2.2 Continuous-Time LTI Systems: The Convolution Integral
2.2 Continuous-Time LTI Systems: The Convolution Integral
2.2 Continuous-Time LTI Systems: The Convolution Integral
The response of a continuous-time LTI system can be computed by convolution of the impulse
response of the system with the input signal, using a convolution integral, rather than a sum.
∞
x(t ) = ∫ x(τ )δ (t − τ )dτ . (2.7)
−∞
The result is obtained by chopping up the signal x(t ) in sections of width ∆ , and taking sum
x (t )
t
−∆ 0 ∆ 2∆ 3∆
Recall the definition of the unit pulse δ ∆ (t ) ; we can define a signal xˆ(t ) as a linear combination
of delayed pulses of height x( k∆ )
∞
xˆ(t ) = ∑ x(k∆)δ
k = −∞
∆ (t − k∆) ∆ (2.8)
Taking the limit as ∆ → 0 , we obtain the integral of Eq. (2.7), in which when ∆ → 0
Eq. (2.7) can also be obtained by using the sampling property of the impulse function. If we
consider t is fixed and τ is time variable, then we have x(τ )δ (t − τ )
= x (τ )δ (−(τ − t )) = x(t )δ (τ − t ) . Hence
7/2 Yao
ELG 3120 Signals and Systems Chapter 2
∞ ∞ ∞
∫−∞
x(τ )δ (t − τ )dτ = ∫ x(τ )δ (τ − t )dτ = x (t )∫ δ (τ − t )dτ = x(t ) .
−∞ −∞
(2.9)
2.2.2 Continuous-Time Unit Impulse Response and the Convolution Integral Representation
of an LTI system
The linearity property of an LTI system allows us to calculate the system response to an input
signal xˆ(t ) using Superposition Principle. Let hˆk∆ (t ) be the pulse response of the linear-varying
system to the unit pulses δ ∆ (t − k∆) for − ∞ < k < +∞ . The response of the system to xˆ(t ) is
∞
yˆ (t ) = ∑ x( k∆ ) h
k = −∞
k∆ (t − k∆ )∆ . (2.10)
Note that the response hˆk∆ (t ) tends to the impulse response hτ (t ) as ∆ → 0 . Then at the limit,
we obtain the response of the system to the input signal x(t ) = lim xˆ (t) :
∆ →0
+∞
y (t ) = lim yˆ (t ) =
∆ →0 ∫
−∞
x (τ )hτ (t )dτ . (2.11)
For an LTI system, the impulse responses hτ (t ) are the same as h0 (t ) , except they are shifted by
τ , that is, hτ (t ) = h0 (t − k ) . Then we may define the unit impulse response of the LTI system
h(t ) = h0 (t ) , (2.12)
So the response to the input signal x(t ) can be written as a convolution integral:
+∞
y (t ) = ∫ x (τ )h(t − τ ) dτ , (2.13)
−∞
y (t ) = x (t ) ∗ h(t ) . (2.14)
8/2 Yao
ELG 3120 Signals and Systems Chapter 2
The output y(t ) is a weighted integral of the input, where the weight on x(τ ) is h(t − τ ) . To
evaluate this integral for a specific value of t ,
• First obtain the signal h(t − τ ) (regarded as a function of τ with t fixed) from h(τ ) by a
reflection about the origin and a shift to the right by t if t >0 or a shift to the left by t is t <0.
• Then multiply together the signals x(τ ) and h(t − τ ) .
• y(t ) is obtained by integrating the resulting product from τ = −∞ to τ = +∞ .
Example: Let x(t ) be the input to an LTI system with unit impulse response h(t ) , where
h (τ )
τ
0
x (τ )
τ
0
h( t − τ )
1 t<0
τ
t 0
9/2 Yao
ELG 3120 Signals and Systems Chapter 2
h( t − τ )
1
t> 0
τ
0 t
Step 2: From the figure we can see that for t < 0 , the product of the product x(τ ) and h(t − τ ) is
zero, and consequently, y(t ) is zero. For t > 0
e − at , t0><0τ < t
x(τ )h (t − τ ) =
0, otherwise
1 1
y (t ) = ∫ e −aτ dτ = − e− aτ 0 = (1 − e −at ) .
t t
0 a a
1
y (t ) = (1 − e −at )u (t) , and is shown in figure below.
a
y (t )
1
a
t
0
For this example, it is convenient to calculate the convolution in separate intervals. x(τ ) is
sketched and h(t − τ ) is sketched in each of the intervals:
10/2 Yao
ELG 3120 Signals and Systems Chapter 2
For t < 0 , and t > 3T , x(τ )h (t −τ ) = 0 for all the values of τ , and consequently y(t ) =0.
