6.

003 Homework #9 Solutions

Problems
1. Fourier varieties

a. Determine the Fourier series coefficients of the following signal, which is periodic in
T = 10.
x1 (t)
1

t
−10 −3 −1 1 3 10

2
a0 = 5

πk
sin 3πk
5 − sin 5 for k 6= 0
ak =
πk

2πk
−1 2πk
3
−1 3
1 e−j 10 t  1 e−j 10 t 
Z Z
1 −j 2πk t 1 −j 2πk t
ak = e 10 dt + e 10 dt =  +
10 10 10 −j 2πk 10 −j 2πk

−3 1 10
 10

−3 1
j3 2πk j 2πk −j 2πk −j3 2πk sin 3πk
− sin πk
e 10 −e 10 e 10 −e 10
5 5
= + =
j2πk j2πk πk

b. Determine the Fourier transform of the following signal, which is zero outside the
indicated range.
x2 (t)
1

t
−3 −1 1 3

2 sin 3ω − 2 sin ω
X2 (jω) =
ω

−1
−1 3
3
e−jωt  e−jωt 
Z Z
−jωt −jωt
X2 (jω) = e dt + e dt = +
−3 1 −jω −3 −jω 1
e j3ω − e jω e−jω − e−j3ω 2 sin 3ω − 2 sin ω
= + =
jω jω ω
6.003 Homework #9 Solutions / Fall 2011 2
c. What is the relation between the answers to parts a and b? In particular, derive an
expression for ak (the solution to part a) in terms of X2 (jω) (the solution to part b).
 
1 2πk
ak = X2 j
10 10

The Fourier series coefficients ak are

1
ak = X2 (jω)|ω= 2πk
10 10

d. Determine the time waveform that corresponds to the following Fourier transform,
which is zero outside the indicated range.
X3 (jω)
1

ω
−3 −1 1 3

sin 3t − sin t
x3 (t) =
πt

−1 3
−1 3
e jωt  e jωt 
Z Z
1 jωt 1 jωt
x3 (t) = e dω + e dω = +
2π −3 2π 1 j2πt −3 j2πt 1
e−jt − e−j3t e j3t − e jt sin 3t − sin t
= + =
j2πt j2πt πt

e. What is the relation between the answers to parts b and d? In particular, derive an
expression for x3 (t) (the solution to part d) in terms of X2 (jω) (the solution to part b).
1
1

x3 (t) = X2 j(−t) or X2 j(t)
2π 2π

The relation between this answer and that of the previous part is duality.
1
x3 (t) = X2 (jω)|ω=−t
2π
1
Since X2 (jω) is real and even, 2π X2 (jt) would also work.
6.003 Homework #9 Solutions / Fall 2011 3
2. Fourier transform properties
Let X(jω) represent the Fourier transform of

−t 0 < t < 1
x(t) = e .
0 otherwise
Express the Fourier Transforms of each of the following signals in terms of X(jω).
x1 (t)
1

t
1

X1 (jω) = X(jω) + X(−jω)

x1 (t) = x(t) + x(−t)

Z ∞ Z ∞
x(−t) ↔ x(−t)e−jωt dt = x(t)ejωt dt = X(−jω)
−∞ −∞

X1 (jω) = X(jω) + X(−jω)

x2 (t)
1

t
1
−1

X2 (jω) = X(jω) − X(−jω)

x2 (t) = x(t) − x(−t)

X2 (jω) = X(jω) − X(−jω)

x3 (t)
1

t
1

X3 (jω) = (1 + ejω )X(jω)

x3 (t) = x(t) + x(t + 1)

Z ∞ Z ∞
−jωt
x(t + 1) ↔ x(t + 1)e dt = x(t)e−jω(t−1) dt = e jω X(jω)
−∞ −∞

X3 (jω) = (1 + e jω )X(jω)
6.003 Homework #9 Solutions / Fall 2011 4
3. Fourier transforms
Find the Fourier transforms of the following signals.

a. x1 (t) = e−|t| cos(2t)

1 1
X1 (jω) = +
1 + (ω − 2)2 1 + (ω + 2)2

1
e−t u(t) ↔
1 + jω
1 1 2
e−|t| ↔ + =
1 + jω 1 − jω 1 + ω2

cos(2t) ↔ πδ(ω − 2) + πδ(ω + 2)

