IERG2310-ESTR2300 Tutorial 2

Principles of Communication Systems

SHI Jiajia
sj022@ie.cuhk.edu.hk
Jan 22, 2025
 2

Contents
✭ Fourier Transform
✭ Convolution
✭ LTI Systems Described by Differential Equations
✭ Amplitude Modulation (AM)
➢ Double Sideband Large Carrier (DSB-LC) Modulation/Common AM
• Modulation Index
• Signal Power
➢ Double Sideband Suppressed Carrier (DSB-SC) Modulation
➢ Single Sideband Suppressed Carrier (SSB-SC) Modulation
➢ Vestigial Sideband Suppressed Carrier (VSB-SC) Modulation
✭ Example: Using a Block Diagram to Generate a Desired AM Signal
 3

Fourier Transform
• Fourier Transform (FT)

• Inverse Fourier Transform (IFT)

• Fourier Transform Pairs

Time Frequency
 4

Properties of FT

• Time Shifting

• Frequency Shifting

• Convolution

• Multiplication

• Differentiation
 5

Convolution
• Why Convolution?

➢ Key
• Properties of LTI system: 𝑥(𝑡) 𝑦(𝑡)
1) Time-invariant:
𝑥(𝑡 − 𝑡0 ) 𝑦(𝑡 − 𝑡0 )
2) Linear:
𝑎𝑥(𝑡) 𝑎𝑦(𝑡)
• Decomposition of function 𝑓(𝑡):
+∞
𝑓 𝑡 =න 𝑓 𝜏 𝛿 𝑡 − 𝜏 𝑑𝜏 = 𝑓 𝑡 ∗ 𝛿(𝑡)
−∞
 6

Convolution
• Formula

• Method:
1) Plot 𝑥(𝜏) and ℎ(𝜏)

2) Flip ℎ(𝜏) around the vertical axis to obtain ℎ(−𝜏)

3) Shift ℎ(𝜏) by 𝑡 to obtain ℎ(𝑡 − 𝜏)

4) Multiply to obtain 𝑥 𝜏 ℎ(𝑡 − 𝜏)

+∞
5) Integrate on 𝜏 to compute‫׬‬−∞ 𝑥 𝜏 ℎ 𝑡 − 𝜏 𝑑𝜏

6) Increase 𝑥 from −∞ to +∞ and repeat steps 3-6.

 7

Convolution
➢ Exercise #1:
 8

Convolution
➢ Solution:
 9

LTI System Described by Differential Equations

• A general LTI system described by N-order differential equations:

• Take FT on both sides, gives

 10

LTI System Described by Differential Equations

➢ Exercise #2:
Consider a LTI system described by

Find the output 𝑦(𝑡) if input 𝑥 𝑡 = 𝑒 −𝑡 𝑢(𝑡).

➢ Solution:
Take FT at both sides, gives:
 11

Some Basic Concepts

➢ Baseband signal/Message signal

• The original message signal has not been modulated and must be modulated to higher
frequencies for radio transmission.
• It is usually a low-frequency signal.

➢ Carrier
• The carrier is higher in frequency than the message signal.
• Usually is denoted by
𝑐 𝑡 = 𝑨𝒄 𝑐𝑜𝑠 2𝜋𝒇𝒄 𝑡 + 𝝓𝒄 .

Amplitude Frequency Phase

Notice that we will set up 𝜙𝑐 = 0 for simplicity, so 𝒄 𝒕 = 𝑨𝒄 𝒄𝒐𝒔(𝟐𝝅𝒇𝒄 𝒕).
 12

Some Basic Concepts

➢ Modulation/Modulator
• Modulation: Impose the baseband signal onto a carrier signal with high frequency.
• How do modulate? Use message signals to change the parameters of the carriers (𝐴𝑐 , 𝑓𝑐 and 𝜙𝑐 ).
• Modulator: A device or circuit performs modulation.

➢ Demodulation/Demodulator
• Demodulation: The reverse process of modulation.
• How do we demodulate? Remove the carrier signal.
• Demodulator: Device or circuit performing demodulation.
 13

Some Basic Concepts

Modulation---Frequency Translation

The original message signal The modulated signal

−𝑊 0 𝑊 𝑓 −𝑓𝑐 0 𝑓𝑐 𝑓

➢ E.g.,
• Multiply the baseband signal 𝑚 𝑡 by the carrier 𝑐𝑜𝑠(2𝜋𝑓𝑐 𝑡), i.e.,
𝑚 𝑡 → 𝐴𝑐 𝑚 𝑡 𝑐𝑜𝑠(2𝜋𝑓𝑐 t).
𝑨𝒄
𝑀 𝑓 → 𝑴 𝒇 + 𝒇𝒄 + 𝑴 𝒇 − 𝒇𝒄 .
𝟐
 14

Amplitude Modulation (AM)

Double Sideband Large Carrier (DSB-LC) Modulation
𝑀(𝑓)
𝑚(𝑡)

−𝑓𝑚 0 𝑓𝑚 𝑓
𝐶(𝑓)
𝑐(𝑡)

Require 𝒇𝒄 ≫ 𝒇𝒎!
−𝑓𝑐 0 𝑓𝑐 𝑓
𝑠(𝑡) 𝑆(𝑓)

−𝑓𝑐 0 𝑓𝑐 𝑓
Upper sideband Lower sideband Upper sideband
➢ Standard form
𝑠 𝑡 = 𝐴𝑐 1 + 𝑎𝑚′ 𝑡 𝑐𝑜𝑠(2𝜋𝑓𝑐 𝑡).
• Where 𝒂 < 𝟏 is the modulation index and 𝒎′(𝒕) is the normalized/pre-scaled version of
𝒎(𝒕) with the maximum magnitude being 1.
 15

Amplitude Modulation (AM)

Double Sideband Large Carrier (DSB-LC) Modulation

➢ Properties
• Can be demodulated by an envelope detector.
• Bandwidth is twice of the original: 𝑊𝐴𝑀 = 2𝑊𝑚 .

➢ Modulation index
• The ratio between the maximum amplitude of the message signal and that of the carrier.

➢ Average power of AM signal

𝐴2𝑐
𝑃𝑠 = 1 + 𝑎2 𝑃𝑚′ .
2
• Where 𝑃𝑚′ is the power of 𝑚′(𝑡).
 16

Amplitude Modulation (AM)

Double Sideband Large Carrier (DSB-LC) Modulation

➢ Example #1
1
Consider the signal 𝑠 𝑡 = 𝐴 1 + 𝑥 𝑡 𝑐𝑜𝑠(2000𝜋𝑡), where 𝑥 𝑡 = 4 𝑐𝑜𝑠 20𝜋𝑡 + 𝑐𝑜𝑠 30𝜋𝑡 .
Determine: (1) The modulation index; (2) The power of 𝑠(𝑡).