For other intervals, the product x(τ )h (t − τ ) can be found in the figure on the next page. Thus for
these three intervals, the integration can be calculated with the result shown below:
11/2 Yao
ELG 3120 Signals and Systems Chapter 2
x (τ )
τ
0 T
h (t − τ )
2T
t< 0
τ
t − 2T t 0
h (t − τ ) x(τ ) h ( t − τ )
2T 2T
h (t − τ ) x(τ ) h ( t − τ )
2T 2T
τ
t − 2T 0 t 0 T
h (t − τ ) x(τ ) h ( t − τ )
2T 2T
t−T
2T < t < 3T 2 T < t < 3T
τ
0 t − 2T t 0 T
h (t − τ )
2T
t > 3T
τ
0 t − 2T t
12/2 Yao
ELG 3120 Signals and Systems Chapter 2
0, t<0 y (t )
1 2
t , 0<t <T
2
1 2
y (t ) = Tt − T , T < t < 2T
2
1 3
− t + Tt + T 2 ,
2
2T < t < 3T
2 2
t
0, t > 3T
0 T 2T 3T
LTI systems can be characterized completely by their impulse response. The properties can also
be characterized by their impulse response.
∞ ∞
x[n ] ∗ h[n] = h[n ] ∗ x[n] = ∑ x[k ]h[n − k ] = ∑ h[k ]k[n − k ] ,
k = −∞ k = −∞
(2.15)
∞ ∞
x(t ) ∗ h(t) = h(t ) ∗ x(t ) = ∫ x(τ )h(t − τ )dτ = ∫ h (τ ) x(t − τ )dτ . (2.16)
−∞ −∞
x h y h x y
x ∗ (h1 + h2 ) = x ∗ h1 + x ∗ h2 (2.17)
for both discrete-time and continuous-time systems. The property means that summing the
outputs of two systems is equivalent to a system with an impulse response equal to the sum of
the impulse response of the two individual systems, as shown in the figure below.
13/2 Yao
ELG 3120 Signals and Systems Chapter 2
h1
x + y
h2
x h 1 +h 2 y
For example, an LTI system has an impulse response h[n ] = u[n ] , with an input
n
1
x[n ] = u[ n] + 2 n u[ −n] . Since the sequence x[n] is nonzero along the entire time axis. Direct
2
evaluation of such a convolution is somewhat tedious. Instead, we may use the distributive
property to express y[n] as the sum of the results of two simpler convolution problems. That is,
n
1
x1 [n] = u[n] , x 2 [n] = 2 n u[− n] , using the distributive property we have
2
x ∗ (h1 ∗ h2 ) = (x ∗ h1 ) ∗ h2 . (2.18)
x h1 h2 y
x h 1*h 2 y
• For LTI systems, the change of order of the cascaded systems will not affect the response.
14/2 Yao
ELG 3120 Signals and Systems Chapter 2
• For nonlinear systems, the order of cascaded systems in general cannot be changed. For
example, a two memoryless systems, one being multiplication by 2 and the other squaring the
input, the outputs are different if the order is changed, as shown in the figure below.
w=2x
x 2 w2 y=4x 2
2
w=x 2
x x 2 y=2x 2
A system is memoryless if its output at any time depends only on the value of its input at the
same time. This is true for a discrete-time system, if h[n ] = 0 for n ≠ 0 . In this case, the impulse
response has the form
h[n ] = Kδ [n ] , (2.19)
where K = h[0] is a constant and the convolution sum reduces to the relation
y[ n] = Kx[ n] . (2.20)
h(t ) = Kδ (t ) , (2.21)
y (t ) = Kx(t ) . (2.22)
Note that if K = 1 in Eqs. (2.19) and (2.21), the systems become identity systems, with output
equal to the input.
