EEE351, Fall 2021

Principles of Communication
Systems

Dr. Saleem Akhtar

Department of Electrical & Computer Engineering
CUI, Lahore
 Principles of Communication Systems

Resource Persons
• Dr. Saleem Akhtar (Power Stream)
• Email: [email protected]

• Mr. M. Usman Iqbal (Electronics Stream)

• Email: [email protected]

[2]
 Principles of Communication Systems

Lecture Outline
❑ 3.4 Signal Transmission through a Linear System

❑ 3.4.2 Distortionless Transmission

❑ 3.5 Ideal Versus Practical filters

❑ 3.7 Signal Energy and Energy Spectral Density

❑ 3.8 Signal Power and Power Spectral Density

[3]
3.4 Signal Transmission through a Linear System
❑ For a Linear Time Invariant (LTI) Continuous Time System, the input and output
are related through the relation
𝑦 𝑡 =𝑥 𝑡 ∗ℎ 𝑡 Eq. 1

Where ℎ 𝑡 = unit impulse response of the LTI system.

Figure 3.24 Signal transmission through a linear time-invariant system.

[4]
3.4 Signal Transmission through a Linear System
▪ In frequency domain, the Eq. 1 can be expressed as
𝑌 𝑓 = 𝑋 𝑓 .𝐻 𝑓 Eq. 2

Where 𝐻 𝑓 = 𝐻(𝑓) 𝑒 𝑗𝜃ℎ 𝑓

Spectral response of the LTI system
𝐻(𝑓) = Amplitude response of the LTI system
𝜃ℎ 𝑓 = Phase response of the LTI system

▪ As
𝑦 𝑡 ⟺ 𝑌 𝑓 = 𝑌 𝑓 𝑒 𝑗𝜃𝑦 𝑓

𝑥(𝑡) ⟺ 𝑋 𝑓 = 𝑋(𝑓) 𝑒 𝑗𝜃𝑥 𝑓

– Therefore, Eq.2 can be expressed as

𝑌 𝑓 𝑒 𝑗𝜃𝑦 𝑓 = 𝑋(𝑓) 𝑒 𝑗𝜃𝑥 𝑓
𝐻(𝑓) 𝑒 𝑗𝜃ℎ 𝑓

𝑌 𝑓 𝑒 𝑗𝜃𝑦 𝑓
= 𝐻(𝑓) 𝑋(𝑓) 𝑒 𝑗 𝜃ℎ 𝑓 +𝜃𝑥 𝑓
Eq. 3

– Comparing LHS and RHS of the Eq. 3

𝑌 𝑓 = 𝐻(𝑓) 𝑋(𝑓) = Ampltude of output
𝜃𝑦 𝑓 = 𝜃ℎ 𝑓 + 𝜃𝑥 𝑓 = Phase of output

[5]
3.4 Signal Transmission through a Linear System
▪ It is clear from the above equations that an LTI system modifies the amplitude
and phase of the input signal to produce the amplitude and phase of the output
signal.

▪ Usually, LTI system response for all frequencies is not same.

▪ The response of an LTI system may be to boost the signal amplitude at some
frequencies and attenuate some other frequencies.

▪ Similarly, the phase response of an LTI system results in phase variation of

different frequency components in different ways.

▪ Thus, the amplitude and phase variations of different frequency components

result in signal distortion and output signal shape may be different from input
signal shape.

[6]
 3.4.2 Distortionless Transmission
▪ A transmission is said to be distortion less if the input and output signals have identical
waveforms within a multiplicative constant.

▪ A delayed output signal that retains the input signal waveform is also considered as
distortionless.

▪ Thus, a distortionless transmission can be expressed as

𝑦 𝑡 = 𝑘. 𝑥 𝑡 − 𝑡𝑑 Eq. 4
or

𝑌 𝑓 = 𝑘𝑋 𝑓 𝑒 −𝑗2𝜋𝑓𝑡𝑑 Eq. 5

▪ Comparing it with output of an LTI system

𝑌 𝑓 =𝐻 𝑓 𝑋 𝑓

▪ Therefore, frequency response 𝐻 𝑓 of the transmission system should be of the form

𝐻 𝑓 = 𝑘𝑒 −𝑗2𝜋𝑓𝑡𝑑 Eq. 6

[7]
 3.4.2 Distortionless Transmission
𝐻 𝑓 = 𝑘𝑒 −𝑗2𝜋𝑓𝑡𝑑 Eq. 6

▪ Amplitude response = 𝐻 𝑓 =𝑘

▪ Phase response = 𝜃ℎ 𝑓 = −2𝜋𝑓𝑡𝑑

Figure 3.25 Linear time invariant system frequency response for

distortionless transmission.