➢ Solution
1 1 𝐴
(1) Maximum magnitude of the message signal 𝐴𝑥(𝑡) is 𝐴 4
+ 4
=2.
𝐴/2 1
Hence the modulation index 𝑎 = 𝐴
=2.
𝐴2𝑐
(2) 𝑃𝑠 = 2
1 + 𝑎2 𝑃𝑚′
𝐴2 1 1 𝟏 2 𝐴2 1 1 17
= 2
1 + 4 ⋅ 𝑇 ‫𝟐 𝑇׬‬ 𝒄𝒐𝒔 𝟐𝟎𝝅𝒕 + 𝒄𝒐𝒔 𝟑𝟎𝝅𝒕 𝑑𝑡 = 2
1 + 4 ⋅ 4 = 32 𝐴2
 17

Amplitude Modulation (AM)

Double Sideband Large Carrier (DSB-LC) Modulation

➢ Exercise #1
9
Consider the sinusoidally modulated signal shown on the right.
(1) Determine the modulation index;
3
(2) Calculate its power. (Hint: 𝑃𝑚′ = 0.5)
0
𝑡
−3

−9
 18

Amplitude Modulation (AM)

Double Sideband Large Carrier (DSB-LC) Modulation

➢ Exercise #3
9
Consider the sinusoidally modulated signal shown on the right.
(1) Determine the modulation index;
3
(2) Calculate its power. (Hint: 𝑃𝑚′ = 0.5)
0
𝑡
➢ Solution −3
𝐴𝑚 3
(1) 𝑎 = 𝐴𝑐
= 6 = 0.5
𝐴2𝑐 62 −9
(2) 𝑃𝑠 = 2
1 + 𝑎2 𝑃𝑚′ = 2
1 + 0.52 ⋅ 0.5 = 18(1 + 0.25)
 19

Amplitude Modulation (AM)

Double Sideband Suppressed Carrier (DSB-SC) Modulation
➢ Definition
• Note that DSB-LC: 𝐴 𝟏 + 𝑎𝑚′ 𝑡 𝑐𝑜𝑠(2𝜋𝑓𝑐 𝑡) contains a carrier term with no information.
• To improve power efficiency, we drop this term and get DSB-SC:
𝑠 𝑡 = 𝐴𝑚 𝑡 𝑐𝑜𝑠(2𝜋𝑓𝑐 𝑡).
𝑀(𝑓)

𝑚(𝑡)
−𝑓𝑚 0 𝑓𝑚 𝑓

𝑆(𝑓)

𝑠(𝑡)
−𝑓𝑐 0 𝑓𝑐 𝑓
Upper sideband Lower sideband Upper sideband

𝐴2
• Average power: 𝑃𝑠 = 𝑃 , where 𝑃𝑚 is the power of the message signal 𝑚(𝑡).
2 𝑚
 20

Amplitude Modulation (AM)

Double Sideband Suppressed Carrier (DSB-SC) Modulation

➢ Coherent detection

𝐴 𝐴
𝑠 𝑡 𝑐𝑜𝑠 2𝜋𝑓𝑐 𝑡 = 𝐴𝑚 𝑡 𝑐𝑜𝑠2 (2𝜋𝑓𝑐 𝑡) = 𝑚 𝑡 + 𝑚 𝑡 𝑐𝑜𝑠(4𝜋𝑓𝑐 𝑡)
2 2
• Here, we use:
1
𝑐𝑜𝑠𝐴 ⋅ 𝑐𝑜𝑠𝐵 = [𝑐𝑜𝑠 𝐴 + 𝐵 + 𝑐𝑜𝑠(𝐴 − 𝐵)]
2
 21

Amplitude Modulation (AM)

Single Sideband Suppressed Carrier (SSB-SC) Modulation

➢ Definition
• To save bandwidth and further improve power efficiency.
• The sidebands are conjugate symmetric.
• We can drop one sideband without loss of useful
information and get SSB-SC.

Reconstructed signal
 22

Amplitude Modulation (AM)

Single Sideband Suppressed Carrier (SSB-SC) Modulation

➢ The resultant spectrum

1 1
𝑆𝑆𝑆𝐵+(𝑓) = 𝑀 𝑓 − 𝑓𝑐 𝑢 𝑓 − 𝑓𝑐 + 𝑀 𝑓 + 𝑓𝑐 𝑢 − 𝑓 + 𝑓𝑐
2 2
1 1
𝑆𝑆𝑆𝐵−(𝑓) = 𝑀 𝑓 − 𝑓𝑐 𝑢[− 𝑓 − 𝑓𝑐 ] + 𝑀 𝑓 + 𝑓𝑐 𝑢(𝑓 + 𝑓𝑐 )
2 2

➢ Time domain signals

𝑠𝑆𝑆𝐵+ 𝑡 = 𝑚 𝑡 𝑐𝑜𝑠 2𝜋𝑓𝑐 𝑡 − 𝑚
ෝ 𝑡 𝑠𝑖𝑛(2𝜋𝑓𝑐 𝑡)
𝑠𝑆𝑆𝐵− 𝑡 = 𝑚 𝑡 𝑐𝑜𝑠 2𝜋𝑓𝑐 𝑡 + 𝑚
ෝ 𝑡 𝑠𝑖𝑛(2𝜋𝑓𝑐 𝑡)

• Where 𝑚ෝ (𝑡) is the Hilbert transform of 𝑚(𝑡).

If 𝒎 𝒕 = 𝒄𝒐𝒔 𝟐𝝅𝒇𝒄 𝒕 , then 𝒎 ෝ 𝒕 = 𝒔𝒊𝒏(𝟐𝝅𝒇𝒄 𝒕).

Reconstructed signal
 23

Amplitude Modulation (AM)

Vestigial Sideband Suppressed Carrier (VSB-SC) Modulation

➢ Motivation
• Too difficult to design the ideal/rigorous sideband filter
(with short cut-off) and the Hilbert transform.
• To mitigate/relieve the requirement of the sideband
filter in SSB.

➢ Definition
• The sideband is suppressed but not entirely.
𝑠 𝑡 = 𝐴𝑚 𝑡 𝑐𝑜𝑠 2𝜋𝑓𝑐 𝑡 ∗ ℎ 𝑡
𝐴
𝑆 𝑓 = 𝑀 𝑓 − 𝑓𝑐 + 𝑀 𝑓 + 𝑓𝑐 𝐻(𝑓)
2
Where 𝐻 𝑓 is not an ideal sideband filter.
 24

Generate a Desired AM Signal

➢ Example #2
Consider the following signal
𝑠 𝑡 = 𝐴𝑚 𝑡 𝑐𝑜𝑠(2𝜋𝑓𝑐 𝑡)
Where 𝐴 is a constant, 𝒎 𝒕 is a baseband message signal with a bandwidth of 𝟐𝟎𝑲𝑯𝒛, and 𝒇𝒄 =
𝟔𝟎𝟎𝑲𝑯𝒛 is the carrier frequency. Use a block diagram to show how you would generate, from 𝑠 𝑡 , a
single sideband signal occupying the frequency band from 𝟑𝟎𝟎𝑲𝑯𝒛 to 𝟑𝟐𝟎𝑲𝑯𝒛. You can only use
𝑠(𝑡) and the following blocks. Each block can be used multiple times, if needed.
(1) Oscillator at 300𝐾𝐻𝑧; (2) Adder; (3) Product modulator (multiplier); (4) Amplifier; (5) Any ideal
lowpass or bandpass filters (Specify the parameters of your filters).

➢ Approach
• By definition, |𝑆 𝑓 | is 580𝐾𝐻𝑧~620𝐾𝐻𝑧. We first need an oscillator at 300𝐾𝐻𝑧 and a multiplier
to translate 𝑆(𝑓) to a lower band 280𝐾𝐻𝑧~320𝐾𝐻𝑧.
• The signal has to further move across a bandpass filter with admission band 300𝐾𝐻𝑧~320𝐾𝐻𝑧,
to cut off its lower sideband 280𝐾𝐻𝑧~300𝐾𝐻𝑧.
 25

Generate a Desired AM Signal

➢ Approach
• Recall:
1
𝑐𝑜𝑠𝐴 ⋅ 𝑐𝑜𝑠𝐵 = [𝑐𝑜𝑠 𝐴 + 𝐵 + 𝑐𝑜𝑠(𝐴 − 𝐵)]
2
• We have 𝑆 𝑓 from 𝟓𝟖𝟎𝑲𝑯𝒛 to 𝟔𝟐𝟎𝑲𝑯𝒛, and we need SSB from 𝟑𝟎𝟎𝑲𝑯𝒛 to 𝟑𝟐𝟎𝑲𝑯𝒛.

