EEE 351

Principles of Communication Systems

Fall 2020

Lecture 04
Analysis and Transmission of Signal – II
- Signal Transmission Through LTI System
- Distortion-less Transmission
- Ideal Filters and Practical Approximations
- Signal Distortion over a Channel
- Energy Spectral Density and Power Spectral Density
Dr. Ghufran Shafiq

Department of Electrical and Computer Engineering

COMSATS University Islamabad
Signal transmission through a LTI System

 Consider a linear time invariant (LTI) system

 Assume the input signal is a Dirac function δ(t). Call
the observed output h(t)
d(t) LTI h(t)
System
 h(t) is called the unit impulse response function
 With h(t), we can relate the input signal to its output
signal through the convolution formula:

2
Signal transmission through a LTI System

 If g (t )  G ( )
h(t )  H ( )
 Then using time convolution property of convolution

y (t ) = g (t )  h(t ) Y ( ) = G ( ) H ( )

 H(ω) is called the system frequency response or the

spectral response

3
Distortion-Less Transmission

 What is the required behavior of an ideal transmission line?

 The output signal from an ideal transmission line may have some time delay
and different amplitude than the input

 It must have no distortion—it must have the same shape as the input

4
Ideal Distortion-less Transmission

 Output signal in time domain

y (t ) = kg (t − t d )
 Output signal in frequency domain

Y ( ) = kG ( )e − jwt d
 Channel’s Frequency Response

H ( ) = ke − jwt d

5
Ideal Distortion-less Transmission

 Channel’s Frequency Response

H ( ) = ke − jwt d
H (w ) = k
 h (w) = − wt d
For distortion-less transmission magnitude of frequency
transfer function must be constant and phase response
must be linear 6
 Intuitive Explanation of Distortion-less
Transmission

 Suppose a signal composed of various sinusoids

being passed through a distortionless system
 Output signal is the input signal multiplied by k
and delayed by td
 System’s frequency response should be such
that each component suffers the same
attenuation k
H (w ) = k
 Each component undergoes same delay of td
seconds
 Time delay is negative of slope of the system phase
response
Constant Time
Delay→ Linear  h (w) = − wt d 7

Phase
Ideal filters

Lowpass
 Distortionless Transmission of certain band of
frequencies
 Suppresses unwanted frequency components

Highpass
 Linear Phase, Constant Magnitude
 Ideal Lowpass Filter

Bandpass
8
Ideal filters

 Not physically realizable because?

 Non-causal impulse response
 Everlasting Response

9
Frequency Impulse
Response Response
Practical (realizable) filters

 Several useful approximations to ideal low-pass filter

 One Approximation: Butterworth Filter
 Popular because they provide almost equal (flat) gain in passband

 Where B is the -3dB cutoff Frequency

10
Practical (realizable) filters

 Frequency (Magnitude) spectrum of Butterworth Filter

11
Comparison: Ideal and Butterworth Filter

12
Signal Distortion over a Communication
Channel

 Linear Distortion
 Distortion by Channel Nonlinearities
 Multipath Distortion
 Fading Channels

13
Linear Distortion

 Consider LTI Channel

 Distortion caused by non-ideal magnitude, phase or both (recall
distortionless transmission)
 Spreading or dispersion of signal will occur

14
Linear Distortion (Example)

15
Distortion caused by Channel Nonlinearities

 Consider a nonlinear channel where output is a nonlinear function of

y = f (g )
input

 Expansion with McLaurin’s series Example

y (t ) = a0 + a1 g (t ) + a2 g (t ) + a3 g (t ) + ......
2 3

 We know that if the Bandwidth of g(t) is B Hz then Bandwidth of gk(t) =

kB Hz
16
Distortion caused by Channel Nonlinearities

 Output spectrum spreads

beyond input spectrum
 Addition of new Frequencies
 High Efficiency Power Amplifiers
for Transmission are non-linear

17
Multipath Effects

 Transmitted signal arrives at receiver from

two or more paths
 Consider two paths: i) unity gain and delay of
td, ii) gain of α and a delay of (td + Δt)

18
 Multipath Effects

 Destructive Interference
when w = nπ/Δt (n odd), cos wΔt = -1
H (w)  0, when,   1

