ECE 438 Homework 3

1. (a)

x[n] = δ[n + 1] − δ[n] + δ[n − 1]

1
y[n] = {x[n] + x[n − 1]}
2
1
= {δ[n + 1] − δ[n] + δ[n − 1] + δ[n] − δ[n − 1] + δ[n − 2]}
2
1 1
= δ[n + 1] + δ[n − 2]
2 2
(b)
1
h[n] = {δ[n] + δ[n − 1]}
2
Convolution of h[n] and x[n]:

y[n] = h[n] ∗ x[n]

1
= {δ[n] + δ[n − 1]} ∗ {δ[n + 1] − δ[n] + δ[n − 1]}
2
1
= {δ[n + 1] + δ[n − 2]}
2
1 1
= δ[n + 1] + δ[n − 2]
2 2
(c) i.
1 1
y[n] = [ejωn + ejω(n−1) ] = [1 + e−jω ]ejωn
2 2
1 −jω
H(ω) = [1 + e ]
2
ii.
∞
X
H(ω) = h[n]e−jωn
n=−∞
∞
X 1
= {δ[n] + δ[n − 1]}e−jωn
n=−∞
2
1
= (1 + e−jω )
2
 (d)
∞
X
X(ω) = x[n]e−jωn
n=−∞
X∞
= δ[n + 1] − δ[n] + δ[n − 1]e−jωn
n=−∞
−jω
=e + ejω − 1
Y (ω) = H(ω)X(ω)
1
= (1 + e−jω )(e−jω + ejω − 1)
2
1 −jω
= (e + ejω − 1 + e−2jω + 1 − e−jω )
2
1 1
= ejω + e−2jω
2 2
1 1
y[n] = δ[n + 1] + δ[n − 2]
2 2

2. (a) x(t) = rep 1 [rect(6000t)]

2000

(b) The sketch is shown below.

(c) Using CTFT Tranform Pairs:

rect(t) ↔ sinc(f )
1
repT [x(t)] ↔ comb 1 [X(f )]
T T

The Fourier Transform is:

1 f
X(f ) = 2000comb2000 [ sinc( )]
6000 6000
∞
1 X k
= sinc( )δ(f − 2000k)
3 k=−∞ 3

2
 This is the k = 2 term in the above summation. The amplitude is
(d) The sketch is shown below. (1/3)sinc(2/3) = (1/3)sin(2pi/3)/(2pi/3) = sqrt(3)/(4pi)

This is the k = 1 term in the

As shown, the amplitudes of
above summation. The
these terms should match
amplitude is correct as shown
those on the other side of the
X(f) axis, since |X(-f)| = |X(f)| The dotted line
shows (sinc(f/6))/3

This is the k = 5 term. The

amplitude should be -sqrt(3)/(10pi)
This is the k = 3 term,
which has amplitude
(sinc(3/3))/3 = 0, as shown

This is the k = 4 term.

(e) From c part, the Fourier Transform is: The amplitude is correct
as shown.
∞
1 X k
X(f ) = sinc( )δ(f − 2000k)
3 k=−∞ 3

The direct inverse CTFT of X(f ):

Z ∞ ∞
1 X k
x(t) = sinc( )δ(f − 2000k)ej2πf t df
−∞ 3 k=−∞ 3
∞ Z ∞
X 1 k
= sinc( ) δ(f − 2000k)ej2πf t df
k=−∞
3 3 −∞
∞
X 1 k
= sinc( )ej2π2000kt
k=−∞
3 3

So, the Fourier coefficient is Xk = 13 sinc( k3 ).

(f) y(t) = x(t) × rect( 3·10t −3 ) = x(t) × rect( 1000t
3
)
(g) The sketch is shown below.

This is not the right place to use

ellipses, since they suggest that y(t)
continues forever, whereas there are
only 6 periods, as stated. It would be
better to draw them explicitly.

3
(h) Using the Fourier Transform Pair rect(t) ↔ sinc(f ) and the convolution theorem
x(t) ∗ h(t) = X(f )H(f ), the Fourier transform:

3 3f
Y (f ) = X(f ) ∗ sinc( )
1000 1000
∞
1 X k 3 3f
= sinc( )δ(f − 2000k) ∗ sinc( )
3 k=−∞ 3 1000 1000
∞
1 X k 3(f − 2000k)
= sinc( )sinc( )
1000 k=−∞ 3 1000

(i) The sinc(3f /1000) function is replicated every 2 kHz. The zeros-crossings occur at
interval 1/3 kHz. So there should be 6 zero-crossings between each replication at
interval 2 kHz. The kth replication of sinc(3f /1000) centered at 2 kHz, is weighted
by sinc(k/3). 2k
The sketch is shown below.
This should be sinc(f/6), which has its first zero-
crossing at f = 6, as shown in the sketch for part
(d). This corresponds to k = 3.