➢ Solution 𝐴
𝐴𝑚 𝑡 𝑐𝑜𝑠(2𝜋𝑓𝑐 𝑡) 𝑚 𝑡 [𝑐𝑜𝑠 3𝜋𝑓𝑐 𝑡 + 𝑐𝑜𝑠 𝜋𝑓𝑐 𝑡 ] SSB Signal
2
Multiplier Bandpass filter

300𝐾𝐻𝑧~320𝐾𝐻𝑧
𝑐𝑜𝑠(𝜋𝑓𝑐 𝑡)

~
Oscillator
 26

Further References

➢ Amplitude Modulation
• By Dr. Hikmat Al-Shamary & Dr. Tariq M. Salman
• https://uomustansiriyah.edu.iq/media/lectures/5/5_2020_03_24!09_42_38_PM.pdf

➢ Communication Principles
• By SHEN Baosuo etc.
• http://www.wenqujingdian.com/Public/editor/attached/file/20180328/20180328120547_79
903.pdf
This is a Chinese version.
 IERG2310 Tutorial 11

Xiangyu Zhong

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 1 / 12
Outline

Error Control Coding

I Basic Concepts

I Linear Binary Block Code

I Exercises

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 2 / 12
Basic Concepts
Error Control Coding
I The coding procedure done to control the occurrences of errors,
including error detection, and/or error correction.

I Basic idea: add redundancy (i.e., extra bits). Or, put another way,
disallow some codewords; if you receive such an invalid codeword,
correct it to the closest valid codeword.

Binary Block Codes

I An (n, k) binary block code contains successive k message bits and
(n − k) check/partiy bits:

I Check bits are computed from the message bits according to the
specific encoding rule.
. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 3 / 12
Basic concepts

Let C is the set of all valid codewords for an (n, k) binary block code,
and c1 , c2 ∈ C are 2 valid codewords.

Concept Definition
Code Rate k/n
Hamming Distance dH (c1 , c2 ) := |c1 − c2 |
d∗H := min dH (c1 , c2 )
Minimum Distance c1 ̸=c2 ,
c1 ,c2 ∈C

Hamming Weight wH (c) := dH (c, 0)

∗ :=
Minimum Weight wH min wH (c)
c̸=0,
c∈C

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 4 / 12
Basic concepts

Table: Error Detection and Correction Capabilities

Description Expression
A code can detect up to d errors
d∗H ≥ d + 1
per codeword if:
A code can correct up to c errors
d∗H ≥ 2c + 1
per codeword if:
A code can detect up to d errors
and correct up to c < d errors per d∗H ≥ d + c + 1
codeword if:

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 5 / 12
Linear Binary Block Code

Linearity: Any linear combination of two valid codewords is also an valid

codeword.

The all-zero sequence 0 is always a codeword.

The minimum distance equals the minimum weight, i.e., d∗H = wH

∗
.

Example:
Hamming codes — can detect up to two-bit errors or correct single-bit
errors.
Repetition codes (p.3) and single parity check bit codes (p.2).

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 6 / 12
Exercise

Exercise 1:
Consider a binary linear block code of the form:
[d1 d2 d3 p1 p2 p3 ]
where d1 , d2 , d3 are message bits and p1 , p2 , p3 are check bits given by
p1 = d1 + d2 mod 2
p2 = d2 + d3 mod 2
p3 = d1 + d3 mod 2
list out all the codewords.

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 7 / 12
Exercise
Exercise 1:
Consider a binary linear block code of the form:
[d1 d2 d3 p1 p2 p3 ]
where d1 , d2 , d3 are message bits and p1 , p2 , p3 are check bits given by
p1 = d1 + d2 mod 2
p2 = d2 + d3 mod 2
p3 = d1 + d3 mod 2
list out all the codewords.

Solution:
[0 0 0 0 0 0], [0 0 1 0 1 1]
[0 1 0 1 1 0], [0 1 1 1 0 1]
[1 0 0 1 0 1], [1 0 1 1 1 0]
[1 1 0 0 1 1], [1 1 1 0 0 0]
. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 8 / 12
Exercise

Exercise 2:
Consider an (8, 4) binary linear block code with minimum distance of 4.
I How many valid codewords are there?
I What is the code rate?
I What is the minimum weight of the code?
I If the code is used for error detection only, how many errors can it
detect?
I If the code is used for error correction, how many errors can it correct?

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 9 / 12
Exercise
Exercise 2:
Consider an (8, 4) binary linear block code with minimum distance of 4.
I How many valid codewords are there?
I What is the code rate?
I What is the minimum weight of the code?
I If the code is used for error detection only, how many errors can it
detect?
I If the code is used for error correction, how many errors can it correct?

Solution:
I 24 = 16.
I 4/8 = 0.5.
∗ ∗
I wH = 4, since wH = d∗H for linear block codes.
∗
I d = 4, by dH ≥ d + 1.
I c = 1, by d∗H ≥ 2c + 1.

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 10 / 12
Exercise

Exercise 3:
Consider a binary linear block code of the form:
[d1 d2 p1 p2 p3 ]
where d1 , d2 are message bits and p1 , p2 , p3 are check bits given by
p1 = d1 + d2 mod 2
p2 = d1 mod 2
p3 = d2 mod 2
Get the error detecting and correcting capabilities.

. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 11 / 12
Exercise
Exercise 3:
Consider a binary linear block code of the form:
[d1 d2 p1 p2 p3 ]
where d1 , d2 are message bits and p1 , p2 , p3 are check bits given by
p1 = d1 + d2 mod 2
p2 = d1 mod 2
p3 = d2 mod 2
Get the error detecting and correcting capabilities.
Solution:
First, we can list all the valid codewords as follows:
[0 0 0 0 0], [0 1 1 0 1]
[1 0 1 1 0], [1 1 0 1 1]
∗
As observed, the minimum weight wH = 3. Since this is a linear block code,
∗ ∗
we have dH = wH = 3.
Thus, the error detection capability is d = 2 and the code correction
capability is c = 1.
. . . . . .
Xiangyu Zhong (CUHK) IERG2310 Tutorial 11 12 / 12
 Assignment 5

Xiangyu Zhong (CUHK) ENGG2310 Tutorial 4 13 / 18

Q16

The low-pass signal x(t) with bandwidth W is sampled with a sampling interval of
Ts , and the signal
∞
X
xp (t) = x(nTs )p(t − nTs )
n=−∞

is generated, where p(t) is an arbitrary shaped pulse (not necessarily time-limited

to the interval [0, Ts ]).
(a) Find the Fourier transform of xp (t) in terms of the Fourier transform of p(t),
namely P (f ).
(b) Find the conditions for perfect reconstruction of x(t) from xp (t), and specify
the required reconstruction filter.

Xiangyu Zhong (CUHK) ENGG2310 Tutorial 4 14 / 18

Q17

Consider a quantizer. The output of the quantizer xQ is given by:



 −0.75, x < −0.5

 −0.25, −0.5 ≤ x < 0
xQ =


 0.25, 0 ≤ x < 0.5
0.75, 0.5 ≤ x


The input samples are distributed according to the probability density function:
(
|x|, 0 ≤ x ≤ 1
fX (x) =
0, else

(a) Determine the signal-to-quantization noise ratio (SQNR).

(b) Write a program to determine the optimal quantizer.

