Exercises Topic I.

2: FOURIER SERIES
Exercise 3-3
(a) x1 (t) = cos2 (100πt) = 21 (1 + cos(200πt)) = 1
2 + 21 cos(200πt)

(b) x2 (t) = ej200πt = cos(200πt) + j sin(200πt)

(c) x3 (t) = sin(2πt) cos2 (10πt) = sin(2πt) 12 (1+cos(20πt)) = 1

2 sin(2πt)+ 21 sin(2πt) cos(20πt)

we then use sin(u) cos(v) = 21 (sin(u − v) + sin(u + v)), so:

1
x3 (t) = 2 sin(2πt) + 41 sin(−18πt) + 14 sin(22πt) = 1
2 sin(2πt) − 41 sin(18πt) + 14 sin(22πt)

(d) Can be ignored.

Exercise 3-4
Usually, first we make a sketch of the signal, to see what it looks like:
x(t)
A

− T20 T0
2 t
0

−A
− T40 T0
4

This signal is real and even, so in the Fourier series we only need the ‘cosine’-terms.
TR
0 /2
a0 = x(t)dt = 0
−T0 /2
TR0 /2
am = T20 x(t) cos(mω0 t)dt
−T0 /2
!
−TR0 /4 TR
0 /4 TR
0 /2
2
= T0 −A cos(mω0 t)dt + A cos(mω0 t)dt + −A cos(mω0 t)dt
−T0 /2 −T0 /4 T0 /4
−T /4 T0 /4 T /2
= mω−2A
0 T0
sin(mω0 t)|−T00 /2 + mω2A
0 T0
sin(mω0 t)|−T 0 /4
−2A
+ mω 0 T0
sin(mω0 t)|T00 /4
 
= 2A
mω0 T0 sin(mω0 T40 ) − sin(mω0 T20 ) + sin(mω0 T40 ) + sin(mω0 T40 ) − sin(mω0 T20 ) + sin(mω0 T40 )
 
= 2A
mω0 T0 4 sin(mω0 T40 ) − 2 sin(mω0 T20 ) (*1)

2π 2A
2 sin(m π2 ) − sin(mπ) .


(*1) can be simplified by noting that ω0 = T0 , so am = mπ

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

The last term on the right is 0 for all values of m and can be discarded. When m is even (including
m = 0), the first term on the right-hand side also equals 0, so am = 0. Furthermore:

For m = 1, 5, 9, 13, . . . : sin(m π2 ) = 1

For m = 3, 7, 11, 15, . . . : sin(m π2 ) = −1

 m’s, see the definition of the trigonometric FS (3-6)!!

We only use the positive
4A
 mπ
 : m = 1, 5, 9, 13, . . .
Concluding: am = :
 − 4A : m = 3, 7, 11, 15, . . .

mπ
We see that we only need the odd coefficients, all even coefficients (including m = 0) are zero, which is
the result of the signal being half-wave odd symmetric, see Exercise 3-16(c).

NEW Exercise 3-6

(waveform 1) See Example 3-6.

(waveform 2) See the extensive explanation of this waveform in the worked-out exam of August
2017, where I show that the Ziemer solution is wrong; the right answer is:

Xn = 2A ∀n.
π(1 − 4n2 )
(waveform 3) See Example 3-5.

(waveform 4) We will do this in two ways. First, let’s use the results we obtained in Exercise 3-4
for the real trigonometric FS (the am ’s and bm ’s, here we will have n = m) and convert them into the
complex FS using (3-32a) and (3-32b):
(
1
Xn = 2 (an − jbn ) : n>0
(3-32a)
1
2 (a−n + jb−n ) : n<0
X0 = a 0 (3-32b)

We directly obtain the desired result by substitution of the an ’s and bn ’s in this formula. It means that
when you are feeling comfortable with working with the real FS, you can easily go to the complex FS.
However, you will miss out some of the important advantages of working with the complex FS, like the
differentiation and time-shift theorems which I will use below a lot (practice these well).

For the second way, let’s look at Example 3-4, the odd square wave form, where after t = 0 the ampli-
tude A is positive. In the lecture slides we extended this example to the complex FS, see Example 3-4
(extended). Let’s call this odd square waveform y(t), we found that its complex FS coefficients are:

−j 2A
(
Yn = nπ : n odd
0 : n even
Now we have the even square wave form, and it is explained in Example 3-7 how to move from the
even waveform to the odd waveform. Let’s move back. The even square wave of Table 3-1, let’s call it
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 3

x(t), is basically the odd square wave y(t) shifted to the left in time with T0 /4: x(t) = y(t + T0 /4).

