AT77.

02 Signals, Systems and Stochastic Processes

Asian Institute of Technology
Handout 31
Mon 23 Aug 2010

1.2.8 Rayleigh’s Energy Theorem

Analogous to the Parseval’s power theorem, we can express the energy of the signal from
its frequency-domain representation. In particular, given that v(t) ↔ V (f ),
∫ ∞ ∫ ∞
E= |v(t)| dt =
2
v(t)v ∗ (t)dt
∫−∞
∞ (∫ ∞ −∞ )
∗ −j2πf t
= v(t) V (f )e df dt
−∞ −∞
| {z }
=v ∗ (t)
∫ ∞ (∫ ∞ )
−j2πf t
= v(t)e dt V ∗ (f )df
−∞ −∞
| {z }
=V (f )
∫ ∞ ∫ ∞
∗
= V (f )V (f )df = |V (f )|2 df.
−∞ −∞

In summary, the Rayleigh’s energy theorem states that

∫ ∞ ∫ ∞
E= |v(t)| dt =
2
|V (f )|2 df
−∞ −∞

1.2.9 Properties of the Fourier Transform

1. Superposition (linearity): If v1 (t) ↔ V1 (f ), v2 (t) ↔ V2 (f ), and a1 , a2 ∈ C, then

a1 v1 (t) + a2 v2 (t) ↔ a1 V1 (f ) + a2 V2 (f )

Proof:2
∫ ∞
F{a1 v1 (t) + a2 v2 (t)} = (a1 v1 (t) + a2 v2 (t))e−j2πf t dt
−∞
∫ ∞ ∫ ∞
−j2πf t
= a1 v1 (t)e dt + a2 v2 (t)e−j2πf t dt
−∞ −∞
= a1 V1 (f ) + a2 V2 (f )

2. Time shifting: If v(t) ↔ V (f ) and t0 ∈ R, then

v(t − t0 ) = V (f )e−j2πf t0
1
Course notes were prepared by Dr. R.M.A.P. Rajatheva and revised by Dr. Poompat Saengudomlert
and Dr. Attaphongse Taparugssanagorn.
2
Let F{·} denote the Fourier transform operation.

1
 Proof: Using the change of variable τ = t − t0 ,
∫ ∞
F{v(t − t0 )} = v(t − t0 )e−j2πf t dt
∫−∞
∞
= v(τ )e−j2πf (τ +t0 ) dτ
(−∞
∫ ∞ )
−j2πf τ
= v(τ )e dτ e−j2πf t0 .
| −∞ {z }
=V (f )

3. Time scaling: If v(t) ↔ V (f ) and a ∈ R with a ̸= 0, then

( )
1 f
v(at) = V
|a| a

Proof: Using the change of variable τ = at,

{ ∫ (f )
1 ∞ −j2π fa τ 1
v(τ )e dτ = V ,
a ( )
a>0
F{v(at)} = a −∞∫
1 ∞ −j2π fa τ
a
− a −∞ v(τ )e dτ = − a V a , a < 0
1 f

NOTE: This property reflects the reciprocal spreading between the time domain
and the frequency domain.

4. Duality: If v(t) ↔ V (f ), then

V (t) ↔ v(−f )
∫∞
Proof: From v(t) = −∞
V (f )ej2πf t dt, we write
∫ ∞
v(−t) = V (f )e−j2πf t dt.
−∞

Interchanging the variables t and f yields

∫ ∞
v(−f ) = V (t)e−j2πf t dt = F{V (t)}.
−∞

Example 1.6 : Recall that v(t) = Arect(t/τ ) ↔ V (f ) = Aτ sinc(f τ ). Consider now

w(t) = A0 sinc(2W t). By setting A = A0 /2W and τ = 2W , we can write
A0
w(t) = 2W sinc(2W t) = Aτ sinc(tτ ) = V (t).
2W
It follows that ( ) ( )
A0 −f A0 f
W (f ) = rect = rect ,
2W 2W 2W 2W
where the last equality follows from the even symmetry of the rectangular function. In
summary,

2
 1.2 1.2
A0=1,2W=1
1 A0 A0/2W
1

V(f)=A0/2W× rect(f/2W)
0.8
0.8
v(t)=A0sinc(2Wt)

0.6
0.6
0.4
0.4
0.2
0.2
0

-0.2 0
-W W
-0.4
1/2W -0.2
-6 -4 -2 0 2 4 6 -1.5 -1 -0.5 0 0.5 1 1.5
t f
Figure 1.15: Sinc pulse and its Fourier transform.

