Digital Signal Processing

Final Exam
2021 - 2022 Fall
07.01.2021

Student Name : SOLUTIONS 1 2 3 4 Total

Student Number : (30p) (20p) (35p) (15p)
Signature :

Time: 90min
 This is a closed books / notes exam.
 Calculators are not allowed.
 Additional / extra answer sheets are not allowed.

Q1. a) Calculate the fast Fourier transform (FFT) for the signal 𝑥[𝑛] = [1 3 5 8] (15)
b) Calculate the inverse discrete Fourier transform (IDFT) for the signal 𝑥[𝑛] = [6 2𝑗 2 − 2𝑗] (15)

a)

𝑥𝑒 [𝑛] = [1 5] , 𝑥𝑜 [𝑛] = [3 8]

⇒ 𝑋𝑒 [𝑘] = [1 + 5 1 − 5] = [6 − 4] , 𝑋𝑜 [𝑘] = [3 + 8 3 − 8] = [11 − 5]

𝑋[𝑘] = 𝑋𝑒 [𝑘] + 𝑊4𝑘 𝑋𝑜 [𝑘] , 𝑘 = 0,1

𝑋[𝑘 + 2] = 𝑋𝑒 [𝑘] − 𝑊4𝑘 𝑋𝑜 [𝑘], 𝑘 = 0,1

𝑋[0] = 6 + 𝑒 −𝑗(2𝜋/4)0 (11) = 6 + 11 = 17

2𝜋
𝑋[1] = −4 + 𝑒 −𝑗( 4 )1 (−5) = −4 + 5𝑗
2𝜋
𝑋[2] = 6 − 𝑒 −𝑗( 4 )0 (11) = 6 − 11 = −5
2𝜋
𝑋[3] = −4 − 𝑒 −𝑗( 4 )1 (−5) = −4 − (−5𝑗) = −4 − 5𝑗

b)
𝑁−1 3
1 1
𝑥[𝑛] = ∑ 𝑋[𝑘]𝑒 𝑗(2𝜋/𝑁)𝑘𝑛 = ∑ 𝑋[𝑘]𝑒 𝑗(2𝜋/4)𝑘𝑛
𝑁 4
𝑘=0 𝑘=0
1
= (6 + (2𝑗)𝑒 𝑗(𝜋/2)𝑛 + 2𝑒 −𝑗(𝜋/2)2𝑛 + (−2𝑗)𝑒 𝑗(𝜋/2)3𝑛 )
4
1 1
𝑥[0] = (6 + (2𝑗)𝑒 𝑗(𝜋/2)0 + 2𝑒 −𝑗(𝜋)0 + (−2𝑗)𝑒 𝑗(3𝜋/2)0 ) = (6 + (2𝑗) + 2 + (−2𝑗)) = 2
4 4
1 1
𝑥[1] = (6 + (2𝑗)𝑒 𝑗(𝜋/2)1 + 2𝑒 −𝑗(𝜋)1 + (−2𝑗)𝑒 𝑗(3𝜋/2)1 ) = (6 + (2𝑗)(𝑗) + 2(−1) + (−2𝑗)(−𝑗)) = 0
4 4
1 1
𝑥[2] = (6 + (2𝑗)𝑒 𝑗(𝜋/2)2 + 2𝑒 −𝑗(𝜋)2 + (−2𝑗)𝑒 𝑗(3𝜋/2)2 ) = (6 + (2𝑗)(−1) + 2(1) + (−2𝑗)(−1))
4 4
=2
1 1
𝑥[3] = (6 + (2𝑗)𝑒 𝑗(𝜋/2)3 + 2𝑒 −𝑗(𝜋)3 + (−2𝑗)𝑒 𝑗(3𝜋/2)3 ) = (6 + (2𝑗)(−𝑗) + 2(−1) + (−2𝑗)(𝑗)) = 2
4 4
Q2. For the system given below, SA corresponds to a zero insertion system that inserts one zero after every input
sample (i.e., 2↑), 𝐻(𝑒 𝑗Ω ) is an ideal low-pass filter with gain 1, and SB corresponds to a decimation system that
extracts every fifth sample of its input (i.e., 5↓).
a) What should be the cut-off frequency of 𝐻(𝑒 𝑗Ω ) to prevent aliasing? Show your calculation or explain your
reasoning.(5)
b) Sketch 𝑇(𝑒𝑗Ω ), 𝑊(𝑒𝑗Ω ) and 𝑌(𝑒𝑗Ω ) based on the input signal’s frequency response provided below, and the cut-
off frequency determined on the previous step. (15)

x[n] 𝑆𝐴 𝐻(𝑒 𝑗Ω ) 𝑆𝐵 y[n]

t [n] w[n]

