Linnæus University

Department of Physics and Electrical Engineering

Sven Nordebo

Written examination in Signal Processing, 4ED344, 7.5 hp

14 January 2021, at 08.00-13.00

Approved accessories: Formulas and tables in mathematics. The written exami-

nation covers 6 problems and a maximum of 6.0 credit points (crp). The following
grading will be used.

E (sufficient) 2 ≤ crp < 2.8

D (satisfactory) 2.8 ≤ crp < 3.6
C (good) 3.6 ≤ crp < 4.4
B (very good) 4.4 ≤ crp < 5.2
A (excellent) 5.2 ≤ crp

Write your name on all submitted papers!

1. Consider the analog signal

xa (t) = 3 cos(2000πt) + 5 sin(6000πt) + 10 cos(12000πt).

a) What is the Nyquist (sampling) rate for this signal (the minimum frequency
that will satisfy the sampling theorem)?
Answer: The highest frequency is 6 kHz so the sample frequency is twice, i.e.,
fs = 12 kHz.
b) Assume now that we sample this signal using a sampling rate fs = 5000 samp-
le/s. What is the discrete-time signal obtained after sampling?
Answer: x(n) = 3 cos 2π 15 n + 5 sin 2π 53 n + 10 cos 2π 65 n.
c) What is the analog signal ya (t) that we can reconstruct from the samples if we
use ideal interpolation?
Answer: Rewrite as follows (with −1/2 < |ν| < 1/2 and ν = f /fs )

1 3 6
x(n) = 3 cos 2π n + 5 sin 2π n + 10 cos 2π n
5 5 5
1 3 6
= 3 cos 2π n + 5 sin 2π( − 1)n + 10 cos 2π( − 1)n
5 5 5
1 2 1
= 3 cos 2π n + 5 sin 2π(− )n + 10 cos 2π n
5 5 5
1 2
= 13 cos 2π n − 5 sin 2π n
5 5
and so x(n) = ya (nT ) where
ya (t) = 13 cos 2π1000t − 5 sin 2π2000t
and T = 1/5000.
 R∞
2. Consider the double sided Laplace transform X(s) = L{x(t)} = −∞ x(t)e−st dt and
1
R σ+i∞
x(t) = 2πi σ−i∞
X(s)est ds. Consider also the double sided Z–transform X(z) =
P∞ −n 1
H
n=−∞ x(n)z and x(n) = 2πi Γ
X(z)z n−1 dz.
Determine the appropriate Laplace transforms X(s) or Z–transforms X(z) and the
related Regions Of Convergence (ROC) for the following signals,
(a) x(t) = e−at u(t)
1
Answer: X(s) = s+a
. ROC: Re s > −a.
(b) x(t) = eat u(−t)
1
Answer: X(s) = − s−a . ROC: Re s < a.
(c) x(t) = e−a|t|
2a
Answer: X(s) = a2 −s2
. ROC: −a < Re s < a.
(d) x(n) = bn u(n)
z
Answer: X(z) = z−b
. ROC: |z| > |b|.
(e) x(n) = b−n u(−n − 1)
−z −1
Answer: X(z) = z−b −1 . ROC: |z| < |b| .
(f) x(n) = b|n|
−z
Answer: X(z) = z
z−b
+ z−b−1
. ROC: |b| < |z| < |b|−1 .
where u(t) denotes the continuous time unit step function and a > 0, and u(n)
denotes the discrete time unit step function and |b| < 1.
R∞
3. Consider the Fourier transform pair X(ω) = F{x(t)} = −∞ x(t)e−iωt dt and
R∞
x(t) = F −1 {X(ω)} = 2π1
−∞
X(ω)eiωt dω.

(a) Show that  

∂x
F = iωX(ω).
∂t
Answer: Check slides.
(b) Show that  
−1 ∂X
F = −itx(t).
∂ω
Answer: Check slides.
R∞
(c) Consider the Fourier transform pair X(ω) = F{x(t)} = −∞ x(t)e−iωt dt and
R∞
x(t) = F −1 {X(ω)} = 2π
1
−∞
X(ω)eiωt dω. Calculate the Fourier transform
n 2
o
X(ω) = F e−at

where a > 0 is a constant, by employing the nonlinear differential equation

∂x
+ 2atx(t) = 0
∂t
−1 ∂X
together with the results F ∂x
 
∂t
= iωX(ω) and F ∂ω
= −itx(t).
Describe all steps in the derivation.
 R∞ 1
R∞
Hint: Use the Parsevals relation −∞ x2 (t) dt = 2π −∞
X 2 (ω) dω to determine
the unknown (scaling) constant in X(ω).
n o √ ω2
2
Answer: F e−at = √πa e− 4a .

