Department of Electronic Systems

Examination paper for TTT4120 Digital Signal Processing

Examination date: 04.08.2020

Examination time (from-to): 09:00 – 13:00

Permitted examination support material: All support material is allowed

Academic contact during examination: Stefan Werner

Phone: +358401424948

Technical support during examination: Orakel support services

Phone: 73 59 16 00

OTHER INFORMATION

If a question is unclear/vague – make your own assumptions and specify in your answer the premises
you have made. Only contact academic contact in case of errors or insufficiencies in the question set.

Saving: Answers written in Inspera are automatically saved every 15 seconds. If you are working in
another program remember to save your answer regularly.

Cheating/Plagiarism: The exam is an individual, independent work. Examination aids are permitted. All
submitted answers will be subject to plagiarism control. Read more about cheating and plagiarism here.

Notifications: If there is a need to send a message to the candidates during the exam (e.g. if there is an
error in the question set), this will be done by sending a notification in Inspera. A dialogue box will appear.
You can re-read the notification by clicking the bell icon in the top right-hand corner of the screen. All
candidates will also receive an SMS to ensure that nobody misses out on important information. Please
keep your phone available during the exam.

All files must be uploaded before the examination time expires. 15 minutes are added to the examination
time to manage the sketches/calculations/files. (The additional time is included in the remaining
examination time shown in the top left-hand corner.)
How to digitize your sketches/calculations
How to create PDF documents
Remove personal information from the file(s) you want to upload

ABOUT SUBMISSION

Your answer will be submitted automatically when the examination time expires and the test
closes, if you have answered at least one question. This will happen even if you do not click “Submit and
return to dashboard” on the last page of the question set. You can reopen and edit your answer as long
as the test is open. If no questions are answered by the time the examination time expires, your answer
will not be submitted.

Withdrawing from the exam: If you wish to submit a blank test/withdraw from the exam, go to the menu
in the top right-hand corner and click “Submit blank”. This can not be undone, even if the test is still open.

Accessing your answer post-submission: You will find your answer in Archive when the examination
time has expired.

1 (4)
Problem 1 (3p + 3p + 3p + 3p = 12p)

For each of the following discrete-time systems, determine whether or not the system is (1)
linear, (2) time-invariant, (3) causal, and (4) stable. The sequences 𝑥[𝑛] and 𝑦[𝑛] are the input
and output sequences of the system, respectively.

a) 𝑦[𝑛] = 0.5𝑥[𝑛] + 0.5𝑥[𝑛 − 1]

b) 𝑦[𝑛] = 𝑥 ! [𝑛]

c) 𝑦[𝑛] = 𝛾 + ∑"#$%" 𝑥[𝑛 − 𝑙], where 𝛾 is a nonzero constant

d) 𝑦[𝑛] = 𝛼𝑥[−𝑛], where α is a nonzero constant

Problem 2 (16p)

Figure 1: Linear time-invariant system.

Figure 1 depicts a linear time-invariant (LTI) system whose overall impulse response is given by

ℎ(𝑛) = 𝛿(𝑛) + 5𝛿(𝑛 − 1) + 9𝛿(𝑛 − 2) + 7𝛿(𝑛 − 3) + 2𝛿(𝑛 − 4)

Compute the impulse response ℎ" [𝑛] when it is known that ℎ& [𝑛] = 𝑢[𝑛] − 𝑢[𝑛 − 2], where
𝑢[𝑛] is the unit step sequence.

Problem 3 (8p

An LTI system with input 𝑥& [𝑛] = 𝛿 [𝑛] + 𝛿 [𝑛 − 1] + 𝛿 [𝑛 − 2] gives the output 𝑦& [𝑛] =
2𝛿 [𝑛 − 1] + 5𝛿 [𝑛 − 2] + 5𝛿 [𝑛 − 3] + 3𝛿 [𝑛 − 4].

What is the output 𝑦" [𝑛] if the input is 𝑥" [𝑛] = 𝛿 [𝑛] + 3𝛿 [𝑛 − 1] + 3𝛿 [𝑛 − 2]?

2 (4)
Problem 4 (5p + 5p = 10p)

Let ℎ& [𝑛] and ℎ" [𝑛] denote impulse responses of two minimum-phase LTI systems.

a) Show that the convolution of two minimum-phase systems is always minimum phase.

b) Show that the sum of two minimum-phase systems is generally not minimum phase.

Hint: To support your claim in b), you can use the following two system functions

1 1
1 − 2 𝑧 %& −2(1 + 3 𝑧 %& )
𝐻& (𝑧) = and 𝐻" (𝑧) =
1 1
1 − 3 𝑧 %& 1 − 3 𝑧 %&

Problem 5 (8p + 4p + 8p + 10p = 30p)

An ARMA process has an autocorrelation 𝛾'' [𝑙 ] whose Z-transform is given by

1
B𝑧 − 3C (𝑧 − 3) 1
Γ'' (𝑧) = 9 , for < |𝑧| < 2
1 2
B𝑧 − 2C (𝑧 − 2)

a) Determine the (causal) filter 𝐻(𝑧) for generating 𝑥 [𝑛] from a white noise input 𝑤 [𝑛].
Plot its pole-zero diagram and indicate the region of convergence (ROC).

b) Determine the difference equation for generating the ARMA process when the excitation
is white noise 𝑤 [𝑛].

c) Determine a stable linear whitening filter 𝐻( (𝑧) for the sequence 𝑥[𝑛]. Plot its pole-zero
diagram and indicate the region of convergence (ROC).

d) The direct-form 1 (DF-I) structure of 𝐻(𝑧) is implemented in finite precision arithmetic.

Assume that the round-off error after every multiplication can be modeled as a zero mean
white noise source with power 𝜎)" . Draw the DF-I structure with noise sources due to
rounding included. Determine the variance of the round-off noise at the filter output.

Problem 6 (12p)

An MA(2) process is described by the difference equation

𝑥 [𝑛] = 𝑤 [𝑛] + 0.81𝑤[𝑛 − 2]

where 𝑤[𝑛] is a white noise sequence with variance 𝜎*" . Determine the parameters of the AR(2)
model that provides a minimum mean-square error fit to the data.

3 (4)
Problem 7 (2p + 2p + 8p = 12p)

𝑋+ (𝐹)

𝐹 (kHz)
Figure 2: Spectrum 𝑋+ (𝐹) of continuous-time signal 𝑥+ (𝑡).

Let 𝑥+ (𝑡) be a bandlimited continuous-time signal whose spectrum 𝑋+ (𝐹) is shown in Fig. 2.
Signal 𝑥+ (𝑡) is sampled with sampling frequency 𝐹, = 8000 Hz to generate the discrete-time
sequence 𝑥[𝑛] = 𝑥+ (𝑡)|- $//1! .

a) With the sampling frequency used, aliasing occurs. What would have been the
smallest sufficient sampling frequency to avoid aliasing?

b) The analog signal 𝑥+ (𝑡) is 0.2 seconds long. How many samples are there in the
sequence 𝑥[𝑛]?

c) Sketch the spectrum |𝑋(𝑓)| of the sampled sequence 𝑥[𝑛].

4 (4)