ELEC50002

IMPERIAL COLLEGE LONDON

DEPARTMENT OF ELECTRICAL AND ELECTRONIC ENGINEERING

EXAMINATIONS 2024

ExamHeader:

COMMUNICATIONS

Monday, 29 April 2:30 pm

Time allowed: hours

There are SIX questions on this paper.

Answer ALL questions.

Any special instructions for invigilators and information for

candidates are on page 1.

Examiners responsible First Marker(s) : G. Li

Second Marker(s) : K.K. Leung

© Imperial College London

 Examination Questions
ELEC50002: Communications, April 2024
Dept. of Electrical and Electronic Engineering, Imperial College London

Instruction:
➢ This is a closed-book examination. But a standard calculator and three pages (8½” X11”) of
handwriting notes are allowed.
o Standard calculator: please check the EEE Dept, Imperial College London, for the
specifications of a standard calculator.
o Handwriting notes: the notes must be in your original handwriting on three pieces of
standard letter size paper (one side only). Please write your name and CID on the top left
on each page of the notes.
➢ Incorrect answers with justification and reasoning can receive partial credit. However, correct
answer without justification or reasoning will receive no credit.
➢ Students need to answer all 6 problems.

Prob I (10 pts)

8
Prob II (9 pts)

8
Prob III (9 pts)

8
Prob IV (11 pts)

8
Prob V (9 pts)

8
Prob VI (12 pts)

8
Total (60 pts)
 Some Fourier Transform Pairs

∞ ∞
𝑥(𝑡) = ∫ 𝑋(𝑓)𝑒 𝑗2𝜋𝑓𝑡 𝑑𝑓 𝑋(𝑓) = ∫ 𝑥(𝑡)𝑒 −𝑗2𝜋𝑓𝑡 𝑑𝑡
−∞ −∞

1 𝑥(𝑡) = 1 𝑋(𝑓) = 𝛿(𝑓)

2 𝑥(𝑡) = 𝛿(𝑡) 𝑋(𝑓) = 1

1, 𝑡≥0 1 1
3 𝑥(𝑡) = 𝑢(𝑡) = { 𝛿(𝑓) +
0, 𝑡<0 2 𝑗2𝜋𝑓
1, 𝑡≥0 1
4 𝑥(𝑡) = 𝑠𝑔𝑛(𝑡) = { 𝑋(𝑓) =
−1, 𝑡<0 𝑗𝜋𝑓
1 −𝑗, 𝑓≥0
5 𝑥(𝑡) = 𝑋(𝑓) = −𝑗𝑠𝑔𝑛(𝑓) = {
𝜋𝑡 𝑗, 𝑓<0
𝑡 1, |𝑡| ≤ 𝑇/2 sin⁡(𝜋𝑓𝑇)
6 𝑥(𝑡) = 𝑟𝑒𝑐𝑡 ( ) = { 𝑋(𝑓) = 𝑇 = 𝑇𝑠𝑖𝑛𝑐(𝑓𝑇)
𝑇 0, |𝑡| > 𝑇/2 𝜋𝑓𝑇
sin⁡(𝜋𝑊𝑡) 𝑓 1, |𝑓| ≤ 𝑊/2
7 𝑥(𝑡) = 𝑊 = 𝑊𝑠𝑖𝑛𝑐(𝑊𝑡) 𝑋(𝑓) = 𝑟𝑒𝑐𝑡 ( ) = {
𝜋𝑊𝑡 𝑊 0, |𝑓| > 𝑊/2
1 1
8 𝑥(𝑡)cos(2𝜋𝑓𝑜 𝑡) 𝑋(𝑓 − 𝑓0 ) + 𝑋(𝑓 + 𝑓0 )
2 2
1 1
9 𝑥(𝑡)sin(2𝜋𝑓𝑜 𝑡) 𝑋(𝑓 − 𝑓0 ) − 𝑋(𝑓 + 𝑓0 )
2𝑗 2𝑗

Fourier Transform Properties

Name of Theorem Function Fourier Transform

Superposition (𝑎1 and 𝑎2

1 𝑎1 𝑥1 (𝑡) + 𝑎2 𝑥2 (𝑡) 𝑎1 𝑋1 (𝑓) + 𝑎2 𝑋2 (𝑓)
are arbitrary constants)

2 Time delay 𝑥(𝑡 − 𝑡0 ) 𝑋(𝑓)𝑒 −𝑗2𝜋𝑓𝑡0

𝑓
3 Scale change 𝑥(𝑎𝑡) |𝑎|−1 𝑋 ( )
𝑎
∗ (𝑓),
𝑋(−𝑓) = 𝑋 * is complex
4 Time reversal 𝑥(−𝑡)
conjugate

