F1.

3YF2

Mathematical Techniques 1
EXAMPLES 1: FOURIER SERIES

1. Find the Fourier series of each of the following functions

(i) f (x) = 1 x2 ,

1 < x < 1.

(ii) g(x) = |x|, < x < .


0 if 2 < x < 0
(iii) h(x) =
1 if 0 x < 2.
In each case sketch the graph of the function to which the Fourier series converges
over an x- range of three periods of the Fourier series.
2. Find the Fourier series for f (x) =
2
(i) 6 = 1 + 212 + 312 + 412 + 512 + . . . .
(ii)
(iii)

2
12
2
8

x2
4,

< x < . Hence deduce that

= 1 212 + 312 412 + 512 . . ..

= 1 + 312 + 512 + 712 + 912 + . . . .

3. Find the Fourier cosine series and the Fourier sine series for the function

f (x) =

1 if 0 < x < 1
0 if 1 x < 2.

4. Find the Fourier cosine series for the function f (x) = sin(x), 0 < x < .
What is the Fourier sine series for f ?
5. Suppose f : R R is a periodic function of period 2L with Fourier series

a0 + an cos(
n=1

nx
nx
) + bn sin(
).
L
L

(i) Using integration by parts, show that, if f 0 exists and is a bounded function on
R, then there exists a constant k such that |an | nk and |bn | nk for all n 1.
(ii) Show that, if f 00 exists and is a bounded function on R, then the Fourier series
for f is absolutely convergent for all x.

F1.3YF2 Fourier Series Solutions

EXAMPLES 1: FOURIER SERIES SOLUTIONS

1. (i) We
must
calculate
the Fourier coefficients.
1R1
1
2
a0 = 2 1 (1 x ) dx = 2 (x 13 x3 j1 1 = 23 .

R1
R1 2
1R1
2
1 (1 x ) cos(nx) dx =
1 cos(nx) dx
1 x cos(nx) dx
1
R
1
2 1
= 0 n x2 sin(nx) 1 1 + n
1 x sin(nx) dx
2
2 R1
1
= 0 0 n2 2 x cos(nx) 1 + n2 2
1 cos(nx) dx
4
=
cos(n) + 0 = ( 1)n+1 n242 :
n2 2
R
Also bn = 11 11 (1 x2 ) sin(nx) dx = 0 as integrand is odd.

an =

Hence
f (x ) =
(ii)

a0 =

1 R

1
1
cos(2x) + cos(3x)
4
9

::: :

jxj dx = 22 R0 x dx = 1 21 x2j0 = 2 .

jxj cos(nx)Rdx = 2 R0 x cos(nx) dx

2
2
= n
x sin(nx)j0 n
sin(nx) dx = 0 + n2 cos(nx)j0
0 4
if n is odd
an =

1 R

2
4
+
fcos(x)
3 2

= n2 [(

1)n

Also bn =

n2

1] =

1 R

0 if n is even.

jxj sin(nx) dx = 0 as integrand is odd.

Hence
g(x) =

4
1
1
f
cos(x) + cos(3x) + cos(5x) + : : :g:

9
25

1R2
1R2
1
2 h(x) dx = 4 0 dx = 2 .
4
R
R
1 2
nx
1 2
nx 2
an = 12 22 h(x) cos( nx
2 ) dx = 2 0 cos( 2 ) dx = 2 n sin( 2 ) 0 =
R
1R2
nx
1 2
nx 2
bn = 12 22 h(x) sin( nx
2 ) dx = 2 0 sin( 2 ) dx =
2 n cos( 2 ) 0

(iii)

a0 =

= n [1

cos(n)] =

0.

0 if n is even
2
n if n is odd.

Hence

1 2
x
1
3x
1
5x
+ fsin(
) + sin(
) + sin(
) + : : :g:
2
2
3
2
5
2
In (i) and (ii), if functions are extended as periodic functions, they are continuous at all points and
so the Fourier series converge to the function value at all points.
h (x ) =

Thus the Fourier series for f converges to the following function

F1.3YF2 Fourier Series Solutions

and the Fourier series for g converges to

In (iii), if function is extended as a periodic function, it is discontinuous at x = 0; 2; 4; thus the
Fourier series converges to 12 at these points and converges to the value of the function at all other
points.

2. Again calculating the Fourier coefficients we have

1 R x2
1 11 3
2
a0 = 2
4 dx = 2 4 3 x j = 12 .
R

1
an = 1 x4 cos(nx) dx = 1 1n x4 sin(nx)j 2n
x sin(nx) dx
R

1
1
1

= 0 + 2n2 x cos(nx)j
cos(nx) dx = n2 cos(n).
2n2

Also bn =

1 R x2
4

sin(nx) dx = 0 as the integrand is odd.