We have seen that a system S is invertible if and only if there exists an inverse system S-1 such
that S -1S is an identity system.
x h h1 y=x
15/2 Yao
ELG 3120 Signals and Systems Chapter 2
Since the overall impulse response in the figure above is h ∗ h1 , h1 must satisfy for it to be the
impulse response of the inverse system, namely h ∗ h1 = δ .
identity
x y=x
system
The impulse response of this system is h(t ) = δ (t − t0 ) , since x(t − t 0 ) = x(t ) ∗ δ (t − t 0 ) , that is,
convolution of a signal with a shifted impulse simply shifts the signal
To recover the signal from the output, that is, to invert the system, all that is required is to shift
the output back. So the inverse system should have a impulse response of δ (t + t0 ) , then
δ (t − t0 ) ∗ δ (t + t0 ) = δ (t )
Example: Consider the LTI system with impulse response h[n ] = u[n ] .
+∞
y[ n] = ∑ x[k ]u[n − k ] .
−∞
n
y[ n] = ∑ x[k ] .
−∞
This is a system that calculates the running sum of all the values of the input up to the present
time, and is called a summer or accumulator. This system is invertible, and its inverse is given as
h1[n] = δ [ n] − δ [n − 1] ,
16/2 Yao
ELG 3120 Signals and Systems Chapter 2
We may check that the two systems are really inverses to each other:
A system is causal if its output depends only on the past and present values of the input signal.
Specifically, for a discrete-time LTI system, this requirement is y[n] should not depend on x[k ]
for k > n . Based on the convolution sum equation, all the coefficients h[n − k ] that multiply
values of x[k ] for k > n must be zero, which means that the impulse response of a causal
discrete-time LTI system should satisfy the condition
h[n ] = 0 , for n < 0 (2.23)
A causal system is causal if its impulse response is zero for negative time; this makes sense as
the system should not have a response before impulse is applied.
Examples: The accumulator h[n ] = u[n ] , and its inverse h[n ] = δ [n] − δ [n − 1] are causal. The
pure time shift with impulse response y (t ) = x(t − t0 ) for t 0 > 0 is causal, but is not causal
for t 0 < 0 .
Recall that a system is stable if every bounded input produces a bounded output.
If this input signal is applied to an LTI system with unit impulse response h[n] , the magnitude of
the output
+∞ +∞ +∞
y[n ] = ∑ h[ k ]x[ n − k ] ≤ ∑ h[k ] x[n − k ] ≤ B ∑ h[k ]
k = −∞ k = −∞ k = −∞
(2.25)
17/2 Yao
ELG 3120 Signals and Systems Chapter 2
+∞
∑ h[k ] < ∞ .
k = −∞
(2.26)
The similar analysis applies to continuous-time LTI systems, for which the stability is equivalent
to
+∞
∫−∞
h (τ ) dτ < ∞ . (2.27)
Example: consider a system that is pure time shift in either continuous time or discrete time.
+∞ +∞
In discrete time, ∑ h[k ] = ∑ δ [ n − n
k = −∞ k = −∞
0 = 1,
+∞ +∞
while in continuous time, ∫
−∞
h(τ ) dτ = ∫
−∞
δ ( t − t 0 ) dτ = 1 ,
+∞ +∞
Example: The accumulator h[n ] = u[n ] is unstable because ∑ h[k ] = ∑ u[n] = ∞ .
k = −∞ k =0
The step response of an LTI system is simply the response of the system to a unit step. It conveys
a lot of information about the system. For a discrete-time system with impulse response h[n] , the
step response is s[n] = u[n ] ∗ h[n] . However, based on the commutative property of convolution,
s[n] = h[n ] ∗ u[n] , and therefore, s[n] can be viewed as the response to input h[n] of a discrete-
time LTI system with unit impulse response. We know that u[n] is the unit impulse response of
the accumulator. Therefore,
n
s[n] = ∑ h[k ] .
k = −∞
(2.28)
From this equation, h[n] can be recovered from s[n] using the relation
It can be seen the step response of a discrete-time LTI system is the running sum of its impulse
response. Conversely, the impulse response of a discrete-time LTI system is the first difference
of its step response.
18/2 Yao