ANALYSIS AND TRANSMISSION OF SIGNALS

Telecommunication I

EE419

Fall 2019

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 1 / 30

Table of Contents

1 Signal Transmission Through a Linear System

2 Signal Distortion during Transmission

3 Ideal and Practical Filters

4 Signal Distortion over a Communication Channel

5 Signal Energy and Energy Spectral Density

6 Signal Power and Power Spectral Density

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 2 / 30

Signal Transmission Through a Linear System

A linear time invarient (LTI) continuous-time system can be characterized

well in either the time domain or the frequency domain.
Communication channels can be illustrated as shown below

A stable LTI system can be characterized in the time domain by its impulse
response h(t), which is the system response to unit impulse input. that is,
y(t) = h(t) when, x(t) = δ(t)

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 3 / 30

Signal Transmission Through a Linear System (cont.)

The system response to an input signal x(t) follows the convolution

relationship:
y (t) = h(t) ∗ x(t)
and the frequency response relationship is:

Y (f ) = H(f )X (f )

where, y (t) ⇐⇒ Y (f ), h(t) ⇐⇒ H(f ), x(t) ⇐⇒ X (f )

H(f) is generally referred to as ’the transfer function’ or ’the frequency
response’ of the LTI system.
H(f) is complex in general

H(f ) = |H(f )|e jθh (f )

where |H(f )| is the amplitude response, and θh (f ) is the phase response.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 4 / 30

Signal Distortion during Transmission

From the previous section we have

Y (f ) = H(f )X (f )

We can rewrite it in a polar form

|Y (f )| e jθy (f ) = |X (f )| |H(f )| e j[θx (f )+θh (f )]

Therefore, |Y (f )| = |X (f )| |H(f )| and θy (f ) = θx (f ) + θh (f )

During transmission, some frequency components may be boosted in

amplitude, while others may be attenuated. The relative phase of the
various component also change.

In generally, the output waveform will be different from the input. where
the input signal amplitude is changed by a factor of |H(f )| and shifted in
phase by angle of θh (f )

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 5 / 30

Distortionless Transmission
In many applications, output have to be a replica of the input waveform.
Therefore, we need to minimize the distortion in communication channel
or amplifier.
Here, we will describe the characteristic (for practical interest) of a system
that allow a signal to pass without distortion ”Distortionless
Transmission”.
For distortionless system the input x(t) and the output y(t) satisfy the
condition:
y (t) = k.x(t − td )
Fourier of the equation is:
Y (f ) = kX (f )e −j2πftd
since,
Y (f ) = H(f )X (f )
then,
H(f ) = ke −j2πftd
Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 6 / 30
Distortionless Transmission (cont.)

The previous equation, H(f ) = ke −j2πftd , is the required transfer function

for distortionless transmission.
LTI system frequncy response
slope equals −td

|H(f )| = k

θh (f ) = −2πftd
|H(f )| must be constant and the phase response must be linear function of
the frequency and going through the origin (f=0).

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 7 / 30

Physical Explaination of Linear Phase

Time delay in a signal causes a linear phase shift in its spectrum.

Imagine g (t) being synthesized by its Fourier components, which are
sinusoids of certain amplitudes and phases. The delayed version g (t − t0 )
can be synthesized by the same sinusoidal component each delayed by t0 .
The amplitudes remain unchanged. However, The time delay of t0 does
change the phase in each sinusoid.

cos ω(t − t0 ) = cos(ωt − ωt0 )

Higher frequency components must

undergo proportionately higher phase
shifts to achieve the same time delay

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 8 / 30

All-pass vs Distortionless Transmission (cont.)

An all pass system is one that has a constant gain for all frequencies [i.e
|H(f )| = k], without the linear phase requirement.
- Distortionless system is always an all pass system, the converse is not
true.
- In practice, many systems have a phase characteristic that may be only
approximately linear. A convenient method is to check the slope for
linearity.
1 dθh (f )
td = −
2π df
for a signal transmission to be distortionless, td (f ) should be a constant td
over the frequency band of interest.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 9 / 30

All-pass vs Distortionless Transmission (cont.)