[8]
 3.4.2 Distortionless Transmission
▪ Therefore, for the distortion less transmission, the amplitude response 𝐻 𝑓 must be a
constant and

▪ Phase response 𝜃ℎ 𝑓 must be linear function of 𝑓 going through the origin.

▪ The slope of 𝜃ℎ 𝑓 with respect to 𝑓 must be constant −𝑡𝑑 , where 𝑡𝑑 is the delay of
output with respect to the input.

▪ It is a common mistake to think that flatness of amplitude response 𝐻 𝑓 alone can

guarantee signal quality. A system that has a flat amplitude response may yet distort a
signal beyond recognition if the phase response is not linear (𝑡𝑑 not constant).

▪ Audio: Human ear is sensitive to amplitude variation/ distortions and relatively in sensitive
to phase variations/distortions.

▪ Video: In contrast, human eye is sensitive to phase variations/distortions and insensitive to

amplitude phase variations/distortions.

[9]
 3.4.2 Distortionless Transmission
❑ Problem 3.4.1: Signals 𝑔1 𝑡 = 104 Π 104 𝑡 and 𝑔2 𝑡 = 𝛿 𝑡 are applied at the inputs of
the ideal low-pass filters 𝐻1 𝑓 = Π 𝑓Τ20,000 and 𝐻2 𝑓 = Π 𝑓Τ10,000 . The outputs
𝑦1 𝑡 and 𝑦2 𝑡 of these filters are multiplied to obtain the signal 𝑦 𝑡 = 𝑦1 𝑡 𝑦2 𝑡 .

▪ Sketch 𝐺1 𝑓 and 𝐺2 𝑓

▪ Sketch 𝐻1 𝑓 and 𝐻2 𝑓 .

▪ Sketch 𝑌1 𝑓 and 𝑌2 𝑓 .

▪ Find the bandwidths of 𝑦1 𝑡 , 𝑦2 𝑡 and 𝑦 𝑡 .

Figure P.3.4-1

[10]
 3.5 Ideal Versus Practical filters
❑ Ideal Filters allow distortion less transmission for a certain band of frequencies
and suppress all other frequencies.

Figure 3.26 (a) Ideal low-pass filter (LPF) frequency response and (b) its impulse response.

Figure 3.27 Ideal high-pass (HPF) and bandpass filter (BPF) frequency responses.

[11]
 3.5 Ideal Versus Practical filters
▪ Let us consider an ideal LPF, its output should be of the form with 𝑘 = 1
𝑦 𝑡 = 𝑥 𝑡 − 𝑡𝑑

To obtain this distortionless output, the frequency response should be of the form

𝑓
𝐻 𝑓 = 𝑟𝑒𝑐𝑡 𝑒 −𝑗2𝜋𝑓𝑡𝑑
2𝐵
𝑓
𝐻 𝑓 = 𝑟𝑒𝑐𝑡 𝜃ℎ 𝑓 = −2𝜋𝑓𝑡𝑑
2𝐵

Therefore, the unit impulse response should be of the form: Figure 3.26 (b)
ℎ 𝑡 = 2𝐵𝑠𝑖𝑛𝑐 2𝜋𝐵 𝑡 − 𝑡𝑑

▪ From Figure 3.26 (b), it is clear that an ideal LPF that can transmit the input signal
𝑥 𝑡 without distortion but with time delay 𝑡𝑑 is noncasual and therefore physically
unrealizable in real time.

▪ Similarly, other ideal filters (such as the ideal high-pass or the ideal bandpass are also
physically unrealizable.
[12]
 3.5 Ideal Versus Practical filters
▪ For physically realizable systems in real time, ℎ 𝑡 must be casual.

ℎ 𝑡 =0 𝑓𝑜𝑟 𝑟 < 0 Time domain criteria (Eq. 1)

In frequency domain, the criteria is known as Paley-Wiener criterion for an amplitude

response to be realizable

∞ ln 𝐻 𝑓
‫׬‬−∞ 1+ 𝑑𝑓 <∞ Frequency domain criteria (Eq. 2)
2𝜋𝑓 2

▪ According to Paley-Wiener criterion , 𝐻 𝑓 cannot be zero over any finite band because
if 𝐻 𝑓 = 0 then this implies that ln 𝐻 𝑓 = ∞ over that band and the condition of
Paley−Wiener criterion is not fullfiled.

▪ Therefore, for a physically realizable system, 𝐻 𝑓 may be zero at some discrete

frequencies, but it cannot be zero over any finite band. According to this criterion, ideal
filter characteristics of LPF, BPF and HPF are clearly unrealizable.