( )
A0 f
A0 sinc(2W t) ↔ rect
2W 2W

which is illustrated in figure 1.15. 

5. Complex modulation: If v(t) ↔ V (f ) and fc > 0, then

v(t)ej2πfc t ↔ V (f − fc )

Proof:
∫ ∞ ∫ ∞
j2πfc t −j2πf t
F{v(t)e j2πfc t
}= v(t)e e dt = v(t)e−j2π(f −fc )t dt = V (f − fc )
−∞ −∞

6. Modulation: If v(t) ↔ V (f ) and fc ∈ R, then

1 1
v(t) cos(2πfc t) ↔ V (f − fc ) + V (f + fc )
2 2

Proof:
{ }
1 1
F{v(t) cos(2πfc t)} = F v(t)ej2πfc t + v(t)e−j2πfc t
2 2
1 1
↔ V (f − fc ) + V (f + fc )
2 2

3
 1.5 1.2
A=1,τ=2,fc=5
1 Aτ/2
1 A
0.8
0.5
0.6

V(f)
v(t)

0 0.4
-τ/2 τ/2
0.2
-0.5
0
-1
-0.2
-fc fc
-1.5 -0.4
-2-1.5-1-0.5 0 0.5 1 1.5 2 -20-15-10 -5 0 5 10 15 20
t f
Figure 1.16: Fourier transform of a modulated signal.

Example 1.7 : Using the modulation property

( )
t 1 1
Arect cos(2πfc t) ↔ Aτ sinc((f − fc )τ ) + Aτ sinc((f + fc )τ ),
τ 2 2

which is illustrated in figure 1.16. 

7. Differentiation: If v(t) ↔ V (f ), then

dn
v(t) ↔ (j2πf )n V (f )
dtn

Proof:3 We begin with the first derivative dv(t)/dt.

(∫ ∞ ) ∫ ∞
d d j2πf t d ( j2πf t )
v(t) = V (f )e df = V (f ) e df
dt dt −∞ −∞ dt
∫ ∞
= (j2πf V (f ))ej2πf t df = F −1 {j2πf V (f )}
−∞

Equivalently, { }
d
F v(t) = j2πf V (f ).
dt
dn
Iterating the above argument n times yields the desired property, i.e. dtn
v(t) ↔
(j2πf )n V (f ).

3
Let F −1 {·} denote the inverse Fourier transform operation.

4
 8. Integration: If v(t) ↔ V (f ) and V (0) = 0, then
∫ t
1
v(τ )dτ ↔
V (f )
−∞ j2πf
(∫ )
t
Proof: From the relationship v(t) = dtd −∞ v(τ )dτ and the differentiation prop-
erty,
{ (∫ t )} {∫ t }
d
F{v(t)} = F v(τ )dτ = j2πf · F v(τ )dτ .
dt −∞ −∞

If V (0) is nonzero, then the Fourier transform of the integral of v(t) includes an
impulse function (to be discussed more later), i.e.
∫ t
1 V (0)
v(τ )dτ ↔ V (f ) + δ(f ).
−∞ j2πf 2

9. Convolution: If v1 (t) ↔ V1 (f ) and v2 (t) ↔ V2 (f ), then4

v1 (t) ∗ v2 (t) ↔ V1 (f )V2 (f )

Proof: We first write

{∫ ∞ }
F {v1 (t) ∗ v2 (t)} = F v1 (τ )v2 (t − τ )dτ
−∞
∫ ∞∫ ∞
= v1 (τ )v2 (t − τ )e−j2πf t dτ dt.
−∞ −∞

Using the change of variable η = t − τ to substitute for t,

∫ ∞∫ ∞
F {v1 (t) ∗ v2 (t)} = v1 (τ )v2 (η)e−j2πf (η+τ ) dτ dη
−∞ −∞
(∫ ∞ ) (∫ ∞ )
−j2πf τ −j2πf η
= v1 (τ )e dτ v2 (η)e dη = V1 (f )V2 (f ).
−∞ −∞
| {z }| {z }
=V1 (f ) =V2 (f )

Signals with Even/Odd Symmetry

Note that we can write
∫ ∞
V (f ) = v(t)e−j2πf t dt
(−∞
∫ ∞ ) ( ∫ ∞ )
= v(t) cos(2πf t)dt +j − v(t) sin(2πf t)dt ,
−∞ −∞
| {z } | {z }
=Ve (f ) =Vo (f )

where we define
∫ ∞ ∫ ∞
Ve (f ) = v(t) cos(2πf t)dt and Vo (f ) = − v(t) sin(2πf t)dt.
−∞ −∞

The following observations can be made.

