You are on page 1of 10

CONFIDENTIAL

FINAL EXAMINATION SEMESTER I


SESSION 2016/2017

COURSE CODE : SEL4223

COURSE NAME : DIGITAL SIGNAL PROCESSING 1

LECTURER : DR. MUSA B. MOHD MOKJI

PROGRAMME : SWB

SECTION : 01

TIME : 2.5 HOURS

DATE :

INSTRUCTION TO CANDIDATE :

SOLUTION

THIS SOLUTION BOOKLET CONSISTS OF 10 PAGES INCLUDING THE FRONT COVER


2
SKEL4223

PART A : ANSWER ALL QUESTIONS

QUESTION 1

(a) What is the difference between discrete-time signal and discrete-time system? (5 marks)

Answer
- Discrete signal is a series consisting of a sequence of quantities. In other words, it is
a series of a function over a domain of discrete integers. Each value in the sequence
is called a sample.
- Discrete system is an entity that changes or transforms a discrete signal into another
form of discrete signal based on specific transfer function

(b) Sketch 𝑦[𝑛] = 𝑥[2𝑛] when 𝑥[𝑛] = [1, −1, 1, −1, 1, −1]. (5 marks)

Answer
𝑦[𝑛] = [1,1,1]

(c) Given 𝑦1 [𝑛] = 𝑥[2𝑛] and 𝑦2 [𝑛] = ∑5𝑘=0 𝑥[𝑘]𝑥[𝑛 − 𝑘]. Which one of these systems is
an LTI system? Explain your answer. (5 marks)

Answer
𝑦2 [𝑛] = ∑5𝑘=0 𝑥[𝑘]𝑥[𝑛 − 𝑘] is the LTI system since this is a convolution operation.
3
SKEL4223

(d) An LTI system is described as 𝑦[𝑛] = 𝑦[𝑛 − 1] + 𝑥[𝑛]. Based on ‘bounded input
bounded output (BIBO)’ rule, show whether the system is stable when 𝑥[𝑛] = 𝑢[𝑛].
(5 marks)
Answer
- The input is bounded.
- The output is not bounded when 𝑛 → ∞
- Thus, the system is not stable.

(e) Give two usage of z-transform in signal processing and explain how each of it is
implemented. (5 marks)

Answer
1. To check stability and causality of an LTI system by identifying the ROC type and
position of the unit circle.
2. To obtain difference equation by manipulating inverse z-transform of the system
function.

(f) Determine the z-transform of signal 𝑥[𝑛] = [1,2,1] and sketch the corresponding pole-
zero plot. (5 marks)

Answer
𝑥[𝑛] = 𝛿[𝑛] + 2𝛿[𝑛 − 1] + 𝛿[𝑛 − 2]
𝐻(𝑧) = 1 + 2𝑧 −1 + 𝑧 −2
𝑧 2 + 2𝑧 + 1
=
𝑧2
4
SKEL4223

(g) List down five type of frequencies and their symbols from both continuous and discrete
domains. (5 marks)

Answer
1. Continuous signal frequency – 𝐹
2. Continuous signal angular frequency – Ω
3. Discrete signal frequency – 𝑓
4. Discrete signal angular frequency – 𝜔
5. Discrete frequency – 𝑘

(h) Given an impulse response ℎ[𝑛] = 𝑢[𝑛 + 1] − 𝑢[𝑛 − 2], find its magnitude frequency
response |𝐻(𝜔)|. (5 marks)

Answer
ℎ[𝑛] = 𝑢[𝑛 + 1] − 𝑢[𝑛 − 2] = 𝛿[𝑛 + 1] + 𝛿[𝑛] + 𝛿[𝑛 − 1]
𝐻(𝜔) = 𝑒 𝑗𝜔 + 1 + 𝑒 −𝑗𝜔 = 1 + 2 cos(𝜔)
|𝐻(𝜔)| = |1 + 2 cos(𝜔)

(i) Compute 𝑋[𝑘] when its time domain signal is 𝑥[𝑛] = [1, −1,1, −1]. (5 marks)

Answer
𝑗2𝜋𝑘𝑛
𝑋[𝑘] = ∑𝑥[𝑛]𝑒 − 𝑁

𝑋[0] = ∑𝑥[𝑛] = 0
𝑗𝜋𝑛 𝑗𝜋 𝑗3𝜋
𝑥[1] = ∑𝑥[𝑛]𝑒 − 2 = 1 − 𝑒 − 2 + 𝑒 −𝑗𝜋 − 𝑒 − 2 =1−𝑗−1+𝑗 = 0
𝑥[2] = ∑𝑥[𝑛]𝑒 −𝑗𝜋𝑛 = 1 − 𝑒 −𝑗𝜋 + 𝑒 −𝑗2𝜋 − 𝑒 −𝑗3𝜋 = 1 + 1 + 1 + 1 = 4
𝑥[3] = 𝑥[1] = 0
𝑋[𝑘] = [0,0,4,0]