Shifting functions along the time axis means that the complex Fourier Series coefficients will be rotating
in the complex plane, see Example 3-7:

x(t) = y(t − t0 ) −→ Xn = Yn · e−jnω0 t0 (3-42)

Here we get: Xn = Yn · e−jnω0 (−T0 /4) = Yn e+jnπ/2 ,

2π T0
where we substitute −T0 /4 because y(t+T0 /4) = y(t−(−T0 /4)), and where ω0 (T0 /4) = T0 4 = π/2.

Since all Yn ’s are zero for even-indexed n we can focus on the odd-indexed n. Just make a table to see
where we are going, for a subset of n’s located around n = 0:

n ejnπ/2 Yn Xn
−9 −j j 2A
9π + 2A
9π
−7 j j 2A
7π − 2A
7π
−5 −j j 2A
5π + 2A
5π
−3 j j 2A
3π − 2A
3π
−1 −j j 2A
π + 2A
1π
1 j −j 2A
π + 2A
1π
3 −j −j 2A
3π − 2A
3π
5 j −j 2A
5π + 2A
5π
7 −j −j 2A
7π − 2A
7π
9 j −j 2A
9π + 2A
9π
Always check that all complex FS coefficients are indeed each other’s complex conjugates!

Hence, we see the pattern emerging of Table 3-1, and get:

2A

 : n = ±1, ±5, ±9, . . .
 |n|π


Xn = −2A : n = ±3, ±7, ±11, . . . qed


 |n|π

0 : n even
(waveform 5) See also Exercise 3-7 below, here we only will do a brief version of what is ex-
plained in more detail there. The odd square wave form is the first derivative of this waveform 5. Let’s
d
call waveform 5 x(t) and the odd square wave form y(t), then y(t) = dt x(t). The derivative is sketched
below with Exercise 3-7, we see that it has magnitude B = 4A/T0 .

An odd waveform with magnitude B has the following complex FS:1

j 2B
(
nπ : n odd
Yn = (*2)
0 : n even
Substitute B = 4A/T0 in (*2) and we get for its complex FS coefficients:

1
Note that this odd waveform, after t = 0 first goes to negative amplitude −B. This means that the signal complex FS
coefficients (as we had when deriving waveform 4 above) also all change their sign.
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 4

8A
(
j nπT : n odd
Yn = 0
0 : n even
d
Now because y(t) = dt x(t), Yn = (jnω0 )Xn , so Xn = Yn /(jnω0 ). Then, because:

1 8A 4A
jnω0 j nπT0 = n2 π 2 ,

we get:

4A
(
: n odd
Xn = n2 π 2 qed
0 : n even

NEW Exercise 3-7

(a) This proof is written-out in the lecture slides on Topic I.2, slide 64/66 (version 2020).

(b) Waveform 5 in Table 4-1 is a triangular wave, let’s call it x(t) and sketch it:

x(t)
A
− T20 T0
t
−T0 2 T0
0
−A

When you would take the first (time-)derivative of this function x(t), we get the function y(t), sketched
next:
y(t)
B

− T20 T0
t
−T0 2 T0
0

−B

The constant B can be easily computed: B = 2A = 4A

T0 .
T0 /2
d
We see that y(t) = dt x(t) is an odd square wave, very similar to the even waveform number 4 in Table
3-1, except that they use A for the magnitude of the signal, rather than B. Let’s call this even waveform,
the even square wave, z(t), and describe it with magnitude B.

Then: y(t) = z(t + T0 /4): y(t) is a time-shifted version of z(t). We will come back to
this later.