∫ ∞ The notation “v1 (t) ∫∗ ∞ v2 (t)” denotes the convolution between v1 (t) and v2 (t), where v1 (t) ∗ v2 (t) =
4

−∞
v 1 (τ )v 2 (t − τ )dτ = −∞ 1
v (t − τ )v2 (τ )dτ .

5
 • If v(t) is even, i.e. v(−t) = v(t), then Vo (f ) = 0, yielding
∫ ∞
V (f ) = Ve (f ) = 2 v(t) cos(2πf t)dt.
0

• If v(t) is odd, i.e. v(−t) = −v(t), then Ve (f ) = 0, yielding

∫ ∞
V (f ) = jVo (f ) = −2j v(t) sin(2πf t)dt.
0

• If v(t) is real and even, then V (f ) is real and even.

• If v(t) is real and odd, then V (f ) is purely imaginary and odd.

Additional Properties of Convolution Operations

1. Commutativity: v1 (t) ∗ v2 (t) = v2 (t) ∗ v1 (t)

2. Distributivity: v1 (t) ∗ [v2 (t) + v3 (t)] = v1 (t) ∗ v2 (t) + v1 (t) ∗ v3 (t)

3. Associativity: [v1 (t) ∗ v2 (t)] ∗ v3 (t) = v1 (t) ∗ [v2 (t) ∗ v3 (t)]

The proofs are simple in the frequency domain as follows.

V1 (f )V2 (f ) = V2 (f )V1 (f )
V1 (f ) [V2 (f ) + V3 (f )] = V1 (f )V2 (f ) + V1 (f )V3 (f )
[V1 (f )V2 (f )] V3 (f ) = V1 (f ) [V2 (f )V3 (f )]

Review of the Convolution Integral

The convolution integral
∫ ∞
v1 (t) ∗ v2 (t) = v1 (τ )v2 (t − τ )dτ
−∞

is computed according to the following steps.

1. Replace t by τ to get v1 (τ ) and v2 (τ ).

2. Reflect v2 (τ ) around the origin to get v2 (−τ ).

3. For t ≥ 0, shift v2 (−τ ) to the right by t to form v2 (t − τ ). For t < 0, shift v2 (−τ )
to the left by −t to form v2 (t − τ ).

4. Multiply v1 (τ ) by v2 (t − τ ) and take the area under the resulting curve as the value
of the convolution integral at time t.

5. The complete result is obtained by repeating steps 3-4 for all t ∈ R. As t increases,
v2 (t − τ ) slides from left to right with respect to v1 (τ ).

6
 0 0

Figure 1.17: Convolution of two rectangular pulses.

Example 1.8 : Consider computing v(t) = v1 (t) ∗ v2 (t) where

( ) ( )
t t
v1 (t) = A1 rect and v2 (t) = A2 rect
τ1 τ2
with τ1 > τ2 .
The signals v1 (t), v2 (t), and v1 (t) ∗ v2 (t) are illustrated in figure 1.17.
There are five ranges of t to consider, as discussed below.
1. t ≤ −( τ21 + τ22 ) (no overlap): In this case, v1 (τ ) and v2 (t − τ ) do not overlap. Hence,
v(t) = 0.
2. −( τ21 + τ22 ) ≤ t ≤ −( τ21 − τ22 ) (partial overlap): In this case, v1 (τ ) and v2 (t − τ )
overlap partially. Accordingly,
∫ t+τ2 /2 ( τ1 τ2 )
v(t) = A1 A2 dτ = A1 A2 t + + .
−τ1 /2 2 2

3. −( τ21 − τ22 ) ≤ t ≤ ( τ21 − τ2

2
) (full overlap): In this case, v1 (τ ) and v2 (t − τ ) overlap
fully. Accordingly,
∫ t+τ2 /2
v(t) = A1 A2 dτ = A1 A2 τ2 .
t−τ2 /2

4. ( τ21 − τ22 ) ≤ t ≤ ( τ21 + τ22 ) (partial overlap): In this case, v1 (τ ) and v2 (t − τ ) overlap
partially. Accordingly,
∫ τ1 /2 ( τ2 τ1 )
v(t) = A1 A2 dτ = A1 A2 −t + + .
t−τ2 /2 2 2

5. t > ( τ21 + τ2
2
) (no overlap): In this case, v1 (τ ) and v2 (t − τ ) do not overlap. Hence,
v(t) = 0.
Finally, the Fourier transform of v(t) is the product of the Fourier transforms of the
two rectangular pulses, i.e.
V (f ) = A1 τ1 sinc(f τ1 ) · A2 τ2 sinc(f τ2 ).