Xiangyu Zhong (CUHK) ENGG2310 Tutorial 4 15 / 18

Q17

clear; clc;

% PDF and range

pdf = @(x) abs(x) .* (x >= -1 & x <= 1); % PDF of X (symmetric |x| o
bins = 4; % Number of quantization levels
thresholds = linspace(-1, 1, bins+1); % Initial decision boundaries
delta = 1e-6; diff = 1; % Convergence criteria

while diff > delta

% Step 1: compute centroids (quantization levels)
for i = 1:bins
a = thresholds(i); b = thresholds(i+1);
num(i) = integral(@(x) x .* pdf(x), a, b); % \int x * f_X(x)
den(i) = integral(@(x) pdf(x), a, b); % \int f_X(x) dx
end
q = num ./ den; % Updated quantization levels

Xiangyu Zhong (CUHK) ENGG2310 Tutorial 4 16 / 18

Q17

% Step 2: update thresholds (except for first and last)

new_thresholds = thresholds;
for i = 2:bins
new_thresholds(i) = (q(i-1) + q(i)) / 2;
end

% Convergence check
diff = max(abs(new_thresholds - thresholds));
thresholds = new_thresholds; % Update thresholds
end

% Display results
disp(’Optimal Quantization Levels:’);
disp(q);
disp(’Optimal Decision Boundaries:’);
disp(thresholds);

Xiangyu Zhong (CUHK) ENGG2310 Tutorial 4 17 / 18

Q18

Simulate the sigma-delta modulation with MATLAB according to the following

steps:
Generate the signal x(t) = sin(πt) + 0.5 cos 3πt


1 2
R
2 Generate y(t) = x(t)dt
3 Apply delta modulation to y(t) at a very high sampling rate.
4 Apply 1-bit D/A, i.e., construct a pulse train using the outputs of the delta
modulator.
5 Smooth the resulting signal.

Xiangyu Zhong (CUHK) ENGG2310 Tutorial 4 18 / 18

 IERG2310 Tutorial 9

Wang Zewei

zwwong@link.cuhk.edu.hk

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310 Tutorial 9 1 / 11
Outline

Binary Communication Systems (signal detection part)

◮ Basic Concepts
⋆ Q function and AWGN channel

◮ Pe of some simple binary signal commun. systems

⋆ NRZ signaling
⋆ RZ signaling

◮ General system
⋆ Given (s1 (t)/s0 (t), h(t), T0 , γ), determine Pe,1 /Pe,0
⋆ Optimal threshold γ: minimax threshold
⋆ Optimal filter h(t): matched filter

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310 Tutorial 9 2 / 11
Basic Concepts
1. Gaussian random variable: X ∼ N (µ, σ 2 ), with pdf:
1 −(x−µ)2
fX (x) = √ e 2σ2
2πσ 2
2. For X ∼ N (µ, σ 2 )
)(
x−µ
Pr(X ≤ x) = Φ
σ
( )
x−µ
Pr(X > x) = Q
σ
3. Q function:
◮ Frequently used in probability of error ’Pe ’
◮ Decreasing with x
◮ Φ(x) + Q(x) = 1
◮ Q-function table check-up, e.g. let Q(x) ≤ 1e−6 , we have x ≥ 4.8

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310 Tutorial 9 3 / 11
Basic concepts

4. Additive white Gaussian noise (AWGN) channel:

◮ N0
Noise: n0 (t), with mean 0 and power spectral density 2

◮ Output of an integrator:
∫ T1
Let η1 = n0 (t)dt; then η1 is a Gaussian random var. with mean 0
0
N 0 T1
and variance 2

◮ Output of a filter:
Let η2 = n0 ∗ h(T2 ); then η2 is a Gaussian random var. with mean 0
and variance (see Slide’s Page 21):
∫
N0 ∞
σ2 = |h(t)|2 dt
2 −∞

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310 Tutorial 9 4 / 11
Pe of some simple binary signal commun. systems
NRZ signaling
◮ Polar (or antipodal) signal: (1, 0) ⇒ (A, −A)
◮ Probability of error:
√  (√ )
2A2 T  2Eb
Pe = Q  =Q
N0 N0

where Eb is the average energy per bit

RZ signaling
◮ On-off binary signal: (1, 0) ⇒ (A, 0)
◮ Probability of error:
√  (√ )
A2 T  Eb
Pe = Q  =Q
2N0 N0

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310A Tutorial 9 5 / 11
General binary communication systems

Assumptions:
◮ k ∈ {1, 0} ⇒ (s1 (t), s0 (t))
◮ n(t) is AWGN, and h(t) is LTI
◮ s0 ∗ h(T0 ) < s1 ∗ h(T0 ); output 1 if Zk (T0 ) ≥ γ, and output 0,
otherwise.

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310A Tutorial 9 6 / 11
General binary communication systems
Determine probability of error for s1 (t):
◮ Given (s1 (t), h(t), T0 , γ), compute Pe,1 :
◮ Formula: ( )
γ − s1 ∗ h(T0 )
Pe,1 = Φ
σ
⋆ Tasks: calculate 1) s1 ∗ h(T0 ), and 2) σ
⋆ σ 2 is the variance of n ∗ h(T0 ), i.e., the noise component of the output
of filter, and
∫ ∞
N0 N0
σ2 = ||h||2 = |h(t)|2 dt
2 2 −∞

Determine probability of error for s0 (t):

◮ Given (s0 (t), h(t), T0 , γ), compute Pe,0 :
◮ Formula: ( )
γ − s0 ∗ h(T0 )
Pe,0 = Q
σ
. . . . . .
zwwong@link.cuhk.edu.hk IERG2310A Tutorial 9 7 / 11
General binary communication systems

Determine optimal decision threshold γ ∗ :

◮ Minimax threshold:
s1 ∗ h(T0 ) + s0 ∗ h(T0 )
γ∗ =
2
◮ Under γ ∗ , the probability of error becomes
( )
s0 ∗ h(T0 ) − s1 ∗ h(T0 )
Pe = Pe,1 = Pe,0 = Φ
2σ

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310A Tutorial 9 8 / 11
General binary communication systems
Determine optimal filter h(t):
◮ Matched filter:
Under the minimax threshold γ ∗ , the optimal h(t) is

h∗ (t) = λ [s1 (T0 − t) − s0 (T0 − t)]

◮ If the minimax threshold and matched filter are used, the choice of
sampling time T0 does not matter, and
∫
1 1
γ∗ = s21 (t) − s20 (t)dt = (E1 − E0 )
2 2
(√ ) (√ )
||s1 −s0 ||2 Ē(1−r)
Pe = Q 2N0 =Q N0

∫
1
where Ē = 12 (E1 + E0 ) and r = s1 (t)s0 (t)dt
Ē

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310A Tutorial 9 9 / 11
General binary communication systems

Exercise:

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310A Tutorial 9 10 / 11
General binary communication systems
Solution:
(a)
∫ T ∫ T ∫ T
2
E1 = B 2 dt = B 2 T, and E0 = A2 dt + (−A)2 dt = A2 T
0 0 T
2

(b) Let T0 = T , then the matched filter:

{
λ(B − A), 0 ≤ t < T2
h(t) = λ[p1 (T − t) − p0 (T − t)] =
λ(B + A), T
2
≤t<T
{
(A,B)=(1,2) λ, 0 ≤ t < T2
=========⇒ h(t) =
3λ, T
2
≤t<T

(c)
1 (A2 +B 2 )T
Ē = 2
(E1 + E0 ) = 2
∫ ∫ T ∫ T
1 1( 2 )
r= p1 (t)P0 (t)dt = ABdt + −ABdt = 0
Ē Ē 0 T
2
(√ ) (√ )
Ē(1−r) (A2 +B 2 )T
⇒ Pe = Q N
=Q 2N
0 0

. . . . . .
zwwong@link.cuhk.edu.hk IERG2310 Tutorial 9 11 / 11
 IERG2310 Tutorial 7

ZHONG Xiangyu

March 12, 2025

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 1 / 12
 Part I
Pulse code modulation (quantization part)

◮ Uniform quantization
⋆ Concepts: Quantization noise, bit rate and bandwidth
⋆ Variations: Non-uniform quantization

◮ Differential PCM (DPCM)

⋆ Simplest case: Delta modulation

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 2 / 12
Pulse Code Modulation

