You are on page 1of 11

Exercises Topic I.

3: FOURIER TRANSFORM
Exercise 4-1
For all signals α > 0.

(a) The Fourier transform follows by direct substitution of the signal into the definition of the
Fourier transform (4-7):
R∞ R∞
Xa (f ) = (Ae−αt u(t))e−j2πf t dt = A e−(j2πf +α)t dt
−∞ 0
∞  
−A −A 1
= (j2πf +α) e−(j2πf +α)t = (j2πf +α) ((0) − (1)) = A (j2πf +α) .
0

(b) By direct substitution we get:

R∞ R0
Xb (f ) = (Aeαt u(−t))e−j2πf t dt = A e−(j2πf −α)t dt
−∞ −∞
0  
−A −A 1
= (j2πf −α) e−(j2πf −α)t = (j2πf −α) ((1) − (0)) = A (−j2πf +α) .
−∞

Note that in the last step we used the fact that e−(j2πf −α)t = e−j2πf t eαt , and when t → −∞ the last
part becomes zero while the magnitude of the first part can never become larger than 1 (it is a complex
exponential function).1

(c) Here we see that for t < 0, xc (t) equals xb (t), and for t > 0, xc (t) equals xa (t). In fact,
xc (t) = xa (t) + xb (t)! So, we can just use the ‘linearity’-property of the Fourier transform, i.e.:
Xc (f ) = Xa (f ) + Xb (f ):
 
A A 2α
Xc (f ) = (α+j2πf ) + (α−j2πf ) =A α2 +(2πf )2
.

This function is real and even, simply because xc (t) is also real and even (sketch it for yourself!).

(d) Since xd (t) = xa (t) − xb (t) we get:


 
A A −A4πf
Xd (f ) = Xa (f ) − Xb (f ) = (α+j2πf ) − (α−j2πf ) =j α2 +(2πf )2
.

This function is imaginary and odd, simply because xd (t) is real and odd (sketch it for yourself).

Exercises 4-2 and 4-3


We generally use MATLAB/Python to do this kind of stuff, but we can also do the first steps just by
analytic calculation. For instance, take Xa (f ):
       
A A (α−j2πf ) Aα −2πf A
Xa (f ) = (α+j2πf ) = (α+j2πf ) (α−j2πf ) = α2 +(2πf )2
+j α2 +(2πf )2
.

1
Another (much shorter!) way to perform this transform is by noting that xb (t) = xa (−t), so Xb (f ) = Xa (−f ) (the ‘time
reversal’ transform, entry 3b in Table 4-1).

1
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 2

The real and imaginary part of Xa (f ) can be used to compute the amplitude and phase of Xa (f ):

A
(Re(Xa (f )))2 + (Im(Xa (f )))2 = √
p
|Xa (f )| = , and
α2 +(2πf )2

6 Xa (f ) = arctan( Im(X a (f )) 2πf 2πf


Re(Xa (f )) ) = arctan(− α ) = − arctan( α ).

From these calculations we can see that when f = 0, |Xa (f )| = A α and Xa (f ) = 0, when f → ∞,
6
|Xa (f )| = 0 and 6 Xa (f ) = − 2 , and when f → −∞, |Xa (f )| = 0 and 6 Xa (f ) = + π2
π

In the following table I summarize the main results for the other three signals:

{xb (t), Xb (f )} {xc (t), Xc (f )} {xd (t), Xd (f )}


Aα 2Aα
real {Xb,c,d (f )} α2 +(2πf )2 α2 +(2πf )2
0
A2πf −A4πf
imag {Xb,c,d (f )} α2 +(2πf )2
0 α2 +(2πf )2
2Aα A4π|f |
|Xb,c,d (f )| √ 2A α2 +(2πf )2 α2 +(2πf )2
α +(2πf )2
f → −∞ 0 0 0
A 2A
f =0 α α 0
f →∞ 0  0 0
6 Xb,c,d (f ) arctan 2πf
α 0 ∀f −π/2 for f > 0, +π/2 for f < 0
f → −∞ −π/2 0 π/2
f =0 0 0 0
f →∞ π/2 0 −π/2

But, as stated above, we generally use programs like MATLAB to do all these calculations for us, using
a typical script included below (for xa (t)).