OK, now let’s first look at the complex Fourier series coefficients of y(t) and see how they can be easily
computed from the triangular waveform x(t). From Table 4-1 we can read that:
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 5

4A
(
: n odd
Xn = n2 π 2
0 : n even
Note that all even-indexed FS coefficients (including n = 0) are zero, which is a consequence of the fact
that this signal x(t) is half-wave odd.

d
Then: y(t) = dt x(t) −→ Yn = (jnω0 )Xn (3-82)

And we rapidly find, since (jn 2π 4A 8A

T0 ) n2 π 2 = j nπT0 , and since B = 4A/T0 , A = BT0 /4:

j 2B
(
nπ : n odd
Yn = (*3)
0 : n even
That was easy, right??!! Much better than solving the integrals again :)

Let’s check this result (*3) using the fact that we know the complex FS coefficients of the even waveform
z(t) to be (Table 4-1, waveform 4):

2B

 : n = ±1, ±5, ±9, . . .
 |n|π


Zn = −2B : n = ±3, ±7, ±11, . . .


 |n|π

0 : n even
Remember that we now use B as the magnitude for z(t), not A. Since y(t) = z(t + T0 /4), we can apply
the phase-shift theorem of the complex FS (study Example 3-7):
−T0
Yn = Zn · e−jnω0 ( 4
)
= Zn e+jnπ/2 (*4)

To rapidly move from Zn to Yn using this relation (*4), just make a table:

n ejnπ/2 Zn Yn
2B 2B
−7 j − 7π − 7π j
2B
−5 −j 5π − 2B
5π j
2B 2B
−3 j − 3π − 3π j
2B
−1 −j π − 2B
1π j
2B 2B
1 j π 1π j
3 −j − 2B 3π
2B
3π j
2B 2B
5 j 5π 5π j
7 −j − 2B 7π
2B
7π j
Note that we do not have to look at the even-indexed coefficients, which all remain to be zero under (*4).
Further note that we only need to look at a small subset of n’s to ‘get the idea of what is happening’.
Finally, note that when dealing with real-vlued phenomena, as we do in this lecture, the complex FS
coefficients X−k and Xk (for positive integer k) should always be each other’s complex conjugate.

As a final result we get:

j 2B
(
nπ : n odd
Yn =
0 : n even
The application of the “FS derivative-theorem”, (3-82), indeed yields the correct result. No need for
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 6

re-doing all calculations when you know what you’re dealing with!

NEW Exercise 3-10

Hey... I am not crazy (well...), this is what we did in Exam Question Q13, August 2018, which has
been fully worked out for you already. Note that here in Exercise 3-10 the average is zero, and the time
multiplier is not 2 but A. The rest is the same :)

NEW Exercise 3-11

(a) The fundamental frequency of the complex exponential FS equals ω0 = 32 π. So f0 = 3
4 Hz and
the period T0 equals 1.3333 sec.

(b) Fill in n = 0 and we get X0 = 1, which equals the signal average.

(c) The third harmonic (n = ±3) equals:

3 3
X−3 e−j3 2 πt + X3 e+j3 2 πt
 
9
Which can be written as a cosine: 2|X3 | cos 2 πt + 6 (X3 ) .

The magnitude and phase of the third component can be computed using the real and imaginary parts:
− j3π = 1 − j3π
X3 = 1 +1j3π = 1 +1j3π 11 − j3π 1 + 9π 2
So Re{X3 } = 1/(1 + 9π 2 ) and Im{X3 } = −3π/(1 + 9π 2 ), with which we can compute the magnitude
and phase:

|X3 | = |X−3 | = p 1
1 + 9π 2
6 X3 = −6 X−3 = − arctan(3π)
√
We see that the amplitude of the third harmonic equals 2/ 1 + 9π 2 .

(d) The phase of the third harmonic equals − arctan(3π).

(e) See also (c) and (d), we obtain:

2
 
9
p cos 2π − arctan(3π) .
1 + 9π 2

NEW Exercise 3-13

We will do the three parts of this exercise in two different ways, which are equivalent.

1
(a) xa (t) = cos2 (20πt) sin(10πt) = 2 sin(10πt) + 12 cos(40πt) sin(10πt)
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 7

Use the trigonometric property: sin(u) cos(v) = 21 (sin(u − v) + sin(u + v)):

1
x(t) = 2 sin(10πt) + 14 sin(−30πt) + 14 sin(50πt) = 1
2 sin(10πt) − 41 sin(30πt) + 14 sin(50πt)

We recognize frequencies of 5, 15 and 25 Hz, so the highest common frequency is 5 Hz. We can write
the equation using the fundamental frequency ω0 = 2π · 5 = 10π rad/s:

1
x(t) = 2 sin(ω0 t) − 41 sin(3ω0 t) + 14 sin(5ωo t)

1 ju
Use Euler’s formula: sin(u) = 2j (e − e−ju ), re-arrange from negative to positive frequencies:

x(t) = 18 je−j5ω0 t − 81 je−j3ω0 t + 41 je−j1ω0 t − 41 je+j1ω0 t + 81 je+j3ω0 t − 81 je+j5ω0 t ,

so we can read-off the complex Fourier Series coefficients Xn quickly, and put them in a table:

n −5 −3 −1 1 3 5
1
Xn 8j − 81 j 1
4j − 41 j 1
8j − 18 j
Check that these are complex conjugates!

