easy to follow that the sinusoidal components cos(2πnt/T), cos(2πmt/T),


80
sin(2πpt/T), and sin(2πqt/T) are orthogonal to each other, given the values n, m, p, and q
are not equal.
navigation symbols

32 / 40
Correlation

• A measure
mm of the extent of similarity between two things.
40 60 80 100 120
• To quantify this mathematically, consider an example of vectors:

d
c
40 b
q3
b
q2 q
q1
a 0 r a
0 r3 r2 r1
60
• The angle between the vectors represent the amount of similarity. The two vectors having
smaller angle have larger similarity.
• The extent of similarity is given by

80 < ~a, ~b >


ρ = cos(θ) =
||~a||||~b||
navigation symbols
where operator < ·, · > represents dot (inner) product.
33 / 40
• The extent of correlation lies between -1 to 1, −1 ≤ ρ ≤ 1.
b
mm 40b 60 80 100 120
b
q
q q
a a r 0
a
0 r r =0
(a) (b) (c)
40

Another example for correlation: In terms of signals


• Consider two signals x(t) and y(t). The similarity index is
Z ∞
60 1
ρ= √ √ x(t)y(t)dt, −1 ≤ ρ ≤ 1.
Ex Ey −∞

Correlation functions

• Two kinds
80 of correlation function
∗ cross-correlation
navigation symbols
∗ autocorrelation.
34 / 40
Cross-correlation

• Cross-correlation
mm between
40 two signals x(t)
60 and y(t) is defined
80 by 100 120
Z ∞
Rxy (τ ) = x(t)y∗ (t − τ )dt or
−∞

Z ∞
40 Rxy (τ ) = x(t + τ )y∗ (t)dt
−∞

y(t)
A
60
0 T1 T t
Transmitted pulse

x(t)
80 A
0 T t Transmitter
Received pulse Receiver
navigation symbols

35 / 40
• A demonstration of cross-correlation

y(t)
mm y(t-t1) y(t-t2) y(t-t3)
A A40 60A 80 A 100 120
0 T1 T t 0 T1 T t 0 T1 T t 0 T1 T t

Shifted versions of the transmitted pulse


40
x(t)
A
0 T t
Received pulse
60
Autocorrelation
• Measure of self-similarity for a large lag τ .
• Autocorrelation of signal x(t) is defined by
80 Z ∞
Rxx (τ ) = x(t)x∗ (t − τ )dt
−∞
navigation symbols

36 / 40
Correlations in terms of convolution operator and corresponding Fourier transform

• Recallmm
that convolution40 60 y(t) is given80
of signals x(t) and by 100 120
Z ∞
h(t) = x(t) ∗ y(t) = x(τ )y(t − τ )dτ
−∞

• This can be equivalently rewritten as


40 Z ∞
h(τ ) = x(τ ) ∗ y(τ ) = x(t)y(τ − t)dt
−∞

• If we compare this with cross-correlation function


Z ∞
60 Rxy (τ ) = x(t)y∗ (t − τ )dτ
−∞

• We get
Rxy (τ ) = x(τ ) ∗ y∗ (−τ )
• If x(t) =
80y(t), then autocorrelation is given by
Rxx (τ ) = x(τ ) ∗ x∗ (−τ )
navigation symbols

37 / 40
Fourier transform relations

• Let a signal x(t) and its Fourier transform are paired as


mm 40 60 80 100 120
x(t) ←→ X(f )
Z ∞ Z ∞
X(f ) = x(t) exp(−2πft)dt and x(t) = X(f ) exp(2πft)df
−∞ −∞
40
• One can verify that
x∗ (−t) ←→ X ∗ (f )
• Also, we have
x(t) ∗ y(t) ←→ X(f )Y(f )
60 transform of Rxy (τ ) = x(τ ) ∗ y∗ (−τ ) can be obtained as
• Then Fourier
Z ∞
Sxy (f ) = x(t) ∗ y∗ (t) exp(−2πft)dt = X(f )Y ∗(f )
−∞

• Similarly, Fourier transform of Rxx (τ ) = x(τ ) ∗ x∗ (−τ ) can be obtained as


80
Z ∞
Sxx (f ) = x(τ ) ∗ x∗ (−τ ) exp(−2πf τ )dτ = |X(f )|2
−∞ navigation symbols

38 / 40
Parseval’s Relation
mm
• Autocorrelation 40x(t) is given by
of signal 60 80 100 120
Z ∞
Rxx (τ ) = x(t)x∗ (t − τ )dt
−∞

• For lag τ = 0, we get


40
Z ∞ Z ∞
Rxx (0) = x(t)x∗ (t)dt = |x(t)|2 dt
−∞ −∞

• And inverse Fourier transform of Sxx (f ) for lag τ = 0 is given by


60 Z Z
∞ ∞
Rxx (0) = Sxx (f ) exp(−2πf (0))df = |X(f )|2 df
−∞ −∞

• These two relations suggest that


80 Z ∞ Z ∞
Rxx (0) = |x(t)|2 dt = |X(f )|2 df
−∞ −∞
navigation symbols

39 / 40
mm 40 60 80 100 120

• Total energy of x(t) in time domain and frequency domain is Rxx (0).
• Energy spectral density: Sxx (f ) versus f . This gives spread of total energy in frequency
domain.
40

60 autocorrelation function and energy spectral density of x(t) = exp(−at).


Practice: Find

80

navigation symbols

40 / 40

You might also like