Listing 1: Exercise 4-2


% Compute Amplitude and Phase spectrum of X_a(f)
%
clear all

% time signal characteristics


alpha = 1; % set alpha
A = 2; % set amplitude

% define frequency axis


f =logspace ( − 2 , 2 , 1 0 0 0 ) ; % set frequency axes from f=0.01 to 100 Hz,
% with 1000 points logarithmically spaced

% compute amplitude spectrum


f h e l p = ( 2 * pi * f ) ;
fhelp2 = fhelp .* fhelp ;
Xmagnitude = A . / ( sqrt ( a l p h a ˆ 2 + f h e l p 2 ) ) ;

% compute phase spectrum


Xphase = atan(− f h e l p / a l p h a ) ;

% plot the result as two plots in a single row, two columns


clf
subplot ( 2 , 1 , 1 )
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 3

ah =loglog ( f , abs ( Xmagnitude ) ) ; % use loglog to have logarithmic x and y axes


set ( ah , ’linewidth’ , 2 ) ;
xlabel ( ’f, Hz’ ) ; title ( ’|X_a(f)|’ ) ;
grid

subplot ( 2 , 1 , 2 )
ah =semilogx ( f , Xphase ) ; % use semilogx to have logarithmic x axis
xlabel ( ’f, Hz’ ) ; title ( ’<X_a(f)’ ) ;
set ( ah , ’linewidth’ , 2 ) ;
grid

print −d e p s c 2 z e 4 2 a

Running this script yields:

10
1 |Xa (f )|

0
10

−1
10

−2
10

−3
10
−2 −1 0 1 2
10 10 10 10 10
f , Hz
0
6 Xa (f )

−0.5

−1

−1.5

−2
−2 −1 0 1 2
10 10 10 10 10
f , Hz

And the same can be done for the other three signals. Obviously, we will not ask you to write and run
MATLAB-scripts on the exam. We do expect, however, that you are able to (1) compute |X(f )| and
6 X(f ) for a given Fourier transform, and (2) look at what the amplitude and phase spectra do in the
limit cases of f = 0 and f → ∞.

Exercise 4-4
Recall that for an even function: x(t) = x(−t), an example is the cosine-function. For an odd function:
x(t) = −x(−t), an example is the sine-function. Note that this exercise is very similar to Exercise 3-16.

(a) We start our proof by just elaborating on the definition of the Fourier transform, (4-7):
R∞
X(f ) = x(t)e−j2πf t dt, using Euler (e−j2πf t = cos(2πf t) − j sin(2πf t)):
−∞
R∞
X(f ) = x(t) (cos(2πf t) − j sin(2πf t)) dt
−∞
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 4

R0 R∞
= x(t) (cos(2πf t) − j sin(2πf t)) dt + x(t) (cos(2πf t) − j sin(2πf t)) dt
−∞ 0

Substitute σ = −t in the first integral on the right-hand side (then: t = −σ, so dt = −(dσ), and the
integral limits change from (t from −∞ to 0) to (σ from ∞ to 0)):

R0 R∞
X(f ) = x(−σ) (cos(−2πf σ) − j sin(−2πf σ)) d(−σ)+ x(t) (cos(2πf t) − j sin(2πf t)) dt
σ=∞ 0

Since cos(−u) = cos(u) (cosine is even) and sin(−u) = − sin(u) (sine is odd), we get:
R∞ R∞
X(f ) = x(−σ) (cos(2πf σ) + j sin(2πf σ)) dσ + x(t) (cos(2πf t) − j sin(2πf t)) dt
0 0

Substitute σ = t in the integral on the left-hand side, and re-arrange yields (if you don’t understand this
substitution, look at the comments given in Exercise 3-16(a)):
R∞ R∞
X(f ) = [x(t) + x(−t)] cos(2πf t)dt − j [x(t) − x(−t)] sin(2πf t)dt (*)
0 0

Clearly then, when x(t) is even, i.e., x(−t) = x(t), the second integral in (*) becomes zero, and the
result is:
R∞
X(f ) = 2 x(t) cos(2πf t)dt qed
0

When x(t) is real (as are all signals in this lecture series) and even, X(f ) is a real and even function of
f : X(f ) = X(−f ).2