(b) Again, let’s continue working with the cosine and sine terms until we get a nice summation of terms:

xb (t) = sin3 (30πt) + 2 cos(25πt) = sin(30πt) 21 (1 − cos(60πt)) + 2 cos(25πt)

= 12 sin(30πt) − 21 sin(30πt) cos(60πt) + 2 cos(25πt)

Use the trigonometric property: sin(u) cos(v) = 12 (sin(u − v) + sin(u + v)):

xb (t) = 12 sin(30πt) − 41 sin(−30πt) − 14 sin(90πt) + 2 cos(25πt)

= 34 sin(30πt) − 41 sin(90πt) + 2 cos(25πt)

We recognize frequencies 15, 45 and 12.5 Hz, so the highest common frequency is 2.5 Hz, ω0 =
2π · 2.5 = 5π rad/s. Let’s use Euler’s formula to move to the complex exponentials:

xb (t) = − 83 j ej6ω0 t − e−j6ω0 t + 18 j ej18ω0 t − e−j18ω0 t + 1 ej5ω0 t + e−j5ω0 t

Arrange from highest negative to highest positive frequency:

xb (t) = − 81 je−j18ω0 t + 83 je−j6ω0 t + e−j5ω0 t + e+j5ω0 t − 38 jej6ω0 t + 81 jej18ω0 t

Now we can read-off the complex Fourier Series coefficients Xn quickly, and put them in a table:

n −18 −6 −5 5 6 18
Xn − 81 j + 38 j 1 1 − 38 j + 18 j
Check that these are complex conjugates!

(c) Let’s try a different way: directly move to the complex exponential basis functions.

xc (t) = sin2 (40πt) cos2 (20πt) + sin(10πt) cos(5πt) = a(t) + b(t)

AE2235ii Exercises Topic I.2 (V2.0, May 2020) 8

1 2

2
2
a(t) = ( 2j ) ej40πt − e−j40πt ( 12 )2 ej20πt + e−j20πt
= 14 − 16
1
e−j120πt + ej120πt − 81 e−j80πt + ej80πt + 1
e−j40πt + ej40πt




16

1
ej10πt − e−j10πt 12 ej5πt + e−j5πt



b(t) = 2j
= − 14 j −e−j15πt − e−j5πt + e+j5πt + ej15πt

Add them together and arrange from highest negative to highest positive frequency:

1 −j120πt
xc (t) = − 16 e − 81 e−j80πt + 16
1 −j40πt
e + 41 je−j15πt + 41 je−j5πt + 1
4
1 j5πt 1 j15πt 1 j40πt
− 4 je − 4 je + 16 e − 81 ej80πt − 16
1 j120πt
e

We see frequencies of 2.5, 7.5, 20, 40 and 60 Hz, the highest common frequency is 2.5 Hz, so we obtain
ω0 = 5π rad/s:

1 −j24ω0 t
xc (t) = − 16 e − 81 e−j16ω0 t + 16
1 −j8ω0 t
e + 14 je−j3ω0 t + 41 je−jω0 t + 1
4
1 jω0 t 1 j3ω0 t 1 j8ω0 t
− 4 je − 4 je + 16 e − 81 ej16ω0 t − 16
1 j24ω0 t
e

so we can read-off the complex Fourier Series coefficients Xn quickly, and put them in a table:

n −24 −16 −8 −3 −1 0 1 3 8 16 24
1
Xn − 16 − 81 1
+ 16 + 14 j + 14 j + 14 − 14 j − 14 j 1
+ 16 − 18 1
− 16
Check that these (except X0 ) are complex conjugates!

Exercise 3-14
(a) y(t) = x(t − τ ), with τ a constant. In that case, see also Example 3-7 and the slides:

2π
Yn = Xn e−j2πnτ /T0 , with ω0 = T0 , we get:

Yn = Xn e−jnω0 τ .