A PCM modulator consists of the following parts:

Sampler: Samples the amplitude of a continuous-time signal at some

discrete time instants
Quantizer: Round off the sampled values of the amplitude into finite
discrete levels
Encoder: Designates each discrete level by a digital code

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 3 / 12
Uniform Quantization
Suppose the amplitude of x(t) ∈ [−xmax , xmax ], we divide the total range (2xmax )
uniformly into N = 2ν regions
The length of each quantization region is ∆ = 2xmax /N
The quantized value xQ ,n is chosen to be the mid-point of each region.
Quantization noise power — the power of average quantization error:
[ ]
E (X − XQ )2
If x(t) is uniformly distributed in [−xmax , xmax ], then
[ ] ∆2
E (X − XQ )2 =
12
and the signal to quantization noise ratio
x2max /3
SQNR = = 4ν SQNRdB ≈ 6ν
∆2 /12
where SQNRdB = 10 log10 SQNR

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 4 / 12
Bit rate and bandwidth of PCM

Suppose the singal bandwidth is W , the sampling rate is fs (> 2W ) and each
sample is translated into a ν-bits code. Then
The bit rate is νfs bits/sec.
If ℓ bits/sec/Hz transmission is permitted, the required bandwidth for PCM
signals is νfs /ℓ Hz.

- Exercise: Consider a signal is band-limited to 3 kHz and converted into a

binary PCM code using 128 equally quantized levels.
a. If the signal amplitude is uniformly distributed within its range,
determine the dB of signal to quantization noise ratio.
b. If 2 bits/sec/Hz transmission is permitted, determine the minimum
required bandwidth.

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 5 / 12
Bit rate and bandwidth of PCM

- Exercise: Consider a signal is band-limited to 3 kHz and converted into a

binary PCM code using 128 equally quantized levels.
a. If the signal amplitude is uniformly distributed within its range,
determine the dB of signal to quantization noise ratio.
b. If 2 bits/sec/Hz transmission is permitted, determine the minimum
required bandwidth.

- Solution:
a. Since each sample is represented by ν = 7 bits, we have
SQNRdB = 6ν = 42.
b. Since the minimum sampling rate (Nyquist rate) is 6 kHz, the
transmission bandwidth required is (6 × 7)/2 = 21 kHz.

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 6 / 12
Non-uniform Quantization
In speech communication, voice amplitude is not uniformly distributed. Smaller
amplitudes (statistically) predominate in speech while larger amplitudes are
relatively rare.
If uniformly quantize, many quantized levels are rarely used (wasteful !)
Idea: Assign more quantized levels for small x and fewer levels for large x

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 7 / 12
Optimal Non-uniform Quantization
Given [xmin , xmax ] and fX (x) of samples,
determine the optimal quantization regions: [x0 , x1 , · · · , xN ] and quantized
values [xQ,1 , xQ,2 , · · · , xQ,N ] that minimizes the quantization noise power:

Via Lloyd-Max quantizer (an iterative approach):

1. Initialize [x0 , x1 , · · · , xN ]
2. Update [xQ,1 , xQ,2 , · · · , xQ,N ] according to

∫ xn
x
xfX (x)dx
xQ,n = ∫ n−1
xn , ∀n.
f (x)dx
xn−1 X

3. Update [x0 , x1 , · · · , xN ] according to

xQ,n + xQ,n+1
xn = , ∀n.
2
4. Repeat steps (2) and (3) until convergence.

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 8 / 12
Differential PCM (DPCM)

If the sampling rate is fast enough, most source signals show significant
correlation between successive samples.

Unlike PCM that quantizes each sample independently, DPCM quantizes the
difference between the actual sample value and its predicted value, taking
the advantage of
|d(k)| = |x(k) − xq (k)| << |x(k)|
Minimize redundant transmission and reduce the bandwidth.

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 9 / 12
Differential PCM (DPCM)

DPCM compression depends on the prediction technique.

In the simplest case, the predictor here can just be a delay element.

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 10 / 12
Delta Modulation
The simplest variant of DPCM, which is also viewed as 1-bit DPCM scheme.
Principle operation:
◮ Message signal is over-sampled, i.e., higher than the Nyquist rate –

aims at increasing the correlation between adjacent samples.

◮ Difference between the input and the staircase approximation is

quantized into two levels (+/− ∆)

◮ Provided signal does not change very rapidly. The approximation

remains within +/− ∆

. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 11 / 12
Delta Modulation

DM system:

Comparator — computes difference between input signal and one interval

delayed version of itself.
Accumulator — produces the approximation mq [n] (final result) at each
step by adding either +∆ or −∆.
. . . . . .
LIANG Zezu (lz017@ie.cuhk.edu.hk) IERG2310A Tutorial 7 12 / 12
 Part II

• Frequency Division Multiplexing

• Time Division Multiplexing

• Gaussian Random Variables

• Signaling Format: NRZ signaling

Transmission
• Multiplexing
• Allow multiple transmissions simultaneously (without affecting each other)

• TDM
• FDM
• (CDMA, etc)
Frequency Division Multiplexing

✓ Enlarge the
system capacity
✓ Less time
consuming
Time Division Multiplexing

5 1
11
Gaussian Random Variables
• A Gaussian random variable is any continuous random variable with a probability density function
of the form
1 𝑥−𝜇 2
−
𝑓𝑋 𝑥 = 𝑒 2𝜎2
2𝜋𝜎 2
where 𝜇 and 𝜎 2 are constants.
• The mean and the variance of 𝑋 are 𝜇 and 𝜎 2 respectively. Hence, a Gaussian random variable is
completely specified by its mean and variance.
• For any real constant 𝑎 ≠ 0, 𝑎𝑋 is also a Gaussian random variable with mean and variance 𝑎𝜇
and 𝑎2 𝜎 2 respectively.

6 1
Zero-mean unit-variance Gaussian random variable

• It has the density function

1 𝑥2
𝑍 𝑥 = 𝑒− 2
2𝜋
• The probability distribution function is given by
𝑥
Φ 𝑥 = න 𝑍(𝑢) d𝑢
−∞
• Very often, it is convenient to use the 𝑄-function defined by
∞
𝑄 𝑥 = න 𝑍(𝑢) d𝑢
𝑥
• Notice that Φ 𝑥 is increasing while 𝑄 𝑥 is decreasing.
Φ 𝑥 +𝑄 𝑥 =1

7 1
Signaling Format
fi
Signaling Format
•
•
•
•
NRZ signaling
• Define the rectangular pulse
1, 0 ≤ 𝑡 < 𝑇,
𝑝𝑇 𝑡 = ቊ
0, 𝑒𝑙𝑠𝑒.
• For 𝐴 > 0, 𝐴𝑝𝑇 𝑡 represents “1”, −𝐴𝑝𝑇 𝑡 represents “0”.
• We assume that the channel is an additive white Gaussian noise
(AWGN) channel.
Thank you!
IERG2310-ESTR2300 Tutorial 6
Principles of Communication Systems
SHI Jiajia
sj022@ie.cuhk.edu.hk
Feb 26, 2025
 2

Contents
✭ Sampling Theorem
✭ Quantization
➢ Uniform Quantization
• An Example of Signal-to-Quantization Noise Ratio (SQNR)
➢ Non-uniform Quantization

✭ Pulse Code Modulation (PCM)

➢ Uniform PCM
➢ Non-uniform PCM
• 𝜇-law and 𝐴-law Companding
➢ Encoding Scheme

✭ Take-Home Assignment 3 (Q9 - Q12)

 3

Sampling Theorem
➢ Principle

➢ Explanation
• Time domain:
+∞ +∞
𝑥∆ 𝑡 = 𝑥(𝑡) ෍ 𝛿 𝑡 − 𝑛𝑇𝑠 = ෍ 𝑥 𝑛𝑇𝑠 𝛿(𝑡 − 𝑛𝑇𝑠 )
−∞ −∞ Aliasing
• Frequency domain:
+∞ +∞
1 1 1 1
𝑋∆ 𝑓 = 𝑋 𝑓 ∗ ෍𝛿 𝑓 − 𝑛 = ෍ 𝑋(𝑓 − 𝑛 )
𝑇𝑠 𝑇𝑠 𝑇𝑠 𝑇𝑠
−∞ −∞
 4

Quantization
➢ Concept
• The process of mapping input values from a large set (often a continuous set) to output values
in a (countable) smaller set, often with a finite number of elements.
• A device or algorithmic function that performs quantization is called a quantizer.
E.g., an analog-to-digital converter (ADC).

