EE 261 The Fourier Transform and its

Applications
Fall 2013
Solutions to Midterm Exam
1
1. (25 points) Periodizing and half-wave symmetry
Let f(t) be a function that is identically 0 outside of the interval (0, T). As a variant
of the usual way to periodize f(t) consider the alternating sum
(1) g(t) =

n=
(1)
n
f(t nT).
(a) (5) For the function f(t) shown below, sketch g(t) when 0 t 3T.
0 T t
()
The following questions pertain to a general function g(t) as dened in (1), not
the particular example in part (a).
(b) (5) For g(t) as in equation (1), show that g(t) has half-wave symmetry,
g(t + T) = g(t).
(c) (5) Show that g(t) has period 2T,
g(t + 2T) = g(t).
You can either use part (b) or show this directly from the denition in equation
(1).
(d) (10) What are the Fourier coecients of g(t) in terms of Ff(s)? As a check on
your work, you should get that half of the Fourier coecients are 0.
Solutions:
(a) The graph of g(t) for 0 t 3T looks like
0 T t 2T 3T

()
2
You can see the periodicity of period 2T, and the half-wave symmetry by shifting
the function by half of the period.
(b) For g(t + T) we have
g(t + T) =

n=
(1)
n
f(t + T nT)
=

n=
(1)
n
f(t (n 1)T)
=

k=
(1)
k+1
f(t kT), (where k = n 1)
=

k=
(1)
k
f(t kT)
= g(t)
(c) We want to show that g(t + 2T) = g(t). From part (b) g(t + T) = g(t) thus
g(t + 2T) = g(t + T + T)
= g(t + T)
= (g(t)) = g(t).
Done. Or, we could work from the denition:
g(t + 2T) =

n=
(1)
n
f(t + 2T nT)
=

n=
(1)
n
f(t (n 2)T)
=

k=
(1)
k+2
f(t kT) (letting k = n 2)
=

k=
(1)
k
(1)
2
f(t kT)
=

k=
(1)
k
f(t kT) = g(t)
(d) We have to integrate g(t) agains a complex exponential over the interval from 0
to 2T. On that interval g(t) is in two parts, f(t) from 0 to T and f(t T) from
3
T to 2T. That is,
g(n) =
1
2T
_
2T
0
g(t)e
2int/2T
dt
=
1
2T
_
T
0
f(t)e
2int/2T
dt
1
2T
_
2T
T
f(t T)e
2int/2T
dt
=
1
2T
_
T
0
f(t)e
2int/2T
dt
1
2T
_
T
0
f(u)e
2i(n/2T)(u+T)
du
(substituting u = t T in the second integral)
=
1
2T
_
T
0
f(t)e
2int/2T
dt
1
2T
e
in
_
T
0
f(u)e
2inu/2T
du
=
1
2T
(1 e
in
)
_
T
0
f(t)e
2int/2T
dt
Now
1 e
in
=
_
1 (1) = 2, n odd,
1 1 = 0, n even.
so, since f(t) = 0 outside (0, T),
g(n) =
_
1
T
_
T
0
e
2int/2T
f(t) dt =
1
T
_

e
2int/2T
f(t) dt =
1
T
Ff(n/2T), n odd
0, n even
4
2. (30 points) Fourier transform values from a graph.
The function F(s) sketched below is the Fourier transform of an unknown function
f(x):
F(s) =
_

f(x)e
2isx
dx
!
"
#
$ ! " ! "
%&'(
Evaluate the following integrals. Your answers should be numbers, not functions! (5
points each)
(a)
_

f(x) dx
(b)
_

f(x)e
2ix
dx
(c)
_

Re{f(x)}e
4ix
dx
(d)
_

(f )(x)dx
(e)
_

f(x) cos(2x) dx
(f)
_

f(x) cos(2x)e
2ix
dx
5
Solution:
(a)
_

f(x) dx = F(0) = 2
(b)
_

f(x)e
2ix
dx = F(1) = 1
(c)
_

Re{f(x)}e
4ix
dx =
_

1
2
f(x)e
4ix
dx +
_

1
2
f(x)e
4ix
dx
=
1
2
(F(2) + F(2)) (use F{f(x)} = F(s))
=
1
2
(F(2) + F(2)) =
1
2
(1 + 2) = 1.5
A common mistake here was to assume f(x) is real, which is incorrect. In fact,
since F(s) is real, f(x) = f(x).
(d)
_