(b) y(t) = ej2πf0 t x(t), with f0 a constant.

1 1
x(t)e−jnω0 t dt, and therefore also that Xn−1 = x(t)e−j(n−1)ω0 t dt.
R R
We know that Xn = T0 T0 T0 T0
Then:

1 1
y(t)e−jnω0 t dt = ej2πf0 t x(t)e−jnω0 t dt
R R
Yn = T0 T0 T0 T0

1
e−j(nω0 t−2πf0 t) x(t)dt
R
= T0 T0

1 1
x(t)e−j(nω0 t−ω0 t) dt = x(t)e−j(n−1)ω0 t dt = Xn−1 .
R R
with ω0 = 2πf0 we get: Yn = T0 T0 T0 T0

Concluding: Yn = Xn−1 .
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 9

NEW Exercise 3-15

(a) See the slides, it is recommended to do this once in your life. Perhaps twice.

(b) These proofs are very much the same, if not exactly the same as what we do in Exercise 3-16(a)
and (b), see below.

Exercise 3-16
Remember: an even function is a function where x(−t) = x(t), an example is the cosine function. An
odd function is a function where x(−t) = −x(t), an example is the sine function.

(a) To prove: Xn is a real and even function when x(t) is real and even. Note that in this lecture
series, all signals are real, we don’t deal with any complex signal.

1 1
x(t)e−jnω0 t dt =
R R
Xn = T0 T0 T0 T0 x(t) (cos(nω0 t) − j sin(nω0 t)) dt

1 R −1 R
Re(Xn ) = T0 T0 x(t) cos(nω0 t)dt, and Im(Xn ) = T0 T0 x(t) sin(nω0 t)dt.

−1 R R
Xn is real when its imaginary part is zero, so T0 T0 x(t) sin(nω0 t)dt = 0, or T0 x(t) sin(nω0 t)dt = 0.

This relation should hold for an integration over an arbitrary period T0 . For a function to be ‘odd’ or
‘even’, its behaviour for positive and negative time t is important, i.e., time relative to t = 0, so we take
the integral to be symmetric around t = 0:
TR
0 /2 R0 TR
0 /2
x(t) sin(nω0 t)dt = x(t) sin(nω0 t)dt + x(t) sin(nω0 t)dt = 0.
−T0 /2 −T0 /2 0

In the integral on the left-hand side, we substitute σ = −t, or dt = −dσ, and the integral limits change
as follows: t = −T0 /2 → σ = T0 /2, and t = 0 → σ = 0:

σ=0
R TR
0 /2
x(−σ) sin(−nω0 σ)d(−σ) + x(t) sin(nω0 t)dt = 0.
σ=T0 /2 0

Now, look at the part of the equation on the left-hand side: sin(−nω0 σ) = − sin(nω0 σ), reversing the
interval limits causes an additional minus sign, and d(−σ) = −dσ, so we get three minus signs on the
left-hand side, resulting in:
σ=T
R 0 /2 TR
0 /2
− x(−σ) sin(nω0 σ)dσ + x(t) sin(nω0 t)dt = 0.
σ=0 0
Then, we simply substitute σ = t on the left-hand side,2 and we obtain:

2
This step raised many questions from students. Why do we first subsititute σ = −t in the step above, and here σ = t again.
Well, the answer is simple. Both substitutions have nothing to do with each other. We could also substitute σ = a in the second
step, and call a the running variable. The ‘name of’ or ‘symbol used’ for the running variable in the integral is arbitrary, so why
don’t we take t? We are usually dealing with time signals, and the second integral also uses t. We could also substitute t = σ
in the second integral and then we get an equality in σ, it does not matter at all. Note that a substitution is just a mathematical
trick to (in this case) change the integral limits to make them more convenient. We can use this trick as many times as we like,
sometimes two or three times after each other. But each substitution at every step has nothing to do really with any of the other
substitutions, we just repeat substituting things until we are satisfied.
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 10

TR
0 /2 TR
0 /2
x(−t) sin(nω0 t)dt = x(t) sin(nω0 t)dt.
t=0 0

This equality holds when x(−t) = x(t), or in other words, when x(t) is even. This proves that when
x(t) is even, the imaginary part of the complex FS will be zero.