(b) From the derivation above we conclude that when x(t) is odd, i.e., x(−t) = −x(t), the first
integral in (*) becomes zero, and the result is:
R∞
X(f ) = −2j x(t) sin(2πf t)dt qed
0

When x(t) is real and odd, X(f ) is an imaginary and odd function of f : X(f ) = −X(−f ).3

Exercise 4-6
(a) Note that this Fourier transform can be found as entry 5 in Table 4-2. It can be computed
through direct substitution:
R∞ R∞
Xa (f ) = (te−αt u(t))e−j2πf t dt = te−(j2πf +α)t dt
−∞ 0

2
Please note that, as Ziemer says on page 152, it is the even cosine function, which is a function of frequency, which causes
the even-ness of X(f ), which is a function of frequency f . The property does not occur because the multiplication of an even
time-function x(t) and the even cosine-function is even!
3
Again, please note that it is the odd sine function, which is a function of frequency, which causes the odd-ness of X(f ),
which is a function of frequency f . The property does not occur because the multiplication of an odd time-function x(t) and
the odd cosine-function is odd, in fact the multiplication of two odd functions is even! The crux lies in that we are looking at a
function of frequency.
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 5

Rb Rb
Use partial integration rule to solve this integral (i.e.: a (df /dx)g(x)dx = f (x)g(x)|ba − a f (x)(dg/dx)dx):
R∞ −1 −(j2πf +α)t
Xa (f ) = td (j2πf +α) e
0
−t −(j2πf +α)t

−1 R∞
= j2πf +α e − (j2πf +α) e−(j2πf +α)t dt
0 0

−1 −1 1
= ((0) − (0)) + (j2πf +α)2
e−(j2πf +α)t = (j2πf +α)2
((0) − (1)) = (j2πf +α)2
qed
0

(b) Note that u(t)u(1 − t) means that the signal t2 only needs to be evaluated from t = 0 to t = 1
s. We compute the Fourier transform through direct substitution:

R∞ R1
Xb (f ) = (t2 u(t)u(1 − t))e−j2πf t dt = t2 e−j2πf t dt
−∞ 0
R1 2 −1

−t2
 1
−1 R1
= t d( j2πf )e−j2πf t = j2πf e−j2πf t − j2πf e−j2πf t dt2
0 0 0

−1 −j2πf

2 R1
= ( j2πf e ) − (0) + j2πf e−j2πf t tdt
0

The integral on the right-hand side requires another integration by parts and equals:

R1 
−1

−t
1
−1 R1
td j2πf e−j2πf t = j2πf e−j2πf t − j2πf e−j2πf t dt
0 0 0
    1
−1 −j2πf 1 −1
= ( j2πf e ) − (0) + j2πf j2πf e−j2πf t
0
 
−1 −j2πf 1
= j2πf e − (j2πf )2
(e−j2πf ) − (1) = △

So we obtain:

−1 −j2πf 2
Xb (f ) = j2πf e + j2πf △,

which after some re-arranging results in:


   
1 −j2πf −1 2
Xb (f ) = j2πf e jπf −1 + (j2πf )3
1 − e−j2πf .

1
Note that the Ziemer solution has a plus sign before the jπf -term in the parentheses of the left-hand part
of the solution. Our answer above is correct.

(c) Again, through direct substitution:

R∞ R1 R1
Xc (f ) = (e−αt u(t)u(1 − t))e−j2πf t dt = e−αt e−j2πf t dt = e−(j2πf +α)t dt
−∞ 0 0
1  
−1 −(j2πf +α)t −1
= (j2πf +α) e = (j2πf +α) (e−(j2πf +α) ) − (1)
0
 
1
Xc (f ) = (j2πf +α) (1 − e−(j2πf +α) ).
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 6

Exercise 4-8
(a) This signal is real and even, so its Fourier transform will also be real and even. We compute
the Fourier transform through direct evaluation:
R∞
X1 (f ) = ( 21 [δ(t + 1) + δ(t + 12 ) + δ(t − 12 ) + δ(t − 1)])e−j2πf t dt
−∞

Using the sifting property of the delta-function we easily obtain:


h i
1
X1 (f ) = 2 ej2πf + ejπf + e−jπf + e−j2πf , and since cos(u) = 12 (eju + e−ju ) we obtain:

X1 (f ) = cos(2πf ) + cos(πf ), which is a real and even function of frequency f .