(j) Given a 4-point DFT 𝑋[𝑘] = [0,0,2,0], what is the frequency value in 𝐻𝑧 for 𝑘 = 2.
Assume the frequency sampling is 𝐹𝑠 = 2000𝐻𝑧. (5 marks)

Answer
𝑁 = 4, 𝐹𝑠 = 2000𝐻𝑧
𝐹𝑆 𝑘 2000×2
𝐹= = = 1000𝐻𝑧
𝑁 4
5
SKEL4223

PART B : ANSWER 2 (TWO) QUESTIONS ONLY.

QUESTION 2
Below is a difference equation of a causal filter with variable K.
𝑦[𝑛] = 𝑥[𝑛] + 𝑥[𝑛 − 𝐾]

(a) Is the filter an LTI system? Explain.


(3 marks)
Answer
Yes, since it is represented by difference equation.

(b) Find 𝐻(𝑧) and state the ROC of the filter when 𝐾 = 0.
(6 marks)
Answer
𝑦[𝑛] = 𝑥[𝑛] + 𝑥[𝑛] = 2𝑥[𝑛]
𝑌(𝑧) = 2𝑋(𝑧)
𝑌(𝑧)
𝐻(𝑧) = =2
𝑋(𝑧)
Since this is a causal filter, then the 𝑅𝑂𝐶 is entire z-plane except 𝑧 = 0

(c) Sketch the signal flow graph of the system when 𝐾 = 2. From the signal flow graph,
determine impulse response ℎ[𝑛] of the system.
(5 marks)
Answer

𝑥[𝑛] ⊕ 𝑦[𝑛]

𝑧 −1

𝑧 −1

ℎ[𝑛] = [1,0,1]
6
SKEL4223

(d) Compute a 4-point DFT of the filter when 𝐾 = 2.


(6 marks)
Answer
𝐻(𝑧) = 1 + 𝑧 −2
𝐻(𝜔) = 1 + 𝑒 −𝑗2𝜔
𝑗2(2)𝜋𝑘
𝐻[𝑘] = 1 + 𝑒 − 4 = 1 + 𝑒 −𝑗𝜋𝑘
𝐻[0] = 1 + 1 = 2
𝐻[1] = 1 + 𝑒 −𝑗𝜋 = 1 − 1 = 0
𝐻[2] = 1 + 𝑒 −𝑗2𝜋 = 1 + 1 = 2
𝐻[3] = 𝐻[1] = 0
𝐻[𝑘] = [2,02,0]

(e) Figure Q2 shows a frequency response of a signal where there is a noise at 𝜔 = 0.5𝜋.
Find 𝐻(0.5𝜋) and prove that the noise will be removed when 𝐾 = 2.

0.5𝜋 𝜋 1.5𝜋 2𝜋

Figure Q2
(5 marks)
Answer
𝐻(𝜔) = 1 + 𝑒 −𝑗2𝜔
𝐻(𝜋/2) = 1 + 𝑒 −𝑗2𝜋/2 = 1 + (−1) = 0
Since 𝐻(𝜋/2) = 0, then it is proven that the frequency component at 𝜔 = 𝜋 is
removed.
7
SKEL4223

QUESTION 3
An impulse response of an FIR filter is described as follow where 𝑤[𝑛] is a window function
for truncation process.
ℎ[𝑛] = 0.5𝑠𝑖𝑛𝑐(0.5(𝑛 − 𝑛𝑑 )). 𝑤(𝑛 − 𝑛𝑑 )

(a) What is FIR filter?


(3 marks)
Answer
FIR filter is a filter whose impulse response is finite in length.

(b) Why the above filter is delayed by 𝑛𝑑 ?


(3 marks)
Answer
To have a causal and linear phase filter.

(c) What is the value of 𝑛𝑑 if the FIR filter is to be designed as a zero-phase filter?
(3 marks)
Answer
For zero-phase filter, symmetricity is at 𝑛 = 0. Thus 𝑛𝑑 = 0.