(b) To prove: Xn is an imaginary and odd function when x(t) is real and odd. This proof is very
similar to the one given just above. Xn is imaginary when its real part equals zero:
TR
0 /2 R0 TR
0 /2
x(t) cos(nω0 t)dt = x(t) cos(nω0 t)dt + x(t) cos(nω0 t)dt = 0.
−T0 /2 −T0 /2 0

In the integral on the left-hand side, we substitute σ = −t, or dt = −dσ, and the integral limits change
as follows: t = −T0 /2 → σ = T0 /2, and t = 0 → σ = 0:

σ=0
R TR
0 /2
x(−σ) cos(−nω0 σ)d(−σ) + x(t) cos(nω0 t)dt = 0, therefore:
σ=T0 /2 0
σ=T
R 0 /2 TR
0 /2
x(−σ) cos(nω0 σ)dσ + x(t) cos(nω0 t)dt = 0, just substitute σ = t on the left-hand
σ=0 0
side, and we obtain:
TR
0 /2 TR
0 /2
x(−t) cos(nω0 t)dt = − x(t) cos(nω0 t)dt.
t=0 0

This equality holds when x(−t) = −x(t), or in other words, when x(t) is odd. This proves that when
x(t) is odd, the real part of the complex FS will be zero.

(c) From the definition of the FS we know that:

1 1
x(t)e−jnω0 t dt =
R R
Xn = T0 T0 T0 T0 [x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt

R0 TR
0 /2
1
So: Xn = T0 [x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt + T10 [x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt.
−T0 /2 0

Again we choose the integration interval to be symmetric around t = 0. But now we substitute in the
integral on the left-hand side σ = t + T0 /2 (or: t = σ − T0 /2). Then dσ = dt and the integration limits
change as follows: t = −T0 /2 becomes σ = 0, and t = 0 becomes σ = T0 /2. We get:
σ=T
R 0 /2
1
Xn = T0 [x(σ − T0 /2) cos(nω0 (σ − T0 /2)) − jx(σ − T0 /2) sin(nω0 (σ − T0 /2))]dσ
σ=0
TR
0 /2
+ T10 [x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt
0

Now substitute in the integral on the left-hand side σ = t (we just change the variable, nothing fancy):
TR
0 /2
1
Xn = T0 [x(t − T0 /2) cos(nω0 (t − T0 /2)) − jx(t − T0 /2) sin(nω0 (t − T0 /2))]dt
t=0
TR
0 /2
+ T10 [x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt
0
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 11

Use the half-wave symmetry property: x(t + T0 /2) = −x(t), so because the signal is periodic with T0 ,
x(t − T0 /2) = −x(t) also holds, we then re-arrange:
TR
0 /2
1
Xn = T0 [x(t) (cos(nω0 t) − cos(nω0 (t − T0 /2))) − jx(t) (sin(nω0 t) − sin(nω0 (t − T0 /2)))]dt.
t=0
(*5)

Now: cos(nω0 (t−T0 /2)) = cos(nω0 t−nω0 T0 /2) = cos(nω0 t) cos(nω0 T0 /2)+sin(nω0 t) sin(nω0 T0 /2)
(*6), with nω0 T0 /2 = n 2π
T0 T0 /2 = nπ. Then, for all n the sine-terms in (*6) equal zero. For even n
we obtain that cos(nω0 (t − T0 /2)) = cos(nω0 t), and for odd n we obtain that cos(nω0 (t − T0 /2)) =
− cos(nω0 t).

Similarly, sin(nω0 (t−T0 /2)) = sin(nω0 t−nω0 T0 /2) = sin(nω0 t) cos(nω0 T0 /2)−cos(nω0 t) sin(nω0 T0 /2)
(*7), again with nω0 T0 /2 = n 2π
T0 T0 /2 = nπ. Then, for all n the sine-term on the right-hand side of (*7)
equals zero. For even n we obtain that sin(nω0 (t − T0 /2)) = sin(nω0 t), and for odd n we obtain that
sin(nω0 (t − T0 /2)) = − sin(nω0 t).

