New Basis Functions

Recall that the Fourier series builds a representation

composed of a weighted sum of the following basis
Section 8 functions.
1 (i.e. a constant term)
cos(t) cos(2t) cos(3t) cos(4t) ...
Complex Fourier Series sin(t) sin(2t) sin(3t) sin(4t) ...
Computing the weights an , bn and c often involves
some nasty integration.
The complex Fourier series is presented first with pe-
riod 2π, then with general period. We now present an alternative representation based
on a different set of basis functions:
The connection with the real-valued Fourier series is 1 (i.e. a constant term)
explained and formulae are given for converting be- eit e2it e3it e4it ...
tween the two types of representation. e−it e−2it e−3it e−4it ...

These can all be represented by the term

Examples are given of computing the complex Fourier
series and converting between complex and real se-
rieses. eint
with n taking integer values from −∞ to +∞. Note
that the constant term is provided by the case when
n = 0.
111 112
 Complex Fourier Series
Series of Complex Exponentials
∞
X
A representation based on this family of functions is f (t) = d + [an cos(nt) + bn sin(nt)]
n=1
called the “complex Fourier series”.
∞ "
eint + e−int eint − e−int
! !#
∞ X
f (t) =
X
cneint =d+ an + bn
n=1 2 2i
n=−∞
The coefficients, cn, are normally complex numbers. ∞ ∞
X (an − ibn) int X (an + ibn) −int
=d+ e + e
n=1 2 n=1 2
It is often easier to calculate than the sin/cos Fourier
∞
series because integrals with exponentials in are usu-
cneint
X
=
ally easy to evaluate. n=−∞

We will now derive the complex Fourier series equa- where

 d
 ,n=0
tions, as shown above, from the sin/cos Fourier series
cn = (an − ibn) /2 , n = 1, 2, 3, . . .
using the expressions for sin() and cos() in terms of 

(a−n + ib−n) /2 , n = −1, −2, −3, . . .
complex exponentials.
Note that a−n and b−n are only defined when n is
negative.

113 114
 1 π cos(nt)f (t) dt
an = π
R
−π
1 Rπ
bn = π −π sin(nt)f (t) dt
d = 2π 1 R π f (t) dt
−π
thus for n positive
1
cn = (an − ibn)
2 Z
1 π
= [cos(nt) − i sin(nt)] f (t) dt Complex Fourier Series Summary
2π Z−π
1 π −int
= e f (t) dt
2π −π 1 π −int
Z
cn = e f (t) dt
2π −π
for n negative
∞
cneint
X
1 f (t) =
cn = (a−n + ib−n)
2 Z n=−∞
1 π
= [cos(−nt) + i sin(−nt)] f (t) dt
2π Z−π
1 π −int
= e f (t) dt
2π −π

and for n = 0
c0 = d
1 π −0
Z
= e f (t) dt
2π −π
115 116
 Complex Series Example 1
Finding the limit as n tends to 1
Find the complex Fourier series to model f (t) = sin(t).
1 π −int
Z
cn = e f (t) dt
2π −π 1 einπ − e−inπ
" #
cn =
1 π −int n2 − 1
Z
= e sin(t) dt 2π
2π −π
1 einπ − e−inπ
" #
Set n = 1 + ǫ and let ǫ tend to zero.
=
2π n2 − 1  
1  eiπ(1+ǫ) − e−iπ(1+ǫ) 
Which is zero when n does not equal 1 or −1. For c1 =
these two special cases we have to set n = 1 + ǫ 2π (1 + ǫ)2 − 1
and calculate the limit of cn as ǫ tends to zero. This 1 −eiπǫ + e−iπǫ
" #

gives us =
2π (1 + ǫ)2 − 1
1
1 −1 − iπǫ + 1 − iπǫ
 
c1 =
2i ≈
2π 1 + 2ǫ − 1
−1
c−1 = 1 −2iπǫ
 
2i ≈
Which means the complex Fourier series for f (t) = 2π 2ǫ

sin(t) is −i
∞ ≈
2
cneint
X
f (t) =
n=−∞ 1
≈
eit − e−it 2i
=
2i
117 118
 Complex Series Example 2 Converting c to a, b and d
Find the complex Fourier f(x) From our example on the previous page.
series to model f (x) that 1  h i
has a period of 2π and is 1 i −inT − 1 , when n 6= 0
 2πn e


when 0 < x < T and zero cn =
when T < x < 2π. x  1 area = T


, when n = 0
T 2π 2π 2π

1 π −int
Z
cn = e f (t) dt We wish to calculate the coefficients for the equivalent
2π −π
i h −inT i Fourier series in terms of sin() and cos().
= e − 1 , when n 6= 0
2πn
T . For n > 0
Clearly d = c0 = 2π
1 T
= area = , when n = 0
2π 2π
cn = (an − ibn)/2
⇒ an = 2 Re{cn}
So the Fourier series is
∞ and bn = −2 Im{cn}
cneint
X
f (t) = converting our expression for cn into sin() and cos():
n=−∞
i