(f )(x)dx = F( f)(0) = F(0)Ff(0) = sinc0F(0) = 2

f(x) cos(2x) dx
=
1
2
__

f(x)e
2ix
dx +
_

f(x)e
2ix
dx
_
=
F(1) + F(1)
2
= 2
(f)
_

f(x) cos(2x)e
2ix
dx
=
1
2
__

f(x)e
2ix
e
2ix
dx +
_

f(x)e
2ix
e
2ix
dx
_
=
F(2) + F(0)
2
=
3
2
6
3. (20 points) Finding Fourier transforms: The following three questions are independent.
(a) (5) In communications theory the analytic signal f
a
(t) of a signal f(t) is dened,
via the Fourier transform, by
Ff
a
(s) =
_
Ff(s), s 0,
0, otherwise .
For a real-valued signal f(t), that is not identically zero, could the corresponding
analytic signal f
a
(t) also be real? Why or why not?
(b) (5) Find the repeated convolution of the sinc function with itself, i.e.
f(x) = (sinc sinc sinc)
. .
n
(x)
(c) (10) Sketch the derivative of the function f(x) shown below. Then using the
derivative theorem for Fourier transforms nd the Fourier transform of f.
Solution
(a) No, the signal cannot be real, unless it is identically zero. If f
a
(t) were real its
Fourier transform would have the property Ff
a
(s) = Ff
a
(s), but we are told
that Ff
a
(s) = 0 for s < 0.
7
(b) We take Fourier transforms using Fsinc = . Using the convolution theorem
F(sinc sinc sinc) = = .
So no matter how many times we convolve sinc with itself we just get !
(c) The horizontal segments have slope 0 and the oblique segments have slopes 1 and
1, respectively. Heres a plot of the derivative:
We see that the derivative consists of two shifted rect functions:
f

(x) = (x
1
2
) (x
5
2
) .
Now the derivative formula for the Fourier transform states
Ff

(x) = 2isFf(s)
and so in the present case
Ff(s) =
1
2is
F((x
1
2
) (x
5
2
))
=
1
2is
(e
is
sincs e
5is
sincs)
Another approach is to dierentiate f(x) twice, and from f

(x) above we see that

(x) = (x) (x 1) (x 2) + (x 3) .
8
For the Fourier transform:
Ff

= (1 e
2i(1)s
e
2i(2)s
+ e
2i(3)s
) .
Now use
Ff

= (2is)
2
Ff = 4s
2
Ff
to obtain
Ff(s) =
1
4
2
s
2
_
1 e
2is
e
4is
+ e
6is
_
.
Combining the complex exponentials into sines will produce the previous answer.
9
4. (10 points) An elementary identity from an advanced standpoint
Use the convolution theorem for Fourier transforms to show that
cos
2
=
1
2
(1 + cos 2).
Hint: Take let f(t) = cos 2at and take the Fourier transform of f(t)
2
= f(t) f(t).
Solution:
As suggested, let f(t) = cos 2at. Then by the convolution theorem
F(cos
2
2at) = F(f(t)f(t)) = Ff(s) Ff(s)
=
1
2
(
a
+
a
)
1
2
(
a
+
a
)
=
1
4
(
a

a
+ 2
a

a
+
a

a
)
=
1
4
(
2a
+ 2
0
+
2a
)
=
1
2

0
+
1
4
(
2a
+
2a
).
Now take the inverse Fourier transform to obtain
cos
2
2at =
1
2
+
1
2
cos 4at =
1
2
(1 + cos 4at).
Finally, put 2at = to write
cos
2
=
1
2
(1 + cos 2).
10
5. (15 points) Conservation of Energy
Suppose u(x, t) satises the wave equation on the real line

2
u
t
2
=

2
u
x
2
with the initial conditions u(x, 0) = f(x) and u
t
(x, 0) = 0 (as in Problem Set 4, #1).
Dene the energy of the wave at time t by
E(t) =
_

u
x
(x, t)

2
dx +
_

u
t
(x, t)

2
dx.
The rst term represents potential energy and the second term kinetic energy.
Prove that the wave conserves energy:
E(t) = E(0).
Thus an ideal traveling wave does not dissipate energy.
Hint: The Fourier transform U(s, t) of u(x, t) with respect to x has the form
U(s, t) =
1
2
F(s)(e
2ist
+ e
2ist
) = F(s) cos 2st where F(s) = Ff(s).
Solution
We convert both expressions to the frequency domain using Parsevals Theorem:
E(t) =
_

|u
x
(x, t)|
2
dx +
_

|u
t
(x, t)|
2
dx
=
_

|2isU(s, t)|
2
ds +
_

|U
t
(s, t)|
2
ds
=
_

4
2
s
2
|F(s)|
2
cos
2
2st ds +
_

|F(s)(2s sin 2st)|

4
2
s
2
|F(s)|
2
(cos
2
(2st) + sin
2
(2st)) ds
=
_

4
2
s
2
|F(s)|
2
ds
11
And for t = 0,
E(0) =
_

|u
x
(x, 0)|
2
+ |u
t
(x, 0)|
2
dx
=
_

|f

(x)|
2
dx
=
_

4
2
s
2
|F(s)|
2
ds
12