5 Duality 𝑋(𝑡) 𝑥(−𝑓)

6 Frequency translation 𝑥(𝑡)𝑒 𝑗2𝜋𝑓𝑜 𝑡 𝑋(𝑓 − 𝑓0 )

1 1
7 Modulation 𝑥(𝑡) cos(2𝜋𝑓0 𝑡) 𝑋(𝑓 − 𝑓0 ) + 𝑋(𝑓 + 𝑓0 )
2 2
𝑑 𝑛 𝑥(𝑡)
8 Differentiation (𝑗2𝜋𝑓)𝑛 𝑋(𝑓)
𝑑𝑡 𝑛
 𝑡
1
9 Integration ∫ 𝑥(𝜏)𝑑𝜏 (𝑗2𝜋𝑓)−1 𝑋(𝑓) + 𝑋(0)𝛿(𝑓)
−∞ 2
∞
∫ 𝑥1 (𝑡 − 𝜏)𝑥2 (𝜏)𝑑𝜏
−∞
10 Convolution ∞ 𝑋1 (𝑓)𝑋2 (𝑓)
= ∫ 𝑥1 (𝜏)𝑥2 (𝑡 − 𝜏)𝑑𝜏
−∞
∞
∫ 𝑋1 (𝑓 − 𝜉)𝑋2 (𝜉)𝑑𝜉
11 Multiplication 𝑥1 (𝑡)𝑥2 (𝑡) −∞
∞
= ∫ 𝑋1 (𝜉)𝑋2 (𝑓 − 𝜉)𝑑𝜉
−∞

Some Trigonometry Formulas

sin(𝑥 + 𝑦) = sin 𝑥 cos 𝑦 + cos 𝑥 sin 𝑦 sin(𝑥 − 𝑦) = sin 𝑥 cos 𝑦 − cos 𝑥 sin 𝑦

cos(𝑥 + 𝑦) = cos 𝑥 cos 𝑦 − sin 𝑥 sin 𝑦 cos(𝑥 − 𝑦) = cos 𝑥 cos 𝑦 + sin 𝑥 sin 𝑦
1 1
sin 𝑥 cos 𝑦 = [sin(𝑥 + 𝑦) + sin(𝑥 − 𝑦)] cos 𝑥 sin 𝑦 = [sin(𝑥 + 𝑦) − sin(𝑥 − 𝑦)]
2 2
1 1
sin 𝑥 sin 𝑦 = [cos(𝑥 − 𝑦) − cos(𝑥 + 𝑦)] cos 𝑥 cos 𝑦 = [cos(𝑥 + 𝑦) + cos(𝑥 − 𝑦)]
2 2
𝑥+𝑦 𝑥−𝑦 𝑥+𝑦 𝑥−𝑦
sin 𝑥 + sin 𝑦 = 2 sin cos sin 𝑥 − sin 𝑦 = 2 cos sin
2 2 2 2
𝑥+𝑦 𝑥−𝑦 𝑥+𝑦 𝑥−𝑦
cos 𝑥 + cos 𝑦 = 2 cos cos cos 𝑥 − cos 𝑦 = −2 sin sin
2 2 2 2

𝑸 Function
∞ −𝑡 2
1
𝑄(𝑥) = ∫ 𝑒 2 𝑑𝑡
√2𝜋 𝑥

Approximation of 𝑸 Function
2 2
1 ∞ −𝑡 1 −𝑥
𝑄(𝑥) = ∫ 𝑒 2 𝑑𝑡 ≈ 𝑒 2 for ⁡𝑥 ≫ 1
√2𝜋 𝑥 √2𝜋𝑥