If g(t) and y(t) are the input and the output for a single RC low-pas filter.
Determine the transfer function H(f ) and sketch |H|(f ), θf (f ) and td (f ).
For distortionless transmission through the filter, what is the requirement
on the bandwidth of g(t) if amplitude response variation within 2% and
time delay variation within 5% are tolerable? What is the transmission
delay?
XC
Y (f ) = G (f )
XR + XC
1 1
jωC RC a
H(f ) = 1
= 1
=
R+ jωC jω + RC
j2πf + a
1
where, a = RC = 106
a
Hence, |H(f )| = √ ' 1 for , 2πf  a
a2 +(2πf )2

|θh (f )| = −tan−1 ( 2πf 2πf

a )'− a for , 2πf  a

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 10 / 30

All-pass vs Distortionless Transmission (cont.)
dθf
td (f ) = − d(2πf ) =
a
(2πf )2 +a2
' 1
a = 10−6 for , 2πf  a
- f0 is the highest frquency
1
|H(0)| = 1 and td (0) = a sec.
0.95
|H(f0 )| ≥ 0.98 and td (f0 ) ≥ a sec.
|H(f0 )| = √ a 2 2 ≥ 0.95
(2πf0 ) +a
2πf0 ≤ 0.203 a = 203, 000 rad/sec
a 0.95
td (f0 ) = (2πf0 )2 +a2
≥ a
2πf0 ≤ 0.2294 a = 229, 400 rad/sec
take the smaller, f0 = 32.3 kHz is the
highest bandwidth that satisfy both
constraints.
The time delay td ≈ 10−6 over this band. Also, the amplitude is almost
unity. Therefore, the output y (t) ≈ g (t − 10−6 ).
Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 11 / 30
Ideal and Practical Filters
Ideal filters allow distortionless transmission of a certain band of
frequencies and suppress all remaining frequencies.
There are three types of filters, depending on the band:
1 Low-pass filter
2 Bandpass filter
3 High pass filter
Ideal low-pass filter: allow all component below f = B Hz to pass without
distortion and suppress all component above B Hz.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 12 / 30

Ideal and Practical Filters (cont.)

y (t) = g (t − td )
The signal g(t) is transmitted by this system without distortion, but with
delay td .
f
H(f ) = Π( )e −j2πftd
2B
hence,
h(t) = 2 B sinc[2πB(t − td )]
which is the impulse response of the filter. You can see the response of
h(t) begins before the input is applied (t=0). Thus, the filter is noncausal.
Therefore, Ideal low-pass, high-pass or bandpass filters are physically
unrealizable.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 13 / 30

Ideal and Practical Filters (cont.)

For physical realizable system, h(t) must be causal. h(t) = 0 for t < 0.
In the frequency domain, this condition is equivalent to the Paley-Wiener
criterion, which states that the necessary and sufficient condition for
|H(f )| to be the amplitude response of a realizable (or causal) system is
Z ∞
|ln|H(f )||
2
df < ∞
−∞ + (2πf )
1
now, for h(t) to be realizable, it needs to be
zero for t < 0.

ĥ(t) = h(t)u(t) causal and physically realizable

- In practice, we can realize a variety of filter characteristics to approach

ideal characteristics.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 14 / 30

Ideal and Practical Filters (cont.)
- Practical filters are gradual without jump discontinuities in the amplitude
response |H(f )|, for example Butterworth filters.
1
|H(f )| = q
1 + ( Bf )2n
The half power bandwidth of the filter is defined
as the bandwidth over which the amplitude
response |H(f )| remains constant within
variations of 3dB (or a ratio of √12 , 0.707).
The half power bandwidth of low-pass filter is
also called the cut-off frequency.
Analog signals can also be processed by digital
means (A/D conversion: sampling, quantizing
and coding). The resulting digital signal is
processed by special-purpose computer
algorithms. The output is converted back into
the desired analog signal (D/A conversion).
Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 15 / 30
Signal Distortion over a Communication Channel

1 Linear distortion: Which is the distortion caused by non-ideal channel

characteristics of magnitude, phase or both.
Eg.: A low-pass filter, its transfer function H(f)
is given 
by
(1 + k cos(2πfT )e −j2πftd ) |f | ≤ B
H(f ) =
0 |f | > 0
A phase g(t) band limited to B Hz is applied to
this filter, find the output y(t).
Y (f ) = G (f )H(f )
f
= G (f )Π( 2B )(1 + kcos(2πfT ))e −j2πftd
= G (f )e −j2πftd + k[G (f )cos(2πfT ))]e −j2πftd
y (t) = g (t −td )+ k2 [g (t −td −T )+g (t −td +T )]
The output consists of g(t) and its echoes shifted by ±T. In short, the
pulse will spread out in time domain. Time dispersion will occur if
amplitude or phase response are non-ideal.
Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 16 / 30
Signal Distortion over a Communication Channel (cont.)