 Constructive Interference
when w = nπ/Δt (n even), cos wΔt = 1

19
 Fading Channel

 Till now we assumed channel characteristics to be constant with

time
 In practice, these characteristics can vary with time
 Changes with season, time of day, hour to hour changes, weather
 One way to reduce fading is to use Automatic gain Control (AGC)

20
 Signal Energy and Energy Spectral Density

 Parseval’s Theorem
 Signal Energy can also be determined from its Fourier Transform

 Changing order of integration

21
Parseval’s Theorem (Example)

22
 Energy Spectral Density (ESD)

 Characterizes distribution of
signal’s energy in Frequency
Domain
 Useful in communication systems
when evaluating the signal and noise

23
 Energy of Modulated Signals

 Recall, AM signal is
 (t ) = g (t ) cos w0t
 Its Fourier Transform is

 (w) = G (w + w0 ) + G (w − w0 )
1
2
 The ESD can be found as

 (w) = G (w + w0 ) + G (w − w0 )
1 2

4 24
 Energy of Modulated Signals

 If w0≥2πB, then G(w+w0) and G(w-w0) are

nonoverlapping and
1
4

 (w) = G (w + w0 ) + G (w − w0 )
2 2

 (w) = g (w + w0 ) + g (w − w0 )
1
4
 Observe that the area under
modulated signal is half the area
under baseband signal
1
E = E g 25

2
 Time Autocorrelation function and Energy
Spectral Density

 Recall, for a real signal the autocorrelation function g(t) is

defined as 
 g ( ) =  g (t ) g (t +  )dt
−

 Note that
 g ( ) =  g (− )

 Autocorrelation is an Even Function

26
 Time Autocorrelation function and Energy
Spectral Density

 Fourier Transform of Autocorrelation Function

 Therefore ESD is the Fourier Transform of Autocorrelation Function

 g ( ) = G ( )
2

 g ( )  Ψ ( )
27
 ESD of Input and Output

 If g(t) and y(t) are the input and the corresponding output of LTI
system, then

 Therefore ESD of output is |H(ω)2| times ESD of input

28
 Signal Power and Power Spectral Density (PSD)

 Recall, the power Pg of a real signal g(t) is given by

T
1
Pg = lim
T → T  2
−T
2
g 2 (t ) dt

 We take a truncated signal gT(t)

 The integral on the right hand side will be the energy of the
truncated signal, thus
E gT
Pg = limT →
T 29
 Signal Power and Power Spectral Density (PSD)

 From Parseval’s Theorem  where

GT (w)
2

S g (w) = limT →
  2

=  g (t )dt =  G (w )
2 1
E gT dw T
2
T T
 Sg(w) is the Power Spectral
− −
Density of Power Signal, Which
 The power of signal is given as is actually the time average of
ESD
E gT  
Pg = limT →
 S (w)dw =   S (w)dw
1 1
T Pg =
2
g g
− 0

GT (w) 
2
1
Pg =
2  limT → T
dw Pg = 2  S g ( f )df 30
−
0
 Time Autocorrelation [Power Signals]

 Time Autocorrelation Function of Power Signals

T
2
1
 g ( ) = lim
T → T  g (t ) g (t +  )d
T
−
2

 Similar to Energy signals, Autocorrelation Function of Power

signals is Even
 g ( ) =  g (− )
31
 Time Autocorrelation [Power Signals]

 As ESD is Fourier transform of autocorrelation function for Energy

Signals  g ( )  Ψ ( )
 Similarly, for power signals
 gT ( )
 g ( ) = lim
T → T
 Because
 gT ( ) = GT ( )
2

 So
GT ( )
2

 g ( ) = lim = S g ( ) 32
T → T
 PSD of INPUT and Output Signals

 Fourier Transform of Output is given as

Y ( ) = H ( )G ( )
 then
Y ( ) = H ( ) G ( )
2 2 2

 Since
GT ( )
2

{ g ( )} = lim = S g ( )
 So T → T
S y ( ) = H ( ) S g ( )
2
33
 PSD of Modulated Signals

 The modulated signal can be represented by

 (t ) = g (t ) cos  0 t
 Its Fourier Transform

S ( ) =
1
4

S g ( + 0 ) + S g ( − 0 ) 
1
P = Pg
2
34
The End

35