4
3. (a) The plot is shown below.

5
(b) The plot is shown below.

6
(c) Apply the Fourier Transform pairs
1
cos(2πf0 t) ↔ (δ(f − f0 ) + δ(f + f0 ))
2
1
combT [x(t)] ↔ rep 1 [X(f )]
T T

1
X1 (f ) = (δ(f − 1000) + δ(f + 1000))
2
1
X2 (f ) = (δ(f − 3000) + δ(f + 3000))
2
1 1
Xs (f ) = 4000rep4000 [ (δ(f − 1000) + δ(f + 1000))]
2
∞
X 1
= 4000 [ (δ(f − 4000k − 1000) + δ(f − 4000k + 1000))]
k=−∞
2
1
Xs2 (f ) = 4000rep4000 [ (δ(f − 3000) + δ(f + 3000))]
2
∞
X 1
= 4000 [ (δ(f − 4000k − 3000) + δ(f − 4000k + 3000))]
k=−∞
2

7
(d) Apply the Fourier Transform pairs

cos(ω0 n) = π × rep2π [δ(ω − ω0 ) + δ(ω + ω0 )]

2000π 2000π
X1 (ω) = π × rep2π [δ(ω − ) + δ(ω + )]
4000 4000
∞
X π π
=π (δ(ω − 2πk − ) + δ(ω − 2πk + ))
k=−∞
2 2
6000π 6000π
X2 (ω) = π × rep2π [δ(ω − ) + δ(ω + )]
4000 4000
∞
X 3π 3π
=π (δ(ω − 2πk − ) + δ(ω − 2πk + ))
k=−∞
2 2

of

(e) In part a, signals x1 (t) and x2 (t) are different from each other. However, both signals
would appear the same after sampling, as shown in part d. This is because the selec-
tion of the sampling frequency. Similarly, when looking at the CTFTs and DTFTs of
the two signals, although both appear to be different, they are actually the same.

8
4. Same as problem 2,
∞
1 X k
X(f ) = sinc( )δ(f − 2000k)
3 k=−∞ 3

and the sketch of X(f ) is shown below.

See corrected sketch

for Prob. 2(d).

sqrt(3)/(4*pi)

-sqrt(3)/(10*pi)

The block diagram of the described process is shown below.

z(t)

9
(a) fc = 7kHz, fs = 14kHz

This is product,
not convolution

(f)

sqrt(3)/(4*pi)
sqrt(3)/14e3, using delta(ax-f_0) = 1/|a|
*delta(x - f_0/a)

sqrt(3)/(28e3)

This should be
(5/7)*sqrt(3)/14e3

This should be
(3/7)*sqrt(3)/14e3

pi omega (rad/sample)

I think that we should

assume that the filter gain
of the D/A converter
restores the frequency
components in z(t) to their
original amplitudes, taking
into account the filter which
(5/7)*sqrt(3)/(2*pi) (3/7)*sqrt(3)/(4*pi) passes the DC component
with no change, and
fs
The cutoff frequency is 2
= 7 kHz. attenuates the 2kHz term by
(5/7)*sqrt(3)/2*pi) 5/7and the 4kHz term by
3/7.
√ √
1 5 3 3 3
Y (f ) = δ(f ) + (δ(f − 2000) + δ(f + 2000)) + (δ(f − 4000) + δ(f + 4000))
3 √ 14π √ 14π
1 5 3 3 3
y(t) = + cos(2π2000t) + cos(2π4000t) (3/7)*sqrt(3)/(4*pi)
3 14π 14π
3/7)*sqrt(3)/pi
(5/7)*sqrt(3)/pi

10
(b) fc = 7kHz, fs = 9kHz Please see pp. 13-15 of this document for a solution to Problem 4(b).

Since the input signal and the low pass filter are the same as part a, the low pass fil-
tered signal for part b is the same as part a.
The cutoff frequency is f2s = 4.5 kHz.

√ √
1 5 3 3
Y (f ) = δ(f ) + (δ(f − 2000) + δ(f + 2000)) + (δ(f − 4000) + δ(f + 4000))
3 √ 18π √ 18π
1 5 3 3
y(t) = + cos(2π2000t) + cos(2π4000t)
3 18π 18π

11
(c) fc = 3kHz, fs = 7kHz Please see pp. 16 and 17 of this PDF document.

This is correct

pi

fs
The curoff frequency is 2
= 3.5 kHz.

√
1 3 3
Y (f ) = δ(f ) + (δ(f − 2000) + δ(f + 2000))
3 √ 14π
1 3 3
y(t) = + cos(2π2000t)
3 14π

12