Exercise 4-9
From the definition of the Fourier transform we obtain:
R∞ R∞ R∞
X(f ) = x(t)e−j2πf t dt = x(t) cos(2πf t)dt − j x(t) sin(2πf t)dt,
−∞ −∞ −∞

giving the real and imaginary part of X(f ) through the first and second integrals on the right, respec-
tively. Now, it is easy to show that:4

Re(X(−f )) = Re(X(f )), that is, the real part is an even function of f , and

Im(X(−f )) = −Im(X(f )), that is, the imaginary part is an odd function of f .
p
The amplitude of X(f ), |X(f )|, equals (Re(X(f )))2 + (Im(X(f )))2 , so:
p
|X(−f )| = (Re(X(−f )))2 + (Im(X(−f )))2
p p
= (Re(X(f )))2 + (−Im(X(f )))2 = (Re(X(f )))2 + (Im(X(f )))2 = |X(f )| qed

Similarly, for the phase of X(f ), 6 X(f ), which equals arctan( Im(X(f ))
Re(X(f )) ), then:

6 X(−f ) = arctan( Im(X(−f )) −Im(X(f )) Im(X(f ))


Re(X(−f )) ) = arctan( Re(X(f )) ) = − arctan( Re(X(f )) ) = − X(f )
6 qed

Concluding, the amplitude of X(f ) is a real and even function of f , and the phase is a real and odd
function of f .

Exercise 4-10
(a) We note that X1 (f ) is a real and even function of f , so we expect that when inverse Fourier-
transforming, we obtain a real and even function of time t. Through direct manipulation:

4
Hint: cos(2π(−f )) = cos(2πf ) and sin(2π(−f )) = − sin(2πf ).
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 7

R∞ R0 R∞
x1 (t) = X1 (f )ej2πf t df = Aeαf ej2πf t df + Ae−αf ej2πf t df = i1 (t) + i2 (t)
−∞ −∞ 0
R∞ R∞ A

i2 (t) = Ae−αf ej2πf t df = A e(j2πt−α)f df = (j2πt−α) e(j2πt−α)f
0 0 0

A A
= (j2πt−α) ((0) − (1)) = (α−j2πt) ,

where in the latter step we used the fact that e(j2πt−α)f = e(j2πt)f e−αf , and for f → ∞ the second ex-
ponential becomes 05 , and the value of the first exponential is never larger than 1 (it is a rotating complex
vector with magnitude 1).

R0
Similarly: i1 (t) = Aeαf ej2πf t df
−∞
0 0
R (j2πt+α)f A
i1 (t) = A e df = (j2πt+α) e(j2πt+α)f
−∞ −∞

A A
= (j2πt+α) ((1) − (0)) = (j2πt+α) .

A A 2Aα
So: x1 (t) = (α−j2πt) + (j2πt+α) = α2 +(2πt)2
, which is a real and even function of t.

Furthermore: x1 (0) = 2A 6
α which equals the value of the integral of X1 (f ) from −∞ to ∞. Similarly,
X1 (0) = A, which equals the value of the integral of x1 (t) from −∞ to ∞.7

Exercise 4-15
F F
Because Π( τt ) −→ τ sinc(f τ ), τ1 Π( τt ) −→ sinc(f τ ).

1 t
τ Π( τ ) is a ‘model’ of the delta-function δ(t), i.e., δ(t) ≈ lim τ1 Π( τt ).
τ →0

Then, its transform sinc(f τ ), how does it behave when τ → 0? The sinc-function equals 1 at f = 0 and
equals zero for f τ = ±1, ±2, ±3, . . ., so for f = ± τ1 , ± τ2 , ± τ3 , . . .

Clearly then, when τ → 0, the first ‘null’ of the sinc-function lies at f → ∞, or, the sinc-function
becomes incredibly broad and equals 1 for a very large range of frequencies.

In the limit-case then, when τ → 0, the Fourier transform of the model of δ(t) equals 1 for all frequencies.

Exercise 4-16
(a) xa (t) = Π( (t+1.5)
1 ) + Π( (t−1.5)
1 ), an even function.