Therefore, by the multiplication property,

1 1
e−|t| cos(2t) ↔ 2
+
1 + (ω − 2) 1 + (ω + 2)2

sin(2πt)
b. x2 (t) =
π(t − 1)

X2 (jω) = e−jω (u(ω + 2π) − u(ω − 2π))

sin(2πt)
↔ u(ω + 2π) − u(ω − 2π)
πt
sin(2πt) sin(2πt − 2π) sin(2π(t − 1))
= = ↔ e−jω (u(ω + 2π) − u(ω − 2π))
π(t − 1) π(t − 1) π(t − 1)
6.003 Homework #9 Solutions / Fall 2011 5

2
c. x3 (t) = t 0 < t < 1
0 otherwise

j −jω 2 2j
+ 2 e−jω + 3 1 − e−jω


X3 (jω) = e
ω ω ω

x3 (t) = t2 (u(t) − u(t − 1))

1
u(t) ↔ + πδ(ω)
jω

1 1 1 − e−jω
u(t) − u(t − 1) ↔ + πδ(ω) − e−jω − e−jω πδ(ω) =
jω jω jω
d
tf (t) ↔ j F (jω)
dω
d2
t2 f (t) ↔ − F (jω)
dω 2
d2 1 − e−jω
 
t2 u(t) − u(t − 1) ↔ − 2


dω jω

j −jω 2 2j
+ 2 e−jω + 3 1 − e−jω


x3 (t) ↔ e
ω ω ω

d. x4 (t) = (1 − |t|) u(t + 1)u(1 − t)

2(1 − cos ω)
X4 (jω) =
ω2

Let p(t) = u(t + 0.5) − u(t − 0.5). Then

x4 (t) = p(t) ∗ p(t)

and
X4 (jω) = P 2 (jω) .
1 1
e−jωt  2 2 sin ω2
Z
2
−jωt
P (jω) = e dt = =
− 12 −jω − 1 ω
2

ω
4 sin2 2 2(1 − cos ω)
X4 (jω) = =
ω2 ω2
You could also solve this by differentiating twice in the time domain to get a sequence
1
of delta functions, computing the transform, and multiplying twice by jω [the delta
functions in the integration property have zero weight].
6.003 Homework #9 Solutions / Fall 2011 6

Engineering Design Problem

4. Parseval’s theorem
Parseval’s theorem relates time- and frequency-domain methods for calculating the av-
erage energy of a signal as follows:
Z ∞
1 2
X
|x(t)| dt = |ak |2
T T
k=−∞

where ak represents the Fourier series coefficients of the periodic signal x(t) with period
T.

a. We can derive Parseval’s theorem from the properties of CT Fourier series.

1. Let y(t) = |x(t)|2 . Find the Fourier series coefficients bk of y(t).

[Hint: |x(t)|2 = x(t)x∗ (t).]
∞
X 2π
x(t) = ak e j T kt

k=−∞

∞
X 2π
x∗ (t) = a∗k e−j T kt

k=−∞
Z Z
1 −j 2π 1 2π
bk = y(t)e T
kt
dt = x(t)x∗ (t)e−j T kt
dt
T T T T
Z ∞ ∞
1 X 2π X 2π
mt −j 2π
= al e j T lt
a∗m e−j T e T
kt
dt
T T l=−∞ m=−∞
∞ ∞ Z
X X 1 2π
= al a∗m e−j T (k−l+m)t
dt
T T
l=−∞ m=−∞
X∞ ∞
X
= al a∗m δ[k − l + m]
l=−∞ m=−∞
∞
X
= ak+m a∗m
m=−∞

2. Use the result from the previous part to derive Parseval’s theorem.
Z ∞ ∞
1 X X
2
|x(t)| dt = b0 = ak a∗k = |ak |2
T T k=−∞ k=−∞

b. Let x1 (t) represent the input to an LTI system, where

∞
X π
x1 (t) = α|k| e jk 4 t
k=−∞

for 0 < α < 1. The frequency response of the system is

H(jω) = 1 |ω| < W

n
0 otherwise.
6.003 Homework #9 Solutions / Fall 2011 7
What is the minimum value of W so that the average energy in the output signal will
be at least 90% of that in the input signal.
The signal x1 (t) is periodic with period T = 8 and has Fourier series coefficients
ak = α|k| . The average energy in the input signal is
∞ ∞
1 + α2
Z
1 X X 1 1
|x(t)|2 dt = |ak |2 = α2|k| = 2
+ 2
−1=
T T 1−α 1−α 1 − α2
k=−∞ k=−∞