Since f is continuous when extended as a periodic function with period 2, the Fourier series for f
converges to f (x) for all x 2 R.

Hence

x 2 2
=
4
12
Setting x = , we have

cos(x) +
2
4

1
cos(2x)
22

1
1
cos(3x) + 2 cos(4x)
2
3
4

2
1
1
1
+1+
+
+
+ ::::
12
22 32 42

:::

F1.3YF2 Fourier Series Solutions

i.e.,
1+

1
1
1
2
+
+
+::: =
22 32 42
4

Setting x = 0, we have
0=

2
12

1+

2
12

1
22

1
1
+
2
2
32

(1) + (2)

1
1
+
2
4
52

gives
2(1 +

2
6

1
1
+
2
3
42

i.e.,
1

:::

(1)

1
+:::
52

2
12

(2 )

1
1
2 2
+
+ : : :) =
+
32 52
6
12

2
:
4

Hence

1
1
2
+
+::: =
:
32 52
8
nx
3. The Fourier cosine series of f is given by a0 +
n=1 an cos( 2 ) where
1+

a0 =
and
an

1
4

Z 2

1R2
2
2

f (x) dx =
2

Hence
f (x) =

Z 2

f (x) dx =
0

f (x) cos( nx
2 ) dx =

8
<
2
n
sin
(
)
=
n
2
:

1
2

R1
0

1
2

Z 1

dx =
0

cos( nx
2 ) dx =

2
n

1
2

1
sin( nx
2 )j0

0 if n is even
2
n if n = 4k + 1
2
n if n = 4k + 3:

1 2
x
+ fcos(
)
2
2

1
3x
1
5x
cos(
) + cos(
)
3
2
5
2

::: :

nx
The Fourier sine series of f is given by
n=1 bn sin( 2 ) where

bn

1R2
2
2

2
n (1

f (x) sin( nx
2 ) dx =

8
<
cos( n
))
=
2
:

R1
0

sin( nx
2 ) dx =

2
n

1
cos( nx
2 )j0

2=n if n is odd
0 if n = 4k
4=n if n = 4k + 2:

Hence
f (x) =

2
x
2
2x
1
3x
5x
2
6x
f
sin( ) + sin(
) + sin(
) + sin(
) + sin(
) + : : :g:

2
2
2
3
2
2
6
2

F1.3YF2 Fourier Series Solutions

4. The Fourier cosine series of f is given by a0 +

n=1 an cos(nx) where
a0 =

1
2

and

f (x) dx =

sin(x) dx = (
0

1
2

) : ( cos(x))j0 =

1
2
an =
f (x) cos(nx) dx =
sin(x) cos(nx) dx:

0
R
R
If n = 1, a1 = 2 0 sin(x) cos(x) dx = 1 0 sin(2x) dx = 0.
R
In n 6= 1, an = 1 0 fsin[(n + 1)x] + sin[(1 n)x]g dx
1
1
1
= f n+1 cos[(n + 1)x] + n 1 cos[(n 1)x]gj
0
1

= n+1 (1

cos[(n + 1)]) + n 1 1 (cos[(n

an = 2 ( n+1 1

Thus an = 0 if n is odd and

Hence
2
sin(x) =

1)]

1)g:

n 1) =

4
(n2 1)

if n is even.

4 1
1
1
f
cos(2x) + cos(4x) + cos(6x) + : : :g:
3
15
35

The Fourier sine series is simply f (x) = sin(x)

5(i) Since f 0 is a bounded function on R, there exists a constant c > 0 such that j f 0 (x)j  c for all
x 2 R. Now
Z
Z
1 L
nx
1
nx L
1 L 0
nx
f (x) sin(
an =
f (x) cos(
) dx =
)j L
f (x) sin(
) dx:
L L
L
n
L
n L
L
nx L
As x ! f (x) sin( nx
L ) has period 2L (or alternatively as sin(n) = 0), f (x) sin( L j L = 0.

Hence an =

1 RL
L
n

f 0 (x) sin( nx
L ) dx and so

nx
2Lc k
max LxL j f 0 (x) sin(
janj  2L
)j 
=
n
L
n
n
k
A similar argument shows that jbn j  n .

Since f 00 is a boundedR function on R, there exists a constant c1 such that j f 00 (x)j  c1 for all
1 L
nx
0
x 2 R. By above an = n
L f (x) sin( L ) dx

(ii)

= n2 2
=

L
f 0 (x) cos( nx
L )j L

L RL
L
n2 2

f 00 (x) cos( nx
L ) dx

L RL
nx
00
L f (x) cos( L ) dx (as x
n2 2

Hence

! f 0(x) cos( nxL ) has period 2L).