5
The exercise should have defined α > 0.
R∞ R∞ ∞
6
Ae−α|f | = 2A e−αf df = 2A
−α
e−αf 0
= 2A
−α
((0) − (1)) = 2A
α
.
∞ 0

R R∞ R
2Aα 1 1
7
x1 (t)dt = 2 α2 +(2πt)2
dt. From Calculus we know that a2 +b2 u2
du = ab
arctan( ab u), we use this result here,
−∞ 0
R∞ 2Aα 1 ∞ 
with a = α and b = 2π to obtain that 2 α2 +(2πt)2
dt = 4Aα α2π arctan( 2π
α
t) 0
= 2A
π
( π2 ) − (0) = A.
0
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 8

We know that F Π( τt ) = τ sinc(f τ ), and we can use the time-delay theorem (Table 4-1, entry 2) to


compute that F Π( t−t


τ ) = τ sinc(f τ )e
0 −j2πf t0 .

     
So: Xa (f ) = F Π( t+1.5 t−1.5
τ ) + F Π( τ ) = sinc(f ) e
j3πf + e−j3πf = 2sinc(f ) cos(3πf ), a real

and even function of f .

(b) xb (t) = −Π( (t+1.5)


1 ) + Π( (t−1.5)
1 ), an odd function.
 
Similar as in (a) we get: Xb (f ) = sinc(f ) −ej3πf + e−j3πf = −2jsinc(f ) sin(3πf ), an imaginary
and odd function of f .

(c) xc (t) = Π( 4t ) + Π( 2t ), an even function.

We readily obtain: Xc (f ) = 4sinc(4f ) + 2sinc(2f ), a real and even function.8


   
(t−1.5)
(d) xd (t) = Π 3 + Π (t−1.5)
1 , so:
 
sin(πf )
Xd (f ) = 3sinc(3f )e−j3πf +1sinc(1f )e−j3πf = e−j3πf (sinc(f ) + 3sinc(3f )) = e−j3πf πf + 3 sin(3πf
3πf
)

1 −j3πf
= πf e (sin(πf ) + sin(3πf )),

where we use the fact that sin(u) cos(v) = 21 (sin(u − v) + sin(u + v)), so with u − v = πf and
u + v = 3πf we get u = 2πf and v = πf , so:

2 −j3πf
Xd (f ) = πf e sin(2πf ) cos(πf ).

Exercise 4-17
The signal x(t) is recorded at 15 inches/second, and is played back at 7 12 inches/second. This is twice as
slow as what is needed to get the original signal back.

Then, if signal x(t) has Fourier transform X(f ), then signal y(t) = x( 12 t) has Fourier transform
Y (f ) = | 11 | X( f1 ) = 2X(2f ), i.e., when time goes twice as slow, frequency runs twice as fast.
2 2

Let’s sketch what happens, in the frequency domain. The signal that is being recorded has an approxi-
mately rectangular spectrum extending from 100 to 10000 Hz.
Another, more cumbersome but still educational way would be to start with defining xc (t) = Π( (t+1.5)
8
1
) + 2Π( 2t ) +
Π( (t−1.5)
1
), and Fourier transforming, which yields: Xc (f ) = 2sinc(f ) cos(3πf )+4sinc(2f ), which after some manipulation
results in the same answer. Of course, students are recommended to use the fast (and easier!) approach stated above.
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 9

X(f )

−10000 −100 100 10000


f

Y (f )
2A

−5000 −50 50 5000


f

Clearly, the sound will appear to be much lower-frequency than the original signal.

Note when you have studied part I.4, look back at this exercise and see what happens to the energy of
the signal...

Exercise 4-18
   
f f
The Fourier transform X(f ) is shown in Figure P4-18, we see that X(f ) = Π 4 +Π 2 .

The duality theorem states that when a time-signal x(t) has a Fourier transform X(f ), then a time-signal
X(t) will have a Fourier transform equal to x(−f ).

t t
 
From Exercise 4-16(c) we know that xc (t) = Π 4 +Π 2 has as Fourier transform Xc (f ) = 4sinc(4f )+
2sinc(2f ).

Then here: since X(f ) equals xc (f ), the time signal that X(f ) originates from, x(t) must equal Xc (−t),
i.e.:

x(t) = 4sinc(−4t) + 2sinc(−2t) which equals 4sinc(4t) + 2sinc(2t) since the sinc-function is
even.