2 Distortion caused by channel non-linearity

The channel we have considered so far is linear, y(t) = a g(t). If y(t) and
g(t) are related by a non-linear function y = f(g). then this equation can
be extended in a McLaurin’s series as
y (t)a0 + a1 g (t) + a2 g 2 (t) + ... + ak g k (t) + ... and if g(t) bandwidth is B
Hz, then the bandwidth of g k (t) is kB Hz. Consequently, the output
spectrum spreads well beyond the input spectrum, and the output signal
contains new frequency components.
Amplifiers used in broadcast communications are non-linear.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 17 / 30

Signal Distortion over a Communication Channel (cont.)
Eg.: The input x(t) and output y(t) are related
by
y (t) = x(t) + 0.000158x 2 (t)
find y(t) and its spectrum if,
x(t) = 2000 sinc(2000πt)
f
x(t) = 2000sinc(2000πt) ⇐⇒ X (f ) = Π( 2000 )

x(t) + 0.000158x 2 (t) = 2000sinc(2000πt)

+ 0.316x2000sinc 2 (2000πt)

f f
Y (f ) = Π( ) + 0.316∆( )
2000 4000

Distortion can be reduced by passing y(t) through a LPF with BW of 1000

Hz. This distortion causes an interference problem with signals between
1000 and 2000 Hz.
Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 18 / 30
Signal Distortion over a Communication Channel (cont.)

3 Distortion caused by multipath effects

Multipath transmission occurs when a transmitted signal arrives at the
receiver by two or more paths of different delay.

The overall transfer function of such a channel is

H(f ) = e −jωtd + αe −jω(td +∆)

−1 αsinω∆t
p
= 1 + α2 + 2αcosω∆t e −j[ωtd +tan 1+αcosω∆t ]
|H(f )| θh (f )

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 19 / 30

Signal Distortion over a Communication Channel (cont.)

4 Fading channels
They are channels in which the effective channel transfer functions varies
semiperiodically and randomly, causing random attenuation of the signal.
There are different types of fading. For example, Slow fading which can be
reduced using Automatic Gain Control (AGC).
Fading may be frequency dependent, such fading is known as
frequency-selective fading. Multipath propagation can cause
frequency-selective fading.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 20 / 30

Signal Energy and Energy Spectral Density

For a signal g(t), Z ∞

Eg = g (t)g ∗ (t).dt
−∞
Parseval’s theorem:
Z ∞ Z ∞ Z ∞
∗
Eg = g (t)g (t).dt = g (t) G ∗ (f )e −j2πft .df .dt
−∞ −∞ −∞
Z ∞ Z ∞  Z ∞
∗ −j2πft
= G (f ) g (t)e .dt df = G (f )G ∗ (f )df
−∞ −∞ −∞
Z ∞
Eg = |G (f )|2 .df
−∞

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 21 / 30

Signal Energy and Energy Spectral Density (cont.)

Eg.: Verify Parseval’s theorem for g(t) = e −at u(t), a > 0

Z ∞
1 ∞ 1
Eg = e −2at .dt = − e −2at =
0 2a 0 2a
1
G (f ) =
a + j2πf
using Parseval’s,
Z ∞ Z ∞
1 1 1
Eg = x df = df
−∞ a + j2πf a − j2πf −∞ a2 + (2πf )2

1 2πf ∞ 1
Eg = tan−1 ( ) =
2πa a −∞ 2a

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 22 / 30

Energy Spectral Density (ESD)

From the previous section we can conclude that the energy of a signal g(t)
is the result of energies contributed by all spectral component of the signal
g(t).
Thus |G (f )|2 is the energy spectral density (per unit bandwidth in hertz)
of g(t).
The Energy Spectral Density (ESD) Ψ(t) is thus defined as

Ψg (f ) = |G (f )|2
Z ∞
Eg = Ψg (f )df
−∞

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 23 / 30

Essential Bandwidth of a Signal
It is the band in which most of the signal energy is contained.
The criterion of selecting B depends on the error tolerance in a particular
application.
Eg.: Estimate the essential BW W (in rad/sec) of the signal e −at u(t) if
the essential band is required to contain 95% of the signal energy.
1
G (f ) =
a + j2πf
1
ESD Ψg (f ) = |G (f )|2 =
a2 + (2πf )2
1
from previous example, Eg =
2a
The BW that contains 95% of the energy may be found using
Z W
0.95 2π 1
= df
2a W
− 2π a2 + (2πf )2

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 24 / 30

Essential Bandwidth of a Signal (cont.)

