Lecture 24 Continuous-Time Signals
and LTI Systems 2
Fundamentals of Digital Signal Processing
Spring, 2012
Wei-Ta Chu
2012/5/25
1 DSP, CSIE, CCU
Time-Invariance
Suppose that a continuous-time system is represented
by
This system is time-invariant if, when we delay the
input signal by an arbitrary amount t0, the result is only
to delay the output by that same amount.
This condition must be true for any signal x(t) and for
any real number t0.
Test of time-invariance for
a discrete-time system
2 DSP, CSIE, CCU
Squaring System is Time-Invariant
When the input is x(t), the output is y(t) = [x(t)]2
If the input is x(t-t0), the corresponding output will be
3 DSP, CSIE, CCU
Integrator is Time-Invariant
Changing the “dummy variable” of integration
to .
is replaced by , the upper limit becomes
4 DSP, CSIE, CCU
A Time-Varying System
Amplitude modulator
Such a system is a fundamental component of many
radio systems.
It’s a time-varying system
5 DSP, CSIE, CCU
Linearity
A continuous-time system is linear if when
and , then
Principle of superposition
Test of linearity for a
discrete-time system
6 DSP, CSIE, CCU
Squaring is Nonlinear
The corresponding output is
7 DSP, CSIE, CCU
Integrator is Linear
The corresponding output is
The integrator system is
linear
8 DSP, CSIE, CCU
The Convolution Integral
Every LTI system can be described by
a convolution integral
This operation is usually written as ,
meaning x(t) is convolved with h(t).
A system is LTI if and only if its output can be
represented as a convolution.
9 DSP, CSIE, CCU
The Convolution Integral
Assume that
The corresponding output is
Thus, the operation of convolution of an input x(t) with
h(t) is a linear operation.
10 DSP, CSIE, CCU
The Convolution Integral
Make a substitution
Therefore, convolution is time-invariant as well as
linear.
For every LTI system, the output y(t) is always equal to
, the convolution of the input signal x(t) with
the system impulse response h(t).
11 DSP, CSIE, CCU
Convolve Unit Steps
When the impulse response is a unit step, h(t) = u(t), and
the input is also a unit step, x(t) = u(t), the convolution
integral becomes
The upper limit becomes t because when
, or
is called a unit ramp because it is linearly increasing
with a slope of one.
12 DSP, CSIE, CCU
Properties of Convolution
The operation of convolution is commutative,
associative, and distributive over addition.
Commutativity:
Recall that reversing the sign of a definite intergal
reverses the order of the limits
13 DSP, CSIE, CCU
Properties of Convolution
Associativity:
Distributivity Over Addition:
Convolution is a linear operation. Therefore, convolution
of x(t) with the sum must be the sum of the
individual convolutions.
Identity Element of Convolution:
When will the following equation works?
The answer is the unit-impulse signal by substituting
14 DSP, CSIE, CCU
Properties of Convolution
The impulse is the identity signal for the operation of
convolution, much as 1 is the identity element for
ordinary multiplication.
15 DSP, CSIE, CCU
Impulse Responses – Integrator
Integrator
The superscript (-1) means the first anti-derivative, i.e.,
integral
The impulse response of this system is the unit step as
shown previously in Equation (9.21).
Thus we can represent the integrator operationally as
follows:
A running integral is equivalent to
convolution with a unit step.
16 DSP, CSIE, CCU
Impulse Responses – Differentiator
It’s an LTI system.
Substitute for x(t)
is the impulse response of the differentiator
system, the following operational definition must be
true:
is called the doublet.
Differentiation is convolution with
derivative of an impulse.
17 DSP, CSIE, CCU
Example: Convolution with Doublet
The convolution of the unit step with the doublet
can be evaluated by using (9.47)
The derivative property of the unit step (9.22); i.e.,
18 DSP, CSIE, CCU
Impulse Responses – Ideal Delay
Substituting gives the impulse response of
the ideal delay system as
The impulse response of the ideal delay is a delayed
impulse
Time shift is the same as
convolution with a shifted impulse.
19 DSP, CSIE, CCU
Convolution of Impulses
Convolution of impulses
The convolution of two shifted impulses at t1 and t2
gives a shifted impulse located at the sum
20 DSP, CSIE, CCU
Convolution of Impulses
When combined with the linearity property, we can
work problems like:
So we end up with impulses at
The area of the impulses at is -3.
21 DSP, CSIE, CCU
Impulse Convolution Causes Shifting
Consider the pulse input x(t) and the impulse response
The equation for the output is simply
We simply shift a scaled copy of
the continuous signal to the
location of each impulse and
sum all the shifted and scaled
copies.
22 DSP, CSIE, CCU
Evaluating Convolution Integrals
Delayed Unit-Step Input
Suppose we wish to evaluate a convolution integral
is what we call the “dummy variable” of integration
in the integral because it disappears when we evaluate
at the upper and lower limits.
23 DSP, CSIE, CCU
Evaluating Convolution Integrals
t is the independent variable of y(t)
To compute each value of the function y(t), we must
form the product and then evaluate the
integral for each different value of t.
Make substitution
This form does not make it obvious how the limits of
integration depend on t.
24 DSP, CSIE, CCU
Evaluating Convolution Integrals
Key idea: draw an auxiliary sketch of the two
functions whose product is the integrand of the
convolution integral.
The sketch of is the same as
For the sketch of
The desired is
obtained by first time-reversing
and then shifting the result
by t.
25 DSP, CSIE, CCU
Evaluating Convolution Integrals
Show how and interact in evaluating the
convolution integral.
There are two different regions
On the left, we see a “typical” plot that is representative of
all values of t such that t – 1 < 0.
The nonzero parts of the two
functions do not overlap and
hence, their product is zero
for t < 1.
26 DSP, CSIE, CCU
Evaluating Convolution Integrals
On the right, this plot is representative for all values of
t such that t -1>0. Since the flipped and shifted
extends infinitely to the left, the nonzero parts of
and will overlap for all t such that t > 1.
The output for t>1 is given by
27 DSP, CSIE, CCU
Evaluating Convolution Integrals
This is an example of the more general case where
for and for . (In this case T0 = 1
and T1 = 0.)
If t < T0 + T1, then the nonzero parts of and
do not overlap. It is always true that y(t) = 0 for t < T0
+ T1 .
Even if one or both of the signals involved in the
convolution is discontinuous, the result of the
convolution will be a continuous function of time.
28 DSP, CSIE, CCU
Exercise 9.4
29 DSP, CSIE, CCU