(d) Supposed that 𝑤[𝑛] = 0.54 + 0.46 cos(𝜋𝑛) for −1 ≤ 𝑛 ≤ 1, compute ℎ[𝑛]. Choose 𝑛𝑑
value that makes the filter causal.
(8 marks)
Answer
𝑁−1
For causal filter, 𝑛𝑑 = = 1.
2

𝑤[𝑛 − 1] = 0. .54 + 0.46 cos(𝜋(𝑛 − 1)) = [0.08, 1, 0.08] for 0 ≤ 𝑛 ≤ 2


ℎ[𝑛] = 0.5𝑠𝑖𝑛𝑐(0.5(𝑛 − 1)). 𝑤[𝑛 − 1]
= [0.3183, 0.5, 0.3183]. [0.08, 1, 0.08]
ℎ[𝑛] = [0.025, 0.5, 0025] for 0 ≤ 𝑛 ≤ 2
8
SKEL4223

(e) It is known that 𝑦[𝑛] = 𝑥[𝑛] ∗ ℎ[𝑛]. Find 𝑦[𝑛] when 𝑥[𝑛] = [1, 1], 𝑛𝑑 = 0 and 𝑤[𝑛] is
a rectangular window function of length three.
(8 marks)
Answer
For 𝑛𝑑 = 0 and 𝑤[𝑛] equals to a rectangular window of length 3,
ℎ[𝑛] = 0.5𝑠𝑖𝑛𝑐(0.5𝑛) for −1 ≤ 𝑛 ≤ 1
= [0.3183, 0.5, 0.3183] for −1 ≤ 𝑛 ≤ 1

To find 𝑦[𝑛], do convolution between 𝑥[𝑛] and ℎ[𝑛] as follow:


𝑛 -1 0 1 2
ℎ[𝑛] 0.3183 0.5 0.3183
𝑥[𝑛] 1 1
0.3183 0.5 0.3183
0.3183 0.5 0.3183
𝑦[𝑛] 0.3183 0.8138 0.8138 0.3183
9
SKEL4223

QUESTION 4
A digital IIR filter is to be designed based on continuous IIR Butterworth filter described as
below where Ω𝑐 = 1.
1
𝐻(𝑠) =
𝑠2 + 1.4142𝑠 + 1

(a) List down three parameters to be considered in designing a digital IIR filter.
(6 marks)
Answer
1. Cutoff frequency
2. Filter order
3. Transition bandwidth

(b) What is the order of the above Butterworth filter?


(2 marks)
Answer
2nd order.

(c) How many poles and zeros that the above filter has?
(2 marks)
Answer
2 poles and 2 zeros.

(d) Find 𝐻(𝑧), the digital system function of the above filter using bilinear transformation
1+𝑧
method where 𝑠 = 𝑐 (1−𝑧). Assume 𝑐 = 1.

(8 marks)
Answer
1
𝐻(𝑧) = 2
1− 𝑧 −1 1 − 𝑧 −1
( −1 ) + 1.4142 ( )+1
1+𝑧 1 + 𝑧 −1

(1 + 𝑧 −1 )2
=
(1 − 𝑧 −1 )2 + 1.4142(1 − 𝑧 −1 )(1 + 𝑧 −1 ) + (1 + 𝑧 −1 )2

(1 + 𝑧 −1 )2
=
3.4142 + 0.5858𝑧 −2
10
SKEL4223

1 + 2𝑧 −1 + 𝑧 −2
= 0.2929
1 + 0.1716𝑧 −2

1+𝑧
(e) By setting 𝑐 = 0.5 to the bilinear transformation of 𝑠 = 𝑐 (1−𝑧), the frequency response

𝐻(𝜔) of the digital IIR filter is denoted as below. Will this filter remove a 50𝐻𝑧 power
line noise on a digital device operated with frequency sampling of 𝐹𝑠 = 100𝐻𝑧. Show
all calculation on your explanation.
1 + 2𝑒 −𝑗𝜔 + 𝑒 −𝑗2𝜔
𝐻(𝜔) =
1.96 + 1.5𝑒 −𝑗𝜔 + 0.54𝑒 −𝑗2𝜔

(7 marks)
Answer
At 60𝐻𝑧 frequency,
𝐹 50
𝜔 = 2𝜋 = 2𝜋 ( )=𝜋
𝐹𝑠 100
To find out whether the filter will remove the noise, compute 𝐻(𝜋) as follow
1 + 2𝑒 −𝑗𝜋 + 𝑒 −𝑗2𝜋
𝐻(𝜋) =
1.96 + 1.5𝑒 −𝑗𝜋 + 0.54𝑒 −𝑗2𝜋
1−2+1
=
1.96 − 1.5 + 0.54
=0
Since 𝐻(𝜋) = 0, the noise is removed.

You might also like