CEE490b Feb.

11, 2002

Time Series Analysis Concepts

Any periodic function x(t) can be represented by a Fourier Series

x(t)

π/2
π/2
∞
x(t ) = ao + ∑ (ar cos rω 1t + br sin rω 1t ) (1)
r =1
where

2π ω 1
ω1 = ; since ω 1 = 2πf1 , then f1 = 1 =
T 2π T
T ≡ the period of x(t)

The constants ao , ar and br are evaluated as:

T /2
1
T −T∫/ 2
ao = x (t )dt , i.e the mean value (2a)
T /2
2
T −T∫/ 2
ar = x (t ) cos rω 1dt (2b)
T /2
2
T −T∫/ 2
br = x(t ) sin rω 1dt (2c)

The mean square value of x(t) can also be expressed in terms of the coefficients
ao , ar and br , squaring and averaging by the period T.

T /2
1
∫
2 2
σ = x (t ) =
x x 2 (t )dt (3)
T −T / 2

substituting Eq. 1 in Eq. 3 we obtain

T /2 2
1  ∞ ∞

2
x
2
σ = x (t ) = ∫ 
T −T / 2 
a o + ∑ a r cos rω 1t + ∑ br sin rω 1t  dt (4)
r =1 r =1 

Page 1 of 3
CEE490b Feb. 11, 2002

All terms involving a product of sin and cosine terms will be zero and the only ones
left will be those containing ao2 and the terms involving either the product of cosines
or sines, i.e. (ak cos kω 1t )(al cos lω 1t ) or (bk sin kω 1t )(bl sin lω 1t ) and only when k= l .
So the mean square value of x(t) becomes:

1 2 ∞
 2T 2 T 
σ x2 = x 2 (t ) = 
T
a o T + ∑  a r + br  
r =1  2 2 
(5)
1 ∞
= ao2 + ∑ (ar2 + br2 )
2 r =1

Complex Fourier Series

We can alternatively express a periodic function using complex notation since:

e irω1t = cos rω 1t + i sin rω 1t (6)

Therefore:

∞
x (t ) = ∑c e
r = −∞
r
irω1t
(see Eq. 8.23) (7)

where:

T /2
1
T −T∫/ 2
cr = x (t )e − irω1t dt (8)

so the mean square amplitude (variance) becomes:

T /2 2
1 ∞ irω t 
∫ ∑ c r e 1  dt
2
x (t ) = (9)
T −T / 2  −∞ 

This can be rewritten as:

T /2
1 ∞ ∞
x (t ) = ∑
2
∑ ∫c n c m e inω1t e imω1t dt
T n = −∞ m = −∞ −T / 2
(10)
∞
= ∑ cr
2

−∞
2
where c r = c r c r* , the modulus of c r , and c r* is the complex conjugate of c r

Page 2 of 3
CEE490b Feb. 11, 2002

Fourier Integrals and Fourier Transforms

We can extend the Fourier Series concept to a random or a non-periodic function by

assuming that the period becomes very large. In the limit,

T → very large
2π
ω = → very small
T
= ∆ω
1 ∆ω
so =
T 2π

Now, we can express x(t) as:

∞  T /2

∆ω
x (t ) = ∑  ∫ x(t )e −ir∆ωt dt e ir∆ωt (11)
− ∞  2π −T / 2
144424443
Cr

Now we assume that T → ∞, ∆ω = dω and r∆ω → ω . This allows a continuous

description with ω , rather than in terms of discrete multiples or harmonics of ω 1 . The
summation now becomes an integration and x(t) is now described by a Fourier
Integral:

∞  dω  iωt
∫  2π ∫
− iω t
x (t ) =  x ( t )e dt e (12)
− ∞ 
∞
1
∫ A(ω )e
iω t
x (t ) = dω (13)
2π −∞

where

∫ x(t )e
− iω t
A(ω ) = dt (14)
−∞

x (t ) and A(ω ) are referred to as a Fourier Transform Pair

Recall that we have introduced R(τ ) and G(τ ) as a Fourier Transform Pair and
R(τ ) and S(f ) as a Cosine Transform Pair

Page 3 of 3