Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Handout 4: Course Notes Were Prepared by Dr. R.M.A.P. Rajatheva and Revised by Dr. Poompat Saengudomlert

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

AT77.

02 Signals, Systems and Stochastic Processes


Asian Institute of Technology
Handout 41
Thu 26 Aug 2010

1.2.10 The Unit Impulse


The integrals of the Fourier transforms for constant, unit step, and sinusoidal signals do
not converge. Therefore, we cannot calculate the Fourier transforms for these signals.
However, the Fourier transforms for these signals can be written in terms of the unit
impulse.
The unit impulse or Dirac delta function can be defined as a function denoted by δ(t)
such that Z ∞
v(t)δ(t − t0 )dt = v(t0 )
−∞

for any given signal v(t). The above property is also referred to as the sifting property.
Although the unit impulse does not exist physically, it can be described as a limiting
form of different unit-area pulses as follows.
 2
 aeπ(at) Gaussian pulse
δ(t) = lim arect(at) rectangular pulse
a→∞ 
asinc(at) sinc pulse

Since the duration of δ(t) is infinitesimally small, the pulse shape is not important;
only the integration of the pulse matters. In fact, the delta function can only have physical
meanings when it appears under integration.

Properties of the Unit Impulse


1. Scaling:
1
δ(at) = δ(t)
|a|
2. Time reversal: The unit impulse is an even function, i.e.

δ(−t) = δ(t).

3. Fourier transform: Z ∞
F{δ(t)} = δ(t)ej2πf t dt = 1,
−∞

where the last equality follows from the sifting property. From the duality of the
Fourier transform,
F{1} = δ(−f ) = δ(f ).
In summary,
δ(t) ↔ 1
1 ↔ δ(f )
1
Course notes were prepared by Dr. R.M.A.P. Rajatheva and revised by Dr. Poompat Saengudomlert.

1
4. Multiplication: The sifting property can also be expressed as

x(t)δ(t − t0 ) = x(t0 )δ(t − t0 ).

5. Convolution:
Z ∞
x(t) ∗ δ(t − t0 ) = x(τ )δ(t − t0 − τ )dτ = x(t − t0 ),
−∞

where the last equality follows form the evenness of δ(t) and the sifting property.

6. Relation to the unit step: The unit step function is defined as2

 0, t<0
u(t) = 1/2, t = 0

1, t>0

The unit impulse is related to u(t) as follows.

d
δ(t) = u(t)
Z dt
t
δ(τ )dτ = u(t)
−∞

We can also evaluate the Fourier transform of the unit impulse as a transform in the
limit. For example, consider the Gaussian pulse in the limit τ → 0.
 
1 −π(t/τ )2
δ(t) = lim e
τ →0 τ
| {z }
v(t)
2
no
F{δ(t)} = lim e−π(f τ ) = 1
τ →0
| {z }
V (f )

We emphasize the following properties:

• The Fourier transform of the unit impulse has a constant value.

• The unit impulse contains all frequency components in equal proportions.

• The infinitely narrow time-domain signal has an infinitely wide spectrum.

Fourier Transforms Involving the Unit Impulse


Starting from the definition of δ(t) and treating δ(t) as other functions whose Fourier
transforms exist, we can derive the following results.

1. Time-domain impulse:
Aδ(t − t0 ) ↔ Ae−j2πf t0
2
In the literature, the value u(0) is sometimes not specified.

2
2. Constant:
A ↔ Aδ(f )
3. Sinusoidal functions:
1 1
cos(2πfc t) ↔ δ(f − fc ) + δ(f + fc )
2 2
1 1
sin(2πfc t) ↔ δ(f − fc ) − δ(f − fc )
2j 2j

4. Unit step:
1 1
u(t) ↔ + δ(f )
j2πf 2

5. Signum function: The signum function is defined as3



 −1, t < 0
sgn(t) = 0, t=0

1, t>0

From sgn(t) = 2u(t) − 1 and duality, we can write


1
sgn(t) ↔
jπf
1
↔ −jsgn(f )
πt

Example 1.9 (Rectangular pulse with DC offset): Consider v(t) = Arect(t/τ ) −


A/2, as illustrated in figure 1.18. The Fourier transform is computed as follows
    
t A
V (f ) = F Arect −F
τ 2
A
= Aτ sinc(f τ ) − δ(f )
2
and is illustrated in figure 1.18. 

Example 1.10 (Another look at rectangular pulse): Using the unit impulse, we
can express the rectangular pulse v(t) = Arect(t/τ ) as the integration of the following
signal  τ  τ
Aδ t + − Aδ t − .
2 2
Figure 1.19 illustrates this relationship. Using the integration property of the Fourier
transform, we can write
  n 
1 d 1 τ  τ o
V (f ) = F v(t) = F Aδ t + − Aδ t −
j2πf dt j2πf 2 2
A  A
= ej2πf τ /2 − e−j2πf τ /2 = · 2j sin(πf τ )
j2πf j2πf
= Aτ sinc(f τ ).