4. A Moving Average (MA) stochastic process is given by the following FIR (Finite
Impulse Response) filtering

y(n) = x(n) + cx(n − 3)

where c is a real constants and x(n) is white noise with zero mean and variance
σx2 and autocorrelation function φx (k) = E{x(n + k)x(n)} = σx2 δ(k). Assume that
the input signal x(n) as well as the output signal y(n) are real stationary stochastic
processes. Derive an explicit expression for the output autocorrelation function

φy (k) = E{y(n + k)y(n)}

as well as the output power spectral density

∞
X
Φy (Ω) = F{φy (k)} = φy (k)e−iΩk
k=−∞

in terms of the constants σx2 and c.

Answer: Note that
h(n) = δ(n) + cδ(n − 3)
and
H(ν) = 1 + ce−i6πν
Hence

Φy (ν) = |H(ν)|2 Φx (ν) = σx2 (1 + ce−i6πν )(1 + cei6πν )

= σx2 cei6πν + 1 + c2 + ce−i6πν

and
φy (k) = σx2 cδ(k + 3) + (1 + c2 )δ(k) + cδ(k − 3)

5. The Fourier integral representation of the continuous-time signal xa (t) is given by

Z ∞
xa (t) = Xa (f )ei2πf t df, (1)
−∞

where the Fourier transform is

Z ∞
Xa (f ) = xa (t)e−i2πf t dt, (2)
−∞

and where t is time (in sec) and f is frequency (in Hz).

The sampling process is defined in the time domain as

x(n) = xa (nT ), for all integers n, (3)

 where T is the sampling time and
1
fs = , (4)
T
the sampling frequency. The Fourier integral representation of the discrete-time
signal x(n) is given by
Z 1/2
x(n) = X(ν)ei2πνn dν, (5)
−1/2

where the discrete-time Fourier transform is

∞
X
X(ν) = x(n)e−i2πνn . (6)
n=−∞

Here, ν is the normalized frequency and X(ν) is periodic with period 1, i.e., X(ν) =
X(ν + m) where m is an integer.
Based on the sampling process defined in the time domain as in (3), derive the
relationship between the discrete-time Fourier transform (6) and the Fourier trans-
form (2), also known as the Poisson summation formula. Hint: Let the normalized
frequency be given by ν = f T .
Answer: ∞
1 X
X(f T ) = Xa (f − kfs ), (7)
T k=−∞
where the normalized frequency ν = f T has been used.

6. An autoregressive (AR) stochastic process is given by the following IIR (Infinite

Impulse Response) filtering

y(n) = ay(n − 1) + bx(n)

where a and b are real constants with |a| < 1 and x(n) is white noise with zero
mean and variance σx2 and autocorrelation function φx (k) = E{x(n + k)x(n)} =
σx2 δ(k). Assume that the input signal x(n) as well as the output signal y(n) are
real stationary stochastic processes. Derive an explicit expression for the output
autocorrelation function
φy (k) = E{y(n + k)y(n)}
as well as the output power spectral density
∞
X
Φy (Ω) = F{φy (k)} = φy (k)e−iΩk
k=−∞

in terms of the constants σx2 , a and b.

Answer: Note that
b
H(ν) =
1 − ae−i2πν
and
h(n) = b · an u(n)
and that ∞
|n|
X 1 − a2
F{a } = a|n| e−i2πνn =
n=−∞
1 − 2a cos(2πν) + a2

Hence
1 1
Φy (ν) = |H(ν)|2 Φx (ν) = b2 σx2 −i2πν
(1 − ae ) (1 − aei2πν )
1 1
= b2 σx2 = b2 σx2
1 − ae−i2πν − ae i2πν +a 2 1 − 2a cos(2πν) + a2
σx2 1 − a2
= b2
1 − a2 1 − 2a cos(2πν) + a2

and
2 σx2 |k|
φy (k) = b a
1 − a2
GOOD LUCK!