[13]
 3.5 Ideal Versus Practical filters
▪ Therefore, for all ideal filters whether they are LPF, BPF or HPF, one practical approach to
have ℎ 𝑡 = 0 for 𝑡 < 0 is to cut off the tail in case of ℎ 𝑡 ≠ 0 for 𝑡 < 0 and to have
𝑡𝑑 sufficiently large.
ℎ෠ 𝑡 = ℎ 𝑡 𝑢 𝑡

– ℎ෠ 𝑡 can be physically realizable in real time because it is causal. ℎ෠ 𝑡 can closely match
ℎ 𝑡 if 𝑡𝑑 is large.

Figure 3.28 Approximate realization of an ideal low-pass filter by truncating its impulse response.

[14]
 3.7 Signal Energy and Energy Spectral Density
2
▪ The energy 𝐸𝑔 of a signal 𝑔 𝑡 is defined as the area under 𝑔(𝑡) .
∞
𝐸𝑔 = ‫׬‬−∞ 𝑔 𝑡 𝑔∗ 𝑡 𝑑𝑡 𝑔∗ 𝑡 = complex conjugate of 𝑔(𝑡)

for real signals 𝑔∗ 𝑡 = 𝑔(𝑡)

▪ We can also determine the signal energy from its Fourier transform 𝐺 𝑓 through
Parseval's theorem.

∞
2
𝐸𝑔 = න 𝐺 𝑓 𝑑𝑓
−∞

▪ Therefore, the energy of a signal 𝑔 𝑡 can be obtained from either the time
domain specification 𝑔 𝑡 and its frequency domain specification 𝐺 𝑓 .

▪ Example 3.16

[15]
 3.7 Signal Energy and Energy Spectral Density
❑ Energy Spectral Density (ESD) is defined as Energy per unit bandwidth (BW)
𝜓𝑔 𝑓 = 𝐺 𝑓 2

Given the ESD, one can also find the energy from ESD as
∞
𝐸𝑔 = න 𝜓𝑔 𝑓 𝑑𝑓
−∞

❑ Essential Bandwidth of a Signal

▪ The spectra of most signals extend to infinity. However, because the energy of a practical
signal is finite, the signal spectrum must approach 0 as 𝑓 → ∞.

▪ Most of the signal energy is contained within a certain band of 𝐵 Hz, and the energy
content of the components of frequencies greater than 𝐵 Hz is negligible.

▪ We can therefore suppress the signal spectrum beyond 𝐵 Hz with little effect on the signal
shape and energy. The bandwidth 𝐵 is called the essential bandwidth of the signal. The
criterion for selecting 𝐵 depends on the error tolerance in a particular application.

▪ Examples 3.17 and 3.18

[16]
 3.7 Signal Energy and Energy Spectral Density
❑ Energy of Modulated Signals:
▪ Let 𝑔 𝑡 be a baseband signal band-limited to 𝐵 Hz. This signals can be amplitude
modulated by multiplying this signal with high frequency carrier (sinusoidal) signal
cos 2𝜋𝑓0 𝑡.

▪ The amplitude-modulated signal 𝜑 𝑡 can be expressed as

𝜑 𝑡 = 𝑔 𝑡 cos 2𝜋𝑓0 𝑡
▪ The spectrum (Fourier transform) of the modulated signal 𝜑 𝑡
1
Φ 𝑓 = 𝐺 𝑓 + 𝑓0 + 𝐺(𝑓 − 𝑓0 )
2
▪ The ESD Φ 𝑓 of the modulated signal 𝜑 𝑡 is Φ 𝑓 2 , that is
1
Ψ𝜑 𝑓 = 𝜙 𝑓 2 = 𝐺 𝑓 + 𝑓0 2 + 𝐺 𝑓 − 𝑓0 2 + 2𝐺 𝑓 + 𝑓0 𝐺(𝑓 − 𝑓0 )
4
▪ For proper modulation scheme, 𝑓0 ≥ 𝐵, 𝐺 𝑓 + 𝑓0 and 𝐺(𝑓 − 𝑓0 ) are always non
overlapping, the above relationship can be simplified as
1 1
Ψ𝜑 𝑓 = 𝐺 𝑓 + 𝑓0 2 + 𝐺 𝑓 − 𝑓0 2 = Ψ𝑔 𝑓 + 𝑓0 + Ψ𝑔 𝑓 − 𝑓0
4 4
1 𝐸𝑔
𝐸𝜑 = 2𝐸𝑔 = 𝑓0 ≥ 𝐵
4 2

[17]
 3.7 Signal Energy and Energy Spectral Density
❑ Energy of Modulated Signals:
▪ The ESDs of both 𝑔 𝑡 and the modulated signal 𝜑 𝑡 are shown in Fig. 3.35.

Figure 3.35 Energy spectral densities of (a) modulating and (b) modulated signals.
▪ It is clear that modulation shifts the ESD of 𝑔 𝑡 by ±𝑓0 . Observe that the area under
Ψ𝜑 𝑓 is half the area under Ψ𝑔 𝑓 .
▪ Because the energy of a signal is proportional to the area under its ESD, it follows that the
energy of 𝜑 𝑡 is half the energy of 𝑔 𝑡 .