−1 2cn = [cos(nT ) − i sin(nT ) − 1]
1  i h −inT πn
− 1 eint
X i
= T+ e 1
2π  n=−∞ n = [sin(nT ) + i(cos(nT ) − 1)]
∞

πn
i −inT 
− 1 eint
X h i
+ e sin(nT ) 1 − cos(nT )
n=1 n
 so an = and bn = .
nπ nπ
119 120
 Converting from Real to Complex
Convert the real Fourier se-
f(t)
1
ries of the square wave
Complex Fourier Series
f (t) to a complex series. −1 t
 −2π −π π 2π
−1
1  i h −inT
− 1 eint
X i
f (t) = T+ e For the real series, we know that d = an = 0 and
2π  n=−∞ n
1 π 4
 Z
∞ bn = sin(nt)f (t) dt = , n odd
i h −inT 
− 1 eint
X i
+ e π −π nπ
n=1 n

4 sin(3t) sin(5t)
h i
giving f (t) = π sin(t) + 3 + 5 + . . .

Real Fourier Series To convert

 to a complex series, use
∞  d ,n=0
T sin(nT )

X
f (t) = + cos(nt) cn = (an − ibn) /2 , n = 1, 2, 3, . . .
2π n=1 nπ 

(a−n + ib−n) /2 , n = −1, −2, −3, . . .
∞
X 1 − cos(nT ) so we have
+ sin(nt)
n=1 nπ c0 = 0
cn = −2i/(nπ) , n positive and odd
cn = 2i/(−nπ) , n negative and |n| odd

e−5it e−3it e−it

"
Both serieses converge as 1/n. −2i
⇒ f (t) = ... + + +
π −5 −3 −1
eit e3it e5it
#
+ + + + ...
1 3 5
121 122
 Example 1
General Complex Series
A even function f (t) is periodic with period L = 2,
For period of 2π and f (t) = cosh(t − 1) for 0 ≤ t ≤ 1. Find a
1 2π −int
Z complex Fourier series representation for f (t).
cn = e f (t) dt
2π 0
∞ f(t)
cneint
X
f (t) =
n=−∞
1
Similarly, for period L t
1 L −inx 2π −2 −1 0 1 2 3 4
Z
cn = e L f (x) dx
L 0
∞ 2π
cneinx L
X
f (x) =
1 L −int 2π
Z
n=−∞ cn = e L f (t) dt
L 0
1 2 −intπ
Z
= e cosh(t − 1) dt
2 0
The fraction 2π
L is often written as ω0 and called the
sinh(1)
fundamental angular frequency. =
1 + n2 π 2

123 124
 Example 2
Hence the complex Fourier series is
Find the complex Fourier
f(x)
∞ 2π 1
cneint L
X
f (t) = series of the the square
n=−∞ wave f (x). −L −1 x
∞
sinh(1)eintπ
X
L
= 2 2
n=−∞ 1 + n π Note that the mean of the function is zero, so c0 = 0.
1 L −inx 2π
Z
cn = e L f (x) dx
We can check this answer by computing the equiv- L 0
alent real Fourier series which we calculated at the
"Z #
1 L/2 Z L
−inx 2π −inx 2π
start of section 7. = e L dx − e L dx
L 0 L/2
an = 2 Re{cn} , n = 1, 2, 3, . . . 1 h −2inπ
+ 1 − 2e−inπ
i
bn = −2 Im{cn} , n = 1, 2, 3, . . . = e
2inπ
d = c0 h i
∞ 1 − e−inπ 2π
einx L
X
In this case, as cn is entirely real, f (x) =
n = −∞ inπ
2 sinh(1) n 6= 0
an = 2cn = , n = 1, 2, 3, . . .
1 + n2 π 2  2π 2π 2π
2  e−5ix L e−3ix L e−ix L
bn = 0 f (x) = ... + + +
iπ −5 −3 −1
ix 2π 3ix 2π 5ix 2π

d = sinh(1) e L e L e L
+ + + + . . .
1 3 5

125 126
 Converting to a Real Series
Section 8: Summary
We wish to convert the complex general range square
wave series into a series with real coefficients. For period L
1 L −inx 2π
( Z
2/(inπ) , |n| odd cn = e L f (x) dx
cn = L 0
0 , |n| even
∞ 2π
cneinx L
X
Clearly d = c0 = 0. For a and b use: f (x) =
n=−∞
cn = (an − ibn)/2
⇒ an = 2 Re{cn} = 0 Relationship with the cos/sin Fourier series.
4
and bn = −2 Im{cn} = , n odd

nπ  d
 ,n=0
cn = (an − ibn) /2 , n = 1, 2, 3, . . .
Which gives us the real series:
(a−n + ib−n) /2 , n = −1, −2, −3, . . .



sin 3x 2π
  
4 2π
 
L
f (t) = sin x +
π L 3 an = 2 Re{cn} , n = 1, 2, 3, . . .
sin 5x 2π
 
bn = −2 Im{cn} , n = 1, 2, 3, . . .

L
+ + . . . d = c0
5

127 128