𝒙 ∈ 𝓧𝒏 Quantizer (𝑸) ෡𝒏
ෝ∈𝓧
𝒙

• E.g., Suppose that the input source information is a real signal. Thus, the output is also the set
of some real numbers and is defined as, 𝑅1 , 𝑅2 , … , 𝑅𝑁 , as follows,

𝑥𝑖 → 𝑥ො𝑘 𝑅𝑘 = [𝑎𝑘 , 𝑎𝑘+1 ]

where 𝑁 is called the quantized level number.

 5

Quantization

➢ Not invertible
• Quantization is not invertible. Thus, some information is lost in the quantization process, and
they are not recoverable.
• The difference between an input value and its quantized value is referred to as quantization
error. This distortion is often called quantization noise.

➢ Types
• Can be classified into uniform quantization and non-uniform quantization.
• To make efficient use of levels, non-uniform quantization is employed.

➢ Differences between the two types

• Uniform quantization has equal step sizes (∆) .
• While, in non-uniform quantization, the step sizes (∆) are not equal.
 6

Quantization
➢ Signal-to-Quantization-Noise Ratio (SQNR)
𝐸[𝑥 2 ] 𝐸[𝑥 2 ] 𝑃𝑥
𝑆𝑄𝑁𝑅 = = = .
𝑬[𝒙෥𝟐 ] 𝑬 𝑿 − 𝑸 𝒙 𝟐 𝑃𝑥෤
• Where 𝑄(⋅) is quantization function, and
1 𝑇/2
𝑃𝑥 = 𝑙𝑖𝑚 න 𝐸 𝑥 2 𝑡 𝑑𝑡 ,
𝑇→∞ 𝑇 −𝑇/2

1 𝑇/2 2
𝑃𝑥෤ = 𝑙𝑖𝑚 න 𝐸 𝑥 𝑡 −𝑄 𝑥 𝑡 𝑑𝑡 .
𝑇→∞ 𝑇 −𝑇/2

➢ Example #1
• Consider that 𝑥 ∈ 𝒳 𝑛 is uniformly quantized to 𝒳෠ 𝑛 . Determine its SQNR (Suppose that 𝑥 is
uniformly distributed).
 7

Quantization
➢ Solution
• Assume that we choose a symmetric interval as the quantization range, i.e.,
𝑎1 , 𝑎𝑁 = −𝑥𝑚𝑎𝑥 , 𝑥𝑚𝑎𝑥 .
• Further assume the quantized level number is 𝑁 = 2𝜈 , where 𝜈 is quantized bit number.
• If we choose the middle point of each interval as the quantized level, i.e.,
𝑥Ƹ𝑘 = (𝑎𝑘 + 𝑎𝑘+1 )/2.
∆ ∆
• Thus, quantization error 𝑥෤ = 𝑥 − 𝑄(𝑥) is a random variable that evaluates in (− 2 , 2].
• Hence, we get:
𝟐𝒙𝒎𝒂𝒙 𝒙𝒎𝒂𝒙
∆/2 ∆= = 𝝂−𝟏
21 ∆2 𝑵 𝟐
2
𝑥𝑚𝑎𝑥
𝐸 𝑥෤ 2 = න 𝑥෤ 𝑑 𝑥෤ =
−∆/2 ∆ 12 3 ⋅ 4𝜈
𝑺𝑸𝑵𝑹 = 𝐸 𝑥 2 /𝐸 𝑥෤ 2 = 3 ⋅ 4𝜈 ⋅ 𝐸 𝑥ҧ02 = 𝑷𝒙ഥ𝟎 + 𝟔𝝂 + 𝟒. 𝟖 𝒅𝑩
2
𝑥 1 𝑇/2 𝑥 𝑡
𝑥ҧ0 = , 𝑃𝑥ҧ0 = 𝑙𝑖𝑚 න 𝐸 𝑑𝑡
𝑥𝑚𝑎𝑥 𝑇→∞ 𝑇 −𝑇/2 𝑥𝑚𝑎𝑥
SQNR can be improved by 6dB for every 1 bit increase in quantization.
 8

Pulse Code Modulation (PCM)

A technique of digital transmission for analogue signals.

 9

Uniform PCM
Sampling-Quantizing-Encoding

➢ Block Diagram

Fig. Block diagram of uniform PCM

• Sampling: 𝑓𝑠 = 2𝐵.
• Quantizing: We have 𝑁 quantized levels in uniform quantization, 𝑁 = 2𝜈 .
• Encoding: Each quantized level corresponds to one code. For 𝑁 quantized levels, number of
bits per code is 𝜈 bits, ν = 𝑙𝑜𝑔2 𝑁.
• The total rate required to transmit the PCM signal is 𝑅 = 𝜈𝑓𝑠 (𝑏𝑖𝑡𝑠/𝑠).
 10

Non-uniform PCM
Using Compressor and Expander

➢ Block Diagram
Non-linear Unit

Opposite Non-linear Unit

Fig. Block diagram of non-uniform PCM

• E.g., Speech signal (the probability of small amplitude is bigger than that of large amplitude, and
the former has more information).
• Hence, we can design more quantization intervals for small amplitude signals while less for large
amplitude signals.
 11

Non-uniform PCM
Using Compressor and Expander

➢ Compressor and Expander

• Compressor: To compress the signals and decrease the range of the signal’s amplitude at the
transmitting end.
• Expander: At the receiving end, it performs the opposite non-linear operation to get the sampled
values.

• Companding: The technique of compressing and expanding is called companding.

 12

Non-uniform PCM
𝝁-law and 𝑨-law Companding

➢ 𝝁-law Companding
𝑙𝑜𝑔(1 + 𝜇|𝑥|)
𝑦=𝑔 𝑥 = 𝑠𝑔𝑛 𝑥 , 𝑥 < 1,
𝑙𝑜𝑔(1 + 𝜇)
• Where 𝜇 is the companding parameter.
• Commonly used in the U.S. and Canada
• Standard PCM: 𝜇 = 255, and 𝑁 = 27 = 128.
• Implementation: 15-fold approach
1 2 3 4 5 6 7
𝑌: 𝑦𝑖 = 0, , , , , , , , 1,
8 8 8 8 8 8 8
𝑦𝑖 − 𝑦𝑖−1
256𝑦−1 256𝑖/8 −1 2𝑖−1 𝑘𝑖 =
𝑥𝑖 − 𝑥𝑖−1
, 𝑖 = 1,2, … , 8
𝑋: 𝑥𝑖 = = = .
255 255 255

It has 7 + 1 𝟏 + 𝟏 + 7 = 15 lines in total.