𝑋(𝑒 𝑗Ω )
… 1 …

π Ω
-π -π/2
−2𝜋/3 2𝜋/3 4𝜋/3 2π 8𝜋/3

a) The signal’s maximum frequency component is at the frequency 2𝜋/3. After the upsampling, it will be at
𝜋/3. If the filter does not cut out any frequency components out of this, after the downsampling the
maximum frequency component will be at 5𝜋/3, which will cause aliasing. In order to prevent aliasing, the
maximum frequency component should be limited to 𝜋 after downsampling, so the filter’s cut-off
frequency should be 𝜋/5.

b)

𝑇(𝑒 𝑗Ω )
… 1 …

−𝜋/3 𝜋 𝜋/3 2𝜋/3 𝜋π 4𝜋/3 5𝜋/3 2𝜋 7𝜋/3 Ω

-π

𝑊(𝑒 𝑗Ω )
… 1 …

−𝜋/5 𝜋 𝜋/5 π𝜋 2𝜋 Ω
-π

𝑌(𝑒 𝑗Ω )
… …

1/5

π𝜋 Ω
-π 2𝜋
Q3. For the system with the LCCDE expression of 𝑦[𝑛] − 2.5𝑦[𝑛 − 1] + 𝑦[𝑛 − 2] = 3𝑥[𝑛]

a) Determine the transfer function of the system. Show your calculations! (10)
b) Draw the pole-zero graph (5)
c) Determine the impulse response of the system if the system is causal. (10)
d) Determine the impulse response of the system if the system is stable. (10)

a)
𝑌(𝑧) − 2.5𝑧 −1 𝑌(𝑧) + 𝑧 −2 𝑌(𝑧) = 3𝑋(𝑧)

𝑌(𝑧) 3 3 3𝑧 2
𝐻(𝑧) = = = =
𝑋(𝑧) 1 − 2.5𝑧 −1 + 𝑧 −2 (1 − 2𝑧 −1 )(1 − 0.5𝑧 −1 ) (𝑧 − 2)(𝑧 − 0.5)

4 1
𝐻(𝑧) = −
1 − 2𝑧 −1 1 − 0.5𝑧 −1
b)
Poles: 0.5, 2
Zeros: 0, 0

Unit circle

c) ROC: |𝑧| > 2

4 1
𝐻(𝑧) = −
1 − 2𝑧 −1 1 − 0.5𝑧 −1

ℎ[𝑛] = 4(2)𝑛 𝑢[𝑛] − (0.5)𝑛 𝑢[𝑛] = ((2)𝑛+2 − (0.5)𝑛 )𝑢[𝑛]

d)

ROC: 0.5 < |𝑧| < 2

4 1
𝐻(𝑧) = −
1 − 2𝑧 −1 1 − 0.5𝑧 −1

ℎ[𝑛] = −4(2)𝑛 𝑢[−𝑛 − 1] − (0.5)𝑛 𝑢[𝑛]

Q4. Draw the amplitude responses |𝐻(𝑒𝑗Ω )| for the filters designed with the pole-zero positions given below.
Make sure that your drawings are consistent / compatible with each other.

a) Poles: 0.7 , Zeros: - 1

b) Poles: 0.7 , - 0.7 , Zeros: j , -j

c) Poles: 0.7𝑒𝑗𝜋/3 , 0.7𝑒−𝑗𝜋/3 , 0 , Zeros: 1 , 𝑒𝑗2𝜋/3, 𝑒−𝑗2𝜋/3

a) b)

c)