W
0.95 1 2πf 2π
= tan−1 ( ) W
2a 2πa a − 2π
0.95 1 h −1 W W i
= tan ( ) − tan−1 (− )
2a 2πa a a
0.95 1 W
= tan−1 ( )
2a πa a
0.95π W
= tan−1 → W = 12.7a rad/sec
2 a
W 12.7a
B= = = 2.02 a Hz
2π 2π
Whenever we talk about bandwidth, we mean in trigonometric sense.
Hence, the essential bandwidth is from 0 to B Hz (or W rad/sec), not
from -B to B.

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 25 / 30

 Energy of Modulated Signal

If g(t) is the baseband signal band-limited to B Hz. The AM signal Φ(t) is

Φ(t) = g (t)cos(2πf0 t)
1
Φ(f ) = [G (f + f0 ) + G (f − f0 )]
2
The ESD of the modulated signal Φ(t) is |Φ(t)|2

1
ΨΦ (f ) = |G (f + f0 ) + G (f − f0 )|2
4

If f0 ≥ B, then G(f+f0 ) and G(f-f0 ) are non-overlapping

1 1
ΨΦ (f ) = Ψg (f + f0 ) + Ψg (f − f0 )
4 4
1
Hence, EΦ = Eg
2

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 26 / 30

Time Autocorrelation Function and the ESD
From chapter 2, the correlation of a signal with itself is called the
autocorrelation. The autocorrelation function ψg (τ ) of a signal g(t) is
Z ∞
ψg (τ ) = g ∗ (t) g (t + τ ) dt
−∞

For a real g(t), the autocorrelation is an even function of τ .

We now show that the ESD Ψg (f ) = |G (f )|2 is the Fourier transform of
the autocorrelation function ψg (τ )
Z ∞ hZ ∞ i
−jωτ
F[ψg (τ )] = e g (t)g (τ + t)dt dτ
−∞ −∞
Z ∞ hZ ∞ i
−jωτ
= g (t) g (t + τ )e dτ dt
−∞ −∞
Z ∞
F[ψg (τ )] = G (f ) g (t)e jωt dt = G (f )G (−f ) = |G (f )|2
−∞
ψg (τ ) ⇐⇒ Ψg (f ) = |G (f )|2
Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 27 / 30
ESD of the input and output

for x(t) and y(t) are input and output of an LTI system

Y (f ) = H(f ) X (f )

Therefore,
|Y (f )|2 = |H((f )|2 |X (f )|2
This shows that
Ψy (f ) = |H(f )|2 Ψx (f )

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 28 / 30

Signal Power and Power Spectral Density
For a power signal g(t)
Z T
1 2
Pg = lim |g 2 (t)|dt
T →∞ T − T2

EgT
P = lim
T →∞ T

The signal power is just the time average energy

Power Spectral Density (PSD)
if , gT (t) ⇐⇒ GT (f )
Z ∞ Z ∞
2
then, EgT = gT (t)dt = |GT (f )|2 df
−∞ −∞
Z ∞
EgT 1h i
hence, Pg = lim = lim |GT (f )|2 df
T →∞ T T →∞ T −∞

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 29 / 30

Power Spectral Density (PSD)

From the last equation, the signal is a power signal and the right hand side
converges: Z ∞
|GT (f )|2
Pg = lim df
−∞ T →∞ T
Power Spectral Density is defined as

|GT (f )|2
Sg = lim
T →∞ T
Z ∞ Z ∞
Pg = Sg (f )df = 2 Sg (f )df
−∞ 0

Input and Output Power Spectral Densities

Sy (f ) = |H(f )|2 Sg (f )

Telecommunication I (EE419) ANALYSIS AND TRANSMISSION OF SIGNALS Fall 2019 30 / 30