 13

Non-uniform PCM
𝝁-law and 𝑨-law Companding

➢ 𝑨-law Companding
1 + log 𝐴|𝑥|
𝑦=𝑔 𝑥 = 𝑠𝑔𝑛 𝑥 , 𝑥 < 1,
1 + log 𝐴
• Commonly used in China and Europe.
• Standard PCM: 𝐴 = 87.5, and 𝑁 = 27 = 128.
• Implementation: 13-fold approach
1 1 1 1 1 1 1
𝑋: 𝑥𝑖 = 0, , , , , , , , 1,
128 64 32 16 8 4 2
1 2 3 4 5 6 7
𝑌: 𝑦𝑖 = 0, , , , , , , , 1. 𝑦𝑖 − 𝑦𝑖−1
8 8 8 8 8 8 8 𝑘𝑖 = , 𝑖 = 1,2, … , 8
𝑥𝑖 − 𝑥𝑖−1

It has 6 + 1 𝟏 + 𝟏 + 𝟏 + 𝟏 + 6 = 13 lines in total.

 14

Pulse Code Modulation (PCM)

➢ Encoding Scheme
• NBC: Natural binary codes.
• FBC: Folded binary codes.
Easy to decode, and more robust for small signals compared with NBC.

𝑥(𝑡) NBC FBC 𝑖𝑡ℎ quantized level

Positive ⋮ ⋮ ⋮
Polarity

Negative ⋮ ⋮ ⋮
Polarity
 15

Take-Home Assignment 3 (Q9)

➢ Solution: (Hint: Page 4 of Chapter 4)

1002
a. Transmitted power: 2 = 5000
b. Peak phase deviation (modulation index): 𝛽𝑝 = 𝑘𝑝 max 𝑚 𝑡 = max |4sin(2000𝜋𝑡)| = 4
𝑘𝑓 max 𝑚 𝑡 1 𝑑
c. Peak frequency deviation: 𝛽𝑓 = 𝑊
= max 2𝜋 𝑑𝑡
4 sin 2000𝜋𝑡 = 4kHz
d. Both are possible, depending on the interpretation of the message signal.
4
If 𝑢(𝑡) is a PM signal, 𝑘𝑝 𝑚 𝑡 = 4 sin 2000𝜋𝑡 , and thus 𝑚 𝑡 = 𝑘 sin(2000𝜋𝑡)
𝑝
𝑡 4000
If 𝑢(𝑡) is a FM signal, 2𝜋𝑘𝑓 ‫׬‬0 𝑚 𝑡 𝑑𝑢 = 4sin(2000𝜋𝑡), and thus 𝑚 𝑡 = 𝑘𝑓
cos(2000𝜋𝑡)
 16

Take-Home Assignment 3 (Q10)

➢ Solution: (Hint: Page 7 & 8 of Chapter 4)

A general FM signal can be alternately written as
∞
𝑠 𝑡 = 𝐴 ෍ 𝐽𝑛 𝛽 cos[2𝜋 𝑓𝑐 + 𝑛𝑓𝑚 𝑡]
𝑛=−∞
where cos(2𝜋 𝑓𝑐 + 𝑛𝑓𝑚 𝑡) with 𝑛 = 0 refers to the carrier frequency.
To satisfy no transmitted power at carrier frequency, 𝐽02 𝛽 = 0 should hold. Thus, the smallest
𝛽 ≈ 2.4
 17

Take-Home Assignment 3 (Q11)

➢ Solution:
In order to see the amplitude and phase of various frequency components of the provided PM
signal, we need to represent the signal as a sum-of-cos functions form:
∞
𝑠 𝑡 = ෍ 𝐴𝑘 cos[ 2𝜋𝑓𝑐 + 𝑓𝑘 𝑡 + 𝜃𝑘 ]
𝑘=−∞
Specifically, it goes through the following steps:
𝑠 𝑡 = 𝐴cos 2𝜋𝑓𝑐 𝑡 + 𝑚 𝑡 = Re 𝐴𝑒 𝑗 2𝜋𝑓𝑐 𝑡+𝑚 𝑡 = Re[𝐴𝑒 𝑗2𝜋𝑓𝑐 𝑡 𝑒 𝑗𝑚 𝑡 ]
Since 𝑚(𝑡) is periodic, 𝑒 𝑗𝑚 𝑡 is also periodic and its Fourier series is given by
 18

Take-Home Assignment 3 (Q11) (Cond’t)

➢ Solution:
𝑇𝑚 𝑇𝑚 0, 𝑛 𝑖𝑠 𝑒𝑣𝑒𝑛
1 2 2 −1 𝑛−1
𝑐𝑛 = න 𝑒 𝑗 𝑒 −𝑗2𝜋𝑛𝑓𝑚 𝑡 𝑑𝑡 + න 𝑒 −𝑗 𝑒 −𝑗2𝜋𝑛𝑓𝑚 𝑡 𝑑𝑡 = − 𝑒 𝑗 − 𝑒 −𝑗 = ቐ2 sin 1
𝑇𝑛 0 𝑇𝑚 2𝑗𝜋𝑛 , 𝑛 𝑖𝑠 𝑜𝑑𝑑
2 𝜋𝑛

Hence,
∞ ∞
2 sin 1
𝑠 𝑡 = Re 𝐴𝑒 𝑗2𝜋𝑓𝑐 𝑡 ෍ 𝑐𝑛 𝑒 𝑗2𝜋𝑛𝑓𝑚 𝑡 = Re 𝐴 ෍ 𝑒 𝑗2𝜋 𝑓𝑐 + 2𝑘+1 𝑓𝑚 𝑡
𝜋(2𝑘 + 1)
𝑛=−∞ 𝑘=−∞
∞ ∞
2 sin 1 2 sin 1
=𝐴 ෍ cos[2𝜋 𝑓𝑐 + 2𝑘 + 1 𝑓𝑚 𝑡] = 𝐴 ෍ cos[2𝜋 𝑓𝑐 + 2𝑘 + 1 𝑓𝑚 𝑡 + 𝜃𝑘 ]
𝜋(2𝑘 + 1) 𝜋(2𝑘 + 1)
𝑘=−∞ 𝑘=−∞

Where the phase and amplitude of the frequency component 𝑓𝑐 + 2𝑘 + 1 𝑓𝑚 are:

0, 𝑘 ≥ 0 2 sin 1
𝜃𝑘 = ቊ and 𝐴
𝜋, 𝑘 < 0 𝜋(2𝑘+1)
 19

Take-Home Assignment 3 (Q12)

 20

Take-Home Assignment 3 (Q12)

➢ Solution:
a. Maximum instantaneous frequency: max 𝑓𝑐 + 𝑘𝑓 𝑚1 𝑡 = 106 + 5 × 105 = 1.5 MHz
𝑘𝑝 𝑑 3
b. Maximum instantaneous frequency:max 𝑓𝑐 + 2𝜋 𝑑𝑡 𝑚1 𝑡 = 106 + 2𝜋 × 105 = 1.05 MHz
𝑘 𝑑 3
Minimum instantaneous frequency: min 𝑓𝑐 + 2𝜋𝑝 𝑑𝑡 𝑚1 𝑡 = 106 − 2𝜋 × 105 = 0.95 MHz
c. Maximum instantaneous frequency: max 𝑓𝑐 + 𝑘𝑓 𝑚2 𝑡 = 106 +103 × 𝑚2 0 = 1.001MHz
By the Fourier transform pair

1
1 𝑓 1, 𝑥 < 2
sinc 𝑎𝑡 𝑎
rect 𝑎
where rect(𝑥) = ൞ 1
0, 𝑥 ≥ 2
1
We know that 𝑚2 (𝑡) is bandlimited within 𝑓𝑚 = 2 × 2 × 104 = 10 kHz. Since 𝑘𝑓 = 103 , according to
the Carson’s Rule for FM bandwidth, the bandwidth of the modulated signal is
𝐵 = 2 ∆𝑓 + 𝑓𝑚 = 2 𝑘𝑓 max 𝑚2 𝑡 + 𝑓𝑚 = 2 × 103 + 104 = 22 kHz
where ∆𝑓 is the maximum frequency deviation of an FM signal.
 IERG2310 Tutorial 5

Wang Zewei

zwwong@link.cuhk.edu.hk

. . . . . .
Wang Zewei IERG2310 Tutorial 5 1 / 14
Outline

Test Checklist

◮ Amplitude Modulation
⋆ Power of frequency components, modulation index
⋆ Use block diagram to generate a desired AM signal

◮ Angle Modulation
⋆ Instantaneous frequency, peak phase/frequency deviation, modulation
index
⋆ Bessel functions representation and effective bandwidth of FM
⋆ Use block diagram to generate a desired FM signal

. . . . . .
Wang Zewei IERG2310 Tutorial 5 2 / 14
Amplitude Modulation

AM signal:
s(t) = [Ac + am(t)] cos(2πfc t)
where Ac is the maximum amplitude of carrier signal, m(t) is the message
signal, and fc is the carrier frequency.