[18]
 3.7 Signal Energy and Energy Spectral Density
❑ Time autocorrelation Function and the Energy Spectral Density

▪ The time autocorrelation function of a signal and its energy spectrum density of
the signal form Fourier transform pair.

▪ Time autocorrelation of a real signal is defined as

∞
𝜓𝑔 𝜏 = න 𝑔 𝑡 𝑔 𝑡 + 𝜏 𝑑𝑡 𝜓𝑔 𝜏 = 𝜓𝑔 −𝜏
−∞
2
𝜓𝑔 𝜏 ⟺ Ψ𝑔 𝑓 = 𝐺 𝑓

❑ ESD of the Input and the Output

▪ If 𝑔 𝑡 and are 𝑦 𝑡 the input and the corresponding output of a linear time-
invariant (LTI) system, then
𝑌 𝑓 =𝐻 𝑓 𝐺 𝑓
Ψ𝑦 𝑓 = 𝐻 𝑓 2 Ψ𝑔 𝑓
2
▪ The output signal ESD is 𝐻 𝑓 times the input signal ESD.

[19]
 3.8 Signal Power and Power Spectral Density
▪ For a power signal, a meaningful measure of its size is its power as the time average of the
signal energy averaged over the infinite time interval.

▪ The power 𝑃𝑔 of a real-valued signal 𝑔 𝑡 is given by

1 ∞
𝑃𝑔 = lim ‫׬‬−∞ 𝑔2 𝑡 𝑑𝑡 𝑔 𝑡 is defined over −∞<𝑡 <∞
𝑇
𝑇→∞

▪ The signal power and the related concepts can be understood by defining a truncated
𝑇
𝑔 𝑡 𝑡 ≤
𝑔𝑇 𝑡 = 2
𝑇
0 𝑡 >
2
𝐸𝑔𝑇
𝑃𝑔 = lim
𝑇→∞ 𝑇

[20]
 3.8 Signal Power and Power Spectral Density
❑ Power Spectral Density (PSD)

▪ Let 𝑆𝑔 𝑓 = Power Spectral Density

2
𝐺𝑇 𝑓
𝑆𝑔 𝑓 = lim
𝑇→∞ 𝑇
Given the PSD, one can also find the energy from PSD as
∞ ∞
𝑃𝑔 = න 𝑆𝑔 𝑓 𝑑𝑓 = 2 න 𝑆𝑔 𝑓 𝑑𝑓
−∞ 0

▪ As is the case with ESD, the PSD is also a positive, real, and even function of 𝑓.

[21]
 3.8 Signal Power and Power Spectral Density
❑ Time autocorrelation Function of Power Signals

▪ The (time) autocorrelation function 𝑅𝑔 𝜏 of a real power signal 𝑔 𝑡 is defined as

𝑇 𝑇
1 2 1 2
𝑅𝑔 𝜏 = lim න 𝑔 𝑡 𝑔 𝑡 − 𝜏 𝑑𝑡 = lim න 𝑔 𝑡 𝑔 𝑡 + 𝜏 𝑑𝑡
𝑇→∞ 𝑇 −𝑇 𝑇→∞ 𝑇 −𝑇
2 2
𝑅𝑔 𝜏 = 𝑅𝑔 −𝜏

▪ Similar to ESD, The time autocorrelation function of a power signal and its power
spectrum density of the signal form Fourier transform pair.
𝐺𝑇 𝑓 2
𝑅𝑔 𝜏 ⟺ lim = 𝑆𝑔 𝑓
𝑇→∞ 𝑇

❑ PSD of the Input and the Output

▪ If 𝑔 𝑡 and are 𝑦 𝑡 the input and the corresponding output of a linear time-
invariant (LTI) system, then
2
𝑌 𝑓 = 𝐻(𝑓)𝐺 𝑓 𝑆𝑦 𝑓 = 𝐻 𝑓 𝑆𝑔 𝑓

[22]
 3.8 Signal Power and Power Spectral Density
❑ Power Spectral Density (PSD) and Power of Modulated Signals:
▪ Following the similar arguments ESD and Energy of modulated signals of energy signals,
we can have
𝜑 𝑡 = 𝑔 𝑡 cos 2𝜋𝑓0 𝑡

1
𝑆𝜑 𝑓 = 𝑆 𝑓 + 𝑓0 + 𝑆𝑔 𝑓 − 𝑓0
4 𝑔
1 𝑃𝑔
𝑃𝜑 = 2𝑃𝑔 = 𝑓0 ≥ 𝐵
4 2

▪ The power of 𝜑 𝑡 is half the power of 𝑔 𝑡 .

[23]