2L c1
k1
janj  n2L2 2 max LxL j f 00(x) cos( nx
)j 
=
2
2
L
n
n2
k
A similar argument shows that jbn j  n .

1
2

Thus

nx
2k1
jan cos( nx
) + bn sin(
)j  jan j + jbn j 
L
L
n2

nx
nx
1
Since
n=1 n2 converges, it follows from the comparison test that n=0 jan cos( L ) + bn sin( L )j
converges.
nx
nx
Hence
n=0 (an cos( L ) + bn sin( L )) converges absolutely.

Fourier series for output voltages of inverter waveforms.

The Fourier series for a periodic function vo(t) can be expressed as

vo (t ) = a o + a n cos(nt ) + bn sin(nt )
n =1

For an odd quarter-wave symmetry waveform,

ao = 0

an = 0

and
2
4
bn = vo sin(nt )d (t )
0
0

for odd n
for even n

Therefore, vo(t) can be written as

vo (t ) =
(i)

n = odd

sin( nt )

Square-wave

/2

bn =

+Vdc

sin(nt )d (t )

dc

sin(nt )d (t )

4Vdc
[ con(nt ]02
n
4V
= dc
n

-Vdc

(ii)

Quasi square-wave
+Vdc

bn =

-Vdc

/2

sin(nt )d (t )

dc

sin( nt )d (t )

4Vdc
[
]
=
con(nt 2
n
4V
= dc cos(n )
n

(iii)

Notched waveform (Harmonics Elimination PWM)

Vdc
1 2 3

/2

Vdc

bn =

sin(nt )d (t )

2
4

= Vdc sin(nt )d (t ) + Vdc sin(nt )d (t )

1
3

4Vdc
2
=
[ con(nt ]1 + [ con(nt ]23
n

4V
,
= dc cos(nt ) 1 3
2 ,
n
2

4Vdc
n
4V
= dc
n
=

(iv)

cos(n 1 ) + cos(n 3 ) cos(n 2 ) cos(n 2 )

[cos(n1 ) + cos(n 3 ) cos(n 2 )]

Sinusoidal PWM (unipolar and bipolar)

Unipolar SPWM waveform as an example

sin n

+ (k ) cos(t )
2
vo (t ) = M aVdc sin(t ) +

n
n =1
= M aVdc sin(t ) + Bessel Function for harmonic terms
4Vdc

The tables are required to resolve for the Bessel function for harmonic terms. The harmonics
in the inverter output appear as sidebands, centered around the switching frequency, that is,
around mf, 2mf, 3mf and so on. This general pattern hold true for all values of ma in the range
0 1 and mf>9. The unipolar SPWM switching scheme has the advantage of effectively
doubling the switching frequency as far as the output harmonics are concerned, compared to
the bipolar SPWM switching scheme. Because of that the harmonics in the inverter output of
unipolar SPWM are centered around 2mf, 4mf, 6mf and so on.
TABLE 8.3 NORMALIZED FOURIER COEFFICIENTS Vn/Vdc FOR BIPOLAR SPWM
Ma = 1 0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
n=1
1.00
0.90
0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10
n = mf
0.60
0.71
0.82 0.92 1.01 1.08 1.15 1.20 1.24 1.27
0.27
0.22 0.17 0.13 0.09 0.06 0.03 0.02 0.00
n = mf1 0.32
Table 8.3 shows the first harmonic frequencies in the output spectrum at and around mf for
the bipolar SPWM switching scheme. The harmonics at and around 2mf, 3mf, 4mf and so on
are not indicated.

TABLE 8.5 NORMALIZED FOURIER COEFFICIENTS Vn/Vdc FOR UNIPOLAR SPWM

Ma = 1 0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
n=1
1.00
0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10
0.25 0.31 035
0.37 0.36 0.33 0.27 0.19 0.10
n = 2mf 1 0.18
0.18 0.14 0.10 0.07 0.04 0.02 0.01 0.00 0.00
n = 2mf 3 0.21
Table 8.5 shows the first harmonic frequencies in the output spectrum at and around 2mf for
the uipolar SPWM switching scheme. The harmonics at and around 4mf, 6mf, 8mf and so on
are not indicated.
Table 8.3 and 8.5 can be used to predict the THD for the ouput current of the inverter
connected to RL load. Higher order harmonics are assumed to contribute little power and
effect, so they can be neglected. To evaluate the THD for the output voltage of the inverter,
higher order harmonics should be taken into account.

-tammat-