Joint Gaussian Density

The joint probability density function (pdf) of two correlated Gaussian random variables 𝑋 and 𝑌 is
given by
1 (𝑥−𝜇𝑋 )2 (𝑥−𝜇𝑋 )(𝑦−𝜇𝑌 ) (𝑦−𝜇𝑌 )2
1 −
2(1−𝜌2 )
[ 2
𝜎𝑋
⁡−⁡2𝜌
𝜎𝑋 𝜎𝑌
⁡+⁡
𝜎𝑌2
]
𝑓𝑥𝑦 (𝑥, 𝑦) = 𝑒
2𝜋𝜎𝑋 𝜎𝑌 √1 − 𝜌2
2
where 𝜇𝑋 = 𝐸[𝑋], 𝜇𝑌 = 𝐸[𝑌] are the mean values, 𝜎𝑋 = √𝐸 [(𝑋 − 𝜇𝑥 )2 ] and 𝜎𝑌 = √𝐸 [(𝑌 − 𝜇𝑦 ) ]
are the standard deviations of 𝑋 and 𝑌, respectively, and 𝜌 is the correlation coefficient defined as 𝜌 =
Cov(𝑋,𝑌)
𝜎 𝜎
.
𝑋 𝑌
Problem I: (Noise) Consider the canonical form of passband noise, 𝑁(𝑡) = 𝑁𝐼 (𝑡) cos(300𝜋𝑡) −
𝑁𝑄 (𝑡) sin(300𝜋𝑡), where N I (t ) and N Q (t ) are the corresponding in-phase and quadrature components
and are independent and wide-sense stationary (WSS) Gaussian random processes with zero mean and
the same power spectrum density. Denote the power spectrum density of N I (t ) and N Q (t ) as
1
, |𝑓| ≤ 4,
𝑆(𝑓) = {4
0, |𝑓| > 4.
a) [2 points] Find the variance of N I (t ) ;

b) [2 points] Find 𝐸(𝑁(𝑡));

c) [2 points] Find 𝐸(𝑁 2 (𝑡));

d) [2 points] Find the probability density function of N (0) ;

e) [2 points] Find 𝑆𝑁 (𝑓), the power spectrum density of N (t ) ;

Problem II. (Down-Conversion) The receiver block diagram of a communication system is shown as
in the following figure.

In the above figure, carrier frequency 𝑓𝑐 = 100 Hz and the received signal before the passband filter is
𝑥(𝑡) = cos(190𝜋𝑡) − sin⁡(230𝜋𝑡).
a) [2 points] Express 𝑥(𝑡) into the form, 𝑥(𝑡) = 𝑥𝑖 (𝑡) cos(2𝜋𝑓𝑐 𝑡) − 𝑥𝑄 (𝑡)sin⁡(2𝜋𝑓𝑐 𝑡), where 𝑥𝐼 (𝑡)
and 𝑥𝑄 (𝑡) are the in-phase and quadrature components corresponding 𝑓𝑐 = 100 Hz.

b) [3 points] Find 𝑦(𝑡), 𝑦𝐼 (𝑡), and 𝑦𝑄 (𝑡) if the frequency responses of the bandpass filter and the
lowpass filters are
1, if⁡|𝑓 − 100| ≤ 20⁡Hz⁡or⁡⁡|𝑓 + 100| ≤ 20⁡Hz, 1, if⁡|𝑓| ≤ 20⁡Hz,
𝐻BP (𝑓) = { and 𝐻LP (𝑓) = {
0, otherwise, 0, otherwise.

c) [2 points] Find 𝑦𝐼 (𝑡) and 𝑦𝑄 (𝑡) if the frequency responses of the bandpass filter and the lowpass
filters are
1, if⁡|𝑓 − 100| ≤ 20⁡Hz⁡or⁡⁡|𝑓 + 100| ≤ 20⁡Hz, 1, if⁡|𝑓| ≤ 10⁡Hz,
𝐻BP (𝑓) = { and 𝐻LP (𝑓) = {
0, otherwise, 0, otherwise.

d) [2 points] Find 𝑦𝐼 (𝑡) and 𝑦𝑄 (𝑡) if the frequency responses of the bandpass filter and the lowpass
filters are
1, if⁡|𝑓 − 100| ≤ 10⁡Hz⁡or⁡⁡|𝑓 + 100| ≤ 10⁡Hz, 1, if⁡|𝑓| ≤ 10⁡Hz,
𝐻BP (𝑓) = { and 𝐻LP (𝑓) = {
0, otherwise, 0, otherwise.
Problem III. (Analoge Modulation) Let 𝑚(𝑡) be a real message signal with 𝐸[𝑚2 (𝑡)] = 1, 𝑚𝑃 =
𝑚𝑎𝑥|𝑚(𝑡)| = 2, and bandwidth 𝑊 = 15⁡Hz. Denote 𝑠𝐷𝑆𝐵−𝑆𝐶 (𝑡) = 𝑚(𝑡)⁡𝑐𝑜𝑠⁡(1000⁡π𝑡) as the signal
of double sideband-suppressed carrier (DSB-SC) modulation.

a) [3 points] If the coherent detector in the above figure is used at the receiver, find carrier frequency
𝑓𝑐 , the expressions of the bandpass filter, 𝐻BP (𝑓), and the lowpass filter, 𝐻LP (𝑓).