We substitute these results into our main result so far, equation (*5), to see that, for even n:
TR
0 /2
1
Xn = T0 [x(t)(cos(nω0 t) − cos(nω0 t)) − jx(t)(sin(nω0 t) − sin(nω0 t))]dt = 0. qed
t=0

For odd n we get:

TR
0 /2
1
Xn = T0 [x(t)(cos(nω0 t) + cos(nω0 t)) − jx(t)(sin(nω0 t) + sin(nω0 t))]dt, so:
t=0
TR
0 /2
2
Xn = T0 [x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt.
t=0

Note that the ‘half-wave’ symmetry says nothing about the fact whether a signal is even or odd.3
3
From the proof in (c) we can also find another way to state the proofs in (a) and (b). We start with:

R0 T0 /2
R
1 1
Xn = T0
[x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt + T0
[x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt
−T0 /2 0

In the first integral, we substitute σ = −t, so dt = −dσ, the integral limits change from (t = −T0 /2 to 0) to (σ = T0 /2 to 0):

R0 T0 /2
R
1 1
Xn = T0
[x(−σ) cos(−nω0 σ) − jx(−σ) sin(−nω0 σ)](−dσ) + T0
[x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt
T0 /2 0

We re-arrange:
T0 /2
R T0 /2
R
1 1
Xn = T0
[x(−σ) cos(nω0 σ) + jx(−σ) sin(nω0 σ)]dσ + T0
[x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt
0 0

And then substitute σ = t on the left-hand side:

T0 /2
R T0 /2
R
1 1
Xn = T0
[x(−t) cos(nω0 t) + jx(−t) sin(nω0 t)]dt + T0
[x(t) cos(nω0 t) − jx(t) sin(nω0 t)]dt
0 0
T0 /2
R T0 /2
R
1
Re-arranging gives: Xn = T0
[x(t) + x(−t)] cos(nω0 t)dt − j T10 [x(t) − x(−t)] sin(nω0 t)dt
0 0

From which we can see that if x(t) is even, the right-hand side becomes zero, and when x(t) is odd, the left-hand side
becomes zero. qed.
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 12

Exercise 3-17
First of all, let’s closely look at Figure P3-17. From these figures we can already deduce that:

• ... all signals have average 0, so for all of them X0 = 0, and

• ... all signals are half-wave odd symmetric (x(t + T0 /2) = −x(t)), and therefore Xn = 0 for even
n (see Exercise 3-17(c)).

Second, furthermore we see that:

• Signal xa (t) is real and odd, so all its Fourier series coefficients are purely imaginary, we do not
need the ‘cosine’-terms.

• Signal xb (t) is real and even, so all its Fourier series coefficients are real, we do not need the
‘sine’-terms (see also Exercise 3-4 where we compute the FS analytically).

• Signal xc (t) is neither odd, nor even, so its Fourier series coefficients are all complex.

• Signal xd (t) is real and odd, so all its Fourier series coefficients are purely imaginary, we do not
need the ‘cosine’-terms.

• Signal xe (t) is real and even, so all its Fourier series coefficients are real, we do not need the
‘sine’-terms.

• Signal xf (t) is neither odd, nor even, so its Fourier series coefficients are all complex.

So, we conclude with the following table:

Waveform
Property a b c d e f
Purely real coefficients X X
Purely imaginary coefficients X X
Complex coefficients X X
Even-indexed coefficients zero X X X X X X
X0 = 0 X X X X X X

Exercise 3-18
Given the waveform x(t) illustrated in Figure P3-18.

(a) The coefficient a0 represents the mean of a signal. In this case, since the mean of x(t) equals
zero, a0 = 0.

(b) The bm -coefficients belong to the ‘sine’-components in the trigonometric Fourier series. Since
x(t) is a real and even function, we do not need the ‘sine’-components to construct the signal model, and
therefore all bm ’s must be zero.

(c) Since x(t) is a real and even function, we only need the ‘cosine’-components in the FS to
model this signal, i.e., the am ’s. From equations (3-32a,b) we know the relation between the complex
FS coefficients Xm and the trigonometric (real) FS:
AE2235ii Exercises Topic I.2 (V2.0, May 2020) 13

(
1
Xm = 2 (am − jbm ) : m>0
(3-32a)
1
2 (a−m + jb−m ) : m<0
X0 = a 0 (3-32b)

Clearly, since all bm ’s are zero, Xm equals am /2 for positive m and a−m /2 for negative m. All am ’s are
real, and therefore also all complex FS coefficients Xm will be real. It is simply a consequence of the
fact that x(t) is even.

(d) If it does, then x(t + T0 /2) must equal −x(t). From Figure P3-18 we can see that this is indeed
the case. As a consequence, all even-indexed coefficients in the FS will be zero (see Exercise 3-16(c) for
the proof).