3
As with u(t), the value sgn(0) is sometimes not specified in the literature.

3
0.8 1.2
A=1,τ=1
0.6 1

0.8
0.4
0.6
0.2
0.4

V(f)
v(t)

0
0.2
-0.2
0
-0.4
-0.2

-0.6 -0.4

-0.8 -0.6
-2-1.5-1-0.5 0 0.5 1 1.5 2 -6 -4 -2 0 2 4 6
t f
Figure 1.18: Rectangular pulse with DC offset for zero mean.

Figure 1.19: Rectangular pulse as integration of unit impulses.

1.2.11 Fourier Transform of Periodic Signals


Consider a periodic signal that can be expressed as a Fourier series

X
v(t) = Vn ej2πnf0 t .
n=−∞

Its Fourier transform is computed as follows.


( ∞ ) ∞ ∞
X X  X
V (f ) = F Vn ej2πnf0 t = Vn F ej2πnf0 t = Vn δ(f − nf0 ).
n=−∞ n=−∞ n=−∞

In summary,

X ∞
X
j2πnf0 t
Vn e ↔ Vn δ(f − nf0 ).
n=−∞ n=−∞

4
COMMENTS:
1. Periodicity in the time domain results in impulses in the spectrum, i.e. frequency
domain. From duality, periodicity in the frequency domain results in impulses in
the time domain.
2. A line spectrum of a Fourier series can be transformed to a continuous spectrum
of a Fourier transform by converting the spectral lines into impulses whose weights
are equal to the line heights. The phase portion of the line spectrum is taken care
of by letting the impulse weights be complex numbers.
3. Through the use of the unit impulse and its Fourier transform, we now have a
unified framework to analyze both periodic and nonperiodic signals.

Example 1.11 (Fourier transform of rectangular pulse train): Consider again a


rectangular pulse train
∞  
X t − nT0
v(t) = Arect ,
n=−∞
τ
From its Fourier series coefficients
A
sin(πnf0 τ ) = Af0 τ sinc(nf0 τ )
πn
with f0 = 1/T0 , we can write

X
V (f ) = Af0 τ sinc(nf0 τ )δ(f − nf0 ).
−∞


Another Look at Fourier Tranforms of Periodic Signals


We can express a periodic signal v(t) with period T0 = 1/f0 in terms of its generating
function g(t) such that

X
v(t) = g(t − mT0 ),
m=−∞

where g(t) is defined as 


v(t), |t| ≤ T0 /2
g(t) =
0, otherwise
We can write the Fourier series coefficient as
Z T0 /2 Z ∞
1 −j2πnf0 t 1 1
Vn = v(t)e dt = g(t)e−j2πnf0 t dt = G(nf0 ),
T0 −T0 /2 T0 −∞ T0
where G(nf0 ) is the Fourier transform of g(t) evaluated at f = nf0 . Hence, in terms of
G(f ), the Fourier transform of v(t) is
∞ ∞
X 1 X
v(t) = g(t − mT0 ) ↔ V (f ) = G(nf0 )δ(f − nf0 )
m=−∞
T0 n=∞

5
Example 1.12 (Triangular pulse train): Consider a triangular pulse train

X
v(t) = g(t − mT0 ),
m=−∞

where g(t) is illustrated in figure 1.20.

Figure 1.20: Triangular pulse train.

As illustrated in figure 1.20, we can view g(t) as a result of integrating d2 g(t)/dt2


twice, where
d2 A 2A A
g(t) = δ(t + τ ) − δ(t) + δ(t − τ ).
dt2 τ τ τ
It follows that
 2 
d A 2A A −j2πf τ 2A 4A 2
F 2
g(t) = ej2πf τ − + e =− (1 − cos(2πf τ )) = − sin (πf τ ).
dt τ τ τ τ τ

Using the differentiation property, i.e. d2 g(t)/dt2 ↔ (j2πf )2 G(f ), we can write

1 4A 2 A
G(f ) = − sin (πf τ ) = sin2 (πf τ ) = Aτ sinc2 (f τ ).
(j2πf )2 τ τ π2f 2

Since Vn = G(nf0 )/T0 , we can write



1 X
V (f ) = G(nf0 )δ(f − nf0 )
T0 n=−∞

Aτ X
= sinc2 (nf0 τ )δ(f − nf0 ).
T0 n=−∞

The Fourier transform of the traingular pulse train is illustrated in figure 1.21.

6
6
A=1,T0=1,τ=1/5
5

3
V(f)

0
1/T0 1/τ
-1
-20 -15 -10 -5 0 5 10 15 20
f
Figure 1.21: Fourier transform of a triangular pulse train.

Example 1.13 (Fourier transform of impulse train): Consider again an impulse


train with period T0 = 1/f0 , i.e.

X
v(t) = δ(t − mT0 ).
m=−∞

The generating function g(t) in this case is g(t) = δ(t). Hence, G(f ) = 1, yielding

∞ ∞
X 1 X
δ(t − mT0 ) ↔ δ(f − nf0 )
m=−∞
T0 n=−∞

Thus, the Fourier transform of a train of impulses uniformly spaced apart by T0 is another
train of impulses weighted by 1/T0 and uniformly spaced spart by 1/T0 in the frequency
domain.
Note that, for T0 = 1, the impulse train is its own transform. 

You might also like