- Frequency spectrum:
F Ac a
s(t) ⇐⇒ [δ(f − fc ) + δ(f + fc )] + [M (f − fc ) + M (f + fc )]
2 2
- Power of each component: Consider m(t) = cos(2πf1 t),
Ac 2 A2
Pfc = ( ) ×2= c
2 2
a 2 a2
Pfc −f1 = Pfc +f1 = ( ) × 2 =
4 8

. . . . . .
Wang Zewei IERG2310 Tutorial 5 3 / 14
Amplitude Modulation

AM signal:
s(t) = [Ac + am(t)] cos(2πfc t)
- Modulation index: The radio between the maximum amplitude of the
message signal and that of the carrier

Consider m(t) = 0.5 cos(2πf1 t) + cos(4πf1 t).

1. The maximum amplitude of am(t) is Am = a(0.5 + 1) = 1.5a

2. The maximum amplitude of carrier is Ac

3. Modulation index= Am = 1.5a

Ac Ac
4. To ensure that m(t) can be demodulated by envelope detector, we
1.5a 2Ac
have ≤ 1 and get a ≤
Ac 3

. . . . . .
Wang Zewei IERG2310 Tutorial 5 4 / 14
Amplitude Modulation
Example 1:

Solution: We know that cos(2πf0 t) cos(2πfc t) can generate 2 frequency

components (i.e., fc + f0 and fc − f0 ):
F 1
cos(2πf0 t) cos(2πfc t) ⇐⇒
= [δ(f +fc +f0 ) + δ(f +fc −f0 ) + δ(f −fc +f0 ) + δ(f − fc − f0 )]
4

The magnitude spectrum of s(t) consists of 5 frequency components:

1 a
|S(f )| = [δ(f −fc )+δ(f +fc )]+ [δ(f +fc +f1 )+δ(f +fc −f1 )+δ(f −fc +f1 )+δ(f − fc −f1 )]
2 2
a
+ [δ(f +fc +f2 )+δ(f +fc −f2 )+δ(f −fc +f2 )+δ(f − fc −f2 )]
4

. . . . . .
Wang Zewei IERG2310 Tutorial 5 5 / 14
Amplitude Modulation
Example 1:

Solution: a. The power of each frequency component:

1 1
Pfc = ( )2 × 2 =
2 2
a2 2 a2
Pfc +f1 = Pfc −f1 = ( ) ×2=
2 2
a2 a2
Pfc +f2 = Pfc −f2 = ( )2 × 2 =
4 8
1
b. Since Ac = 1 and Am = 3a, the maximum value of a is 3 so that s(t)
can be demodulated by an envelop detector.
. . . . . .
Wang Zewei IERG2310 Tutorial 5 6 / 14
Amplitude Modulation
Example 2:

Solution:
a. The magnitude spectrum of the SSB signal s(t) is

We can see that s(t) has frequencies fc + f1 and fc + f2 .

A2 9A2
b. Pfc +f1 = 2 , Pfc +f2 = 2 .
. . . . . .
Wang Zewei IERG2310 Tutorial 5 7 / 14
Amplitude Modulation
Use block diagram to generate a desired AM signal

Example 3:

Solution:

. . . . . .
Wang Zewei IERG2310 Tutorial 5 8 / 14
Angle Modulation
AM: use the angle of a carrier to carrier information:
s(t) = A cos [2πfc t + ∅(t)]
- If s(t) is PM, ∅(t) = kp m(t);
∫t
- If s(t) is FM, ∅(t) = 2πkf 0 m(u)du or 2π dt ∅(t)
1 d
= kf m(t)

Instantaneous frequency:
{
k d
1 d fc + 2πp dt m(t), PM
fi (t) = [2πfc t + ∅(t)] =
2π dt fc + kf m(t), FM
Maximum phase deviation= max |∅(t)|

2π dt ∅(t) ⇒ kf max |m(t)|(if FM)

1 d
Maximum frequency deviation= max
Modulation index:

βp = kp max |m(t)|, PM
βf = kf max |m(t)| , FM
W
. . . . . .
Wang Zewei IERG2310 Tutorial 5 9 / 14
Angle Modulation

A general FM signal can be represented as (Page 10, Chapter 4)

∞
∑
s(t) = A Jn (β) cos(2π(fc + nfm )t)
n=−∞

where β is the modulation index of FM.

- Infinite number of frequency component fc + nfm , n = 0, ±1, ±2, ...
- Power of each component fc + nfm is A2 Jn2 (β)/2 1
∑∞
- Total power of the whole signal n=−∞ A2 Jn2 (β)/2 = A2 /2.

- Carson’ rule: the effective bandwidth of an FM:

B = 2(β + 1)W or B = 2(∆f + W )
where W is message bandwidth and ∆f is maximum frequency deviation.

1
Values of Jn (x) can be found by checking the table . . . . . .
Wang Zewei IERG2310 Tutorial 5 10 / 14
Angle Modulation
Example 4 (Assignment Set, P12): If m(t) = sinc(2 × 104 t) is frequency
modulated on a carrier with frequency 106 Hz and with kf = 1000.
- What is the maximum instantaneous frequency of the modulated signal?
- What is the bandwidth of the modulated signal?

Solution:
- max |fi (t)| = max |fc + kf m(t)| = 106 +103 ×m(0) = 1.001 MHz
- According to the Fourier transform pair:
( ) {
F 1 f 1, |x| < 1
sinc(at) ⇐⇒ rect , with rect(x) , 2 ,
|a| a 0, |x| ≥ 1
2

m(t) is bandlimited to W = 21 ×2×104 = 10 kHz. Using Carson’s rule, the

bandwidth of the modulated signal with kf = 103 is
B = 2(∆f + W ) = 2(kf max |m(t)| + W ) = 2(103 + 104 ) = 22kHz
where ∆f is the maximum frequency deviation.

. . . . . .
Wang Zewei IERG2310 Tutorial 5 11 / 14
Angle Modulation

Generation of FM
A block diagram for generating an FM usually includes the following blocks:

- Frequency multiplier: scale the whole 2πfc t + ∅(t).

- Oscillator+product modulator: frequency shifting of the carrier.
- Bandpass filter: filter out the unwanted frequency caused by the frequency
modulation.

. . . . . .
Wang Zewei IERG2310 Tutorial 5 12 / 14
Angle Modulation

Example 5:

. . . . . .
Wang Zewei IERG2310 Tutorial 5 13 / 14
Angle Modulation

How to generate s(t) from u(t)?

u(t) = cos[2π(106 )t + 0.2 sin(2πfm t)]
s(t) = A cos[2π(10 × 106 )t + 4 sin(2πfm t)]

Solution:

- u1 (t) = cos[2π(10 × 106 )t + 2 sin(2πfm t)]

- u2 (t) = cos[2π(5 × 106 )t + 2 sin(2πfm t)]
- u3 (t) = cos[2π(10 × 106 )t + 4 sin(2πfm t)]

. . . . . .
Wang Zewei IERG2310 Tutorial 5 14 / 14