The lowpass filter passes some number K of the harmonic components of x(t) so that
the average energy in the output signals is
K
1 − α2K+2 1 − α2K+2 1 + α2 − 2α2K+2
Z
1 2
X
|y(t)| dt = α2|k| = + − 1 =
T T 1 − α2 1 − α2 1 − α2
k=−K

To make the energy in the output at least 90% of that in the input
1 + α2 − 2α2K+2 1 + α2
 
≥ 0.9
1 − α2 1 − α2

1 + α2 − 2α2K+2 ≥ 0.9(1 + α2 )

2α2K+2 ≤ 0.1(1 + α2 )

1 + α2
(K + 1) log α2 ≤ log
20
2
log 1+α
20
K> −1 (inequality switches because the logs are negative)
log α2
2π
Because the harmonics are spaced at T intervals in frequency
2
!
log 1+α
20 2π
W > 2
−1
log α 8
6.003 Homework #9 Solutions / Fall 2011 8
5. Filtering
The point of this question is to understand how the magnitude of a filter affects the
output and how the angle of a filter affects the output. Consider the following RC circuit
as a “filter.”
R
+
vi + vo
− C
−

Assume that the input vi (t) is the following square wave.

vi (t)
1
2
t
0 T
− 21

If the fundamental frequency of the square wave ( 2π T ) is equal to the cutoff frequency of
1
the RC circuit ( RC ) then the output vo (t) will have the following form.
vo (t)
1
2
t
0 T
− 12

We can think of the RC circuit as “filtering” the square wave as shown below.
1
|H(jω)|

0.1

0.01 ω
0.01 0.1 1 10 100 1/RC
∠H(jω)|

− π2 ω
0.01 0.1 1 10 100 1/RC
The RC filter has two effects: (1) The amplitudes of the Fourier components of the
input (vertical red lines in upper panel) are multiplied by the magnitude of the frequency
response (|H(jω)|). (2) The phase of the Fourier components (red dots in lower panel)
are shifted by the phase of the frequency response (∠H(jω)).

a. Determine (using whatever method you find convenient) the output that would result
if vi (t) were passed through a filter whose magnitude is |H(jω)| (as above) but whose
phase function is 0 for all frequencies. Compare the result with vo (t) above.
6.003 Homework #9 Solutions / Fall 2011 9
The following plot shows the sum of the first 46 terms of the series expansion for the
square wave, with each term filtered by the magnitude (but not the phase) of the RC
lowpass filter. The original output is also shown (dashed green) for reference.
vo (t)
1
2
t
0 T
− 12

The asymmetry in the “charging” and “discharging” portions of the RC response is

gone. The effect of the filter is to reduce the magnitudes of the higher harmonics
without adding phase delay.

b. Determine (using whatever method you find convenient) the output that would result
if vi (t) were passed through a filter whose phase function is ∠H(jω) (as above) but
whose magnitude function is 1 for all frequencies. Compare the result with vo (t)
above.
The following plot shows the sum of the first 46 terms of the series expansion for the
square wave, with each term filtered by the phase (but not the magnitude) of the RC
lowpass filter.

vo (t)
1
2
t
0 T
− 12

The asymmetry in the “charging” and “discharging” portions of the RC response is

even more pronounced than before. Because the magnitudes of the higher harmonics
are not attenuated, they now accumulate to make a substantial peak that was not seen
in the RC response. We can understand the large overshoot as follows. Each harmonic
component in the square wave is a sinusoid:
∞  
X 2 2πkt
vi (t) = sin .
πk T
k=1
k odd
Except for the fundamental, the phase shifts are nearly a quarter cycle. Delaying each
of the harmonic components by a quarter cycle aligns the peaks of each component at
t = T2 as shown below
6.003 Homework #9 Solutions / Fall 2011 10

vx (t)
1
2
t
0 T
− 21

where
∞  
X 2 2πkt
vx (t) = sin − π/2 .
πk T
k=1
k odd