b) [3 points] Find the signal-to-noise ratio (SNR) at the output of the coherent detector if the channel
𝑁
noise is white Gaussian with power spectral density (PSD), 𝑜 = 10−4 Watt/Hertz.
2

c) [3 points] If received signal⁡𝑟(𝑡)⁡in the above figure is added by 3 cos(1000⁡𝜋𝑡), then

𝑟̂ (𝑡) = 𝑟(𝑡) + 3 cos(1000⁡𝜋𝑡) = m(t)cos(1000𝜋𝑡) + n(t) + 3cos(1000⁡π𝑡).
Can we use an envelope detector to substitute the coherent detector in the figure to recover message
signal 𝑚(𝑡) from 𝑟̂ (𝑡)? State the reason if the answer is “No”; otherwise find the SNR at the output
of the envelope detector.
Problem IV (Digital Communication) The following digital communication system is used to
transmit speech signal 𝑚(𝑡) with 𝐸{𝑚2 (𝑡)} = 1, max|𝑚(𝑡)| = 2, and bandwidth of 3.5 kHz. Assume
that sampling frequency 𝑓𝑠 = 8 kHz and 16-level quantisation are used.

a) [3 points] Find the minimum sampling frequency in the above system to ensure accurate recovery
of 𝑚(𝑡) from the sampling sequence without distortion.

b) [3 points] If 16-level uniform quantisation is used, what is the SNR (due to quantisation noise) at
the quantiser output?

c) [3 points] What is the required data transmission rate at the output of the encoder (the encoder here
just converts the 16-level discrete signal into a binary data sequence, without adding any
redundancy)?

d) [2 points] If the data transmission rate is 50⁡kbits/sec and the binary digital modulation in the
figure is used, what is the symbol duration and the required channel bandwidth?
Problem V. (Digital Modulation) For ASK, the modulated signal can be expressed as
0,⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡⁡𝑖𝑓⁡𝑠𝑦𝑚𝑏𝑜𝑙⁡"0"⁡𝑖𝑠⁡𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑡𝑒𝑑,
𝑠(𝑡) = {
𝐴 cos(2π𝑓𝑐 𝑡) , 𝑖𝑓⁡𝑠𝑦𝑚𝑏𝑜𝑙⁡"1"⁡𝑖𝑠⁡𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑡𝑒𝑑,
where amplitude 𝐴 = 1 volt, symbol duration 𝑇 = 10 msec (0.01⁡sec), 𝑓𝑐 = 200 Hz.
a) [3 points] Draw the waveform of the modulated signal if the transmitted symbol is {1,0,0,1}.

b) [3 points] Find the probability of error if coherent detection is used and the power spectral density
𝑁
(PSD) of the channel noise is 𝑜 = 10−3 Watt/Hz.
2

c) [3 points] Find the probability of error if non-coherent detection is used and the PSD of the channel
𝑁
noise is 2𝑜 = 10−3 Watt/Hz.
Problem VI. (Information Theory) In the following figure, the discrete memoryless source (DMS)
randomly generates a symbol every second with Pr(𝑠 = 𝑎1 ) = 0.4 and Pr(𝑠 = 𝑎2 ) = ⁡ Pr(𝑠 =
𝑎3 ) = 0.3, the binary symmetric channel (BSC) transmits two (binary) symbols every second with
probability of error 𝑃𝑒 = 0.1, that is, Pr{𝑦 = 1|𝑥 = 0} = Pr{𝑦 = 0|𝑥 = 1} = 𝑃𝑒 = 0.1. The encoder in
the following figure converts source symbols into binary symbols through a code scheme. (Hint:
−0.1 log 2 0.1 = 0.3322, −0.2 log 2 0.2 = 0.4644, −0.3 log 2 0.3 = 0.5211, −0.4 log 2 0.4 = 0.5288,
−0.5 log 2 0.5 = 0.5000, −0.6 log 2 0.6 = 0.4422, −0.7 log 2 0.7 = 0.3602, −0.8 log 2 0.8 = 0.2575,
and −0.9 log 2 0.9 = 0.1368)

a) [3 points] Find the entropy of the DMS (bits/second).

b) [3 points] Find the Huffman coding of the DMS and its average length (per source symbol).

c) [3 points] Find the capacity of the BSC (bits/second)

d) [3 points] Are there an encoder and a decoder such that the probability of decoding error, Pr⁡{𝑠 ≠
𝑠̂ }, can be arbitrarily small? State the reason if the answer is “No”; otherwise design the encoder if
the answer is “yes”.