Digital Signal Processing - week5 quiz solutions

June 15, 2018

1 Question 1
Which of the following are linear systems:

1.1

1. Envelope detection (via squaring), i.e. y[n] = |[n]|2 ∗ h[n] where h[n] is the impulse response
of a lowpass lter such as the moving average lter.
2. This is obviously not liner since squaring is not linear

1.2

1. AM radio modulation, i.e. multiply a signal x[n] by a cosine at the carrier frequency :
y[n] = x[n] cos(2πωc n)
2. It is linear: (ax1 [n] + bx2 [n]) cos(2πωc n) = ax1 [n] cos(2πωc n) + bx2 [n] cos(2πωc n) = ay1 [n] +
by2 [n]

1.3

1. Second derivative, i.e. y(t) = d2

dt2 x(t)

2. It is easy to see that derivative operation is linear

1.4

1. Clipping, i.e. enforce a maximum signal amplitude MM,e.g:

x[n] if x[n] ≤ M
(
y[n] =
M otherwise

2. take x1 [n] = x2 [n] = M +1 then x3 [n] = x1 [n]+x2 [n] = 2M +2 but y1 [n] = y2 [n] = y3 [n] = M
So y3 [n] 6= y1 [n] + y2 [n]

2 Question 5
2.1 Q

n if n ≤ 0
(
Consider the lter h[n] = δ[n] − δ[n − 1], and the input x[n] =
0 otherwise
and the output y[n] = x[n] ∗ h[n].
Compute y[−1], y[0], y[1], y[2]

1
2.2 A

y[n] = x[n] ∗ h[n] = x[n] ∗ (δ[n] − δ[n − 1]) = x[n] ∗ δ[n] − x[n] ∗ δ[n − 1] = x[n] − x[n − 1]. Therefore:
1. y[−1] = x[−1] − x[−2] = 0
2. y[0] = x[0] − x[−1] = 0
3. y[1] = x[1] − x[0] = 1
4. y[2] = x[2] − x[1] = 1

3 Question 6
3.1 Q

Which of the following lters are BIBO-stable? Assume N ∈ N and 0 < ω < π

3.2 A

1. Any lter h[n] with nite support and bounded coecients. Is BIBO-stable - it is absolutly
summable
2. the moving avarage. Is BIBO stable - if the input is bounded so is the output.
3. The following smoothing lter: h[n] = ∞ k=0 k+1 δ[n − k]. is NOT BIBO-stable - this is the
1
P
harmonic series which is not absolutly summble.
4. The ideal low pass lter with a cuto frequency ω :

1 if |ω| ≤ ωc
(
jω
H(e ) =
0 otherwise
is NOT BIBO stable since sinc is not absolutly summble.

4 Question 7
4.1 Q

See the gure for question 7.

4.2 A

We need to nd the impules response h[n] from the given input and output.
y[n] = x[n] ∗ h[n] = Σx[k]h[n − k] = x[0]h[n] + x[1]h[n − 1] = 1h[n] + (−1)h[n − 1] = h[n] − h[n − 1]

1. for n ≤ −2, n ≥ 2 we have y[n] = h[n] − h[n − 1] = 0 so h[n] = h[n − 1]

2. for n = −1: y[−1] = h[−1] − h[−2] = 1 → h[−1] = h[−2] + 1
3. for n = 0: y[0] = h[0] − h[−1] = −1 → h[0] = h[−1] − 1 = h[−2] + 1 − 1 = h[−2]
4. for n = 1: y[1] = h[1] − h[0] = 1 → h[1] = h[0] + 1 = h[−2] + 1
dene A := h[−2] so h[−1] = A + 1, h[0] = A, h[1] = A + 1.
So h[n] looks: ...A, (A)−2 , (A + 1)−1 , (A)0 , (A + 1)1 , (A + 1)2 , ....
Now we have x1 [n] = ..0, 0, −2−1 , 10 , −21 , 32 , 0, 0, .. so:
1.
y1 [−2] = (x1 [n]∗h[n])(−2) = Σx1 [k]h[−2−k] = x1 [−1]h[−1]+x1 [0]h[−2]+x1 [1]h[−3]+x1 [2]h[−4]

= −2(A + 1) + (1A) + (−2A) + (3A) = −2

2
Figure 1: Question 7

3
 Figure 2: Question 12

2.
y1 [−1] = (x1 [n]∗h[n])(−1) = Σx1 [k]h[−1−k] = x1 [−1]h[0]+x1 [0]h[−1]+x1 [1]h[−2]+x1 [2]h[−3]

= (−2A) + 1(A + 1) + (−2A) + 3A = 1

5 Question 12
5.1 Q

See the gure for question 12.

5.2 A

The way to solve it is to note that (−1)n = cos(πn) so multiply by (−1)n is equivalent to modul-
tation with ω0 = π .
From week 4 we know DT F T {x[n]cos(ω0 n)} = 21 (X(ej(ω−ω0 ) ) + X(ej(ω+ω0 ) )).
Also, one should remember the 2π preiodicity, from there it is easy to note that the system
eliminates x1 and does not modify x2

6 Question 14
6.1 Q

See the gure for question 14.

4
 Figure 3: Question 14

6.2 A

The thing to note here is that the cos( π2 n)2 and the h[n] are not convulated but multiplyed, so it
relates to modulation of the h[n] lter.
From the hint we have cos( π2 n)2 = 12 (1 + cos(πn)) and we have from modulation:
1 1 1
DT F T { (1 + cos(πn))h[n]} = DT F T {h[n]} + (H(ej(ω−π) ) + H(ej(ω+π) ))
2 2 4
1 1
= H(ejω ) + (H(ej(ω−π) ) + H(ej(ω+π) ))
2 4
So the system in the frequency domain is:
1 1
X(ejω )( H(ejω ) + (H(ej(ω−π) ) + H(ej(ω+π) ))(1 − H(ejω ))
2 4
First we multiply with a lter that take the low frequencies and the high frequencies, so we
eliminate the bandpass frequencies. Then we multiply with a highpass lter, so overall the nal
system is a highpass lter. Note that because of perudicity, the high frequencies is a sum of the
modulation from 0 → π and from π ← 2π and since it is multiplied by 14 the overall gain is 21 .

7 Question 15
7.1 Q

See the gure for question 15.

7.2 A

Since x[n] = δ[n] then y[n] is the system itself. Note again that multiplying by (−1)n is a mod-
ulation with cos(πn) so the system is the sum of the lter h[n] and it's modulation with cos(πn)

5
 Figure 4: Question 15

but h[n] is a lowpass lter with cuto freq π2 so the system in the frequency sapce (call it S(ejω ))
is S(ejω ) = 1 for each ω (the sum of H and it's modulation covers the whole range).
Now, note that Σy[n] = (y[k] ∗ 1)[0]. So in the frequency domain it is S(ejω )DT F T {1}. Also
DT F T {1} is an impulse train with period 2π and amplitude 2π .
So we have:
Z π Z π Z π
1 1 1
Σy[n] = (y[k]∗1)[0] = S(ejω )DT F T {1}ejω0 dω = DT F T {1}dω = 2πδ(ω)dω = 1
2π −π 2π −π 2π −π