Hence, the time signal x(t) is real and even, as its Fourier transform (Figure P4-18) is also real and even.

Exercise 4-19
All signals in Exercise 4-16 are multiplied with p(t) = cos(20πt), a 10 Hz sinusoidal signal. The Fourier
transform of this signal equals:

1
P (f ) = 2 (δ(f + 10) + δ(f − 10)).

Multiplication of a signal x(t) in the time domain with p(t), means that in the frequency domain their
Fourier transforms are convolved, that is, when y(t) = x(t)p(t), then Y (f ) = X(f ) ⋆ P (f ).
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 10

Convolution of X(f ) with a δ-function δ(f − f0 ) means that we get a ‘copy’ of the original spectrum
X(f ), but now centered around f = f0 , that is X(f − f0 ).

Multiplication with a cosine-signal therefore yields two copies of the original Fourier transform, one at
f = −f0 and one at f = f0 .

(a) Ya (f ) = Xa (f ) ⋆ P (f ) = 21 2sinc(f + 10) cos(3π(f + 10)) + 12 2sinc(f − 10) cos(3π(f − 10))

= sinc(f + 10) cos(3π(f + 10)) + sinc(f − 10) cos(3π(f − 10)).

(b) Yb (f ) = Xb (f ) ⋆ P (f ) = −jsinc(f + 10) sin(3π(f + 10)) − jsinc(f − 10) sin(3π(f − 10)).

(c) Yc (f ) = Xc (f ) ⋆ P (f ) = 2sinc(4(f + 10)) + sinc(2(f + 10)) + 2sinc(4(f − 10)) + sinc(2(f −


10)).

(d) Yd (f ) = Xd (f ) ⋆ P (f ) = 12 e−j3π(f +10) (sinc(f + 10) + 3sinc(3(f + 10))) +


1 −j3π(f −10)
2e (sinc(f − 10) + 3sinc(3(f − 10))).

Exercise 4-20
See Figure P4-20. xa (t) is (real and) even, so Xa (f ) will be real and even. xb (t) is (real and) odd, so
Xb (f ) will be imaginary and odd.

We will use the differentiation theorem and the time-delay theorem to obtain the Fourier transforms
Xa (f ) and Xb (f ).

(a) Suppose we have a triangle c(t), shown in the left-hand sketch below, and its derivative d(t),
shown in the right-hand sketch:
c(t) d(t)

1 1

−1 1 t −1 t
−1

d
Clearly then: dt c(t) = d(t), so using the differentiation theorem: (j2πf )C(f ) = D(f ), or C(f ) =
D(f )
(*).
(j2πf )
     
Then d(t) = −Π t−0.5 1 + Π t+0.5
1 , so D(f ) = sinc(f ) ejπf − e−jπf and since the entity between
brackets equals 2j sin(πf ), we obtain that D(f ) = 2jsinc(f ) sin(πf ), an imaginary and odd function
of f .9

We can then compute C(f ) from D(f ) using (*):

9
Note that sinc(f ) is even, sin(πf ) is odd, so their product is odd, and that 2j is purely imaginary.
AE2235ii Exercises Topic I.3 (V2.0, May 2020) 11

1 1 sin(πf ) 2
C(f ) = (j2πf ) D(f ) = (j2πf ) 2jsinc(f ) sin(πf ) = sinc(f ) πf = sinc(f )sinc(f ) = sinc (f ),
a real and even function of f , as we expected since c(t) is a real and even function of t.

Now, xa (t) = c(t−1)+c(t+1), so Xa (f ) = C(f )e−j2πf +C(f )ej2πf , thus Xa (f ) = 2sinc2 (f ) cos(2πf ),
a real and even function of f , as we expected since xa (t) is real and even.

(b) The Fourier transform of xb (t) follows directly from what we have computed above. Since
xb (t) = −c(t + 1) + c(t − 1) we get:

Xb (f ) = −C(f )ej2πf + C(f )e−j2πf = −2jsinc2 (f ) sin(2πf ),

an imaginary and odd function of f , as we expected.10

10
Note that sinc2 (f ) is even, sin(2πf ) is odd, so their product is odd, and that 2j is purely imaginary.

You might also like